**Fuzzy Image Processing, Analysis and Visualization Methods for Hydro-Dams and Hydro-Sites Surveillance and Monitoring**

Gordan Mihaela1, Dancea Ovidiu1, Cislariu Mihaela1, Stoian Ioan2 and Vlaicu Aurel1 *1Technical University of Cluj-Napoca, 2S.C. IPA S.A. CIFATT Cluj, Romania* 

### **1. Introduction**

The continuous surveillance, monitoring and operational planning of hydro-dams and hydro-sites is a very important issue, considering the impact of these critical structures on the environment, society, economy and ecology. On one hand, the failure of hydro-dams can dramatically affect the environment and humans; on the other hand, the operating policies must take into account the impact of the water resource exploitation on the hydro-site region and on the regions supplied by the reservoir.

The importance of periodic surveillance and monitoring through both objective measurements and subjective observations is emphasized by existing international standards, which provide the main surveillance and monitoring guidelines for hydro-dams and hydro-sites (CSED, 1983; DSC, 2010). Among other issues, these guidelines clearly state that the visual inspection of the hydro-dams and their surroundings is an important component of the surveillance process, as it aids the decision making process based on direct observations (CSED, 1983, pp. 21-28). Visual inspections complement the other type of data acquired from sensors and transducers placed within the dam body and its surroundings. It is a common practice in hydro-dam surveillance to store the visual observations by human observers in the form of visual observations records. Typically these records regard the state of the reservoir, banks and slopes, concrete structure and downstream valley, and are backed-up by digital image archives of the inspected structures (CSED, 1983; Bradlow *et al.*, 2002).

In respect to the water resource exploitation policy related to the hydro-sites, it is important to develop tools for water resource management evaluation and planning. However these should not be fully automated decision systems, but rather decision support components, to assist the human specialists in establishing the best operation policy. According to the EU Water Framework Directive (2000/60/EC), the water management plan must take into account the natural geographical and hydrological unit rather than the administrative or political boundaries (European Parliament, 2000). This assumes a thorough analysis of the

Fuzzy Image Processing, Analysis and

temperature loss, or poor isolations in buildings.

images of dam walls.

Visualization Methods for Hydro-Dams and Hydro-Sites Surveillance and Monitoring 5

affecting the structure of the dam. The main interest in processing was to create a 3-D dam map, to be further investigated by the human operator, and even in this step, human intervention is often required. Taking into account the wide variety of computer vision algorithms currently available, is fair to consider that the automation of the visual inspection process of the dam, aiming to detect, diagnose and predict possible faults, can be further increased. Some of the methods presented in this chapter provide solutions to perform specific image processing and analysis tasks in the particular case of infrared and visible

Bimodal analysis of optical and infrared images is a problem still needed to be tackled with. Few such applications have been reported, mainly in the fields of surveillance, people counting and tracking, robust skin detection (face detection), forest fires detection, or land mines detection (Ollero *et al*., 1998; O'Conaire *et. al*., 2006). However, for the diagnosis of dams such works are scarce, although infrared imaging is used extensively in assessing

Thermal images can provide information about the scene being scanned which is not available from a visual image. Although much work has been performed for finding various image segmentation techniques in both imaging modalities, little efforts have been made for integration of complementary information extracted from the two imaging modalities.

**2.2 The role of artificial intelligence techniques in hydro-sites operation monitoring**  The significant development of the information systems puts nowadays its fingerprint on the hydro-sites surveillance and monitoring as well, with a strong emphasize on the design and implementation of intelligent systems to assist the specialists in the above mentioned areas. The artificial intelligence methods play a significant role in the development of systems devoted to dam surveillance and dam monitoring, especially in the form of decision support components and knowledge-based expert systems; among these methods, the wellknown fuzzy theory and machine learning solutions (especially neural networks) are commonly employed. Some examples of such artificial intelligence based solutions for hydro-dams and hydro-sites surveillance, monitoring and assessments are briefly mentioned herein. Knowledge-based systems have been employed to assist the diagnosis of seepage from different types of hydro-dams (Asgian *et al*., 1988; Sieh *et al*., 1998). Neural networks are also employed in the investigation of seepage under concrete dams founded on rock (Ohnishi & Soliman, 1995) or in the estimation of the dam permeability (Najjar *et al*., 1996). The joint use of fuzzy mathematics and neural networks is also reported by (Wen *et al*., 2004), in the development of a bionics model of dam safety monitoring composed of integration control, inference engine, database, model base, graphics base, and input/output modules. Fuzzy logic and artificial neural networks were employed in the inference models

building stage, needed to analyze and evaluate the run characteristics of dams.

**2.3 Overview of the integrated hydro-site surveillance and monitoring system** 

Artificial intelligence techniques (including fuzzy logic, fuzzy knowledge based systems, neural networks and other supervised classifiers) have been extensively employed recently in hydro-dam and hydro-sites surveillance applications, as diagnostic tools and policy

associated complex and heterogeneous data, to perform both the analysis of the current resource management policy and to predict the impact of some management policy on the environment, economy and society. Such a complex task is best performed by a computer decision support system, considering the amount and diversity of the required data/information to be processed. However since the decision on the best water resource management policy to be adopted is to be made by specialists, it is important to provide the decision support system with a human-compliant interface, both for introducing the input information and for displaying the assessment and prediction results in a meaningful and intuitive form to the end-user (this includes, besides numerical data, linguistic and qualitative assessments and, of course, a visual description of the results and recommendations, wherever this is possible). While adopting some existing fuzzy reasoning strategies for the evaluation of the water resource management policy, we mainly emphasize here on our contribution in the enhancement of the results presentation form – particularly on the visual presentation of the future effect of some particular policy, as a geotypically textured map of the region, using image processing methods to transpose the numerical and qualitative assessment results into a suggestive visual representation.

Most of the solutions presented in this chapter were integrated in a hydro-dam and hydrosite surveillance system, devoted to the monitoring of the Tarnita hydro-site on the Somes River in Transilvania County, Romania. The details of the fuzzy image processing and analysis tools proposed are presented in the remaining of this chapter.

### **2. Problem formulation**

Prior to the introduction of the proposed fuzzy image processing and analysis methods suitable to the visual examination of the concrete hydro-dams surface condition and to the visual rendering of the water resource management policy assessment in a hydro-site region, we consider necessary to give a description of the addressed problems. This should allow the reader to understand and acknowledge the fact that image processing methods may indeed play an important role in the assessment and evaluation of hydro-dams and hydro-sites, although this type of strategy is not so commonly encountered in the field. The following three subsections briefly point the roles of image processing and analysis methods, the role of artificial intelligence approaches and finally present the structure of the system we designed for hydro-dams/hydro-sites monitoring and surveillance, with an emphasize on the role of visual surveillance. Some of the significant references in the scientific literature related to the subject are also outlined.

### **2.1 The role of image processing and analysis methods in hydro-dams surveillance**

In order to enhance the visual observations made by human experts, computer vision techniques may be employed. The approach is to acquire images and then, by the means of specific image processing algorithms, enhance and analyse them. Also, the periodical recording of these images into a database could prove very useful when monitoring the overall condition of the dam walls during time.

Less interest was oriented on incorporating image processing and analysis algorithms to automatically detect, diagnose and predict the behaviour of the dam and the possible faults

associated complex and heterogeneous data, to perform both the analysis of the current resource management policy and to predict the impact of some management policy on the environment, economy and society. Such a complex task is best performed by a computer decision support system, considering the amount and diversity of the required data/information to be processed. However since the decision on the best water resource management policy to be adopted is to be made by specialists, it is important to provide the decision support system with a human-compliant interface, both for introducing the input information and for displaying the assessment and prediction results in a meaningful and intuitive form to the end-user (this includes, besides numerical data, linguistic and qualitative assessments and, of course, a visual description of the results and recommendations, wherever this is possible). While adopting some existing fuzzy reasoning strategies for the evaluation of the water resource management policy, we mainly emphasize here on our contribution in the enhancement of the results presentation form – particularly on the visual presentation of the future effect of some particular policy, as a geotypically textured map of the region, using image processing methods to transpose the

numerical and qualitative assessment results into a suggestive visual representation.

analysis tools proposed are presented in the remaining of this chapter.

scientific literature related to the subject are also outlined.

overall condition of the dam walls during time.

**2. Problem formulation** 

Most of the solutions presented in this chapter were integrated in a hydro-dam and hydrosite surveillance system, devoted to the monitoring of the Tarnita hydro-site on the Somes River in Transilvania County, Romania. The details of the fuzzy image processing and

Prior to the introduction of the proposed fuzzy image processing and analysis methods suitable to the visual examination of the concrete hydro-dams surface condition and to the visual rendering of the water resource management policy assessment in a hydro-site region, we consider necessary to give a description of the addressed problems. This should allow the reader to understand and acknowledge the fact that image processing methods may indeed play an important role in the assessment and evaluation of hydro-dams and hydro-sites, although this type of strategy is not so commonly encountered in the field. The following three subsections briefly point the roles of image processing and analysis methods, the role of artificial intelligence approaches and finally present the structure of the system we designed for hydro-dams/hydro-sites monitoring and surveillance, with an emphasize on the role of visual surveillance. Some of the significant references in the

**2.1 The role of image processing and analysis methods in hydro-dams surveillance**  In order to enhance the visual observations made by human experts, computer vision techniques may be employed. The approach is to acquire images and then, by the means of specific image processing algorithms, enhance and analyse them. Also, the periodical recording of these images into a database could prove very useful when monitoring the

Less interest was oriented on incorporating image processing and analysis algorithms to automatically detect, diagnose and predict the behaviour of the dam and the possible faults affecting the structure of the dam. The main interest in processing was to create a 3-D dam map, to be further investigated by the human operator, and even in this step, human intervention is often required. Taking into account the wide variety of computer vision algorithms currently available, is fair to consider that the automation of the visual inspection process of the dam, aiming to detect, diagnose and predict possible faults, can be further increased. Some of the methods presented in this chapter provide solutions to perform specific image processing and analysis tasks in the particular case of infrared and visible images of dam walls.

Bimodal analysis of optical and infrared images is a problem still needed to be tackled with. Few such applications have been reported, mainly in the fields of surveillance, people counting and tracking, robust skin detection (face detection), forest fires detection, or land mines detection (Ollero *et al*., 1998; O'Conaire *et. al*., 2006). However, for the diagnosis of dams such works are scarce, although infrared imaging is used extensively in assessing temperature loss, or poor isolations in buildings.

Thermal images can provide information about the scene being scanned which is not available from a visual image. Although much work has been performed for finding various image segmentation techniques in both imaging modalities, little efforts have been made for integration of complementary information extracted from the two imaging modalities.

### **2.2 The role of artificial intelligence techniques in hydro-sites operation monitoring**

The significant development of the information systems puts nowadays its fingerprint on the hydro-sites surveillance and monitoring as well, with a strong emphasize on the design and implementation of intelligent systems to assist the specialists in the above mentioned areas. The artificial intelligence methods play a significant role in the development of systems devoted to dam surveillance and dam monitoring, especially in the form of decision support components and knowledge-based expert systems; among these methods, the wellknown fuzzy theory and machine learning solutions (especially neural networks) are commonly employed. Some examples of such artificial intelligence based solutions for hydro-dams and hydro-sites surveillance, monitoring and assessments are briefly mentioned herein. Knowledge-based systems have been employed to assist the diagnosis of seepage from different types of hydro-dams (Asgian *et al*., 1988; Sieh *et al*., 1998). Neural networks are also employed in the investigation of seepage under concrete dams founded on rock (Ohnishi & Soliman, 1995) or in the estimation of the dam permeability (Najjar *et al*., 1996). The joint use of fuzzy mathematics and neural networks is also reported by (Wen *et al*., 2004), in the development of a bionics model of dam safety monitoring composed of integration control, inference engine, database, model base, graphics base, and input/output modules. Fuzzy logic and artificial neural networks were employed in the inference models building stage, needed to analyze and evaluate the run characteristics of dams.

### **2.3 Overview of the integrated hydro-site surveillance and monitoring system**

Artificial intelligence techniques (including fuzzy logic, fuzzy knowledge based systems, neural networks and other supervised classifiers) have been extensively employed recently in hydro-dam and hydro-sites surveillance applications, as diagnostic tools and policy

Fuzzy Image Processing, Analysis and

cracks, and so on.

humidity of the concrete dam walls.

natural presentation to the end user.

compared to 91% with the classical fuzzy c-means).

Visualization Methods for Hydro-Dams and Hydro-Sites Surveillance and Monitoring 7

information representation is often encountered in management evaluation systems. Our contribution in terms of a visually enhanced representation of the water resource

**3. Downstream concrete surface evaluation of hydro-dams by image analysis**  Visual inspection is a key element in dam monitoring process, allowing decisions to be made about dam behavior, based on direct observations. Visual inspections complement the data analysis process concerning different sensors and transducers placed within the dam body and it's surroundings, and the observations are filled in a standardized form describing the inspections results about: reservoir, banks and slopes, concrete structure, downstream valley. These records hold, for every feature observed, the procedures utilized during inspection as well as significant images illustrating the observations. Hence, once digital images of the inspected structure are available, a series of aspects are suitable for image analysis: detection and quantification of calcite deposits, detection of areas with humidity, evaluation of concrete surface of the wall in order to reveal structure faults or

It is a known fact that most cracks in dam walls have calcite exuding from them, indicating that moisture traversed the cracks (Abare, 2006). As water seeps through cracks, it leaves calcite deposits at the surface adjacent to the cracks. If the area between concrete layer is porous, the movement of water through them would accelerate the leaching action. Seepage samples may be collected, analyzed and compared to reservoir water to help determine whether soluble minerals pose a structural safety problem (Craft *et al*., 2007). Seepage could be estimated by estimating the volume of water required to precipitate the measured volumes of calcite in the unsaturated zone (Marshall *et al*., 2003). Besides these techniques, we will show that computer vision can also help detect and assess the calcite deposits and

The deterioration of the concrete walls may also be an important concern as it may indicate the degradation of the downstream side, and to give an estimate of this type of degradation we proposed a solution to examine the surface roughness (Gordan *et al*., 2008). Besides an accurate identification of such deteriorations, we show that computer vision techniques help in providing a quantitative and qualitative description of the extent of the deterioration. It is important to note that all the results of the proposed computer vision techniques can easily be transcribed to the visual observation record and offer the advantage of an intuitive and

In terms of downstream concrete surface evaluation of dams, we propose the following:

1. A modified fuzzy c-means segmentation method (semi-supervised through the use of support vector regression) for the detection, localization and quantification of calcite areas in the plots of the downstream concrete surface of a hydro-dam. The difficulty of this image segmentation problem comes from the large variability of calcite deposits appearance, uneven distribution of data, variations of the concrete appearance depending on the acquisition conditions and devices. The proposed solution outperforms the classical segmentation algorithms in terms of accuracy (96% as

management policy assessment results is also presented in the end of this chapter.

recommendation tools. However, most existing solutions use measurements acquired from different sensors, and very few of them integrate visual observations obtained from some image analysis modules applied on digital images acquired during hydro-dams and hydrosites monitoring. In this respect, we describe here a set of image analysis tools developed specifically for the concrete hydro-dams surveillance, which were implemented in the form of an integrated computer vision-based hydro-dam analysis system, capable of providing quantitative, qualitative and linguistic assessments of the concrete surface. The presented visual inspections and expert system components are part of a large hydro-dam and hydrosite surveillance system devoted to the monitoring of the Tarnita hydro-site on the Somes River in Transilvania County, Romania; its block diagram is illustrated in Fig. 1.

Fig. 1. The integrated system for dam safety decision support, using computer vision techniques and integrating the result of image analysis with the results of other data analysis modules

The automatic acquisition equipments collect multi-sensorial data from the sensors placed in the dam body. These equipments are: automatic acquisition station, capacitive sensor telependulum, optical tele-pendulum, tele-limnimeter, laser telemeter, infrared and visible spectra cameras. All these data are stored into a relational multimodal database. The data fusion algorithms are used to extract relevant information regarding water infiltrations in the dam body, based on infrared and visible spectrum image fusion. Other image processing algorithms are applied to dam wall surface roughness examination, which is also likely to be caused by systematic water infiltrations. The dam models are used for dam behavior prediction and utilize the information stored in the database. The expert system use the human expert knowledge in specific domains and metadata resulted upon their own inference. The decision support system links the user with modeling components, image analysis and fusion modules, expert systems, and the database. Its role is to provide synthetic data in graphical, numerical and linguistic format, which would help the dam surveillance personnel in taking the right decisions regarding interventional measures that will prevent dam degradation and will ensure its functioning in good conditions. Another useful component that may be integrated in the system from Fig. 1 is the one devoted to water resource management policy evaluation and prediction in the hydro-site and the surrounding areas. Most commonly, such components are built using fuzzy rule base systems/fuzzy logic systems, as this mathematical framework is very suitable to handle both exact and approximate (qualitative or linguistic) knowledge, and this mixture of

recommendation tools. However, most existing solutions use measurements acquired from different sensors, and very few of them integrate visual observations obtained from some image analysis modules applied on digital images acquired during hydro-dams and hydrosites monitoring. In this respect, we describe here a set of image analysis tools developed specifically for the concrete hydro-dams surveillance, which were implemented in the form of an integrated computer vision-based hydro-dam analysis system, capable of providing quantitative, qualitative and linguistic assessments of the concrete surface. The presented visual inspections and expert system components are part of a large hydro-dam and hydrosite surveillance system devoted to the monitoring of the Tarnita hydro-site on the Somes

River in Transilvania County, Romania; its block diagram is illustrated in Fig. 1.

Fig. 1. The integrated system for dam safety decision support, using computer vision techniques and integrating the result of image analysis with the results of other data

The automatic acquisition equipments collect multi-sensorial data from the sensors placed in the dam body. These equipments are: automatic acquisition station, capacitive sensor telependulum, optical tele-pendulum, tele-limnimeter, laser telemeter, infrared and visible spectra cameras. All these data are stored into a relational multimodal database. The data fusion algorithms are used to extract relevant information regarding water infiltrations in the dam body, based on infrared and visible spectrum image fusion. Other image processing algorithms are applied to dam wall surface roughness examination, which is also likely to be caused by systematic water infiltrations. The dam models are used for dam behavior prediction and utilize the information stored in the database. The expert system use the human expert knowledge in specific domains and metadata resulted upon their own inference. The decision support system links the user with modeling components, image analysis and fusion modules, expert systems, and the database. Its role is to provide synthetic data in graphical, numerical and linguistic format, which would help the dam surveillance personnel in taking the right decisions regarding interventional measures that will prevent dam degradation and will ensure its functioning in good conditions. Another useful component that may be integrated in the system from Fig. 1 is the one devoted to water resource management policy evaluation and prediction in the hydro-site and the surrounding areas. Most commonly, such components are built using fuzzy rule base systems/fuzzy logic systems, as this mathematical framework is very suitable to handle both exact and approximate (qualitative or linguistic) knowledge, and this mixture of

analysis modules

information representation is often encountered in management evaluation systems. Our contribution in terms of a visually enhanced representation of the water resource management policy assessment results is also presented in the end of this chapter.

### **3. Downstream concrete surface evaluation of hydro-dams by image analysis**

Visual inspection is a key element in dam monitoring process, allowing decisions to be made about dam behavior, based on direct observations. Visual inspections complement the data analysis process concerning different sensors and transducers placed within the dam body and it's surroundings, and the observations are filled in a standardized form describing the inspections results about: reservoir, banks and slopes, concrete structure, downstream valley. These records hold, for every feature observed, the procedures utilized during inspection as well as significant images illustrating the observations. Hence, once digital images of the inspected structure are available, a series of aspects are suitable for image analysis: detection and quantification of calcite deposits, detection of areas with humidity, evaluation of concrete surface of the wall in order to reveal structure faults or cracks, and so on.

It is a known fact that most cracks in dam walls have calcite exuding from them, indicating that moisture traversed the cracks (Abare, 2006). As water seeps through cracks, it leaves calcite deposits at the surface adjacent to the cracks. If the area between concrete layer is porous, the movement of water through them would accelerate the leaching action. Seepage samples may be collected, analyzed and compared to reservoir water to help determine whether soluble minerals pose a structural safety problem (Craft *et al*., 2007). Seepage could be estimated by estimating the volume of water required to precipitate the measured volumes of calcite in the unsaturated zone (Marshall *et al*., 2003). Besides these techniques, we will show that computer vision can also help detect and assess the calcite deposits and humidity of the concrete dam walls.

The deterioration of the concrete walls may also be an important concern as it may indicate the degradation of the downstream side, and to give an estimate of this type of degradation we proposed a solution to examine the surface roughness (Gordan *et al*., 2008). Besides an accurate identification of such deteriorations, we show that computer vision techniques help in providing a quantitative and qualitative description of the extent of the deterioration. It is important to note that all the results of the proposed computer vision techniques can easily be transcribed to the visual observation record and offer the advantage of an intuitive and natural presentation to the end user.

In terms of downstream concrete surface evaluation of dams, we propose the following:

1. A modified fuzzy c-means segmentation method (semi-supervised through the use of support vector regression) for the detection, localization and quantification of calcite areas in the plots of the downstream concrete surface of a hydro-dam. The difficulty of this image segmentation problem comes from the large variability of calcite deposits appearance, uneven distribution of data, variations of the concrete appearance depending on the acquisition conditions and devices. The proposed solution outperforms the classical segmentation algorithms in terms of accuracy (96% as compared to 91% with the classical fuzzy c-means).

Fuzzy Image Processing, Analysis and

by a vector <sup>T</sup>

clustering results to identify if any represents calcite or not.

Visualization Methods for Hydro-Dams and Hydro-Sites Surveillance and Monitoring 9

significant variability of the calcite appearance makes almost impossible the derivation of a calcite appearance model to be used in the identification; model-free approaches seem more suitable, trying to identify natural pixels clusters, followed by an interpretation of the

A rather powerful approach to non-supervised image segmentation by pixel clustering is the fuzzy c-means algorithm (FCM) (Dunn, 1973; Bezdek, 1981). Many variations of the FCM algorithm were successfully applied in image segmentation. Actually, various forms of fuzzy clustering have been employed to different image segmentation tasks. In (Chamorro *et al*., 2003), the segmentation of color images is achieved by a nested hierarchy of fuzzy partitions, based on a measure of color similarity. Starting from an initial fuzzy segmentation, a hierarchical approach, based on a similarity relation between regions, is employed to obtain a nested hierarchy of regions at different precision levels. Type 2 fuzzy sets are employed in (Clairet *et al*., 2006), for color images segmentation, to allow a better modeling of the uncertainty. A modified fuzzy c-means segmentation scheme with spatial constraints is introduced in (Hafiane *et al*., 2005), in the form of a two step segmentation method. Another fuzzy clustering method, with no constraints on the number of clusters, aiming to segment an image in homogeneous regions, is presented in (Das *et al*., 2006).

The solution that we proposed for the segmentation of the calcite deposits on the concrete hydro-dam walls images is more application-targeted (and it worth noting that it may be also generalized to other application-specific segmentation tasks, as it provides a framework to incorporate a-priori knowledge in the fuzzy c-means cost function). Details on this approach may also be found in (Dancea *et al*., 2010). The calcite identification on the concrete dam wall can be treated as a pixel classification problem. As there is no prior knowledge regarding the shape of the calcite deposits, the spatial constraints are not really helpful in the segmentation; the colors of the pixels are the only relevant features to consider. An important fact to consider however is the amount of the calcite deposits on each dam wall image, which is significantly smaller than the entire wall region. If we build a data set to be clustered comprising all the pixels in the currently analyzed image, classified into calcite and non-calcite samples, this set will be highly unbalanced among the classes of interest, and this is an unfavorable situation in a classification task, being prone to more errors in the poor represented class. This situation can be partially overcame by defining the classification data as the set of distinct colors in the concrete dam wall image, each color being included only once. From the several possible color spaces, we prefer the natural Red Green Blue (RGB) representation, as it is just as suitable as others for Euclidian distance based classifiers; thus each sample (corresponding to a color from the image) is represented

**x** RGB . The image to be segmented is considered to be a sub-plot

image of a dam wall, as shown in Fig. 2. Therefore the current data set is formed by the colors in this sub-plot color image, X i 1,2,...,N , C C **x**<sup>i</sup> where NC denotes the number of distinct colors in the current image. Our goal is to classify/cluster the data in XC in one of two possible classes of interest: calcite deposit – denoted by CC , and not calcite – denoted here by Cc . Although this is actually nothing else but a binary classification problem, trying to solve it by an unsupervised fuzzy c-means clustering of the data in only two classes will risk to be unable to group all the colors corresponding to the class "anything else

2. Furthermore, since less severe infiltrations may only be visible in the infrared spectrum, we also propose an integration of infrared image analysis with the visible image analysis, using a late decision fusion to integrate the results of the two image analysis modules. The fusion is thought to take into account the spatial and temporal correlation of the two types of images of the same hydro-dam downstream surface. This approach should yield more reliable results in terms of infiltration assessment.

These algorithms and techniques are described in detail in the following sub-sections. The images to be processed are drawn from the multimodal database, which holds digital images of concrete dam walls. Such an image is illustrated in Fig. 2. These are cropped to elementary units, called sub-plots. Each sub-plot image is identified by information that allows later identification and association with the real scene (the identification data is: horizontal, vertical and plot number). Thus, it is easier to extract images from the same subplot taken at different dates or in different modalities (e.g. visible or infrared spectrum).

Fig. 2. Image sample at the input of the visual inspection module

### **3.1 Infiltration assessment by the analysis of calcite deposits using fuzzy segmentation**

Calcite patches are good indicators of significant and time persistent water infiltrations; they are most likely to occur as being transported by the water infiltrations from concrete in the case of a repetitive water infiltration in a certain area of the dam. Therefore the problem of identifying the calcite formations on the concrete wall through an algorithm able to provide maximum accuracy despite the variability of appearance of calcite deposits, the variable lighting conditions on the portion of the wall, without knowing in advance if calcite is or is not present in the current image, or in what amount, must be tackled. These aspects make the calcite identification and assessment a rather difficult image analysis problem: the

2. Furthermore, since less severe infiltrations may only be visible in the infrared spectrum, we also propose an integration of infrared image analysis with the visible image analysis, using a late decision fusion to integrate the results of the two image analysis modules. The fusion is thought to take into account the spatial and temporal correlation of the two types of images of the same hydro-dam downstream surface. This approach

These algorithms and techniques are described in detail in the following sub-sections. The images to be processed are drawn from the multimodal database, which holds digital images of concrete dam walls. Such an image is illustrated in Fig. 2. These are cropped to elementary units, called sub-plots. Each sub-plot image is identified by information that allows later identification and association with the real scene (the identification data is: horizontal, vertical and plot number). Thus, it is easier to extract images from the same subplot taken at different dates or in different modalities (e.g. visible or infrared spectrum).

should yield more reliable results in terms of infiltration assessment.

Fig. 2. Image sample at the input of the visual inspection module

**segmentation** 

**3.1 Infiltration assessment by the analysis of calcite deposits using fuzzy** 

Calcite patches are good indicators of significant and time persistent water infiltrations; they are most likely to occur as being transported by the water infiltrations from concrete in the case of a repetitive water infiltration in a certain area of the dam. Therefore the problem of identifying the calcite formations on the concrete wall through an algorithm able to provide maximum accuracy despite the variability of appearance of calcite deposits, the variable lighting conditions on the portion of the wall, without knowing in advance if calcite is or is not present in the current image, or in what amount, must be tackled. These aspects make the calcite identification and assessment a rather difficult image analysis problem: the significant variability of the calcite appearance makes almost impossible the derivation of a calcite appearance model to be used in the identification; model-free approaches seem more suitable, trying to identify natural pixels clusters, followed by an interpretation of the clustering results to identify if any represents calcite or not.

A rather powerful approach to non-supervised image segmentation by pixel clustering is the fuzzy c-means algorithm (FCM) (Dunn, 1973; Bezdek, 1981). Many variations of the FCM algorithm were successfully applied in image segmentation. Actually, various forms of fuzzy clustering have been employed to different image segmentation tasks. In (Chamorro *et al*., 2003), the segmentation of color images is achieved by a nested hierarchy of fuzzy partitions, based on a measure of color similarity. Starting from an initial fuzzy segmentation, a hierarchical approach, based on a similarity relation between regions, is employed to obtain a nested hierarchy of regions at different precision levels. Type 2 fuzzy sets are employed in (Clairet *et al*., 2006), for color images segmentation, to allow a better modeling of the uncertainty. A modified fuzzy c-means segmentation scheme with spatial constraints is introduced in (Hafiane *et al*., 2005), in the form of a two step segmentation method. Another fuzzy clustering method, with no constraints on the number of clusters, aiming to segment an image in homogeneous regions, is presented in (Das *et al*., 2006).

The solution that we proposed for the segmentation of the calcite deposits on the concrete hydro-dam walls images is more application-targeted (and it worth noting that it may be also generalized to other application-specific segmentation tasks, as it provides a framework to incorporate a-priori knowledge in the fuzzy c-means cost function). Details on this approach may also be found in (Dancea *et al*., 2010). The calcite identification on the concrete dam wall can be treated as a pixel classification problem. As there is no prior knowledge regarding the shape of the calcite deposits, the spatial constraints are not really helpful in the segmentation; the colors of the pixels are the only relevant features to consider. An important fact to consider however is the amount of the calcite deposits on each dam wall image, which is significantly smaller than the entire wall region. If we build a data set to be clustered comprising all the pixels in the currently analyzed image, classified into calcite and non-calcite samples, this set will be highly unbalanced among the classes of interest, and this is an unfavorable situation in a classification task, being prone to more errors in the poor represented class. This situation can be partially overcame by defining the classification data as the set of distinct colors in the concrete dam wall image, each color being included only once. From the several possible color spaces, we prefer the natural Red Green Blue (RGB) representation, as it is just as suitable as others for Euclidian distance based classifiers; thus each sample (corresponding to a color from the image) is represented by a vector <sup>T</sup> **x** RGB . The image to be segmented is considered to be a sub-plot image of a dam wall, as shown in Fig. 2. Therefore the current data set is formed by the colors in this sub-plot color image, X i 1,2,...,N , C C **x**<sup>i</sup> where NC denotes the number of distinct colors in the current image. Our goal is to classify/cluster the data in XC in one of two possible classes of interest: calcite deposit – denoted by CC , and not calcite – denoted here by Cc . Although this is actually nothing else but a binary classification problem, trying to solve it by an unsupervised fuzzy c-means clustering of the data in only two classes will risk to be unable to group all the colors corresponding to the class "anything else

Fuzzy Image Processing, Analysis and

NC

 

uji i i 1 ; <sup>j</sup> NC

**x**

**<sup>v</sup>**

either U or V is under a certain tolerance (error) (in theory, arbitrarily small).

 

sample and the cube of the sample's standard deviation:

uji i 1

membership degrees uji:

Visualization Methods for Hydro-Dams and Hydro-Sites Surveillance and Monitoring 11

whose minimization is done iteratively, as in the standard fuzzy c-means algorithm, using the following equations for the computation of the fuzzy class centers vj and for the fuzzy

<sup>C</sup> wd , j ij uji <sup>2</sup> l 1 wd , l l <sup>i</sup>

In the expressions above, V is the set of the class centers, V={v1,..,vC}*,* <sup>3</sup> vj ; m is a parameter controlling the shape of the resulting clusters (typically m=2); d(·,·) is a distance norm in the RGB space between any two vectors. A common choice for d, used in our approach as well, is the Euclidian distance. The iterative process ends when the change in

The three weights w1, w2 and w3 are estimated roughly using the shape of the histogram of the brightness component of the segmented image; the shape descriptor which proves useful for our case is the skew of the histogram, as it provides a numerical measure of the distribution of the samples to the left and right of their mean. Using solely the brightness and not the color is sufficient for our goal, as our concern is to be able to "differentiate" the light-most class (which accounts for calcite as explained above) from the other two classes. Therefore we give default fixed weights to the non-calcite classes and tune just the calcite class weight as indicated by the histogram's skew. Considering an N sample set formed by the brightness values of the pixels in the currently analyzed sub-plot, y ,y ,...,y 12 N , the sample's skew γ can be estimated as the ratio between the third central moment of the

1 N 3

For a uni-modal histogram having the gray levels are evenly distributed around the mode, the skew is close to zero. If more darker pixels than brighter pixels are present in the examined image, the skew γ will be negative. On the opposite, if the brighter pixels are dominant and outnumber the darker ones, γ will be positive. Based on these considerations, we can perform the following adjustment of the calcite class weight depending on the skew γ (assuming the other two classes have fixed weights). If γ is positive (i.e., the number of light pixels accounted for calcite is large enough), there is no need to enhance the importance of the calcite class in respect to the other two, and we can set the calcite weight equal to the other classes. If γ is negative or near zero, it indicates the areas of calcite are rather small as compared to the examined surface, so the calcite class weight should be increased. Intuitively, the more

negative γ is, the larger the weight assigned to the calcite class should be.

y y <sup>μ</sup> <sup>N</sup> <sup>i</sup> <sup>1</sup> <sup>N</sup> <sup>3</sup> i 1 <sup>γ</sup> , y y . 3 3 <sup>N</sup> <sup>i</sup> i 1 2 2 <sup>1</sup> <sup>N</sup> <sup>2</sup> <sup>μ</sup><sup>2</sup> y y <sup>N</sup> <sup>i</sup> i 1 

**x v**

**x v**

<sup>1</sup> <sup>1</sup>

(2)

(3)

2 m 1

but calcite", since their variance is too large. Therefore a larger number of classes than two only will be needed in the initial clustering, one per dominant color. An examination of the sub-plots shows that generally two dominant colors are present in the non-calcite sub-plot areas: a grayish like color corresponding to the concrete and a brown-black color corresponding to organic deposits. Thus a 3-class clustering should be performed, with two classes for the Cc dataset and one the calcite, CC .

The fuzzy c-means algorithm (Bezdek, 1981) is a very efficient clustering procedure when the number of clusters is known a-priori, aiming to find natural fuzzy groupings of the data according to their similarity in respect to a selected distance metric. In the end of an iterative objective function minimization process, the optimal class centers and membership degrees of the data to be clustered are found, with the optimality defined as the minimization of the classification uncertainty among the data in the classes. However a good clustering result is only achieved if the amount of data in each cluster is relatively balanced; otherwise the expected fuzzy centroid of the class with fewest data can be rather different than the real centroid of the class. This is mainly due to the fact that although the distance between the data and the resulting class center is large (leading to a large cost in the objective function), if the number of these terms is negligible in comparison to the size of the data set, it will contribute insignificantly to the total cost. While we already tried to avoid this case by taking all the colors in the sub-plot only once, this caution might still not be enough to guarantee a balanced data set. Therefore, furthermore, we propose to apply a modified objective function in the fuzzy c-means clustering, which assigns a higher penalty to the misclassification of the expected calcite pixels colors, that is, of the lighter colors in the data set XC . We should mention here that, although the number of pixels colors corresponding to the organic deposits (brown-black, that means – dark-most) is also much smaller than of the grayish pixels, we are not concerned about their misclassification here, as in the worse case, the color of a brown-dark pixel is closer to a grayish pixel than to a calcite one, and then the misclassified data for the organic deposits can never appear in the calcite class CC .

Let us denote by C – the number of classes to which the NC samples x from the set XC are to be assigned in some membership degree; in our case, C=3. The membership degrees of the data to the classes is stored in a matrix UC NC , where the uji element, j=1,...,C and i=1,...,NC, represents the membership degree of the vector i **<sup>x</sup>** to the class j. Each line in U is the discrete representation of the fuzzy set corresponding to a data class. The C fuzzy sets are constrained to form a fuzzy partition of the data set XC. Starting from any initial fuzzy partition of the data set to be fuzzy classified XC, the algorithm aims to optimize the partition in the sense of minimizing the uncertainty regarding the membership of every data xi*,* i=1,…,NC, to each of the classes. In the proposed weighted fuzzy c-means algorithm, we introduce a set of class-specific scalar positive weights wj, j=1,…,C, to assign different relative importance to the distances of the data in XC to each of the classes centers. With these weights, we build a fuzzy c-means weighted objective function in the form:

$$\mathbf{J}\_{\mathbf{W},\mathbf{m}}(\mathbf{U},\mathbf{V}) = \sum\_{i=1}^{N\_{\mathbf{C}}} \sum\_{\mathbf{j}=1}^{C} \mathbf{u}\_{\mathbf{j}\mathbf{i}}^{\mathbf{m}} \cdot \mathbf{w}\_{\mathbf{j}} \cdot \mathbf{d}^{2} \left(\mathbf{x}\_{\mathbf{i}},\mathbf{v}\_{\mathbf{j}}\right) \tag{1}$$

but calcite", since their variance is too large. Therefore a larger number of classes than two only will be needed in the initial clustering, one per dominant color. An examination of the sub-plots shows that generally two dominant colors are present in the non-calcite sub-plot areas: a grayish like color corresponding to the concrete and a brown-black color corresponding to organic deposits. Thus a 3-class clustering should be performed, with two

The fuzzy c-means algorithm (Bezdek, 1981) is a very efficient clustering procedure when the number of clusters is known a-priori, aiming to find natural fuzzy groupings of the data according to their similarity in respect to a selected distance metric. In the end of an iterative objective function minimization process, the optimal class centers and membership degrees of the data to be clustered are found, with the optimality defined as the minimization of the classification uncertainty among the data in the classes. However a good clustering result is only achieved if the amount of data in each cluster is relatively balanced; otherwise the expected fuzzy centroid of the class with fewest data can be rather different than the real centroid of the class. This is mainly due to the fact that although the distance between the data and the resulting class center is large (leading to a large cost in the objective function), if the number of these terms is negligible in comparison to the size of the data set, it will contribute insignificantly to the total cost. While we already tried to avoid this case by taking all the colors in the sub-plot only once, this caution might still not be enough to guarantee a balanced data set. Therefore, furthermore, we propose to apply a modified objective function in the fuzzy c-means clustering, which assigns a higher penalty to the misclassification of the expected calcite pixels colors, that is, of the lighter colors in the data set XC . We should mention here that, although the number of pixels colors corresponding to the organic deposits (brown-black, that means – dark-most) is also much smaller than of the grayish pixels, we are not concerned about their misclassification here, as in the worse case, the color of a brown-dark pixel is closer to a grayish pixel than to a calcite one, and then the misclassified data for the organic deposits can never appear in the calcite class CC . Let us denote by C – the number of classes to which the NC samples x from the set XC are to be assigned in some membership degree; in our case, C=3. The membership degrees of the data to the classes is stored in a matrix UC NC , where the uji element, j=1,...,C and i=1,...,NC, represents the membership degree of the vector i **<sup>x</sup>** to the class j. Each line in U is the discrete representation of the fuzzy set corresponding to a data class. The C fuzzy sets are constrained to form a fuzzy partition of the data set XC. Starting from any initial fuzzy partition of the data set to be fuzzy classified XC, the algorithm aims to optimize the partition in the sense of minimizing the uncertainty regarding the membership of every data xi*,* i=1,…,NC, to each of the classes. In the proposed weighted fuzzy c-means algorithm, we introduce a set of class-specific scalar positive weights wj, j=1,…,C, to assign different relative importance to the distances of the data in XC to each of the classes centers. With

these weights, we build a fuzzy c-means weighted objective function in the form:

 NC <sup>C</sup> m 2 J U,V u w d , w,m ji <sup>j</sup> <sup>i</sup> <sup>j</sup> i 1j 1 

**x v** (1)

classes for the Cc dataset and one the calcite, CC .

whose minimization is done iteratively, as in the standard fuzzy c-means algorithm, using the following equations for the computation of the fuzzy class centers vj and for the fuzzy membership degrees uji:

$$\mathbf{v}\_{\mathbf{j}} = \frac{\sum\_{\begin{subarray}{c} \sum \mathbf{u}\_{\text{j}\mathbf{i}} \mathbf{x}\_{\mathbf{i}} \\ \sum \mathbf{u}\_{\text{j}\mathbf{i}} \end{subarray}}{\sum\_{\begin{subarray}{c} \sum \mathbf{u}\_{\text{j}\mathbf{i}} \end{subarray}}}; \mathbf{u}\_{\mathbf{j}\mathbf{i}} = \left( \frac{\mathbb{C}}{\sum\_{\begin{subarray}{c} \sum \mathbf{u}\_{\text{j}} \cdot \mathbf{d}\left(\mathbf{x}\_{\mathbf{i}}, \mathbf{v}\_{\mathbf{j}}\right)^{2} \\ \mathbf{u}\_{\mathbf{i}} \cdot \mathbf{d}\left(\mathbf{x}\_{\mathbf{i}}, \mathbf{v}\_{\mathbf{l}}\right)^{2} \end{subarray}} \right)^{-1} \tag{2}$$

In the expressions above, V is the set of the class centers, V={v1,..,vC}*,* <sup>3</sup> vj ; m is a parameter controlling the shape of the resulting clusters (typically m=2); d(·,·) is a distance norm in the RGB space between any two vectors. A common choice for d, used in our approach as well, is the Euclidian distance. The iterative process ends when the change in either U or V is under a certain tolerance (error) (in theory, arbitrarily small).

The three weights w1, w2 and w3 are estimated roughly using the shape of the histogram of the brightness component of the segmented image; the shape descriptor which proves useful for our case is the skew of the histogram, as it provides a numerical measure of the distribution of the samples to the left and right of their mean. Using solely the brightness and not the color is sufficient for our goal, as our concern is to be able to "differentiate" the light-most class (which accounts for calcite as explained above) from the other two classes. Therefore we give default fixed weights to the non-calcite classes and tune just the calcite class weight as indicated by the histogram's skew. Considering an N sample set formed by the brightness values of the pixels in the currently analyzed sub-plot, y ,y ,...,y 12 N , the sample's skew γ can be estimated as the ratio between the third central moment of the sample and the cube of the sample's standard deviation:

$$\mathbf{y} = \frac{\mu\_{\mathbf{3}}}{\frac{3}{2}} = \frac{\frac{1}{\mathbf{N}} \sum\_{\mathbf{i}=1}^{\mathbf{N}} \left(\mathbf{y}\_{\mathbf{i}} - \overline{\mathbf{y}}\right)^{3}}{\frac{1}{\mathbf{N}} \sum\_{\mathbf{i}=1}^{\mathbf{N}} \left(\mathbf{y}\_{\mathbf{i}} - \overline{\mathbf{y}}\right)^{2}}, \overline{\mathbf{y}} = \frac{1}{\mathbf{N}} \sum\_{\mathbf{i}=1}^{\mathbf{N}} \mathbf{y}\_{\mathbf{i}}.\tag{3}$$

For a uni-modal histogram having the gray levels are evenly distributed around the mode, the skew is close to zero. If more darker pixels than brighter pixels are present in the examined image, the skew γ will be negative. On the opposite, if the brighter pixels are dominant and outnumber the darker ones, γ will be positive. Based on these considerations, we can perform the following adjustment of the calcite class weight depending on the skew γ (assuming the other two classes have fixed weights). If γ is positive (i.e., the number of light pixels accounted for calcite is large enough), there is no need to enhance the importance of the calcite class in respect to the other two, and we can set the calcite weight equal to the other classes. If γ is negative or near zero, it indicates the areas of calcite are rather small as compared to the examined surface, so the calcite class weight should be increased. Intuitively, the more negative γ is, the larger the weight assigned to the calcite class should be.

Fuzzy Image Processing, Analysis and

Fig. 3. Skew to class weight mapping

c-means segmentation result.

sub-section.

Visualization Methods for Hydro-Dams and Hydro-Sites Surveillance and Monitoring 13

difference between the ground truth segmentation and the segmentation result of our algorithm allows us to assess the segmentation error, the false positives and the false negatives for the calcite class. The segmentation error for the calcite class achieved with our method, expressed as the average percentage of misclassified pixels in the test set for the total 60 sub-plots, is 4.19%, whereas for the standard fuzzy c-means algorithm is 9.64%(more

(a) (b) (c) Fig. 4. Example of the calcite segmentation for a sub-plot: (a) the original (not segmented) sub-plot image; (b) the fuzzy c-means segmentation result; (c) the proposed weighted fuzzy

Despite the good performance of this segmentation procedure in the localization of calcite, one must take into account that less severe infiltrations may not produce significant calcite deposits yet, and may only be visible in the infrared spectrum. Therefore the integration of infrared image analysis results with the visible image analysis results, using a late decision fusion, can bring more valuable information in the infiltration assessment. The fusion is thought to take into account the spatial and temporal correlation of the two types of images of the same hydro-dam downstream surface. This approach is presented in the following

than double). Some segmentation results are illustrated in Fig. 4.

Note that although there is no a-priori association of the class index j, j=1,2 or 3, and the brightness of the colors in the class, we always know that the fuzzy class with light most colors is the fuzzy class whose center is the lightest, and this class will be considered to correspond to the calcite (if any):

$$\mathbf{C\_{C}} = \mathbf{C\_{k}} \left| \mathbf{k} = \operatorname\*{argmax}\_{\mathbf{j} = \mathbf{l}, \mathbf{2}, \mathbf{3}} \text{x} \left( \begin{bmatrix} 0.299 & 0.587 & 0.114 \end{bmatrix} \cdot \mathbf{v\_{j}} \right) . \tag{4} \right|$$

To be able to effectively employ the above considerations into our algorithm, a numerical mapping between the range of values γ and the range of weights of the light-most, i.e. calcite pixels class, must be obtained. Denoting our target weight by wk, with k given by Eq. (4), we search for the mapping wk(γ) that best fits a set of training data, obtained by manually tuning the value wk on a set of statistically significant dam wall images (with enough variability in appearance, to cover as many practical cases as possible). A set of 15 images of several sub-plots, with different aspect, under different lighting conditions and different amounts of calcite (from none to very severe) have been selected and manually analyzed to optimize the calcite class' weight for an accurate calcite identification. The pairs formed by the skew values and the best manually selected weight values wk have been collected, and an interpolation procedure based on support vector regression (SVR) has been applied on this training set to completely define in an automatic fashion the computation of the weight wk. We assumed the other two classes' weights "fixed" to 1.

The reason for using SVR in the interpolation step is its proven good performance when only a relatively sparse set of data points is available. Based on Vapnik and Chervonenkis's statistical learning theory (Vapnik, 1998), support vector learning principle allows handling successfully difficult cases, with better precision and recall than other learning methods. This is mainly due to the structural risk minimization principle implemented by SVMs. SVMs were initially "built" for classification and later extended to the regression issue – SVR – by introducing a loss function (Scholkoph *et al*., 1998; Platt, 2000). Starting from an input data set, represented by a vector x, the SVM learns the functional dependency between input and output, represented in the form of a scalar-valued function f(x). The expression of the regression function provided as a result of learning by an SVM is:

$$\mathbf{f(x)} = \sum\_{\mathbf{i}=1}^{L} \mathbf{(a\_{\hat{i}} - a\_{\hat{i}}^{\*})} \mathbf{K(x, x\_{\hat{i}})},\tag{5}$$

where L denotes the total number of training data, i and \**i* are their associated Lagrange multipliers, and the function K(x,xi) represents a kernel function used for mapping the input data in a higher dimensional input space. In our experiments, a polynomial kernel of degree 7 was considered. According to the observed skew values in our images, its range was limited to [-2;2]. The range of values for the weights wk is chosen to be [1;10]. The resulting mapping wk(γ), after applying SVR on the training set is represented in Fig. 3.

Experiments were run on a set of 15 large, high resolution images, from which we chose 60 manually segmented sub-plots (as illustrated in Fig.2). The performance of the proposed segmentation method was assessed on the test set of 60 sub-plots, using a previously manually drawn ground truth (on which the calcite regions were manually marked). The difference between the ground truth segmentation and the segmentation result of our algorithm allows us to assess the segmentation error, the false positives and the false negatives for the calcite class. The segmentation error for the calcite class achieved with our method, expressed as the average percentage of misclassified pixels in the test set for the total 60 sub-plots, is 4.19%, whereas for the standard fuzzy c-means algorithm is 9.64%(more than double). Some segmentation results are illustrated in Fig. 4.

Fig. 3. Skew to class weight mapping

12 Sustainable Natural Resources Management

Note that although there is no a-priori association of the class index j, j=1,2 or 3, and the brightness of the colors in the class, we always know that the fuzzy class with light most colors is the fuzzy class whose center is the lightest, and this class will be considered to

> C C k ar gma x 0.299 0.587 0.114 . <sup>C</sup> <sup>k</sup> <sup>j</sup> j 1,2,3

To be able to effectively employ the above considerations into our algorithm, a numerical mapping between the range of values γ and the range of weights of the light-most, i.e. calcite pixels class, must be obtained. Denoting our target weight by wk, with k given by Eq. (4), we search for the mapping wk(γ) that best fits a set of training data, obtained by manually tuning the value wk on a set of statistically significant dam wall images (with enough variability in appearance, to cover as many practical cases as possible). A set of 15 images of several sub-plots, with different aspect, under different lighting conditions and different amounts of calcite (from none to very severe) have been selected and manually analyzed to optimize the calcite class' weight for an accurate calcite identification. The pairs formed by the skew values and the best manually selected weight values wk have been collected, and an interpolation procedure based on support vector regression (SVR) has been applied on this training set to completely define in an automatic fashion the computation of

The reason for using SVR in the interpolation step is its proven good performance when only a relatively sparse set of data points is available. Based on Vapnik and Chervonenkis's statistical learning theory (Vapnik, 1998), support vector learning principle allows handling successfully difficult cases, with better precision and recall than other learning methods. This is mainly due to the structural risk minimization principle implemented by SVMs. SVMs were initially "built" for classification and later extended to the regression issue – SVR – by introducing a loss function (Scholkoph *et al*., 1998; Platt, 2000). Starting from an input data set, represented by a vector x, the SVM learns the functional dependency between input and output, represented in the form of a scalar-valued function f(x). The

expression of the regression function provided as a result of learning by an SVM is:

mapping wk(γ), after applying SVR on the training set is represented in Fig. 3.

where L denotes the total number of training data, i and \**i* are their associated Lagrange multipliers, and the function K(x,xi) represents a kernel function used for mapping the input data in a higher dimensional input space. In our experiments, a polynomial kernel of degree 7 was considered. According to the observed skew values in our images, its range was limited to [-2;2]. The range of values for the weights wk is chosen to be [1;10]. The resulting

Experiments were run on a set of 15 large, high resolution images, from which we chose 60 manually segmented sub-plots (as illustrated in Fig.2). The performance of the proposed segmentation method was assessed on the test set of 60 sub-plots, using a previously manually drawn ground truth (on which the calcite regions were manually marked). The

<sup>L</sup> \* f( ) α α K( , ), i 1 ii i **<sup>x</sup> x x** (5)

**v** (4)

the weight wk. We assumed the other two classes' weights "fixed" to 1.

correspond to the calcite (if any):

Fig. 4. Example of the calcite segmentation for a sub-plot: (a) the original (not segmented) sub-plot image; (b) the fuzzy c-means segmentation result; (c) the proposed weighted fuzzy c-means segmentation result.

Despite the good performance of this segmentation procedure in the localization of calcite, one must take into account that less severe infiltrations may not produce significant calcite deposits yet, and may only be visible in the infrared spectrum. Therefore the integration of infrared image analysis results with the visible image analysis results, using a late decision fusion, can bring more valuable information in the infiltration assessment. The fusion is thought to take into account the spatial and temporal correlation of the two types of images of the same hydro-dam downstream surface. This approach is presented in the following sub-section.

Fuzzy Image Processing, Analysis and

Figure 7.

Visualization Methods for Hydro-Dams and Hydro-Sites Surveillance and Monitoring 15

The visible image analysis and segmentation for calcite detection was already presented in the previous sub-section. For the bi-modal analysis we discuss here, a further step is required: the creation of the "water infiltration map" in the visible domain. This map must actually illustrate the severity of the water infiltration, but this severity is (as discussed priorly) correlated to the "amount" or severity of the calcite deposits: in the areas where the water infiltrated on a long period of time, the calcite deposits will appear brighter, as the calcite layer is thicker. We map the severity degree of water infiltration to an intensity range {0,1,...,255}, with 0 for the lack of any infiltration to 255 for maximum severity infiltration. Accordingly we can convert the segmented "visible image" (with calcite areas identified as explained in the previous sub-section) into a visible infiltration severity degree map. To do so, we consider that the brightness component Y is a sufficiently good indicator of the "whiteness" of the calcite – therefore, of the severity of the infiltration also. We represent in matrix form the brightness component of the "visible image" from the current plot pair by IY[H×W], with elements in the range {0,1,...,255}. Let us consider the segmented plot image, with the pixels assigned to one of the two classes: calcite or non-calcite, represented as a binary matrix as well, SVis[H×W]. With these notations we build the water infiltration map of the visible image as the matrix MapVisible[H×W], according to the following expression:

i,j i,j 255

(6)

Map Vis Y

Fig. 7. Plot image segmentation result in the visible domain: Calcite against not calcite

On the other hand, the infrared image analysis and segmentation aims to identify the cold areas, to produce a water infiltration map from the infrared image of the plot. It is known that, at least in the spring/summer, when the ambient temperature is rather high, the areas of the plot with water infiltrations appear colder in the plot's thermal map. The more significant the water infiltration is, the colder is the local part of the plot, thus the lower the temperature on the plot's thermal map. However we can expect that in such areas little evidence of calcite will be identified in the visible image, since the calcite is likely to occur in

segmented image (left) and infiltration severity degree map (right)

Visible YMax,Calcite

where YMax,Calcite is the maximum possible intensity for calcite areas, derived from a set of training images corresponding to calcite patches on the hydro-dam wall. An example of a water infiltration map for the "visible image" in the left side of Figure 6 is illustrated in

**S I**

### **3.2 Bimodal infiltration assessment through the integration of infrared and visible information**

The block diagram of the bimodal fusion based approach for water infiltration assessment is schematically illustrated in Figure 5. Many of the operations involved in the acquisition and low level processing of the visible spectrum and infrared spectrum images are done independently (in a parallel processing fashion). Apart from the acquisition, these operations include: visible spectrum and infrared spectrum images delimitation at plot level (as shown in Figure 2); visible and infrared image segmentation and infiltration severity degree mapping in the two imaging modalities for the quantitative description of water infiltration information. On the output of the corresponding stages, we simultaneously have the two water infiltration degrees maps decided by the two modalities, to integrate these decisions by a simple fusion process.

Fig. 5. Block diagram of the proposed method for infiltration assessment within the dam body

After the acquisition step, a registered pair of sub-plot images is available to be taken from the visual inspections database, each corresponding to the same element. The processing described in the following refers to such an aligned pair. For the infrared image acquisition we used a thermal camera with temperature coding capabilities (providing a thermal map of the corresponding dam wall area). We refer the image of the currently analysed plot in the visible spectrum as "the visible image" and the image of the same plot in the infrared domain will be referred as "the infrared image" and we assumed they are pixel-level registered by scaling and translation compensation. An example of such a registered (visible, infrared) image pair for a plot of the dam wall is shown in Figure 6.

Fig. 6. A pair of images for a hydro-dam wall plot acquired in the two modalities: visible spectrum modality (left) and infrared modality (right)

The block diagram of the bimodal fusion based approach for water infiltration assessment is schematically illustrated in Figure 5. Many of the operations involved in the acquisition and low level processing of the visible spectrum and infrared spectrum images are done independently (in a parallel processing fashion). Apart from the acquisition, these operations include: visible spectrum and infrared spectrum images delimitation at plot level (as shown in Figure 2); visible and infrared image segmentation and infiltration severity degree mapping in the two imaging modalities for the quantitative description of water infiltration information. On the output of the corresponding stages, we simultaneously have the two water infiltration degrees maps decided by the two modalities, to integrate these

Fig. 5. Block diagram of the proposed method for infiltration assessment within the dam body

After the acquisition step, a registered pair of sub-plot images is available to be taken from the visual inspections database, each corresponding to the same element. The processing described in the following refers to such an aligned pair. For the infrared image acquisition we used a thermal camera with temperature coding capabilities (providing a thermal map of the corresponding dam wall area). We refer the image of the currently analysed plot in the visible spectrum as "the visible image" and the image of the same plot in the infrared domain will be referred as "the infrared image" and we assumed they are pixel-level registered by scaling and translation compensation. An example of such a registered

(visible, infrared) image pair for a plot of the dam wall is shown in Figure 6.

Fig. 6. A pair of images for a hydro-dam wall plot acquired in the two modalities: visible

spectrum modality (left) and infrared modality (right)

**3.2 Bimodal infiltration assessment through the integration of infrared and visible** 

**information** 

decisions by a simple fusion process.

The visible image analysis and segmentation for calcite detection was already presented in the previous sub-section. For the bi-modal analysis we discuss here, a further step is required: the creation of the "water infiltration map" in the visible domain. This map must actually illustrate the severity of the water infiltration, but this severity is (as discussed priorly) correlated to the "amount" or severity of the calcite deposits: in the areas where the water infiltrated on a long period of time, the calcite deposits will appear brighter, as the calcite layer is thicker. We map the severity degree of water infiltration to an intensity range {0,1,...,255}, with 0 for the lack of any infiltration to 255 for maximum severity infiltration. Accordingly we can convert the segmented "visible image" (with calcite areas identified as explained in the previous sub-section) into a visible infiltration severity degree map. To do so, we consider that the brightness component Y is a sufficiently good indicator of the "whiteness" of the calcite – therefore, of the severity of the infiltration also. We represent in matrix form the brightness component of the "visible image" from the current plot pair by IY[H×W], with elements in the range {0,1,...,255}. Let us consider the segmented plot image, with the pixels assigned to one of the two classes: calcite or non-calcite, represented as a binary matrix as well, SVis[H×W]. With these notations we build the water infiltration map of the visible image as the matrix MapVisible[H×W], according to the following expression:

$$\text{Map}\_{\text{Visible}} = \frac{\mathbf{S}\_{\text{Vis}}(\text{i}\mathbf{j}) \cdot \mathbf{I}\_{\text{Y}}(\text{i}\mathbf{j}) \cdot 255}{\mathbf{Y}\_{\text{Max,Calcite}}} \tag{6}$$

where YMax,Calcite is the maximum possible intensity for calcite areas, derived from a set of training images corresponding to calcite patches on the hydro-dam wall. An example of a water infiltration map for the "visible image" in the left side of Figure 6 is illustrated in Figure 7.

Fig. 7. Plot image segmentation result in the visible domain: Calcite against not calcite segmented image (left) and infiltration severity degree map (right)

On the other hand, the infrared image analysis and segmentation aims to identify the cold areas, to produce a water infiltration map from the infrared image of the plot. It is known that, at least in the spring/summer, when the ambient temperature is rather high, the areas of the plot with water infiltrations appear colder in the plot's thermal map. The more significant the water infiltration is, the colder is the local part of the plot, thus the lower the temperature on the plot's thermal map. However we can expect that in such areas little evidence of calcite will be identified in the visible image, since the calcite is likely to occur in

Fuzzy Image Processing, Analysis and

maximum infiltration severity, is obtained as:

the infiltration reported to the total area exhibiting infiltration.

current plot.

Visualization Methods for Hydro-Dams and Hydro-Sites Surveillance and Monitoring 17

where IR,IR[H×W] represents the intensity of the red component in the infrared image of the

The segmentation result in "cold" and "not cold" areas for the infrared plot image given in Figure 6, where the segmented image is presented as a binary image (black – "not cold", white – "cold"), and its associated water infiltration severity degree map, are given in Fig. 9.

The final processing step is the bimodal fusion of visible and infrared water infiltration severity information. We use the two individual information sources already provided by the independent image processing and analysis stages: the visible and infrared image processing, to obtain the overall assessment and quantification of the water infiltration amount in the currently analysed plot. Several fusion schemes are available, varying from very simple (pixel-based) to complex ones, to perform the information integration from two or more modalities; the most used in particular for visible and infrared bimodal information fusion can be found in (Yin and Malcolm, 2000; O'Conaire *et al*., 2006). Among these, one of the simplest schemes is by weighted averaging of the decisions given by each modality alone at pixel level, provided that the visible and infrared image registration was previously performed. Let us denote the decision about the plausibility of presence of a certain event in the spatial position (i,j) in the visible spectrum modality by dVis(i,j) and the decision about the plausibility of presence of the same event in the spatial position (i,j) in the infrared modality by dIR(i,j). We also consider the weights (confidences) assigned to each modality denoted by wVis and wIR, chosen to satisfy the constraints: w (0;1); Vis w (0;1); IR w w 1. Vis IR The confidences wVis and wIR assigned to each modality are derived based on expert's knowledge about the relative significance of each modality in assessing the severity of the water infiltration. The presence of calcite shows persistant, longer duration water infiltration in the plot, thus its weight should be higher than the infrared's information source weight. We chose as confidence values in our application: wVis=0.65 and wIR=0.35. As information sources to be weighted aggregated, we use the individual water infiltration severity degrees maps, MapVis and MapIR. The overall water infiltration severity degree map, represented as an intensity image in the range {0,1,…,255}, with 255 –

InfMap(i,j) w Map (i,j) w Map (i,j). Vis Vis IR IR (8)

An example of the resulting water infiltration severity degree map after bimodal image fusion, for the plot presented in Fig. 6, is given in Fig. 10. Then this overall decision map can be used to compute quantitative descriptors of the water infiltration amount and local severity on the plot. Examples of such simple quantitative descriptors are given in (Gordan *et al*., 2007): the percentage of the water infiltration area from the total plot area; the maximum local severity degree of water infiltration, assessed as the accumulated severity of

In order to test this method we used the same multi-modal database containing images acquired from Tarnita dam, near Cluj-Napoca. We selected 5 pairs of plots acquired in both

Map i,j S i,j 255 I i,j Infrared IR R,IR (7)

the region below the wet areas. This gives reason to believe that the two information sources can favourably complement each other. Since in the case of the thermal maps we always have available exactly the color-temperature conversion scale, we can use this scale and apriori knowledge about the numerical range related to the qualifier "cold" to obtain the accurate identification of the water infiltration areas. An example of the selected scale portion, as considered to represent water infiltrations in our application (the severity of the infiltrations is stronger as the color is closer to violet and dark violet than to red), is illustrated in Figure 8. The red color is considered to be already not cold at all, whereas the very dark violet is considered to be the coldest possible. The simplest way to convert this color scale into a scalar scale in the range {0,1,...,255}, with 0 for the minimum coldness and 255 for the maximum coldness, is to use the negative of the red color component intensity of the scale image, as shown in Figure 8.

Fig. 8. The cold temperature part of the infrared scale: Original (left); its red component (right)

The segmentation process of the infrared image of the plot into cold areas and not cold areas is done pixel wise, based on the pixel color. The RGB space is uniformly quantized with only 4 bits per color component to guarantee the color match of the "infrared image" pixels with the infrared scale. We also gather and denote by SCold the set of the quantized color intensities in the RGB representation of the IR scale corresponding to cold from Figure 8, and simply assign the pixels in the infrared image of the plot the label 1 if their quantized color is found in SCold, and 0 otherwise. As a result, we obtain the segmented infrared image of the plot into cold against not cold areas, described by the matrix SIR [H×W].

Fig. 9. Plot image segmentation result in the infrared domain: Cold against not cold segmented image (left) and infiltration severity degree map (right)

Afterwards, we build the water infiltration severity degree map in the infrared domain – similar to the one in the visible domain. However in the infrared case, we consider as infiltration severity degree indicator – the negative of the red color component in each pixel position previously classified as cold, as discussed earlier. Let us denote by MapInfrared[H×W] – the severity degree map of the water infiltration in the infrared modality, represented in the range {0,1,...,255}. The values in this matrix are computed as:

the region below the wet areas. This gives reason to believe that the two information sources can favourably complement each other. Since in the case of the thermal maps we always have available exactly the color-temperature conversion scale, we can use this scale and apriori knowledge about the numerical range related to the qualifier "cold" to obtain the accurate identification of the water infiltration areas. An example of the selected scale portion, as considered to represent water infiltrations in our application (the severity of the infiltrations is stronger as the color is closer to violet and dark violet than to red), is illustrated in Figure 8. The red color is considered to be already not cold at all, whereas the very dark violet is considered to be the coldest possible. The simplest way to convert this color scale into a scalar scale in the range {0,1,...,255}, with 0 for the minimum coldness and 255 for the maximum coldness, is to use the negative of the red color component intensity of

Fig. 8. The cold temperature part of the infrared scale: Original (left); its red component (right)

The segmentation process of the infrared image of the plot into cold areas and not cold areas is done pixel wise, based on the pixel color. The RGB space is uniformly quantized with only 4 bits per color component to guarantee the color match of the "infrared image" pixels with the infrared scale. We also gather and denote by SCold the set of the quantized color intensities in the RGB representation of the IR scale corresponding to cold from Figure 8, and simply assign the pixels in the infrared image of the plot the label 1 if their quantized color is found in SCold, and 0 otherwise. As a result, we obtain the segmented infrared

image of the plot into cold against not cold areas, described by the matrix SIR [H×W].

Fig. 9. Plot image segmentation result in the infrared domain: Cold against not cold

represented in the range {0,1,...,255}. The values in this matrix are computed as:

Afterwards, we build the water infiltration severity degree map in the infrared domain – similar to the one in the visible domain. However in the infrared case, we consider as infiltration severity degree indicator – the negative of the red color component in each pixel position previously classified as cold, as discussed earlier. Let us denote by MapInfrared[H×W] – the severity degree map of the water infiltration in the infrared modality,

segmented image (left) and infiltration severity degree map (right)

the scale image, as shown in Figure 8.

$$\text{Map}\_{\text{Infraared}}(\text{i.j}) = \text{S}\_{\text{IR}} \, (\text{i.j}) \cdot \left(255 - \text{I}\_{\text{R}, \text{IR}} \, (\text{i.j})\right) \tag{7}$$

where IR,IR[H×W] represents the intensity of the red component in the infrared image of the current plot.

The segmentation result in "cold" and "not cold" areas for the infrared plot image given in Figure 6, where the segmented image is presented as a binary image (black – "not cold", white – "cold"), and its associated water infiltration severity degree map, are given in Fig. 9.

The final processing step is the bimodal fusion of visible and infrared water infiltration severity information. We use the two individual information sources already provided by the independent image processing and analysis stages: the visible and infrared image processing, to obtain the overall assessment and quantification of the water infiltration amount in the currently analysed plot. Several fusion schemes are available, varying from very simple (pixel-based) to complex ones, to perform the information integration from two or more modalities; the most used in particular for visible and infrared bimodal information fusion can be found in (Yin and Malcolm, 2000; O'Conaire *et al*., 2006). Among these, one of the simplest schemes is by weighted averaging of the decisions given by each modality alone at pixel level, provided that the visible and infrared image registration was previously performed. Let us denote the decision about the plausibility of presence of a certain event in the spatial position (i,j) in the visible spectrum modality by dVis(i,j) and the decision about the plausibility of presence of the same event in the spatial position (i,j) in the infrared modality by dIR(i,j). We also consider the weights (confidences) assigned to each modality denoted by wVis and wIR, chosen to satisfy the constraints: w (0;1); Vis w (0;1); IR w w 1. Vis IR The confidences wVis and wIR assigned to each modality are derived based on expert's knowledge about the relative significance of each modality in assessing the severity of the water infiltration. The presence of calcite shows persistant, longer duration water infiltration in the plot, thus its weight should be higher than the infrared's information source weight. We chose as confidence values in our application: wVis=0.65 and wIR=0.35. As information sources to be weighted aggregated, we use the individual water infiltration severity degrees maps, MapVis and MapIR. The overall water infiltration severity degree map, represented as an intensity image in the range {0,1,…,255}, with 255 – maximum infiltration severity, is obtained as:

$$\text{InfMap}(\text{i}, \text{j}) = \mathbf{w}\_{\text{Vis}} \cdot \mathbf{Map}\_{\text{Vis}}(\text{i}, \text{j}) + \mathbf{w}\_{\text{IR}} \cdot \mathbf{Map}\_{\text{IR}}(\text{i}, \text{j}). \tag{8}$$

An example of the resulting water infiltration severity degree map after bimodal image fusion, for the plot presented in Fig. 6, is given in Fig. 10. Then this overall decision map can be used to compute quantitative descriptors of the water infiltration amount and local severity on the plot. Examples of such simple quantitative descriptors are given in (Gordan *et al*., 2007): the percentage of the water infiltration area from the total plot area; the maximum local severity degree of water infiltration, assessed as the accumulated severity of the infiltration reported to the total area exhibiting infiltration.

In order to test this method we used the same multi-modal database containing images acquired from Tarnita dam, near Cluj-Napoca. We selected 5 pairs of plots acquired in both

Fuzzy Image Processing, Analysis and

visualization tools (Shi *et al*., 2006).

Visualization Methods for Hydro-Dams and Hydro-Sites Surveillance and Monitoring 19

form. The originality of the presented solution consists in the presentation of the water resource management evaluation "grading" in the form of a geotypical textured map of the region, where the natural texture changes according to the evaluation result for a specific category and according to the qualifier assigned to the management policy (varying from worst to very good). Therefore, in this sub-section we primarily emphasize on this visualization enhanced results presentation part. The interested reader may find more

To achieve a meaningful graphical representation, we propose to employ fuzzy alphablending, image morphology and fuzzy image inpainting algorithms, which allow the production of high quality and meaningful geotypically textured maps of the hydro-site region. This allows the user to get multiple clues on the results of the water resource management evaluation, and have a stronger impact than the numerical assessment alone. The advancements and new application tracks of image processing algorithms and display devices provide the means for advanced graphical representations to be easily integrated in decision support software tools. These components are not so widely employed in the existing systems, but some implementations exist, as e.g. the integrated information management and simulation system combining WebGIS, database and hydrological model in (Shaomin *et al*., 2009) – which integrates a flood simulator and visually presents the flooded areas; or, the GIS based integrated system, which also incorporates hydrological analysis and cascade hydroelectric station dispatching functions, with powerful

The case of water resource management assessment may significantly benefit from a visualization module provided in the form of a geotypically textured map of the evaluated region. This can easily embed digital maps and natural images specific to the site, combined with specifically designed rendering tools. The fuzzy evaluation process results should

Adopting the terminology in (Zhou & Huang, 2007), the factors involved in the assessment of the water resource management are called indexes. Each index represents a relevant attribute in the water resource management evaluation, and it must allow either a numerical or a qualitative description. During the system's setup, a weight must be assigned to each index, showing its relevance in the assessment of the water resource management. The weights may vary depending on the available water resources in the region and on the overall regional conditions. As the water resource management may impact several facets of life (the natural resources of the region, the ecology and the environment, the society and the economy of the region), a group of indexes is defined for each category individually. This will allow an independent evaluation of the water resource management policy's impact on each category. So far we implemented the decision support component only for the category of natural resources. This implies the definition of the appropriate set of relevant indexes for the natural resources, influenced by the water management policy.

As shown in the literature, five indexes are most relevant for the natural resources category in the framework of water resource management: the total water resources; the water resources per capita; the utilization rate of the water resources; the annual rainfall; the water shortage rate (Zhou & Huang, 2007). These five indexes are grouped into the index layer of the component. Based on their current values and on the management evaluation procedure, the

drive the rendering of the appropriate textures on the digital map of the region.

details of the implementation of the tool in (Gordan *et al*., 2010).

modalities (visible and infrared). As shown earlier in this section, a ground truth for visible image segmentation into calcite areas and non-calcite areas can be easily obtained, and the same – a ground truth for pixel classification into cold areas for the infrared images. Thus we can assess the functionality of these processing stages very accurately. However, this is not the case for the assessment of water infiltrations severity, which in general can only be subjectively estimated by human observers. Therefore we can only roughly compare the results provided by our algorithm, converted to subjective scales, to subjective (human) evaluation of the water infiltrations based on the visible and infrared plot image evaluation. These comparative results for the 5 pairs of plots are presented in Table 1. The only difference from the human expert's opinion is in the 4th line in Table 1, for a plot exhibiting water infiltration in a very small area, in respect to the local severity of the water infiltration: although the numerical results show a large local value, the human expert identifies it as not significant, and this could be explained by the overall assessment done by the human expert, with almost no attention to local details when the water infiltration region size is not significant. The segmentation results, both for the visible and infrared plot images show in all cases good accuracy.

Although we employ here one of the most simple fusion schemes, we can see how the use of the two modalities can lead to better results than the analysis of each imaging modality alone. Also, the implementation of the joint analysis of visible and infrared images into the visual inspections module we described at the beginning of this chapter, has the advantage of providing numerical estimates of the extension of the water infiltrations and severity of the water infiltrations in the plots, reducing the risk of human observer subjectivity and image display quality.


Table 1. Quantitative results of our algorithm against subjective human expert's opinion

### **4. Assessment of the water resources management policy in a hydro-site region**

As the hydro-dams reservoirs are also the main water supply resources for the geographical region, the assessment of the water management policy in the operation of the hydro-dam in respect to various economical and environmental factors is also an issue of significant interest. In this respect, we propose and implement a fuzzy decision support component to help in assessing the water resource management. Whereas the evaluation strategy itself is inspired by the work of (Zhou & Huang, 2007), employing a hierarchical process analysis strategy with qualitative reasoning, the presentation of the assessment results is novel, as we aim to display the evaluation not only in numerical and linguistic form, but also in a visual

modalities (visible and infrared). As shown earlier in this section, a ground truth for visible image segmentation into calcite areas and non-calcite areas can be easily obtained, and the same – a ground truth for pixel classification into cold areas for the infrared images. Thus we can assess the functionality of these processing stages very accurately. However, this is not the case for the assessment of water infiltrations severity, which in general can only be subjectively estimated by human observers. Therefore we can only roughly compare the results provided by our algorithm, converted to subjective scales, to subjective (human) evaluation of the water infiltrations based on the visible and infrared plot image evaluation. These comparative results for the 5 pairs of plots are presented in Table 1. The only difference from the human expert's opinion is in the 4th line in Table 1, for a plot exhibiting water infiltration in a very small area, in respect to the local severity of the water infiltration: although the numerical results show a large local value, the human expert identifies it as not significant, and this could be explained by the overall assessment done by the human expert, with almost no attention to local details when the water infiltration region size is not significant. The segmentation results, both for the visible and infrared plot images show in

Although we employ here one of the most simple fusion schemes, we can see how the use of the two modalities can lead to better results than the analysis of each imaging modality alone. Also, the implementation of the joint analysis of visible and infrared images into the visual inspections module we described at the beginning of this chapter, has the advantage of providing numerical estimates of the extension of the water infiltrations and severity of the water infiltrations in the plots, reducing the risk of human observer subjectivity and

Infiltration amount

Infiltration severity

(subjective)

(subjective)

Infiltration Severity

1 32.05% 81% Medium/Large Severe 2 23.63% 58% Medium Moderate 3 24.46% 64.7% Medium/Small Moderate 4 2.4% 72% Almost none Reduced 5 43.7% 78% Large Severe

Table 1. Quantitative results of our algorithm against subjective human expert's opinion

**4. Assessment of the water resources management policy in a hydro-site** 

As the hydro-dams reservoirs are also the main water supply resources for the geographical region, the assessment of the water management policy in the operation of the hydro-dam in respect to various economical and environmental factors is also an issue of significant interest. In this respect, we propose and implement a fuzzy decision support component to help in assessing the water resource management. Whereas the evaluation strategy itself is inspired by the work of (Zhou & Huang, 2007), employing a hierarchical process analysis strategy with qualitative reasoning, the presentation of the assessment results is novel, as we aim to display the evaluation not only in numerical and linguistic form, but also in a visual

all cases good accuracy.

image display quality.

Water Infiltration Area

Plot pair Number

**region** 

form. The originality of the presented solution consists in the presentation of the water resource management evaluation "grading" in the form of a geotypical textured map of the region, where the natural texture changes according to the evaluation result for a specific category and according to the qualifier assigned to the management policy (varying from worst to very good). Therefore, in this sub-section we primarily emphasize on this visualization enhanced results presentation part. The interested reader may find more details of the implementation of the tool in (Gordan *et al*., 2010).

To achieve a meaningful graphical representation, we propose to employ fuzzy alphablending, image morphology and fuzzy image inpainting algorithms, which allow the production of high quality and meaningful geotypically textured maps of the hydro-site region. This allows the user to get multiple clues on the results of the water resource management evaluation, and have a stronger impact than the numerical assessment alone. The advancements and new application tracks of image processing algorithms and display devices provide the means for advanced graphical representations to be easily integrated in decision support software tools. These components are not so widely employed in the existing systems, but some implementations exist, as e.g. the integrated information management and simulation system combining WebGIS, database and hydrological model in (Shaomin *et al*., 2009) – which integrates a flood simulator and visually presents the flooded areas; or, the GIS based integrated system, which also incorporates hydrological analysis and cascade hydroelectric station dispatching functions, with powerful visualization tools (Shi *et al*., 2006).

The case of water resource management assessment may significantly benefit from a visualization module provided in the form of a geotypically textured map of the evaluated region. This can easily embed digital maps and natural images specific to the site, combined with specifically designed rendering tools. The fuzzy evaluation process results should drive the rendering of the appropriate textures on the digital map of the region.

Adopting the terminology in (Zhou & Huang, 2007), the factors involved in the assessment of the water resource management are called indexes. Each index represents a relevant attribute in the water resource management evaluation, and it must allow either a numerical or a qualitative description. During the system's setup, a weight must be assigned to each index, showing its relevance in the assessment of the water resource management. The weights may vary depending on the available water resources in the region and on the overall regional conditions. As the water resource management may impact several facets of life (the natural resources of the region, the ecology and the environment, the society and the economy of the region), a group of indexes is defined for each category individually. This will allow an independent evaluation of the water resource management policy's impact on each category. So far we implemented the decision support component only for the category of natural resources. This implies the definition of the appropriate set of relevant indexes for the natural resources, influenced by the water management policy.

As shown in the literature, five indexes are most relevant for the natural resources category in the framework of water resource management: the total water resources; the water resources per capita; the utilization rate of the water resources; the annual rainfall; the water shortage rate (Zhou & Huang, 2007). These five indexes are grouped into the index layer of the component. Based on their current values and on the management evaluation procedure, the

Fuzzy Image Processing, Analysis and

*A*4=0.8; *ABest*=*A*5=1

given by:

R*k*: If *Qualifier* is *q* then

5

1 5

*k*

 

visualization purposes

1

*k k*

*u*

*u A*

. *k k*

Visualization Methods for Hydro-Dams and Hydro-Sites Surveillance and Monitoring 21

where is some blending factor ranging between 0 and 1, generated at the output of a

2. Output fuzzy sets: five singletons – one for each reference value , corresponding to the reference case of a *Worst*, *Bad*, *Moderate*, *Good* or *Best* management. These values were chosen intuitively and empirically to: *AWorst*=*A*1=0; *ABad*=*A*2=0.2; *AModerate*=*A*3=0.5; *AGood*=

3. Rule base: five fuzzy rules, associating each qualifier to an output singleton, in the form:

The results of the assessment are illustrated for two different cases of indexes' values: close to best management and between bad and moderate, but not worst management, as shown

Fig. 10. Illustration of the Somes River Basin marked map, to be further processed in

As the result of the Takagi-Sugeno inference, the output value for the blending factor

=*Ak*, *k*=1,2,...,5; *q =* { *Worst*, *Bad*, *Moderate*, *Good*, *Best*}.

is

1. Input fuzzy sets: the five linguistic qualifiers (*Worst*, *Bad*, *Moderate*, *Good*, *Best*)

Takagi-Sugeno fuzzy system, with the following configuration:

in Figure 12. The results are compliant to the observer's expectations.

quality of the water resource management policy in respect to the natural resources preservation is expressed in terms of five fuzzy qualifiers: Worst, Bad, Moderate, Good and Best, grouped in the output layer of the component – known as the "condition layer".

The decision support component for the evaluation of water resource management policy in respect to the natural resources preservation must include a so-called training phase, in which the specialist helps defining the fuzzy sets membership functions associated to each index and each linguistic qualifier in a set Q={*Worst*, *Bad*, *Moderate*, *Good*, *Best*} (in respect to the specific category), and the weights of the indexes in the evaluation. Then, in the evaluation phase, the current values of the indexes – let us consider them given in the form of a vector x - are provided to the input of the system. Based on the values in x, the evaluation algorithm computes a membership degrees vector u[1×5], showing the confidence in assigning the currently examined water management policy to the fuzzy categories from Q, in the *Worst* to *Best* order. The vector u of confidence degrees in the suitability of each linguistic qualifier for the current water management policy in respect to the resource category is also used in the visual rendering sub-system.

The visual rendering of the evaluation results is achieved as follows. Assume that, for the current geographical region, we have its geographical map, with some manual marking of the interest categories, as e.g. the one shown in Figure 10, for the Somes river basin in Romania, corresponding to a good operation situation. Starting from this image, we would like to generate two geotypically textured images: one corresponding to the *Worst* resource management case, in which the exploitation was not proper, and one corresponding to the *Best* resource management case, with a very good water resource management policy.

In principle, the *Best* case geotypically textured map simply needs some texture synthesis applied on the image in Figure 10, using suitable natural textures for the forest, water, rock – and the approach we employed to generate the natural looking textured map was a modified version of the exemplar-based image inpainting approach of (Criminisi *et al*., 2003). An example of inpainting the forest region over the map from Figure 10 is shown in Figure 11. However, in the *Worst* case image, it would be good to also apply some additional processing; a suitable choice is to perform some morphological operations – as: erosion of the rivers; dilation of the mountain area, to enhance the visual effect of a very bad policy, prior to inpainting the map with the suitable textures.

Once the two geotypically textured images corresponding to the two extreme water resource management qualifiers are created, we would like to display any intermediate results as given by our assessment fuzzy system. Consider the two images represented as three-dimensional matrices IWorst and IBest, of size WI×HI×3 each, where WI is the image width, HI - the image height, and 3 is the number of color components per image. We already have available the degrees in which the management of the water resources can be considered *Worst* (the value of the first component from the vector u), *Bad* (the value of the second component from u), *Moderate* (the value of the third component from u), *Good* (the value of the fourth component from u) and *Best* (the value of the last component from u). The only thing to be done is to combine the two images IWorst and IBest to obtain the correct visualization as a new image IResult, according to:

$$I\_{Result} = \alpha \cdot I\_{Best} + \left(1 - \alpha\right) \cdot I\_{Norst,} \tag{9}$$

where is some blending factor ranging between 0 and 1, generated at the output of a Takagi-Sugeno fuzzy system, with the following configuration:


As the result of the Takagi-Sugeno inference, the output value for the blending factor is

$$\text{given by: } \alpha = \frac{\sum\_{k=1}^{5} \mu\_k \cdot A\_k}{\sum\_{k=1}^{5} \mu\_k}.$$

20 Sustainable Natural Resources Management

quality of the water resource management policy in respect to the natural resources preservation is expressed in terms of five fuzzy qualifiers: Worst, Bad, Moderate, Good and

The decision support component for the evaluation of water resource management policy in respect to the natural resources preservation must include a so-called training phase, in which the specialist helps defining the fuzzy sets membership functions associated to each index and each linguistic qualifier in a set Q={*Worst*, *Bad*, *Moderate*, *Good*, *Best*} (in respect to the specific category), and the weights of the indexes in the evaluation. Then, in the evaluation phase, the current values of the indexes – let us consider them given in the form of a vector x - are provided to the input of the system. Based on the values in x, the evaluation algorithm computes a membership degrees vector u[1×5], showing the confidence in assigning the currently examined water management policy to the fuzzy categories from Q, in the *Worst* to *Best* order. The vector u of confidence degrees in the suitability of each linguistic qualifier for the current water management policy in respect to

The visual rendering of the evaluation results is achieved as follows. Assume that, for the current geographical region, we have its geographical map, with some manual marking of the interest categories, as e.g. the one shown in Figure 10, for the Somes river basin in Romania, corresponding to a good operation situation. Starting from this image, we would like to generate two geotypically textured images: one corresponding to the *Worst* resource management case, in which the exploitation was not proper, and one corresponding to the *Best* resource management case, with a very good water resource management policy.

In principle, the *Best* case geotypically textured map simply needs some texture synthesis applied on the image in Figure 10, using suitable natural textures for the forest, water, rock – and the approach we employed to generate the natural looking textured map was a modified version of the exemplar-based image inpainting approach of (Criminisi *et al*., 2003). An example of inpainting the forest region over the map from Figure 10 is shown in Figure 11. However, in the *Worst* case image, it would be good to also apply some additional processing; a suitable choice is to perform some morphological operations – as: erosion of the rivers; dilation of the mountain area, to enhance the visual effect of a very bad policy,

Once the two geotypically textured images corresponding to the two extreme water resource management qualifiers are created, we would like to display any intermediate results as given by our assessment fuzzy system. Consider the two images represented as three-dimensional matrices IWorst and IBest, of size WI×HI×3 each, where WI is the image width, HI - the image height, and 3 is the number of color components per image. We already have available the degrees in which the management of the water resources can be considered *Worst* (the value of the first component from the vector u), *Bad* (the value of the second component from u), *Moderate* (the value of the third component from u), *Good* (the value of the fourth component from u) and *Best* (the value of the last component from u). The only thing to be done is to combine the two images IWorst and IBest to obtain the correct

, 1 *Result Best Worst II I*

  (9)

Best, grouped in the output layer of the component – known as the "condition layer".

the resource category is also used in the visual rendering sub-system.

prior to inpainting the map with the suitable textures.

visualization as a new image IResult, according to:

The results of the assessment are illustrated for two different cases of indexes' values: close to best management and between bad and moderate, but not worst management, as shown in Figure 12. The results are compliant to the observer's expectations.

Fig. 10. Illustration of the Somes River Basin marked map, to be further processed in visualization purposes

Fuzzy Image Processing, Analysis and

**5. Conclusion** 

challenging issue.

**6. Acknowledgment** 

Romanian Government

repair.html

**7. References** 

Visualization Methods for Hydro-Dams and Hydro-Sites Surveillance and Monitoring 23

This chapter aimed to present a series of novel fuzzy image processing methods and algorithms implemented in the rather general framework of hydro-dams and hydro-sites surveillance, monitoring and assessments, emphasizing on their theoretical motivation and results. Most of these methods have been employed in a hydro-dam and hydro-site integrated system for the safety decision support of these critical structures, thus the results are verified on real image data. Future work still needs to be done in this field, as the integration of the presented fuzzy image analysis algorithms (especially for the visible and infrared modalities) is just in its beginning; other imaging modalities as e.g. sonar, as well as the underwater examination of the hydro-dam structure would be of significant interest. Furthermore, the integration of the hydro-sites surveillance systems with water resource management policy assessment in the region operated by the dam reservoirs is another

Part of the work described in this chapter, as well as the implementation of the system in Romania and the image acquisition, was performed with the support of the project no. 705/2006 in the frame of CEEX research and development programme, financed by

Abare, R. (2006). Shotcrete done right. Failed repair teaches lessons about shotcrete. *Public* 

Asgian, M.I., Arulmoli, K., Miller, W.O., and Sanjeevan, K. (1988). An expert system for

Bezdek, J. C. (1981). *Pattern Recognition with Fuzzy Objective Function Algorithms,* Plenum

Bradlow, D., Palmieri, A., and Salman, S. (2002). *Regulatory Frameworks for Dam Safety: A Comparative Study*, World Bank Publications, ISBN 0821351915, Washington, D.C. Chamorro-Martinez, J., Sanchez, D., Prados-Suarez, B., Galan-Perales, E., and Vila, M.A.

Clairet, J., Bigand, A., and Colot, O. (2006). Color Image Segmentation using Type-2 Fuzzy

diagnosis and treatment of dam seepage problems, In: *Microcomputer knowledgebased expert systems in Civil Engineering*, Adeli, H., pp. 118-126, American Society of

(2003). A hierarchical approach to fuzzy segmentation of colour images, *Proceedings of the 12th IEEE International Conference on Fuzzy Systems,* Vol. 2, pp. 966 – 971, ISBN

Sets, *Proceedings of the 1st IEEE International Conference on E-Learning in Industrial Electronics*, pp. 52 – 57, ISBN 1-4244-0324-3, Hammamet, Tunisia, December 2006 National Research Council (U.S.) Committee on the Safety of Existing Dams (CSED) (1983).

*Safety of existing dams: evaluation and improvement*, National Academy Press, ISBN

http://goliath.ecnext.com/coms2/gi\_0199-5268853/Shotcrete-done-right-failed-

Civil Engineers (ASCE), ISBN 9780872626539, New York, N.Y.

*Works Magazine*, January 1, Retrieved from

Press, ISBN 0306406713, New York, N.Y.

0-7803-7810-5, St. Louis, MO, May 2003

030903387X, Washington, D.C.

Fig. 11. Illustration of an inpainting result for the forest region with a natural texture, corresponding to the best resource management case


Fig. 12. Illustration of the water resource management policy assessment for the *Resource* category for: The *Best* management case (confidence 0.926) (left); The *Bad* to *Moderate*  management case (confidence 0.41 for *Bad* and 0.39 to *Moderate*) (right)

### **5. Conclusion**

22 Sustainable Natural Resources Management

Fig. 11. Illustration of an inpainting result for the forest region with a natural texture,

Fig. 12. Illustration of the water resource management policy assessment for the *Resource* category for: The *Best* management case (confidence 0.926) (left); The *Bad* to *Moderate* 

management case (confidence 0.41 for *Bad* and 0.39 to *Moderate*) (right)

corresponding to the best resource management case

This chapter aimed to present a series of novel fuzzy image processing methods and algorithms implemented in the rather general framework of hydro-dams and hydro-sites surveillance, monitoring and assessments, emphasizing on their theoretical motivation and results. Most of these methods have been employed in a hydro-dam and hydro-site integrated system for the safety decision support of these critical structures, thus the results are verified on real image data. Future work still needs to be done in this field, as the integration of the presented fuzzy image analysis algorithms (especially for the visible and infrared modalities) is just in its beginning; other imaging modalities as e.g. sonar, as well as the underwater examination of the hydro-dam structure would be of significant interest. Furthermore, the integration of the hydro-sites surveillance systems with water resource management policy assessment in the region operated by the dam reservoirs is another challenging issue.

### **6. Acknowledgment**

Part of the work described in this chapter, as well as the implementation of the system in Romania and the image acquisition, was performed with the support of the project no. 705/2006 in the frame of CEEX research and development programme, financed by Romanian Government

### **7. References**


Fuzzy Image Processing, Analysis and

ISSN 0169-7722

ISSN 0266-352X

September 1995

New York, N.Y.

0471030031, New York

Florence, Italy, July 2006

0262194481, Cambridge, MA

Genoa, Italy, September 2005

Visualization Methods for Hydro-Dams and Hydro-Sites Surveillance and Monitoring 25

Marshall, B. D., Neymark, L.A., and Peterman, E.Z. (2003). Estimation of past seepage

Najjar, Y. M., Basheer, I.A., and Naouss, W.A. (1996). On the Identification of Compaction

O'Conaire, C., O'Connor, N., Cooke, E., and Smeaton, A.F. (2006). Comparison of Fusion

Ohnishi, Y., and Soliman, M. (1995). Seepage Under Concrete Dam Founded on Rock

Ollero, A., Martinez-De Dios, J.R., and Arrúe, B.C. (1998). Integrated systems for early

Platt, J. (2000). Probabilistic outputs for support vector machines and comparisons to

Shaomin, L., Hai, S., and Cheng, W. (2009). Flood Simulation and Information Management

Shi, S., Ye, X., Dong, Z., and Zhou, H. (2006). Research on the Integration of GIS-Based

Sieh, D., King, D., and Gientke, F. (1988). Dam Seepage Analysis Using Artificial

Vapnik, V.N. (1998). *Statistical Learning Theory* (1st Edition), Wiley-Interscience, ISBN

Wen, Z., Wu, Z., and Su, H. (2004). Safety monitoring system of dam based on bionics,

2, pp. 1099 – 1104, ISBN 0-7803-8403-2, Shanghai, China, August 2004

ISBN 972-97973-0-7, Luso, Portugal, November, 1998

978-1-4244-3581-4, Wuhan, Hubei, China, March 2009

7695-2581-4, Hangzhou, Zhejiang, China, June 2006

*Conference on Image Processing (ICIP 2005)*, Vol. 3, pp. 840-843, ISBN 0-7803-9134-9,

volumes from calcite distribution in the Topopah Spring Tuff, Yucca Mountain, Nevada. *Journal of contaminant hydrology,* Vol. 62-63 (April-May 2003), pp. 237-247,

Characteristics by Neuronets. *Computers and Geotechnics*, Vol. 18, No. 3, pp. 167-187,

Methods for Thermo-Visual Surveillance Tracking, *Proceedings of the 9th International Conference on Information Fusion (FUSION 2006)*, pp. 1-7, ISBN 1-4244-0953-5,

Formation using Artificial Neural Networks, *Proceedings of the International Workshop on Rock Foundation of Large Scale Structures,* pp. 355-360, Tokyo, Japan,

forest-fire detection, *Proceedings of the 3rd International Conference on Forest Fire Research and 14th Conference on Fire and Forest Meteorology*, Vol. 2, pp. 1977-1988,

regularized likelihood methods, In: *Advances in Large Margin Classifiers*, Smola, A., Bartlett, P., Schölkopf, B., and Schuurmans, D., pp. 61-74, MIT Press, ISBN

System's Design and Implement, *Proceedings of the 1st International Workshop on Education Technology and Computer Science (ETCS 2009)*, Vol. 1, pp.737-740, ISBN

Digital Valley System, *Proceedings of the 1st International Multi-Symposiums on Computer and Computational Sciences (IMSCCS 2006*), Vol. 1, pp. 452-457, ISBN 0-

Intelligence, In: *Planning Now for Irrigation and Drainage in the 21st Century*, DeLynn, H., pp 417-422, American Society of Civil Engineers (ASCE), ISBN 9780872626669,

*Proceedings of 2004 International Conference on Machine Learning and Cybernetics*, Vol.


Craft, C.D., Pearson, R.M., and Hurcomb, D. (2007). Mineral Dissolution and Dam Seepage

Criminisi, A., Pérez, P., and Toyama, K. (2003). Object Removal by Exemplar-Based

Dam Safety Committee (DSC) (2010). *Surveillance Reports for Dams*, Retrieved from

Dancea, O., Tsatos, O., Gordan, M., and Vlaicu, A. (2010). Adaptive fuzzy c-means through

Das, S., Konar, A., and Chakraborty, U.K. (2006). Automatic Fuzzy Segmentation of Images

Dunn, J. C. (1973). A Fuzzy Relative of the ISODATA Process and Its Use in Detecting

European Parliament (2000). Directive 2000/60/EC (The EU Water Framework Directive).

Gordan, M., Dancea, O., Vlaicu, A., Stoian, I., Tsatos, O., and Oltean, G. (2007). Hydro-dams

Gordan, M., Dancea, O., Vlaicu, A., Stoian, I., and Tsatos, O. (2008). Computer Vision

Gordan, M., Florea, C., Dancea, O., Stancel, E., and Tsatos, O. (2010). A visualization

Hafiane, A., Zavidovique, B., and Chaudhuri, S. (2005). A modified FCM with optimal

pdf

Madison, WI, June 2003

Napoca, Romania, May 2010

Cluj-Napoca, Romania, July 2007

Napoca, Romania, May 2010

Canada, July 2006

ISSN 0022-0280

2423

8658

DSC2C.pdf

Chemistry – The Bureau of Reclamation Experience, *Proceedings of the 2007 National Meeting, Dam Safety 2007*, Austin Texas, Lexington, Retrieved from http://www.craftgeochemistry.com/CraftMineralDissolutionASDSOmeeting2007.

Inpainting, *Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR) 2003,* Vol. 2, No. 2, pp.721-728, ISBN 0769519008,

http://www.damsafety.nsw.gov.au/DSC/Download/Info\_Sheets\_PDF/General/

support vector regression for segmentation of calcite deposits on concrete dam walls, *Proceedings of the 2010 IEEE International Conference on Automation, Quality and Testing, Robotics (AQTR 2010)*, Vol. 3, pp. 293-298, ISBN 978-1-4244-6724-2, Cluj-

with Differential Evolution, *Proceedings of the IEEE Congress on Evolutionary Computation (CEC 2006),* pp. 2026 – 2033, ISBN 0-7803-9487-9, Vancouver, BC,

Compact Well-Separated Clusters. *Journal of Cybernetics*, Vol. 3, No. 3, pp. 32-57,

*Official Journal of the European Communities*, Vol. OJ L 327, pp. 1-72, ISSN 1725-

Security Assessment by Visible and Infrared Image Fusion, *Proceedings of the 1st IFAC Workshop on Convergence of Information Technologies and Control Methods with Power Plants and Power Systems (ICPS'07)*, pp. 234-239, ISBN 978-973-713-180-5,

Support Tool for Assessing Concrete Hydro-Dams Surface Deterioration. *Journal of Control Engineering and Applied Informatics*, Vol. 10, No. 2, pp. 68-75, ISSN 1454-

enhanced fuzzy decision support tool for water resource management tasks, *Proceedings of the 2010 IEEE International Conference on Automation, Quality and Testing, Robotics (AQTR 2010)*, Vol. 3, pp. 105-110, ISBN 978-1-4244-6724-2, Cluj-

Peano scans for image segmentation, *Proceedings of the 2005 IEEE International* 

*Conference on Image Processing (ICIP 2005)*, Vol. 3, pp. 840-843, ISBN 0-7803-9134-9, Genoa, Italy, September 2005


**2** 

**Upstream Landscape Dynamics of US National** 

The mission of the United States (US) National Park Service (NPS) is to "conserve the scenery and the natural and historic objects and the wild life therein and to provide for the enjoyment of the same in such manner and by such means as will leave them unimpaired for the enjoyment of future generations" (NPS, 1916). NPS currently manages 397 parks covering about 358,200 km2, or approximately 4% of all US states and territories. The National Park system includes approximately 300 parks that are considered to contain significant natural resources. These parks are key components of a larger network of protected areas that anchor the conservation of natural resources in the US. They also afford direct protection for a number of important and defining resources in the US, including 421 species of threatened or endangered plants and animals, nearly two-thirds of native fishes in the 50 states (Lawrence *et al*., 2011), the highest point in North America (Mt. McKinley in Denali National Park, 6,194 m), the longest cave system in the world (Mammoth Cave National Park with more than 587 mapped km of caves), the country's deepest lake (Crater Lake in Crater Lake National Park, 589 m), the lowest terrestrial point in the western hemisphere (Badwater Basin in Death Valley National Park at 86 m below sea level), and – within these extremes – many other natural resources that are significant at local, regional,

While protected areas are foundational to a strong natural resource conservation network, ecologists have long recognized that virtually all parks are too small to be self-sustaining ecosystems, and activities outside park boundaries can profoundly impact park resources (Newmark, 1985; US General Accounting Office, 1994; Parks & Harcourt, 2002; Wiersma *et al*., 2004; Hansen & DeFries, 2007; Hansen *et al*., 2011). Chief among these activities is the intensification of land use and the appropriation of ecological services. Land use intensification leads to the conversion of natural habitat, which generally results in an overall loss of habitat, fragmentation of remaining natural areas, increases in edge zones, changes in the runoff of water, sediments, and nutrients, and follow-on modification of physical and ecological processes in terrestrial and aquatic ecosystems. Depending on the

**1. Introduction** 

and national scales.

**Parks with Implications for Water Quality and** 

**Watershed Management** 

William B. Monahan and John E. Gross

*Inventory and Monitoring Program United States National Park Service* 

*Fort Collins, Colorado* 

*USA* 


## **Upstream Landscape Dynamics of US National Parks with Implications for Water Quality and Watershed Management**

William B. Monahan and John E. Gross *Inventory and Monitoring Program United States National Park Service Fort Collins, Colorado USA* 

### **1. Introduction**

26 Sustainable Natural Resources Management

Yin, Z., and Malcolm, A. (2000). Thermal and Visual Image Processing and Fusion, *SIMTech* 

Zhou, Y., and Huang, J. (2007). An AHP-Based Fuzzy Evaluation Approach to Management

– 5028, ISBN 978-1-4244-1311-9, Shanghai, China, September 2007

star.edu.sg/Research/TechnicalReports/TR0630.pdf

*Technical Report AT/00/016/MVS*, Retrieved from http://www.simtech.a-

of Sustainable Water Resources, *Proceedings of the 2007 International Conference on Wireless Communications, Networking and Mobile Computing* (WiCom 2007), pp. 5025

> The mission of the United States (US) National Park Service (NPS) is to "conserve the scenery and the natural and historic objects and the wild life therein and to provide for the enjoyment of the same in such manner and by such means as will leave them unimpaired for the enjoyment of future generations" (NPS, 1916). NPS currently manages 397 parks covering about 358,200 km2, or approximately 4% of all US states and territories. The National Park system includes approximately 300 parks that are considered to contain significant natural resources. These parks are key components of a larger network of protected areas that anchor the conservation of natural resources in the US. They also afford direct protection for a number of important and defining resources in the US, including 421 species of threatened or endangered plants and animals, nearly two-thirds of native fishes in the 50 states (Lawrence *et al*., 2011), the highest point in North America (Mt. McKinley in Denali National Park, 6,194 m), the longest cave system in the world (Mammoth Cave National Park with more than 587 mapped km of caves), the country's deepest lake (Crater Lake in Crater Lake National Park, 589 m), the lowest terrestrial point in the western hemisphere (Badwater Basin in Death Valley National Park at 86 m below sea level), and – within these extremes – many other natural resources that are significant at local, regional, and national scales.

> While protected areas are foundational to a strong natural resource conservation network, ecologists have long recognized that virtually all parks are too small to be self-sustaining ecosystems, and activities outside park boundaries can profoundly impact park resources (Newmark, 1985; US General Accounting Office, 1994; Parks & Harcourt, 2002; Wiersma *et al*., 2004; Hansen & DeFries, 2007; Hansen *et al*., 2011). Chief among these activities is the intensification of land use and the appropriation of ecological services. Land use intensification leads to the conversion of natural habitat, which generally results in an overall loss of habitat, fragmentation of remaining natural areas, increases in edge zones, changes in the runoff of water, sediments, and nutrients, and follow-on modification of physical and ecological processes in terrestrial and aquatic ecosystems. Depending on the

Upstream Landscape Dynamics of

watershed context?

vary geographically?

**2. Assessing watershed condition** 

**2.1 Ecological foundation** 

watersheds?

US National Parks with Implications for Water Quality and Watershed Management 29

2. Which major landscape factors explain most among-park variation in upstream

3. What can we infer about the condition of park freshwater resources, and how do these

4. What are the major challenges and opportunities for managing park upstream

We first describe the ecological foundation and general approach to evaluating park upstream watersheds. We proceed to describe the selection of variables and data sources used in the analyses, and briefly review the ecological basis for including those variables. We then outline the analytical techniques and criteria for including or omitting parks from the study. The final sections of the chapter present the results of our watershed and water quality analyses, and our interpretation of these results. We conclude with a summary of

Figure 1 illustrates our overall conceptual model for assessing parks at a landscape scale. The model acknowledges key anthropogenic drivers (or stressors), important attributes of the natural landscape, and contextual elements that affect conservation and management actions. Analyses that consider these elements can evaluate geospatially explicit broad-scale vulnerabilities and opportunities for conservation and management. Our model is founded on more comprehensive analyses by Hansen & DeFries, (2007) of the mechanisms that link

our principal findings and recommendations for future research.

land use intensification to the resources within protected areas.

Fig. 1. Conceptual model used as a basis for landscape-scale assessment of parks.

location, extent, and magnitude of these anthropogenic changes, the effects may propagate over very large areas and have important consequences for resource management in protected areas.

While the NPS mission is to protect all natural resources, water is perhaps the most universally important resource to parks and to protected areas worldwide. Provision of fresh water is a key ecosystem service provided by many parks, and wetlands and riparian habitats are often biological 'hot spots' that support disproportionately high levels of biodiversity (Stein *et al.*, 2000; Scott *et al*., 2001). Because fresh water resources are so important to parks, the focus of this chapter is on landscape-scale factors that affect water resources and associated values. Flowing water directly connects water resources inside and outside park boundaries. Landscape-scale activities beyond park boundaries can particularly affect water resources and the ability of parks to manage and protect these resources. A means to identify and quantify imposing threats is thus important to designing and implementing effective park management strategies.

To manage an extensive network of protected areas like the NPS system of parks, there is a clear need to assess the system-wide context and status of parks relative to their goals (Scott *et al.*, 2001; Svancara *et al*., 2005). Results from broad-scale analyses can identify patterns and trends that may be undetectable at the individual unit scale, and provide guidance for changes in regional or national-level policy. Scott *et al*., (2001), for example, noted that US protected areas (parks, refuges, etc) disproportionately represent lands characterized by high elevation, low productivity, and low rainfall – the places that are cold, dry, and barren. Areas in highly productive river valleys – the location of many biodiversity hot spots – were disproportionately under-represented in the network of protected areas. The widespread availability of broad-coverage, geospatial data on environmental conditions and landscape attributes has facilitated new and sophisticated analyses of the geographical context and anthropogenic impacts to terrestrial, freshwater, coastal, and marine ecosystems at regional, national, and global scales (Sanderson *et al*., 2002; Halpern *et al*., 2007, 2008; Leu *et al*., 2008; Woolmer *et al*., 2008; Lawrence *et al*., 2011).

While a few studies have measured and assessed the landscape characteristics of US National Parks (Scott *et al.*, 2001; Svancara *et al*., 2009; Davis & Hansen, in press; Wade *et al.*, 2011), these efforts focused on the broader landscape context or specific components of the landscape, rather than watersheds, even though water is one of the most defining resources for parks and other protected areas (Dixon & Sherman, 1991; Hawkins *et al.*, 2003). To our knowledge, only Lawrence *et al.*, (2011) have rigorously evaluated system-wide the upstream landscape dynamics of US National Parks, specifically from the perspective of maintaining protection for freshwater fish diversity. They used broad-scale datasets to assess both threats to the use of parks as 'freshwater protected areas' and the potential capacity to manage activities in the contributing watersheds. Based on a relatively simple set of analyses, but involving computationally intensive operations, they were able to identify a subset of parks that could serve as the foundation for a system that would likely preserve a large proportion of US freshwater fish.

To guide the analyses in this chapter, we asked a series of questions:

1. Based on established ecological principles and landscape-scale data, what is the general context of park upstream watersheds?


We first describe the ecological foundation and general approach to evaluating park upstream watersheds. We proceed to describe the selection of variables and data sources used in the analyses, and briefly review the ecological basis for including those variables. We then outline the analytical techniques and criteria for including or omitting parks from the study. The final sections of the chapter present the results of our watershed and water quality analyses, and our interpretation of these results. We conclude with a summary of our principal findings and recommendations for future research.

### **2. Assessing watershed condition**

### **2.1 Ecological foundation**

28 Sustainable Natural Resources Management

location, extent, and magnitude of these anthropogenic changes, the effects may propagate over very large areas and have important consequences for resource management in

While the NPS mission is to protect all natural resources, water is perhaps the most universally important resource to parks and to protected areas worldwide. Provision of fresh water is a key ecosystem service provided by many parks, and wetlands and riparian habitats are often biological 'hot spots' that support disproportionately high levels of biodiversity (Stein *et al.*, 2000; Scott *et al*., 2001). Because fresh water resources are so important to parks, the focus of this chapter is on landscape-scale factors that affect water resources and associated values. Flowing water directly connects water resources inside and outside park boundaries. Landscape-scale activities beyond park boundaries can particularly affect water resources and the ability of parks to manage and protect these resources. A means to identify and quantify imposing threats is thus important to designing

To manage an extensive network of protected areas like the NPS system of parks, there is a clear need to assess the system-wide context and status of parks relative to their goals (Scott *et al.*, 2001; Svancara *et al*., 2005). Results from broad-scale analyses can identify patterns and trends that may be undetectable at the individual unit scale, and provide guidance for changes in regional or national-level policy. Scott *et al*., (2001), for example, noted that US protected areas (parks, refuges, etc) disproportionately represent lands characterized by high elevation, low productivity, and low rainfall – the places that are cold, dry, and barren. Areas in highly productive river valleys – the location of many biodiversity hot spots – were disproportionately under-represented in the network of protected areas. The widespread availability of broad-coverage, geospatial data on environmental conditions and landscape attributes has facilitated new and sophisticated analyses of the geographical context and anthropogenic impacts to terrestrial, freshwater, coastal, and marine ecosystems at regional, national, and global scales (Sanderson *et al*., 2002; Halpern *et al*., 2007, 2008; Leu *et al*., 2008;

While a few studies have measured and assessed the landscape characteristics of US National Parks (Scott *et al.*, 2001; Svancara *et al*., 2009; Davis & Hansen, in press; Wade *et al.*, 2011), these efforts focused on the broader landscape context or specific components of the landscape, rather than watersheds, even though water is one of the most defining resources for parks and other protected areas (Dixon & Sherman, 1991; Hawkins *et al.*, 2003). To our knowledge, only Lawrence *et al.*, (2011) have rigorously evaluated system-wide the upstream landscape dynamics of US National Parks, specifically from the perspective of maintaining protection for freshwater fish diversity. They used broad-scale datasets to assess both threats to the use of parks as 'freshwater protected areas' and the potential capacity to manage activities in the contributing watersheds. Based on a relatively simple set of analyses, but involving computationally intensive operations, they were able to identify a subset of parks that could serve as the foundation for a system that would likely preserve a

1. Based on established ecological principles and landscape-scale data, what is the general

and implementing effective park management strategies.

Woolmer *et al*., 2008; Lawrence *et al*., 2011).

large proportion of US freshwater fish.

context of park upstream watersheds?

To guide the analyses in this chapter, we asked a series of questions:

protected areas.

Figure 1 illustrates our overall conceptual model for assessing parks at a landscape scale. The model acknowledges key anthropogenic drivers (or stressors), important attributes of the natural landscape, and contextual elements that affect conservation and management actions. Analyses that consider these elements can evaluate geospatially explicit broad-scale vulnerabilities and opportunities for conservation and management. Our model is founded on more comprehensive analyses by Hansen & DeFries, (2007) of the mechanisms that link land use intensification to the resources within protected areas.

Fig. 1. Conceptual model used as a basis for landscape-scale assessment of parks.

Upstream Landscape Dynamics of

Housing Spatially Explicit

Roads Environmental

Land cover National Land

Conservation status

**Measure Source data Years Spatial** 

Regional Growth Model (SERGoM)

Systems Research Institute (ESRI)

Cover Data (NLCD)

(PAD-US)

Protected Areas Database of the US

Land cover Impervious surface %, area

development

development

development

Table 2. NPScape variables used in the present analyses.

Low intensity

High intensity

Conservation status

Medium intensity

Population US Census Bureau 2000 Census block

US National Parks with Implications for Water Quality and Watershed Management 31

Varies, up to 2005

Varies, up to 2010

**Measure Variable Units Comments**

 Exurban housing % area 7-145 units/km2 Suburban housing % area 146-1,234 units/km2

Developed % area Anderson Level I

 Agriculture % area Anderson Level I Cultivated crops % area Anderson Level II Hay/pasture % area Anderson Level II

Landowner density count/km2 All owners of conservation

Table 1. NPScape data sources used to compute the landscape variables (listed in Table 2).

Rural housing % area 0-6 units/km2

Urban housing % area >1,234 units/km2 Commercial/industrial % area Business – non-residential Roads Weighted road density km/km2 Highway weighted by a factor

weighted

Developed open space % area Anderson Level II

Population Population density count/km2 Based on population totals Housing Housing density # units/km2 Based on mid-points of rural,

groups

**resolution Reference** 

2010 100 m cells Theobald, 2005; NPS, 2010b

2006 30 m cells NPS, 2010d; Fry *et al*., 2011

US Census Bureau, 2001;

Analysis Program, 2011

exurban, suburban , and urban

of 3, interstates by 10

(GAP) codes 1 and 2

lands, including NGO & private

% area Anderson Level II

% area Anderson Level II

% area Anderson Level II

Protected % area Based on Gap Analysis Program

Anthropogenic sources only

NPS, 2010a

Varies ESRI, 2010; NPS, 2010c

Varies NPS, 2011a; USGS Gap

Broad-scale data generally available include the human drivers represented in the model, and all of these drivers are well known to influence biodiversity and other park resources. Natural systems can be characterized in many ways, and the types of attributes in Figure 1 are a subset of important attributes that can be used to assess the context and condition of natural systems. The conservation context provides information that may be essential to decisions on land management. Svancara *et al.*, (2009) conducted a national-level analysis by county of the conservation context of US national parks and refuges, and they discuss the use of this information to achieve conservation goals.

### **2.2 Landscape variables and data sources**

Using the conceptual model as an overarching framework, we approached our analyses within the broader goals of the NPS landscape dynamics monitoring project, NPScape (http://science.nature.nps.gov/im/monitor/npscape). NPScape is designed to support landscape-scale monitoring conducted by the NPS Inventory and Monitoring Program (Fancy *et al*., 2009). Key NPScape objectives are to provide: a coherent conceptual and analytical framework for conducting landscape-scale analyses and evaluations that can inform decisions; Geographic Information System (GIS) data and maps at broad spatial scales that transcend the bounds of park-level data; well-documented methods founded on strong science that are readily repeatable and extensible with local data; and, assistance to parks in interpreting results.

In support of these objectives, NPScape produces and delivers a suite of landscape-scale datasets, methods, GIS scripts and tools, maps, and guidance reports to approximately 300 natural resource parks in the NPS system. Results from NPScape are intended to inform resource management and planning at multiple scales. Because the overarching goal of NPScape is to deliver information to parks across the entire NPS system, inputs are limited to data sources that cover broad spatial extents (i.e., regional to national).

NPScape incorporates a large number of datasets that can be roughly categorized into 'base layers' and 'variables'. Base layers are such things as topography, jurisdictional boundaries, hydrography, and the other geospatial data that are relatively static and that generally serve as covariates or provide a geopolitical context. The NPScape variables considered here address two major elements in our conceptual model: stressors and conservation context.

Important stressors are evaluated by measures of population, housing, roads, and impervious surface (a category of land cover). Conservation context was evaluated from the percentage of land in a protected status, and potential management partnerships from the number of different agency or institutional owners of conservation lands. The NPScape data sources and variables used in our analyses are described in Tables 1 and 2. Although NPScape includes a variety of other metrics related to natural land cover and landscape pattern, we did not use these in our national-level assessment because they require a more in-depth analysis at ecoregional scales. Our present focus on human drivers and conservation context is designed to serve as a foundation to these future studies.

Broad-scale data generally available include the human drivers represented in the model, and all of these drivers are well known to influence biodiversity and other park resources. Natural systems can be characterized in many ways, and the types of attributes in Figure 1 are a subset of important attributes that can be used to assess the context and condition of natural systems. The conservation context provides information that may be essential to decisions on land management. Svancara *et al.*, (2009) conducted a national-level analysis by county of the conservation context of US national parks and refuges, and they discuss the

Using the conceptual model as an overarching framework, we approached our analyses within the broader goals of the NPS landscape dynamics monitoring project, NPScape (http://science.nature.nps.gov/im/monitor/npscape). NPScape is designed to support landscape-scale monitoring conducted by the NPS Inventory and Monitoring Program (Fancy *et al*., 2009). Key NPScape objectives are to provide: a coherent conceptual and analytical framework for conducting landscape-scale analyses and evaluations that can inform decisions; Geographic Information System (GIS) data and maps at broad spatial scales that transcend the bounds of park-level data; well-documented methods founded on strong science that are readily repeatable and extensible with local data; and, assistance to

In support of these objectives, NPScape produces and delivers a suite of landscape-scale datasets, methods, GIS scripts and tools, maps, and guidance reports to approximately 300 natural resource parks in the NPS system. Results from NPScape are intended to inform resource management and planning at multiple scales. Because the overarching goal of NPScape is to deliver information to parks across the entire NPS system, inputs are limited

NPScape incorporates a large number of datasets that can be roughly categorized into 'base layers' and 'variables'. Base layers are such things as topography, jurisdictional boundaries, hydrography, and the other geospatial data that are relatively static and that generally serve as covariates or provide a geopolitical context. The NPScape variables considered here address two major elements in our conceptual model: stressors and

Important stressors are evaluated by measures of population, housing, roads, and impervious surface (a category of land cover). Conservation context was evaluated from the percentage of land in a protected status, and potential management partnerships from the number of different agency or institutional owners of conservation lands. The NPScape data sources and variables used in our analyses are described in Tables 1 and 2. Although NPScape includes a variety of other metrics related to natural land cover and landscape pattern, we did not use these in our national-level assessment because they require a more in-depth analysis at ecoregional scales. Our present focus on human drivers and conservation context is designed to serve as a foundation to these future

to data sources that cover broad spatial extents (i.e., regional to national).

use of this information to achieve conservation goals.

**2.2 Landscape variables and data sources** 

parks in interpreting results.

conservation context.

studies.



Table 1. NPScape data sources used to compute the landscape variables (listed in Table 2).

Table 2. NPScape variables used in the present analyses.

Upstream Landscape Dynamics of

considered in a larger NPS context.

Wild and Scenic River.

US National Parks with Implications for Water Quality and Watershed Management 33

We used these four datasets (flowlines, flow accumulation grids, flow direction grids, and park boundaries) as inputs to the NPScape ArcGIS watershed toolbox (NPS, 2011b) to generate upstream catchments for all 261 natural resource parks in the contiguous US. Most parks had multiple upstream catchments originating from different sets of pour points around park boundaries. We dissolved catchment boundaries by park to derive final park upstream watersheds. Importantly, watersheds were computed with respect to the entire network of parks, meaning that upstream catchments were delineated based on the most proximate park in the NPS system. This decision was made in order to evaluate the landscape factors that relate directly to each park. Furthermore, because many parks occur in major river systems, it helped ensure that upstream watersheds were small enough to be practical for park management considerations, yet still ecologically relevant when

From these outputs we applied a series of quality-control filters to eliminate park upstream watersheds with obvious inaccuracies (NPS, 2011b). We eliminated park watersheds where there were obvious errors with the source NHD Plus data, parks that were too small in relation to the spatial resolution and mapping accuracy of the source data, and parks that were in areas with complex hydrography (e.g., coastal, marine). These filters eliminated 110 of the 261 natural resource parks in the contiguous US, leaving a total of 151 focal parks and

their contributing upstream watersheds that were considered in the analyses (Fig. 2).

Fig. 2. Focal parks and upstream watersheds considered in the present analyses. Labelled parks are referred to in the text. NP = National Park, NRA = National Recreation Area, NMP = National Military Park, NM = National Monument, NMem = National Memorial, WSR =

In addition to NPScape variables, we included the percentage of each watershed in private ownership (derived from USGS Gap Analysis Program, 2011), and we obtained data on river impoundments and nitrogen (N) deposition. The US National Inventory of Dams database now contains more than 83,000 records of significant impoundments in US states and territories (US Army Corps of Engineers, 2010). These impoundments have dramatically altered hydrological flow regimes, sedimentation processes, inhibited or prevented biological migrations and other movements, and influenced virtually every ecological process in some catchments (Ward & Stanford, 1979). Following Sabo *et al*., (2010) and Lawrence *et al*., (2011) we used the density of impoundments (count/km2) in the contributing watershed as an indicator of river fragmentation. A future refinement to our analyses is to include additional information on the characteristics of dams and their effects on aquatic resources. For example, watersheds in the eastern US tend to have a higher density of dams than western watersheds, but because the size of dams varies, the storage capacity as a portion of annual flow is nearly the same in the east and west (Sabo *et al*., 2010). Thus western rivers generally have fewer but larger dams, so fragmentation is greater in the east but dams in the west alter hydrological dynamics more.

Anthropogenic activities now contribute more nitrogen (N) to the global cycle than all natural sources combined (Vitousek, 1997). Nitrogen is, or was, a key limiting element in many aquatic systems, but these limits are now exceeded in many parks (Baron *et al*., 2011). Baron *et al*., (2011) reviewed studies on N limitations in North American lakes. These studies revealed a consistent pattern of historical N limitations, especially in nutrient-poor environments typical of high elevations, and undisturbed temperate and boreal forests. Broad patterns of response to atmospheric N deposition further supported assertions that the majority of lakes in the Northern Hemisphere may have been N-limited prior to increased N deposition from anthropogenic sources. Atmospheric deposition of N has sufficiently altered the balance of N and Phosphorous (P) so that P limitations are now more commonly observed in North American lakes. These results emphasize the need to incorporate aspects of global change in broad-scale studies. Our analyses of N deposition are based on measurement of inorganic N wet deposition from the National Atmospheric Deposition Program (NADP), corrected for topographic precipitation differences using PRISM (Parameter-elevation Regressions on Independent Slopes Model) climate data as described in Baron *et al*., (2011). These data underestimate total N deposition because they do not account for dry deposition. We did not attempt to account for terrestrial runoff or other N inputs.

### **2.3 Upstream watersheds and the headwater index**

Watersheds that are either upstream or downstream with respect to a particular management unit may be calculated from Digital Elevation Models (DEMs) using GIS (Djokic & Ye, 1999). The park upstream watersheds considered here were calculated by NPScape using this basic methodology (NPS, 2011b). However, rather than relying on source DEM data, NPScape was able to take advantage of published DEM-derived datasets that serve as standardized pre-processed inputs to watershed calculations: the National Hydrology Dataset Plus (NHD Plus, 2010) vector flowlines, NHD Plus flow accumulation rasters, and NHD Plus flow direction rasters. In addition to these inputs, our calculations required NPS current administrative vector boundaries (NPS, 2011c) to determine pour points from the flowlines.

In addition to NPScape variables, we included the percentage of each watershed in private ownership (derived from USGS Gap Analysis Program, 2011), and we obtained data on river impoundments and nitrogen (N) deposition. The US National Inventory of Dams database now contains more than 83,000 records of significant impoundments in US states and territories (US Army Corps of Engineers, 2010). These impoundments have dramatically altered hydrological flow regimes, sedimentation processes, inhibited or prevented biological migrations and other movements, and influenced virtually every ecological process in some catchments (Ward & Stanford, 1979). Following Sabo *et al*., (2010) and Lawrence *et al*., (2011) we used the density of impoundments (count/km2) in the contributing watershed as an indicator of river fragmentation. A future refinement to our analyses is to include additional information on the characteristics of dams and their effects on aquatic resources. For example, watersheds in the eastern US tend to have a higher density of dams than western watersheds, but because the size of dams varies, the storage capacity as a portion of annual flow is nearly the same in the east and west (Sabo *et al*., 2010). Thus western rivers generally have fewer but larger dams, so fragmentation is greater

Anthropogenic activities now contribute more nitrogen (N) to the global cycle than all natural sources combined (Vitousek, 1997). Nitrogen is, or was, a key limiting element in many aquatic systems, but these limits are now exceeded in many parks (Baron *et al*., 2011). Baron *et al*., (2011) reviewed studies on N limitations in North American lakes. These studies revealed a consistent pattern of historical N limitations, especially in nutrient-poor environments typical of high elevations, and undisturbed temperate and boreal forests. Broad patterns of response to atmospheric N deposition further supported assertions that the majority of lakes in the Northern Hemisphere may have been N-limited prior to increased N deposition from anthropogenic sources. Atmospheric deposition of N has sufficiently altered the balance of N and Phosphorous (P) so that P limitations are now more commonly observed in North American lakes. These results emphasize the need to incorporate aspects of global change in broad-scale studies. Our analyses of N deposition are based on measurement of inorganic N wet deposition from the National Atmospheric Deposition Program (NADP), corrected for topographic precipitation differences using PRISM (Parameter-elevation Regressions on Independent Slopes Model) climate data as described in Baron *et al*., (2011). These data underestimate total N deposition because they do not account for dry deposition. We did not

Watersheds that are either upstream or downstream with respect to a particular management unit may be calculated from Digital Elevation Models (DEMs) using GIS (Djokic & Ye, 1999). The park upstream watersheds considered here were calculated by NPScape using this basic methodology (NPS, 2011b). However, rather than relying on source DEM data, NPScape was able to take advantage of published DEM-derived datasets that serve as standardized pre-processed inputs to watershed calculations: the National Hydrology Dataset Plus (NHD Plus, 2010) vector flowlines, NHD Plus flow accumulation rasters, and NHD Plus flow direction rasters. In addition to these inputs, our calculations required NPS current administrative vector boundaries (NPS, 2011c) to determine pour

in the east but dams in the west alter hydrological dynamics more.

attempt to account for terrestrial runoff or other N inputs.

**2.3 Upstream watersheds and the headwater index** 

points from the flowlines.

We used these four datasets (flowlines, flow accumulation grids, flow direction grids, and park boundaries) as inputs to the NPScape ArcGIS watershed toolbox (NPS, 2011b) to generate upstream catchments for all 261 natural resource parks in the contiguous US. Most parks had multiple upstream catchments originating from different sets of pour points around park boundaries. We dissolved catchment boundaries by park to derive final park upstream watersheds. Importantly, watersheds were computed with respect to the entire network of parks, meaning that upstream catchments were delineated based on the most proximate park in the NPS system. This decision was made in order to evaluate the landscape factors that relate directly to each park. Furthermore, because many parks occur in major river systems, it helped ensure that upstream watersheds were small enough to be practical for park management considerations, yet still ecologically relevant when considered in a larger NPS context.

From these outputs we applied a series of quality-control filters to eliminate park upstream watersheds with obvious inaccuracies (NPS, 2011b). We eliminated park watersheds where there were obvious errors with the source NHD Plus data, parks that were too small in relation to the spatial resolution and mapping accuracy of the source data, and parks that were in areas with complex hydrography (e.g., coastal, marine). These filters eliminated 110 of the 261 natural resource parks in the contiguous US, leaving a total of 151 focal parks and their contributing upstream watersheds that were considered in the analyses (Fig. 2).

Fig. 2. Focal parks and upstream watersheds considered in the present analyses. Labelled parks are referred to in the text. NP = National Park, NRA = National Recreation Area, NMP = National Military Park, NM = National Monument, NMem = National Memorial, WSR = Wild and Scenic River.

Upstream Landscape Dynamics of

**3. Results and discussion** 

US National Parks with Implications for Water Quality and Watershed Management 35

For water quality, we used Pearson's product moment correlation to characterize the association between the percentage of park waterway impaired (arcsine transformed) and PC1. We limited this correlation to PC1 because it explained the majority of landscape variation among park upstream watersheds. Meanwhile, an initial examination of STORET water quality data revealed implausible observations (outliers) for some variables. To reduce the effect of outliers in our analyses, we calculated the 95th percentile of the distribution for each variable and then multiplied this value by 3 (P3) and by 20 (P20). We removed all observations with values greater than P20. For observations with values between P3 and P20, we changed the observed value to the value of P3 (i.e., we truncated the distribution to ± P3). To obtain a single value for each variable and each park, we first calculated the median value of the observation for each site within a specific park area of analysis. We then calculated the mean of the site-specific medians for that area. We used linear regression with the park-specific mean values and our predictor variables (i.e., PC1, PC2, and a subset of NPScape variables in Table 2) to explore relationships between water quality and landscape attributes. After filtering the STORET data for date, location, and

outliers, our analyses were based on usable data for 29-117 parks (mean = 78).

developed land cover, and 30% are greater than 10% agricultural land cover.

below in the context of watershed management opportunities for parks (Section 3.4).

Park upstream watersheds are potentially threatened by a number of landscape-level factors related to park-watershed geometry, housing development, habitat conversion and resource extraction, and N deposition (Fig. 3). Of the 151 park upstream watersheds considered, 81% have more than 50% of their watersheds extending beyond park boundaries, 77% have less than 50% area formally protected, 61% have greater than 10% rural development, and 37% have values for N deposition exceeding 3.5 kg N ha-1 yr-1 – a conservative critical load for most parts of the contiguous US (Baron *et al*., 2011). Taken in combination, these numbers suggest that most parks do not directly control most of their watersheds, and that both physical and chemical stressors originating beyond park boundaries will likely affect water resources inside park boundaries. However, despite these challenges, it is equally noteworthy that park upstream watersheds are relatively unthreatened by converted land cover, including high-intensity human land use (Fig. 3). Of the 151 park upstream watersheds, just 12% are greater than 50% converted land cover, 17% are greater than 10%

Several of these patterns merit further discussion. The low-level of protection afforded most park upstream watersheds is due in large part to the working definition of 'protected'. We consider parks and other areas 'protected' if they have permanent protection from conversion of natural land cover and a mandated management plan to maintain a primarily natural state. This definition follows from the US Geological Survey (USGS) Gap Analysis Program (GAP), which uses a series of four codes to rank areas based on their level of protection (USGS Gap Analysis Program, 2011). Our definition is based on GAP status codes 1 and 2 and includes most parks and all wilderness areas, but it excludes most lands managed by the Bureau of Land Management (BLM) and US Forest Service (USFS). These two Federal agencies combined manage approximately 1.8 million km2, which irrespective of their use and reduced level of protection represent significant areas for natural resource conservation. We revisit this subject

**3.1 What is the general context of park upstream watersheds?** 

We used the park boundaries and upstream watersheds to compute a geometric index of the degree to which a park includes its own headwaters. The headwater index was calculated by intersecting each park with its upstream watershed, then dividing that area by the total area of the upstream watershed. The resulting proportion ranged from zero (i.e., all upstream areas flowing into the park) to one (i.e., all upstream areas flowing out of the park).

### **2.4 Water quality variables and data sources**

We derived estimates of water quality inside focal parks from two sources: the NPS Hydrographic and Impairment Statistics (HIS) database (http://nature.nps.gov/water/his/) and the Environmental Protection Agency (EPA) Storage and Retrieval System (STORET; EPA, 2011). The HIS provided data for each park on the total length of waterway (rivers, streams, canals, etc), as well as the total length of 'impaired' waterway identified by states according to the federal Clean Water Act sections 303(d) and 305(b). We used these two variables to estimate the percentage of total waterway in each focal park that was impaired (impairment data were not available for the Rio Grande Wild and Scenic River). In addition, we downloaded water chemistry data from STORET for the area within a 3 km buffer outside park boundaries for all parks in this study, restricting the data to observations from 1995 present, and to samples from rivers/streams, lakes, and reservoirs. Although STORET provides access to a very large number of chemical and biological variables, we restricted our analyses to acid neutralizing capacity, ammonia-nitrogen as N, dissolved oxygen, nitrogennitrate, pH, phosphate-phosphorus as P, dissolved solids, and specific conductance.

### **2.5 Analyses**

We used a combination of univariate and multivariate methods to address our starting questions. Where possible we tried to emphasize univariate approaches, which are methodologically more intuitive and straightforward to interpret in a management context. However, because we considered a large number of landscape variables, we also needed a means to simplify analysis of the many correlated variables. To do so, we used principal component analysis (PCA) to identify broad orthogonal groupings of variables that explained most variation in park upstream watershed context. All statistical analyses were performed in R (R Development Core Team, 2011). Corresponding maps of select results were produced in ArcMap (ESRI, 2011); all maps are Albers equal area conic, NAD83.

The PCA was conducted using the landscape variables in Table 2, plus mean N deposition and dam density. Owing to non-normal distributions of the raw variables, arcsine transformations were first applied to all percentage (proportion) variables, and log transformations were used on all density variables. We excluded the headwater index from the PCA because we wanted to evaluate the major factors responsible for landscape-level change and management response (i.e., human drivers and conservation context) in park upstream watersheds, irrespective of their relatively static spatial geometries. We used the loadings of each transformed variable on principal components 1 and 2 (PC1, PC2) to interpret the meaning of each axis. Park-specific scores on PC1 and PC2 were then evaluated both geographically and in relation to the headwater index. For the latter, we regressed each principal component (dependent variable) on the arcsine transformed headwater index in order to explore the residuals and characterize the management potentials of non-headwater parks.

For water quality, we used Pearson's product moment correlation to characterize the association between the percentage of park waterway impaired (arcsine transformed) and PC1. We limited this correlation to PC1 because it explained the majority of landscape variation among park upstream watersheds. Meanwhile, an initial examination of STORET water quality data revealed implausible observations (outliers) for some variables. To reduce the effect of outliers in our analyses, we calculated the 95th percentile of the distribution for each variable and then multiplied this value by 3 (P3) and by 20 (P20). We removed all observations with values greater than P20. For observations with values between P3 and P20, we changed the observed value to the value of P3 (i.e., we truncated the distribution to ± P3). To obtain a single value for each variable and each park, we first calculated the median value of the observation for each site within a specific park area of analysis. We then calculated the mean of the site-specific medians for that area. We used linear regression with the park-specific mean values and our predictor variables (i.e., PC1, PC2, and a subset of NPScape variables in Table 2) to explore relationships between water quality and landscape attributes. After filtering the STORET data for date, location, and outliers, our analyses were based on usable data for 29-117 parks (mean = 78).

### **3. Results and discussion**

34 Sustainable Natural Resources Management

We used the park boundaries and upstream watersheds to compute a geometric index of the degree to which a park includes its own headwaters. The headwater index was calculated by intersecting each park with its upstream watershed, then dividing that area by the total area of the upstream watershed. The resulting proportion ranged from zero (i.e., all upstream areas

We derived estimates of water quality inside focal parks from two sources: the NPS Hydrographic and Impairment Statistics (HIS) database (http://nature.nps.gov/water/his/) and the Environmental Protection Agency (EPA) Storage and Retrieval System (STORET; EPA, 2011). The HIS provided data for each park on the total length of waterway (rivers, streams, canals, etc), as well as the total length of 'impaired' waterway identified by states according to the federal Clean Water Act sections 303(d) and 305(b). We used these two variables to estimate the percentage of total waterway in each focal park that was impaired (impairment data were not available for the Rio Grande Wild and Scenic River). In addition, we downloaded water chemistry data from STORET for the area within a 3 km buffer outside park boundaries for all parks in this study, restricting the data to observations from 1995 present, and to samples from rivers/streams, lakes, and reservoirs. Although STORET provides access to a very large number of chemical and biological variables, we restricted our analyses to acid neutralizing capacity, ammonia-nitrogen as N, dissolved oxygen, nitrogen-

flowing into the park) to one (i.e., all upstream areas flowing out of the park).

nitrate, pH, phosphate-phosphorus as P, dissolved solids, and specific conductance.

We used a combination of univariate and multivariate methods to address our starting questions. Where possible we tried to emphasize univariate approaches, which are methodologically more intuitive and straightforward to interpret in a management context. However, because we considered a large number of landscape variables, we also needed a means to simplify analysis of the many correlated variables. To do so, we used principal component analysis (PCA) to identify broad orthogonal groupings of variables that explained most variation in park upstream watershed context. All statistical analyses were performed in R (R Development Core Team, 2011). Corresponding maps of select results were produced in ArcMap (ESRI, 2011); all maps are Albers equal area conic, NAD83.

The PCA was conducted using the landscape variables in Table 2, plus mean N deposition and dam density. Owing to non-normal distributions of the raw variables, arcsine transformations were first applied to all percentage (proportion) variables, and log transformations were used on all density variables. We excluded the headwater index from the PCA because we wanted to evaluate the major factors responsible for landscape-level change and management response (i.e., human drivers and conservation context) in park upstream watersheds, irrespective of their relatively static spatial geometries. We used the loadings of each transformed variable on principal components 1 and 2 (PC1, PC2) to interpret the meaning of each axis. Park-specific scores on PC1 and PC2 were then evaluated both geographically and in relation to the headwater index. For the latter, we regressed each principal component (dependent variable) on the arcsine transformed headwater index in order to explore the

residuals and characterize the management potentials of non-headwater parks.

**2.4 Water quality variables and data sources** 

**2.5 Analyses** 

### **3.1 What is the general context of park upstream watersheds?**

Park upstream watersheds are potentially threatened by a number of landscape-level factors related to park-watershed geometry, housing development, habitat conversion and resource extraction, and N deposition (Fig. 3). Of the 151 park upstream watersheds considered, 81% have more than 50% of their watersheds extending beyond park boundaries, 77% have less than 50% area formally protected, 61% have greater than 10% rural development, and 37% have values for N deposition exceeding 3.5 kg N ha-1 yr-1 – a conservative critical load for most parts of the contiguous US (Baron *et al*., 2011). Taken in combination, these numbers suggest that most parks do not directly control most of their watersheds, and that both physical and chemical stressors originating beyond park boundaries will likely affect water resources inside park boundaries. However, despite these challenges, it is equally noteworthy that park upstream watersheds are relatively unthreatened by converted land cover, including high-intensity human land use (Fig. 3). Of the 151 park upstream watersheds, just 12% are greater than 50% converted land cover, 17% are greater than 10% developed land cover, and 30% are greater than 10% agricultural land cover.

Several of these patterns merit further discussion. The low-level of protection afforded most park upstream watersheds is due in large part to the working definition of 'protected'. We consider parks and other areas 'protected' if they have permanent protection from conversion of natural land cover and a mandated management plan to maintain a primarily natural state. This definition follows from the US Geological Survey (USGS) Gap Analysis Program (GAP), which uses a series of four codes to rank areas based on their level of protection (USGS Gap Analysis Program, 2011). Our definition is based on GAP status codes 1 and 2 and includes most parks and all wilderness areas, but it excludes most lands managed by the Bureau of Land Management (BLM) and US Forest Service (USFS). These two Federal agencies combined manage approximately 1.8 million km2, which irrespective of their use and reduced level of protection represent significant areas for natural resource conservation. We revisit this subject below in the context of watershed management opportunities for parks (Section 3.4).

Upstream Landscape Dynamics of

US National Parks with Implications for Water Quality and Watershed Management 37

Park upstream watersheds are bimodally distributed with respect to N deposition (Fig. 3). This bimodality is strongly influenced by a combination of longitude and elevation. Based on critical loads from Baron *et al*., (2011), all park upstream watersheds in the east exceed the critical load of 3.5 kg N ha-1 yr-1 reported for the northeast; Yosemite, Sequoia, and Kings Canyon National Parks all exceed the critical load of 1.5 kg N ha-1 yr-1 reported for the Sierra Nevada; and, all parks in the Central Rockies exceed the critical load of 1.0 kg N ha-1 yr-1 reported for the Rocky Mountains (Fig. 4). Hence, despite geographic variation in N deposition across parks in the contiguous US, most park upstream watersheds considered here have values exceeding critical loads for their respective geographies. Future work is needed to incorporate more detailed geographic estimates of critical loads for N deposition.

Fig. 4. N deposition in park upstream watersheds, with legend categories reflecting critical

Rural development (<7 housing units km-2; Theobald, 2005) has already occurred over extensive areas in most park upstream watersheds, and there is great concern about the rate of development of rural landscapes around parks (Hansen *et al*., 2005; Wade & Theobald, 2009; Radeloff *et al*., 2010). Increases in the extent of low-density housing in previously undeveloped areas has numerous biological impacts (Hansen *et al*., 2002, 2005) and housing development is increasingly recognized as a primary driver of ecological processes and as a threat to biodiversity (McKinney, 2002; Miller & Hobbs, 2002). In the Greater Yellowstone Ecosystem, which is threatened by exurban development, riparian habitat, elk winter range, migration corridors, and other important habitat and biodiversity indices are expected to experience substantial conversion (between 5% and 40%) by 2020 (Gude *et al*., 2007). Hence, this driver will be increasingly important for ongoing and future management of park

Lastly, dam density is characteristically low in most park upstream watersheds (Fig. 3; mean = 0.02 dams/km2), but it is important to note that ecologically relevant thresholds for this

loads described by Baron *et al*., (2011) for different areas of the US.

watersheds.

Fig. 3. Univariate distributions of select landscape variables for upstream contributing watersheds of 151 National Parks in the contiguous US. Dashed lines show means; dotted lines show medians.

Fig. 3. Univariate distributions of select landscape variables for upstream contributing watersheds of 151 National Parks in the contiguous US. Dashed lines show means; dotted

lines show medians.

Park upstream watersheds are bimodally distributed with respect to N deposition (Fig. 3). This bimodality is strongly influenced by a combination of longitude and elevation. Based on critical loads from Baron *et al*., (2011), all park upstream watersheds in the east exceed the critical load of 3.5 kg N ha-1 yr-1 reported for the northeast; Yosemite, Sequoia, and Kings Canyon National Parks all exceed the critical load of 1.5 kg N ha-1 yr-1 reported for the Sierra Nevada; and, all parks in the Central Rockies exceed the critical load of 1.0 kg N ha-1 yr-1 reported for the Rocky Mountains (Fig. 4). Hence, despite geographic variation in N deposition across parks in the contiguous US, most park upstream watersheds considered here have values exceeding critical loads for their respective geographies. Future work is needed to incorporate more detailed geographic estimates of critical loads for N deposition.

Fig. 4. N deposition in park upstream watersheds, with legend categories reflecting critical loads described by Baron *et al*., (2011) for different areas of the US.

Rural development (<7 housing units km-2; Theobald, 2005) has already occurred over extensive areas in most park upstream watersheds, and there is great concern about the rate of development of rural landscapes around parks (Hansen *et al*., 2005; Wade & Theobald, 2009; Radeloff *et al*., 2010). Increases in the extent of low-density housing in previously undeveloped areas has numerous biological impacts (Hansen *et al*., 2002, 2005) and housing development is increasingly recognized as a primary driver of ecological processes and as a threat to biodiversity (McKinney, 2002; Miller & Hobbs, 2002). In the Greater Yellowstone Ecosystem, which is threatened by exurban development, riparian habitat, elk winter range, migration corridors, and other important habitat and biodiversity indices are expected to experience substantial conversion (between 5% and 40%) by 2020 (Gude *et al*., 2007). Hence, this driver will be increasingly important for ongoing and future management of park watersheds.

Lastly, dam density is characteristically low in most park upstream watersheds (Fig. 3; mean = 0.02 dams/km2), but it is important to note that ecologically relevant thresholds for this

Upstream Landscape Dynamics of

A

B

US National Parks with Implications for Water Quality and Watershed Management 39

Fig. 5. Principal component scores 1 (A) and 2 (B) shown for park upstream watersheds. Orange and red colours denote watersheds that have higher PC scores (higher threat), in

units of standard deviations (SD).

variable are also low and likely close to this mean for many natural resources, especially when considered in the context of dam size. For example, a single large dam may affect water temperatures and benthic communities for hundreds of kilometres downstream (Baxter, 1977). In addition, higher densities of small dams may have cumulative effects on physicochemistry and macroinvertebrate diversity that exceed those of large dams (Mantel *et al*., 2010). Single dams may also create serious obstacles to the long-range movement of fish, either upstream (e.g., anadromous salmon) or downstream (e.g., catadromous eels). In brief, dams have pervasive and varied effects on aquatic resources (Ward & Stanford, 1979), and the analyses presented here would greatly benefit from an expanded set of ecologically informative variables and thresholds related to impoundments.

### **3.2 Which landscape factors explain most variation in park upstream watersheds?**

Using the human driver and conservation context variables shown in Figure 3, plus additional physical landscape variables described under landscape variables and data sources (Section 2.2), we conducted a PCA to understand which of the 21 landscape factors explained most of the among-park variation in upstream watershed context. Principal components 1 and 2 (PC1, PC2) explained 51% and 15% (respectively) of the variation (66% total). PC1 loaded positively on several variables related to urban development, while PC2 loaded positively on variables related to both agriculture and N deposition, and negatively on the amount of protected area (Table 3). According to both axes, higher values (denoting higher urban development, agriculture, and N deposition; less protected area) are associated with parks east of the Rocky Mountains (Fig. 5). Dam density loaded most strongly on PC4, but this axis explained only 6% of the variation and is thus not shown.


Table 3. Principal component analysis loadings by variable for axes 1 and 2 (PC1, PC2). Values are grouped on each column to facilitate interpretation of the axes.

variable are also low and likely close to this mean for many natural resources, especially when considered in the context of dam size. For example, a single large dam may affect water temperatures and benthic communities for hundreds of kilometres downstream (Baxter, 1977). In addition, higher densities of small dams may have cumulative effects on physicochemistry and macroinvertebrate diversity that exceed those of large dams (Mantel *et al*., 2010). Single dams may also create serious obstacles to the long-range movement of fish, either upstream (e.g., anadromous salmon) or downstream (e.g., catadromous eels). In brief, dams have pervasive and varied effects on aquatic resources (Ward & Stanford, 1979), and the analyses presented here would greatly benefit from an expanded set of ecologically

**3.2 Which landscape factors explain most variation in park upstream watersheds?**  Using the human driver and conservation context variables shown in Figure 3, plus additional physical landscape variables described under landscape variables and data sources (Section 2.2), we conducted a PCA to understand which of the 21 landscape factors explained most of the among-park variation in upstream watershed context. Principal components 1 and 2 (PC1, PC2) explained 51% and 15% (respectively) of the variation (66% total). PC1 loaded positively on several variables related to urban development, while PC2 loaded positively on variables related to both agriculture and N deposition, and negatively on the amount of protected area (Table 3). According to both axes, higher values (denoting higher urban development, agriculture, and N deposition; less protected area) are associated with parks east of the Rocky Mountains (Fig. 5). Dam density loaded most strongly on PC4,

**Variable PC1 PC2**  Urban development 0.29 -0.12 Low intensity development 0.29 -0.12 Population density 0.28 0.04 Suburban housing 0.28 -0.14 Urban housing 0.27 -0.12 Medium intensity development 0.27 -0.18 Developed open space 0.27 -0.01 High intensity development 0.26 -0.18 Agriculture 0.17 0.42 Cultivated crops 0.09 0.38 Rural housing -0.04 0.38 Pasture/hay 0.18 0.31 Nitrogen deposition 0.16 0.30 Protected area -0.14 -0.28 Commercial/industrial 0.25 -0.21 Housing density 0.24 0.16 Exurban housing 0.23 0.20 Impervious surface 0.22 -0.08 Weighted road density 0.17 -0.08 Owner density 0.06 -0.13 Dam density 0.13 -0.04 Table 3. Principal component analysis loadings by variable for axes 1 and 2 (PC1, PC2). Values

informative variables and thresholds related to impoundments.

but this axis explained only 6% of the variation and is thus not shown.

are grouped on each column to facilitate interpretation of the axes.

Fig. 5. Principal component scores 1 (A) and 2 (B) shown for park upstream watersheds. Orange and red colours denote watersheds that have higher PC scores (higher threat), in units of standard deviations (SD).

Upstream Landscape Dynamics of

phosphate-phosphorus (Fig. 7).

US National Parks with Implications for Water Quality and Watershed Management 41

with approximately 2 million observations. The large number of observations required automated processes to screen data. For these preliminary analyses, we did not attempt to correct for factors such as season, variation in sampling effort, or flow regime. Despite the absence of strong statistical associations between water chemistry and landscape context, regional patterns were apparent for most of the chemistry variables we examined, such as

Fig. 7. Concentration of phosphate-phosphorus in focal parks, with legend categories reflecting conservative critical thresholds described by Van Sickle *et al.*, (2006) for total P. Note that concentrations were calculated for each park with 3 km buffer, but results are symbolized here by park upstream watershed in order to facilitate comparisons with the other maps. Watersheds in light grey denote parks that were not sampled for this metric.

The absence of strong statistical relationships between landscape and water quality variables in our national-scale assessment indicates the need for more sophisticated analyses when working at these very broad scales and with generalized databases. Other studies have found considerably stronger relationships between land cover and water chemistry (e.g., King *et al*., 2005; Wickham *et al*., 2005; Riva-Murray *et al*., 2010). Our future efforts will include more sophisticated processes for screening water chemistry data, and additional analyses. For example, King *et al*., (2005) evaluated a water quality index based on binary (0, 1) values for predictor variables that were above or below a quality threshold. Their index was the sum or four predictor variables. The binary transformation effectively addresses issues with high variance in the predictor variables, and it simplifies estimation and interpretation of the index. In addition, evaluations of the relative contributions of land cover versus broad-scale environmental setting to determining water chemistry are

### **3.3 What can we infer about the condition of park freshwater resources?**

Based on the percentage of impaired waterway, 62%, 64%, and 70% of parks (respectively) have less than 5%, 10%, and 20% of their total waterways in non-compliance with federal Clean Water Act sections 303(d) and 305(b) (Fig. 6). However, in this context it is important to note that 'impairment' standards vary by state and are generally less stringent than critical ecological thresholds in most parks. Furthermore, the sources of waterway 'impairment' do not all originate from park upstream watersheds. For these reasons, one would not *a priori* expect a substantial amount of among-park variation in waterway impairment to be explained solely by the landscape dynamics of upstream watersheds. We find that the percentage of park waterway impaired is positively correlated with PC1, and that this variable explains approximately 26% of variation.

Fig. 6. The percentage of impaired waterway in focal parks. Note that the summary statistic was calculated for each park, but results are symbolized here by park upstream watershed to facilitate comparisons with the other maps.

When compared to ecological threshold values for poor or good water quality (e.g., Van Sickle *et al*., 2006; Wazniak *et al*., 2007; Riva-Murray *et al*., 2010), water quality in and near most parks is in a good range. These results reflect the landscape location of most park watersheds, which tend to include a high portion of natural land cover and a relatively small area of cropland or intensive development. Using simple regression analyses, we generally found weak relationships between STORET water chemistry variables and watershed landscape variables. Certain attributes of the data likely contributed to our inability to link these factors. We wished to evaluate the ability of large, broad-extent databases to inform regional and national-scale analyses, and we thus began our analyses

Based on the percentage of impaired waterway, 62%, 64%, and 70% of parks (respectively) have less than 5%, 10%, and 20% of their total waterways in non-compliance with federal Clean Water Act sections 303(d) and 305(b) (Fig. 6). However, in this context it is important to note that 'impairment' standards vary by state and are generally less stringent than critical ecological thresholds in most parks. Furthermore, the sources of waterway 'impairment' do not all originate from park upstream watersheds. For these reasons, one would not *a priori* expect a substantial amount of among-park variation in waterway impairment to be explained solely by the landscape dynamics of upstream watersheds. We find that the percentage of park waterway impaired is positively correlated with PC1, and

Fig. 6. The percentage of impaired waterway in focal parks. Note that the summary statistic was calculated for each park, but results are symbolized here by park upstream watershed

When compared to ecological threshold values for poor or good water quality (e.g., Van Sickle *et al*., 2006; Wazniak *et al*., 2007; Riva-Murray *et al*., 2010), water quality in and near most parks is in a good range. These results reflect the landscape location of most park watersheds, which tend to include a high portion of natural land cover and a relatively small area of cropland or intensive development. Using simple regression analyses, we generally found weak relationships between STORET water chemistry variables and watershed landscape variables. Certain attributes of the data likely contributed to our inability to link these factors. We wished to evaluate the ability of large, broad-extent databases to inform regional and national-scale analyses, and we thus began our analyses

**3.3 What can we infer about the condition of park freshwater resources?** 

that this variable explains approximately 26% of variation.

to facilitate comparisons with the other maps.

with approximately 2 million observations. The large number of observations required automated processes to screen data. For these preliminary analyses, we did not attempt to correct for factors such as season, variation in sampling effort, or flow regime. Despite the absence of strong statistical associations between water chemistry and landscape context, regional patterns were apparent for most of the chemistry variables we examined, such as phosphate-phosphorus (Fig. 7).

Fig. 7. Concentration of phosphate-phosphorus in focal parks, with legend categories reflecting conservative critical thresholds described by Van Sickle *et al.*, (2006) for total P. Note that concentrations were calculated for each park with 3 km buffer, but results are symbolized here by park upstream watershed in order to facilitate comparisons with the other maps. Watersheds in light grey denote parks that were not sampled for this metric.

The absence of strong statistical relationships between landscape and water quality variables in our national-scale assessment indicates the need for more sophisticated analyses when working at these very broad scales and with generalized databases. Other studies have found considerably stronger relationships between land cover and water chemistry (e.g., King *et al*., 2005; Wickham *et al*., 2005; Riva-Murray *et al*., 2010). Our future efforts will include more sophisticated processes for screening water chemistry data, and additional analyses. For example, King *et al*., (2005) evaluated a water quality index based on binary (0, 1) values for predictor variables that were above or below a quality threshold. Their index was the sum or four predictor variables. The binary transformation effectively addresses issues with high variance in the predictor variables, and it simplifies estimation and interpretation of the index. In addition, evaluations of the relative contributions of land cover versus broad-scale environmental setting to determining water chemistry are

Upstream Landscape Dynamics of

National Memorial (PC2).

US National Parks with Implications for Water Quality and Watershed Management 43

In the face of these monumental challenges, opportunities for managing park upstream watersheds are generally positively related to the headwater index – the proportion of the upstream watershed that exists within park boundaries. Managers of headwater parks obviously have the greatest direct management control over upstream watershed issues. Examples in this category include large National Parks in the western US: Yellowstone, Rocky Mountain, Sequoia and Kings Canyon, Yosemite, and Mount Rainier. In addition, owing to the strong human land cover and land use variables loading into PC1 and PC2, the two axes are negatively related to the headwater index (Fig. 8). From these relationships we can identify particular parks that – based on their headwater index – have characteristically low values for PC1 or PC2. In effect, these are parks with upstream watersheds that are relatively unchallenged by human drivers of landscape change, at least considering that significant portions of their upstream watersheds lie beyond park boundaries. Management opportunities for these parks lie in working collaboratively with other land owners to maintain protection of the upstream watershed. Example parks in this regard include Guilford Courthouse National Military Park (PC1), Chattahoochee River National Recreation Area (PC1), Effigy Mounds National Monument (PC2), and Arkansas Post

Fig. 9. Dominant owners of conservation land in focal park upstream watersheds. NPS = National Park Service, BLM = Bureau of Land Management, USFS = US Forest Service,

Conservation partnerships are challenging to promote, in part due to varied and sometimes conflicting missions of the partners, and perhaps also due to an insular history of managing for resources within ownership boundaries. Nevertheless, partnership opportunities may

USFWS = US Fish & Wildlife Service, DOD = Department of Defense.

ambiguous and clearly scale-dependent (e.g., compare King *et al.*, 2005; Wickham *et al.*, 2005; Goldstein *et al.*, 2007). Our analyses combined all samples for a given park so we could reach conclusions at the scale of an aggregated park upstream watershed. In some cases, this procedure merged samples from contributing watersheds that differed dramatically in land use patterns and threats to small-scale watersheds (e.g., Delaware Water Gap; Gross *et al*., 2011). Future analyses of land use effects on park water resources will likely need to resolve data at a finer spatial scale, perhaps in the form of hierarchical models. An appropriate management-relevant improvement would be to conduct local and regional-scale analyses on watersheds upstream of sampling sites, and then extrapolate these results to park watersheds within relevant ecological regions (Rohm *et al*., 2002).

### **3.4 What are the major challenges and opportunities for managing park upstream watersheds?**

Among the 151 focal parks, the big challenges identified by these analyses for managing park upstream watersheds relate to three major categories: urban development (PC1, Fig. 5A), agriculture and diffuse rural development (PC2, Fig. 5B), and N deposition (Fig. 4). Habitat fragmentation and alteration due to dams is undoubtedly a fourth major challenge, but one that we were unable to adequately quantify in this analysis. Nonetheless, the assessment of landscape context revealed that practically all parks are threatened in their respective geographies by N deposition, and parks east of the Rocky Mountains are especially threatened by development and agriculture. Importantly, this is not to say that parks in the western US are unthreatened by historical changes in land cover and land use. When compared to eastern parks the upstream watersheds of western parks are not presently as impacted by these factors, but critical ecological thresholds may still be exceeded in certain areas (e.g., Porter *et al*., 2005; Porter & Johnson, 2007). Furthermore, human population and housing projections suggest that many western parks will be increasingly challenged by development pressures in the coming decades (Theobald, 2005; Radeloff *et al*., 2010).

Fig. 8. Principal component scores 1 (A) and 2 (B), versus the headwater index (arcsine transformed), for 151 focal park upstream watersheds.

ambiguous and clearly scale-dependent (e.g., compare King *et al.*, 2005; Wickham *et al.*, 2005; Goldstein *et al.*, 2007). Our analyses combined all samples for a given park so we could reach conclusions at the scale of an aggregated park upstream watershed. In some cases, this procedure merged samples from contributing watersheds that differed dramatically in land use patterns and threats to small-scale watersheds (e.g., Delaware Water Gap; Gross *et al*., 2011). Future analyses of land use effects on park water resources will likely need to resolve data at a finer spatial scale, perhaps in the form of hierarchical models. An appropriate management-relevant improvement would be to conduct local and regional-scale analyses on watersheds upstream of sampling sites, and then extrapolate these results to park

**3.4 What are the major challenges and opportunities for managing park upstream** 

Fig. 8. Principal component scores 1 (A) and 2 (B), versus the headwater index (arcsine

transformed), for 151 focal park upstream watersheds.

Among the 151 focal parks, the big challenges identified by these analyses for managing park upstream watersheds relate to three major categories: urban development (PC1, Fig. 5A), agriculture and diffuse rural development (PC2, Fig. 5B), and N deposition (Fig. 4). Habitat fragmentation and alteration due to dams is undoubtedly a fourth major challenge, but one that we were unable to adequately quantify in this analysis. Nonetheless, the assessment of landscape context revealed that practically all parks are threatened in their respective geographies by N deposition, and parks east of the Rocky Mountains are especially threatened by development and agriculture. Importantly, this is not to say that parks in the western US are unthreatened by historical changes in land cover and land use. When compared to eastern parks the upstream watersheds of western parks are not presently as impacted by these factors, but critical ecological thresholds may still be exceeded in certain areas (e.g., Porter *et al*., 2005; Porter & Johnson, 2007). Furthermore, human population and housing projections suggest that many western parks will be increasingly challenged by development pressures in the coming decades (Theobald, 2005;

watersheds within relevant ecological regions (Rohm *et al*., 2002).

**watersheds?** 

Radeloff *et al*., 2010).

In the face of these monumental challenges, opportunities for managing park upstream watersheds are generally positively related to the headwater index – the proportion of the upstream watershed that exists within park boundaries. Managers of headwater parks obviously have the greatest direct management control over upstream watershed issues. Examples in this category include large National Parks in the western US: Yellowstone, Rocky Mountain, Sequoia and Kings Canyon, Yosemite, and Mount Rainier. In addition, owing to the strong human land cover and land use variables loading into PC1 and PC2, the two axes are negatively related to the headwater index (Fig. 8). From these relationships we can identify particular parks that – based on their headwater index – have characteristically low values for PC1 or PC2. In effect, these are parks with upstream watersheds that are relatively unchallenged by human drivers of landscape change, at least considering that significant portions of their upstream watersheds lie beyond park boundaries. Management opportunities for these parks lie in working collaboratively with other land owners to maintain protection of the upstream watershed. Example parks in this regard include Guilford Courthouse National Military Park (PC1), Chattahoochee River National Recreation Area (PC1), Effigy Mounds National Monument (PC2), and Arkansas Post National Memorial (PC2).

Fig. 9. Dominant owners of conservation land in focal park upstream watersheds. NPS = National Park Service, BLM = Bureau of Land Management, USFS = US Forest Service, USFWS = US Fish & Wildlife Service, DOD = Department of Defense.

Conservation partnerships are challenging to promote, in part due to varied and sometimes conflicting missions of the partners, and perhaps also due to an insular history of managing for resources within ownership boundaries. Nevertheless, partnership opportunities may

Upstream Landscape Dynamics of

**4. Conclusion** 

US National Parks with Implications for Water Quality and Watershed Management 45

Fig. 10. Percentage of non-conservation private land in focal park upstream watersheds.

in turn may inform interpretations of site-level analyses at policy-relevant scales.

We demonstrate a GIS framework for quantifying broad-scale landscape dynamics of park upstream watersheds and interpreting those analyses in the context of park water resources and watershed management potential. The framework is valuable for assessing NPS-wide opportunities and challenges associated with preserving water resources across an entire network of management units. Because it is founded on publically available data and methods, which were chosen based on mechanistic relationships between landscape-scale factors known to affect protected areas (Hansen & DeFries, 2007) and water resources (Allan, 2004), the framework may be readily applied to other systems of protected areas in the US, and also to protected areas in other parts of the world with comparable landscape and water resource data. When applied to 151 focal parks in the contiguous US, we demonstrate how (1) major anthropogenic stressors upstream from parks vary geographically, both in terms of magnitudes and critical ecological thresholds (e.g., N deposition), (2) water chemistry and impairment observations from most parks are within a good range, reflecting the overall landscape context of parks, (3) certain non-headwater parks are surprisingly unchallenged by upstream stressors that affect water quality, and (4) parks vary dramatically in terms of the public and private partnership opportunities for coordinating watershed management. While these findings do not provide park-specific recommendations for managing water resources, they are foundational to helping us better understand park watersheds and water quality in a comparative NPS-wide context, which

initially be evaluated using a simple landscape metric like the density of landowners that manage lands for conservation. Although this variable did not emerge as a major factor explaining among-park variation in watershed context (Table 3), it can be very useful for particular parks seeking to understand the potential diversity of partners that need to be engaged, as well as the dominant landowners that control most areas upstream (Fig. 9). The recognition that neighbouring landowners share a common responsibility for managing resources in the face of landscape-level anthropogenic change has motivated actions at local to national scales to form new partnerships. It has also recently stimulated the establishment of the Department of Interior (DOI) Landscape Conservation Cooperatives (LCCs) and regional Climate Science Centers (DOI Secretarial Order 3289).

Private lands not held for conservation pose a separate and distinct set of challenges and opportunities for managing park upstream watersheds. Although not shown in Figure 9, non-conservation private lands encompass approximately 61% of the US, and they thus dominate many park upstream watersheds (Fig. 10). While it is challenging to coordinate a large number of different private landowners, such coordination may be facilitated when private lands share a common land use. For example, private landowners in an area dominated by cultivated crops may share problems with ditch erosion (i.e., increased time and costs with ditch maintenance), which also poses sedimentation challenges to water resources in a downstream park. Despite different concerns over the threat, there would be a united interest in identifying creative solutions to the problem. Non-governmental organizations (NGOs) have traditionally played an especially important role in coordinating private-public partnerships. Such partnerships may also be promoted through newly established LCCs.

### **3.5 Next steps**

There are several important ways to build upon the analyses presented here. To evaluate a wider range of anthropogenic landscape stressors and pollutants, it is important to consider other areas of analysis besides upstream watersheds. Other ecologically informative areas of analysis could include downstream watersheds, ecoregions, or a local area that is proximate to the management unit (e.g., 30 km buffer or a protected area centered ecosystem, PACE; Hansen *et al.*, 2011). Using these varied areas of analysis would extend our framework to consider other water quality response variables that are affected by pollutants that traverse the landscape in different ways. In addition to new water resource response variables, it would also be useful to extend the analyses to consider point source drivers and their spatial proximity to parks based on flow length. However, given that these can vary so dramatically by geography, both in terms of point source type and magnitude, the explanatory power of these additional landscape predictors may prove most useful in analyses of parks at local to regional scales. Beyond the human drivers and conservation context of park upstream watersheds considered here, future analyses need to explicitly consider the ecological benefits and buffering potentials of natural systems. While some of these variables – like the percentage of natural land cover types – will be inversely correlated with many of the landscape stressors (e.g., percentage urban, percentage agriculture), others related to landscape pattern (Riitters *et al.*, 1995, 2007, 2009a, 2009b) and habitat connectivity (Hilty *et al.*, 2006; Theobald, 2006; Goetz *et al.*, 2009; Galpern *et al.*, 2011) will provide key management insights at local to regional scales.

Fig. 10. Percentage of non-conservation private land in focal park upstream watersheds.

### **4. Conclusion**

44 Sustainable Natural Resources Management

initially be evaluated using a simple landscape metric like the density of landowners that manage lands for conservation. Although this variable did not emerge as a major factor explaining among-park variation in watershed context (Table 3), it can be very useful for particular parks seeking to understand the potential diversity of partners that need to be engaged, as well as the dominant landowners that control most areas upstream (Fig. 9). The recognition that neighbouring landowners share a common responsibility for managing resources in the face of landscape-level anthropogenic change has motivated actions at local to national scales to form new partnerships. It has also recently stimulated the establishment of the Department of Interior (DOI) Landscape Conservation Cooperatives (LCCs) and

Private lands not held for conservation pose a separate and distinct set of challenges and opportunities for managing park upstream watersheds. Although not shown in Figure 9, non-conservation private lands encompass approximately 61% of the US, and they thus dominate many park upstream watersheds (Fig. 10). While it is challenging to coordinate a large number of different private landowners, such coordination may be facilitated when private lands share a common land use. For example, private landowners in an area dominated by cultivated crops may share problems with ditch erosion (i.e., increased time and costs with ditch maintenance), which also poses sedimentation challenges to water resources in a downstream park. Despite different concerns over the threat, there would be a united interest in identifying creative solutions to the problem. Non-governmental organizations (NGOs) have traditionally played an especially important role in coordinating private-public partnerships. Such partnerships may also be promoted through newly

There are several important ways to build upon the analyses presented here. To evaluate a wider range of anthropogenic landscape stressors and pollutants, it is important to consider other areas of analysis besides upstream watersheds. Other ecologically informative areas of analysis could include downstream watersheds, ecoregions, or a local area that is proximate to the management unit (e.g., 30 km buffer or a protected area centered ecosystem, PACE; Hansen *et al.*, 2011). Using these varied areas of analysis would extend our framework to consider other water quality response variables that are affected by pollutants that traverse the landscape in different ways. In addition to new water resource response variables, it would also be useful to extend the analyses to consider point source drivers and their spatial proximity to parks based on flow length. However, given that these can vary so dramatically by geography, both in terms of point source type and magnitude, the explanatory power of these additional landscape predictors may prove most useful in analyses of parks at local to regional scales. Beyond the human drivers and conservation context of park upstream watersheds considered here, future analyses need to explicitly consider the ecological benefits and buffering potentials of natural systems. While some of these variables – like the percentage of natural land cover types – will be inversely correlated with many of the landscape stressors (e.g., percentage urban, percentage agriculture), others related to landscape pattern (Riitters *et al.*, 1995, 2007, 2009a, 2009b) and habitat connectivity (Hilty *et al.*, 2006; Theobald, 2006; Goetz *et al.*, 2009; Galpern *et al.*, 2011)

regional Climate Science Centers (DOI Secretarial Order 3289).

will provide key management insights at local to regional scales.

established LCCs.

**3.5 Next steps** 

We demonstrate a GIS framework for quantifying broad-scale landscape dynamics of park upstream watersheds and interpreting those analyses in the context of park water resources and watershed management potential. The framework is valuable for assessing NPS-wide opportunities and challenges associated with preserving water resources across an entire network of management units. Because it is founded on publically available data and methods, which were chosen based on mechanistic relationships between landscape-scale factors known to affect protected areas (Hansen & DeFries, 2007) and water resources (Allan, 2004), the framework may be readily applied to other systems of protected areas in the US, and also to protected areas in other parts of the world with comparable landscape and water resource data. When applied to 151 focal parks in the contiguous US, we demonstrate how (1) major anthropogenic stressors upstream from parks vary geographically, both in terms of magnitudes and critical ecological thresholds (e.g., N deposition), (2) water chemistry and impairment observations from most parks are within a good range, reflecting the overall landscape context of parks, (3) certain non-headwater parks are surprisingly unchallenged by upstream stressors that affect water quality, and (4) parks vary dramatically in terms of the public and private partnership opportunities for coordinating watershed management. While these findings do not provide park-specific recommendations for managing water resources, they are foundational to helping us better understand park watersheds and water quality in a comparative NPS-wide context, which in turn may inform interpretations of site-level analyses at policy-relevant scales.

Upstream Landscape Dynamics of

Boca Raton, FL

ISSN 1051-0761

1905, ISSN 1051-0761

153, ISSN 1051-0761

*Letters*, ISSN 1755-263X

pp. 1119-1139, ISSN 1051-0761

495-510, ISSN 0167-6369

US National Parks with Implications for Water Quality and Watershed Management 47

Goldstein, R.M., Carlisle, D.M., Meador, M.R. & Short, T.M. (2007). Can basin land use

Gross, J.E., Hansen, A.J., Goetz, S.J., Theobald, D.M, Melton, F.M., Piekielek, N.B. & Nemani,

Gude, P.H., Hansen, A.J. & Jones, D.A. (2007). Biodiversity consequences of alternative

Halpern, B.S., Selkoe, K.A., Micheli, F. & Kappel, C.V. (2007). Evaluating and ranking the

Hansen, A.J., Davis, C.R., Piekielek, N., Gross, J., Theobald, D.M., Goetz, S., Melton, F. & DeFries,

management. *BioScience*, Vol.61, No.5, (May 2011), pp. 363-373, ISSN 0006-3568 Hansen, A.J. & DeFries, R. (2007). Ecological mechanisms linking protected areas to

Hansen, A.J., Knight, R.L., Marzluff, J.M., Powell, S., Brown, K., Gude, P.H. & Jones, K.

Hansen, A.J., Rasker, R., Maxwell, B., Rotella, J.J., Johnson, J.D., Parmenter, A.W., Langner,

Hawkins, B.A, Field, R., Cornell, H.V., Currie, D.J., Guégan, J.-F., Kaufman, D.M., Kerr, J.T.,

Hilty, J.A., Lidicker Jr, W.Z. & Merenlender, A.M. (2006). *Corridor Ecology: The Science and Practice of Linking Landscapes for Biodiversity Conservation*. Island Press, Washington, DC King, R.S., Baker, M.E., Whigham, D.F., Weller, D.E., Jordan, T.E., Kazyak, P.F. & Hurd,

Lawrence, D.J., Larson, E.R., Liermann, C.A.R., Mims, M.C., Pool, T.K. & Olden, J.D. (2011).

Leu, M., Hanser, S.E. & Knick, S.T. (2008). The human footprint in the west: a large-scale

Vol.84, No.12, (December 2003), pp. 3105-3117, ISSN 0012-9658

*Biology*, Vol.21, No.5, (October 2007), pp. 1301-1315, ISSN 0888-8892 Halpern, B.S., Walbridge, S., Selkoe, K.A., Kappel, C.V., Micheli, F., D'agrosa, C., Bruno, J.F.,

No.4, (June 2007), pp. 1004-1018, ISSN 1051-0761

(February 2008), pp. 948-952, ISSN 0036-8075

(February 2002), pp. 151-162, ISSN 0006-3568

effects on physical characteristics of streams be determined at broad geographic scales? *Environmental Monitoring and Assessment*, Vol.130, No.1-3, (July 2007), pp.

R.R. (2011). Remote sensing for inventory and monitoring of U.S. National Parks. In: *Remote Sensing of Protected Lands*, Y.Q. Yang (Ed.), pp. 29-56, Taylor & Francis,

future land use scenarios in Greater Yellowstone. *Ecological Applications*, Vol.17,

vulnerability of global marine ecosystems to anthropogenic threats. *Conservation* 

Casey, K.S., Ebert, C., Fox, H.E., Fujita, R., Heinemann, D., Lenihan, H.S., Madin, E.M.P., Perry, M.T., Selig, E.R., Spalding, M., Steneck, R. & Watson, R. (2008). A global map of human impact on marine ecosystems. *Science*, Vol.319, No.5865,

R. (2011). Delineating the ecosystems containing protected areas for monitoring and

surrounding lands. *Ecological Applications*, Vol.17, No.4, (June 2007), pp. 974-988,

(2005). Effects of exurban development on biodiversity: patterns, mechanisms, and research needs. *Ecological Applications*, Vol.15, No.6, (December 2005), pp. 1893-

U., Cohen, W.B., Lawrence, R.L. & Kraska, M.P.V. (2002). Ecological causes and consequences of demographic change in the New West. *BioScience*, Vol.52, No.2,

Mittelbach, G.G., Oberdorff, T., O'Brien, E.M., Porter, E.E. & Turner, J.R.G. (2003). Energy, water, and broad-scale geographic patterns of species richness. *Ecology*,

M.K. (2005). Spatial considerations for linking watershed land cover to ecological indicators in streams. *Ecological Applications*, Vol.15, No.1, (February 2005), pp. 137-

National parks as protected areas for U.S. freshwater fish diversity. *Conservation* 

analysis of anthropogenic impacts. *Ecological Applications*, Vol.18, No.5, (July 2008),

### **5. Acknowledgment**

We thank the NPScape team for developing analytical procedures that transformed huge quantities of data into the variables we analysed in this chapter: Peter Budde, Tom Philippi, Lisa Nelson, Brent Frakes, Mike Story, Shepard McAninch, Leona Svancara, Mara Kali, Sean Worthington, Thom Curdts, Dave Hollema, Ursula Glick, Bill Hovanec, Nick Viau, Thomas Flowe, Molly Thomas, and Allison Lundeby. Additional thanks to Lisa Duarte (National Gap Analysis Program), Collin Homer (USGS EROS), Dave Theobald (Colorado State University, CSU), Jill Baron (USGS, CSU), and Dean Tucker (NPS Water Resources Division) for assistance with source data, and to Steve Fancy for supporting NPScape as part of the NPS Inventory and Monitoring Program.

### **6. References**


We thank the NPScape team for developing analytical procedures that transformed huge quantities of data into the variables we analysed in this chapter: Peter Budde, Tom Philippi, Lisa Nelson, Brent Frakes, Mike Story, Shepard McAninch, Leona Svancara, Mara Kali, Sean Worthington, Thom Curdts, Dave Hollema, Ursula Glick, Bill Hovanec, Nick Viau, Thomas Flowe, Molly Thomas, and Allison Lundeby. Additional thanks to Lisa Duarte (National Gap Analysis Program), Collin Homer (USGS EROS), Dave Theobald (Colorado State University, CSU), Jill Baron (USGS, CSU), and Dean Tucker (NPS Water Resources Division) for assistance with source data, and to Steve Fancy for supporting NPScape as part of the

Allan, J.D. (2004). Landscapes and riverscapes: the influence of land use on stream

Baron, J.S., Driscoll, C.T., Stoddard, J.L. & Richer, E.E. (2011). Empirical critical loads of

US lakes. *BioScience*, Vol.61, No.8, (August 2011), pp. 602-613, ISSN 0006-3568 Baxter, R.M. (1977). Environmental effects of dams and impoundments. *Annual Review of* 

Davis, C.R. & Hansen, A.J. (In press).Trajectories in land use change around U.S. national

Dixon, J.A. & Sherman, P.B. (1991). Economics of protected areas. *Ambio*, Vol.20, No.2, (April

Djokic, D. & Ye, Z. (2000). DEM preprocessing for efficient watershed delineation, In:

Environmental Protection Agency (EPA). (2011). *Storage and Retrieval System (STORET /*  WQX). Online database at: www.epa.gov/storet/. Accessed 20 September 2011 Environmental Systems Research Institute (ESRI). (2010). *U.S. and Canada Detailed Streets, Compiled by Tele Atlas North America (2005), Inc.* ESRI, Redlands, CA Environmental Systems Research Institute (ESRI). (2011). *ArcGIS Desktop: Release 10*,

Fancy, S.G., Gross, J.E. & Carter, S.L. (2009). Monitoring the condition of natural resources in

Fry, J., Xian, G., Jin, S., Dewitz, J., Homer, C., Yang, L., Barnes, C., Herold, N. & Wickham, J.

Galpern, P., Manseau, M. & Fall, A. (2011). Patch-based graphs of landscape connectivity: A

*Conservation*, Vol.144, No.1, (January 2011), pp. 44-55, ISSN 0006-3207 Goetz, S.J., Jantz, P. & Jantz, C.A. (2009). Connectivity of core habitat in the Northeastern

*Environment*, Vol.113, No.7, (July 2009), pp. 1421-1429, ISSN 0034-4257

*Ecology and Systematics*, Vol.8, pp. 255-283, ISSN 0066-4162

ecosystems. *Annual Review of Ecology Evolution and Systematics*, Vol.35, pp. 257-284,

atmospheric nitrogen deposition for nutrient enrichment and acidification of sensitive

parks and their challenges and opportunities for management. *Ecological* 

*Hydrologic and Hydraulic Modeling Support with GIS*, D. Maidment & D. Djokic,

US national parks. *Environmental Monitoring and Assessment*, Vol.151, No.1-4, (April

(2011). Completion of the 2006 National Land Cover Database for the conterminous United States. *Photogrammetric Engineering & Remote Sensing*, Vol.77, No.9, pp. 858-

guide to construction, analysis and application for conservation. *Biological* 

United States: parks and protected areas in a landscape context. *Remote Sensing of* 

**5. Acknowledgment** 

**6. References** 

NPS Inventory and Monitoring Program.

*Applications*. ISSN 1051-0761

1991), pp. 68-74, ISSN 0044-7447

2009), pp. 161-174, ISSN 0167-6369

(Eds.), pp. 65-84, ESRI Press, Redlands, CA

ISSN1543-592X

Redlands, CA

864, ISSN 0099-1112


Upstream Landscape Dynamics of

http://www.R-project.org/

1033-1043, ISSN 0921-2973

117, ISSN 1470-160X

(May 2009), pp. 699-709, ISSN 0921-2973

(December 2010), pp. 1489-1503, ISSN 0921-2973

(February 2002), pp. 213-239, ISSN 1093-474X

(October 2002), pp. 891-904, ISSN 0006-3568

*United States*. Oxford University Press

ISSN 0027-8424

US National Parks with Implications for Water Quality and Watershed Management 49

Porter, E., Blett, T., Potter, D.U. & Huber, C. (2005). Protecting resources on federal lands:

Porter, E. & Johnson, S. (2007). Translating science into policy: using ecosystem thresholds to

R Development Core Team. (2011). *R: A language and environment for statistical computing*.

Radeloff, V.C., Stewart, S.I., Hawbaker, T.J., Gimmi, U., Pidgeon, A.M., Flather, C.H.,

Riitters, K.H., O'Neill, R.V., Hunsaker, C.T., Wickham, J., Yankee, D.H., Timmins, S.P., Jones,

Riitters, K.H., Vogt, P., Soille, P. & Estreguil, C. (2009a). Landscape patterns from

Riitters, K.H., Wickham, J.D. & Wade, T.G. (2009b). An indicator of forest dynamics using a

Riva-Murray, K., Riemann, R., Murdoch, P., Fischer, J.M. & Brightbill, R. (2010). Landscape

Rohm C.M., Omernik J.M., Woods A.J. & Stoddard, J.L. (2002). Regional characteristics of

Sabo, J.L., Sinha, T., Bowling, L.C., Schoups, G.H.W., Wallender, W.W., Campana, M.E.,

*of America*, Vol.107, No.50, (December 2010), pp. 21263-21270, ISSN 0027-8424 Sanderson, E.W., Jaiteh, M., Levy, M.A., Redford, K.H., Wannebo, A.V. & Woolmer, G.

Scott, J.M., Davis, F.W., McGhie, R.G., Wright, R.G., Groves, C. & Estes, J. (2001). Nature

Svancara, L.K., Brannon, R., Scott, J.M., Groves, C.R., Noss, R.F. & Pressey, R.L. (2005).

*Applications*, Vol.11, No.4, (August 2001), pp. 999-1007, ISSN 1051-0761 Stein, B., Kutner, L.S. & Adams, J.S. (2000). *Precious Heritage: The Status of Biodiversity in the* 

*BioScience*, Vol.55, No.7, (July 2005), pp. 603-612, ISSN 0006-3568

Vol.149, No.3, (October 2007), pp. 268-280, ISSN 0269-7491

implications of critical loads for atmospheric deposition of nitrogen and sulfur.

protect resources in Rocky Mountain National Park. *Environmental Pollution*,

Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, URL

Hammer, R.B. & Helmers, D.P. (2010). Housing growth in and near United States protected areas limits their conservation value. *Proceedings of the National Academy of Sciences of the United States of America*, Vol.107, No.2, (January 2010), pp. 940-945,

K.B. & Jackson, B.L. (1995). A factor analysis of landscape pattern and structure metrics. *Landscape Ecology*, Vol.10, No.1, (February 1995), pp. 23-39, ISSN 0921-2973 Riitters, K.H., Vogt, P. & Soille, P. (2007). Neutral model analysis of landscape patterns from

mathematical morphology. *Landscape Ecology*, Vol.22, No.7, (August 2007), pp.

mathematical morphology on maps with contagion. *Landscape Ecology*, Vol.24, No.5,

shifting landscape mosaic. *Ecological Indicators*, Vol.9, No.1, (January 2009), pp. 107-

characteristics affecting streams in urbanizing regions of the Delaware River Basin (New Jersey, New York, and Pennsylvania, U.S.). *Landscape Ecology*, Vol.25, No.10,

nutrient concentrations in streams and their application to nutrient criteria development. *Journal of the American Water Resources Association*, Vol.38, No.1,

Cherkauer, K.A., Fuller, P.L., Graf, W.L., Hopmans, J.W., Kominoski, J.S., Taylor, C., Trimble, S.W., Webb, R.H. & Wohl, E.E. (2010). Reclaiming freshwater sustainability in the Cadillac Desert. *Proceedings of the National Academy of Sciences of the United States* 

(2002). The human footprint and the last of the wild. *BioScience*, Vol.52, No.10,

reserves: Do they capture the full range of America's biological diversity? *Ecological* 

Policy-driven versus evidence-based conservation: a review of political targets and


http://www.horizon-systems.com/nhdplus/documentation.php


https://irma.nps.gov/App/Reference/Profile?Code=2165453


https://irma.nps.gov/App/Reference/Profile?Code=2170522


Mantel, S.K., Hughes, D.A. & Muller, N.W.J. (2010). Ecological impacts of small dams on

McKinney, M.L. (2002). Urbanization, biodiversity, and conservation. *BioScience*, Vol.52,

Miller, J.R. & Hobbs, R.J. (2002). Conservation where people live and work. *Conservation* 

National Park Service (NPS). (2010a). *NPScape Population Measure – Phase 1 Metrics Processing* 

National Park Service (NPS). (2010b). *NPScape Housing Measure – Phase 1 Metrics Processing* 

National Park Service (NPS). (2010c). *NPScape Roads Measure – Phase 2 Road Metrics* 

National Park Service (NPS). (2010d). *NPScape Landcover Measure – Phase 1 Metrics Processing* 

National Park Service (NPS). (2011a). *NPScape conservation status measure – Phase 2 Protected* 

National Park Service (NPS). (2011b). *NPScape Upstream Watershed Delineation Processing* 

National Park Service (NPS). (2011c). *NPScape NPS Boundary-derived Areas of Analysis SOP:* 

Newmark, W.D. (1985). Legal and biotic boundaries of western North American national

Parks, S.A. & Harcourt, A.H. (2002). Reserve size, local human density, and mammalian

*SOP: Current population total and density, historic population density, and projected population density*. Natural Resource Report NPS/NRPC/IMD/NRR—2010/254.

*SOP: Current housing density, historic housing density, and projected housing density metrics*. Natural Resource Report NPS/NRPC/IMD/NRR—2010/251. National

*Processing SOP: Road density and distance from roads*. National Park Service, Natural

*SOP: Landcover area per category, natural vs. converted landcover, landcover change, and impervious surface metrics*. Natural Resource Report NPS/NRPC/IMD/NRR— 2010/252. National Park Service, Fort Collins, CO. Available at

*Areas Database of the United States metrics processing SOP: Protected area and ownership category metrics*. National Park Service, Natural Resource Program Center, Fort

*SOP: Upstream watershed analysis for select National Park units*. National Park Service,

*Park, 3km, and 30km areas of analysis*. National Park Service, Natural Resource

parks: a problem of congruence. *Biological Conservation*, Vol.33, No.3, pp. 197- 208,

extinctions in US protected areas. *Conservation Biology*, Vol.16, No.3, (June 2002), pp.

*SA*, Vol.36, No.3, (April 2010), pp. 351-360, ISSN 0378-4738

*Biology*, Vol.16, No.2, (April 2002), pp. 330-337, ISSN 0888-8892

No.10, (October 2002), pp. 883-890, ISSN 0006-3568

National Hydrology Dataset (NHD) Plus. (2010). *NHD Plus User Guide*. http://www.horizon-systems.com/nhdplus/documentation.php National Park Service (NPS). (1916). *The National Park Service Organic Act of 1916*.

National Park Service, Fort Collins, CO. Available at https://irma.nps.gov/App/Reference/Profile?Code=2165453

https://irma.nps.gov/App/Reference/Profile?Code=2165448

Resource Program Center, Fort Collins, CO. Available at https://irma.nps.gov/App/Reference/Profile?Code=2166959

https://irma.nps.gov/App/Reference/Profile?Code=2165449

Natural Resource Program Center, Fort Collins, CO. Available at

https://irma.nps.gov/App/Reference/Profile?Code=2170511

https://irma.nps.gov/App/Reference/Profile?Code=2173077

Program Center, Fort Collins, CO. Available at https://irma.nps.gov/App/Reference/Profile?Code=2170522

Park Service, Fort Collins, CO. Available at

Collins, CO. Available at

ISSN 0006-3207

800-808, ISSN 0888-8892

South African rivers Part 1: Drivers of change – water quantity and quality. *Water* 


**3** 

 *Izmir Turkey* 

**Sustainable Management of Large Scale** 

 *Department of Irrigation and Agricultural Structures, Bornova* 

**for Gediz Basin, Turkey** 

*Ege University, Faculty of Agriculture,* 

Murat Kilic and Suer Anac

**Irrigation Systems: A Decision Support Model** 

While water on a global scale is plentiful, 97% of it is saline and 2.25% is trapped in glaciers and ice, leaving only 0.75% available in freshwater aquifers, rivers and lakes. About 70% of this fresh water is used for agricultural production, 22% for industrial purposes and 8% for domestic purposes. Increasing competition for water for domestic and industrial purposes is likely to reduce the water available for agriculture. Thus, water scarcity is being increasingly accepted as a major limitation on increased agricultural production and food security in the 21st century (Yazar, 2006). Climate change and hydric stress are limiting the availability of clean water. Overexploitation of natural resources has led to environmental unbalance. Present decisions relative to the management of hydric resources will deeply affect the

In developing countries, agriculture continues to be an important economic sector as it makes a significant contribution to national incomes and economic growth. As water scarcity intensifies in many regions of the world, better management of irrigation is becoming an issue of paramount importance (Hussain et al., 2007). Skilled management of irrigation should start from planning at the regional level (Lorite et al., 2007). The main problem in planning the management of deficit resources is how to allocate them among multiple users efficiently and equitably by considering the social, economic and political issues, while considering the heterogeneity in soils, crops and climate and the complexity of the water distribution system (Brumbelow et al., 2007; Chambers, 1988; Kilic & Ozgurel, 2005). Sustainable irrigation water management should simultaneously achieve two objectives: sustaining irrigated agriculture for food security and preserving the associated natural environment. A stable relationship should be maintained between these two objectives now and in the future, while potential conflicts between these objectives should be mitigated through appropriate irrigation practices. Cai et al. (2003) carried out an investigation on sustainability analysis for irrigation water management in the Aral Sea Region. This study presents an integrated modeling framework for sustainable irrigation management analysis and applies it to analyze irrigation water management. Based on the

economy and our future environment (Lermontov et al., 2011).

**1. Introduction** 

biological needs. *BioScience*, Vol.55, No.11, (November 2005), pp. 989-995, ISSN 0006-3568


## **Sustainable Management of Large Scale Irrigation Systems: A Decision Support Model for Gediz Basin, Turkey**

Murat Kilic and Suer Anac *Ege University, Faculty of Agriculture, Department of Irrigation and Agricultural Structures, Bornova Izmir Turkey* 

### **1. Introduction**

50 Sustainable Natural Resources Management

Svancara, L.K., Scott, J.M., Loveland, T.R. & Pidgorna, A.B. (2009). Assessing the landscape

Theobald, D.M. (2006). Exploring the functional connectivity of landscapes using landscape

US Army Corps of Engineers. (2010). *National Inventory of Dams*. http://nid.usace.army.mil.

US Census Bureau. (2001). *Census 2000 Summary File 1 United States*. Prepared by the U.S. Census Bureau, 2001. http://www.census.gov/prod/cen2000/doc/sf1.pdf US General Accounting Office. (1994). *Activities outside park borders have caused damage to* 

USGS Gap Analysis Program. (2011). *Protected Areas Database of the United States (PADUS)* 

Van Sickle, J., Stoddard, J.L., Paulsen, S.G. & Olsen, A.R. (2006). Using relative risk to

*Management*, Vol.38, No.6, (December 2006), pp. 1020-1030, ISSN 0364-152X Vitousek, P.M., Aber, J.D., Howarth, R.W., Likens, G.E., Matson, P.A., Schindler, D.W.,

Wade, A.A. & Theobald, D.M. (2009). Residential development encroachment on US

Wade, A.A., Theobald, D.M. & Laituri, M.J. (2011). A multi-scale assessment of local and

Wickham, J.D., Riitters, K.H., Wade, T.G. & Jones, K.B. (2005). Evaluating the relative roles

Wiersma, Y.F., Nudds, T.D. & Rivard, D.H. (2004). Models to distinguish effects of

Woolmer, G., Trombulak, S.C., Ray, J.C., Doran, P.J., Anderson, M.G., Baldwin, R.F.,

*Urban Planning*, Vol.101, No.3, (June 2011), pp. 215-227, ISSN 0169-2046 Ward, J.V. & J.A. Stanford. (1979). *The Ecology of Regulated Streams*. Plenum Press, New York Wazniak, C.E., Hall, M.R., Carruthers, T.J.B., Sturgis, B., Dennison, W.C. & Orth, R.J. (2007).

*Ecological Applications*, Vol.17, pp. S64-S78, ISSN 1051-0761

*Ecology and Society*, Vol.10, No.1, pp. 32, ISSN 1708-3087

444, Cambridge University Press, NY

Accessed 20 September 2011

Last accessed February 28, 2011

pp. 737-750, ISSN 1051-0761

ISSN 0888-8892

ISSN 0921-2973

786, ISSN 0921-2973

No.1, (July 2008), pp. 42-53, ISSN 0169-2046

0006-3568

biological needs. *BioScience*, Vol.55, No.11, (November 2005), pp. 989-995, ISSN

context and conversion risk of protected areas using satellite data products. *Remote Sensing of Environment*, Vol.113, No.7, (July 2009), pp. 1357-1369, ISSN 0034-4257 Theobald, D.M. (2005). Landscape patterns of exurban growth in the USA from 1980 to 2020.

networks, In: *Connectivity Conservation*, K.R. Crooks & M. Sanjayan (Eds.), pp. 416-

*resources and will likely cause more*. US Government Printing Office, GAO/RCED-94-59

*version 1.2 Geospatial Metadata*. http://gap.uidaho.edu/padus/protectedareas.html.

compare the effects of aquatic stressors at a regional scale. *Environmental* 

Schlesinger, W.H. & Tilman, G.D. (1997). Human alteration of the global nitrogen cycle: sources and consequences. *Ecological Applications*, Vol.7, No.3, (August 1997),

protected areas. *Conservation Biology*, Vol.24, No.1, (February 2010), pp. 151-161,

contextual threats to existing and potential U.S. protected areas. *Landscape and* 

Linking water quality to living resources in a mid-Atlantic lagoon system, USA.

of ecological regions and land-cover composition for guiding establishment of nutrient criteria. *Landscape Ecology*, Vol.20, No.7, (November 2005), pp. 791-798,

landscape patterns and human population pressures associated with species loss in Canadian national parks. *Landscape Ecology*, Vol.19, No.7, (October 2004), pp. 773-

Morgan, A. & Sanderson, E.W. (2008). Rescaling the human footprint: a tool for conservation planning at an ecoregional scale. *Landscape and Urban Planning*, Vol.87, While water on a global scale is plentiful, 97% of it is saline and 2.25% is trapped in glaciers and ice, leaving only 0.75% available in freshwater aquifers, rivers and lakes. About 70% of this fresh water is used for agricultural production, 22% for industrial purposes and 8% for domestic purposes. Increasing competition for water for domestic and industrial purposes is likely to reduce the water available for agriculture. Thus, water scarcity is being increasingly accepted as a major limitation on increased agricultural production and food security in the 21st century (Yazar, 2006). Climate change and hydric stress are limiting the availability of clean water. Overexploitation of natural resources has led to environmental unbalance. Present decisions relative to the management of hydric resources will deeply affect the economy and our future environment (Lermontov et al., 2011).

In developing countries, agriculture continues to be an important economic sector as it makes a significant contribution to national incomes and economic growth. As water scarcity intensifies in many regions of the world, better management of irrigation is becoming an issue of paramount importance (Hussain et al., 2007). Skilled management of irrigation should start from planning at the regional level (Lorite et al., 2007). The main problem in planning the management of deficit resources is how to allocate them among multiple users efficiently and equitably by considering the social, economic and political issues, while considering the heterogeneity in soils, crops and climate and the complexity of the water distribution system (Brumbelow et al., 2007; Chambers, 1988; Kilic & Ozgurel, 2005). Sustainable irrigation water management should simultaneously achieve two objectives: sustaining irrigated agriculture for food security and preserving the associated natural environment. A stable relationship should be maintained between these two objectives now and in the future, while potential conflicts between these objectives should be mitigated through appropriate irrigation practices. Cai et al. (2003) carried out an investigation on sustainability analysis for irrigation water management in the Aral Sea Region. This study presents an integrated modeling framework for sustainable irrigation management analysis and applies it to analyze irrigation water management. Based on the

Sustainable Management of Large Scale

real-time conditions at the system level.

Irrigation Systems: A Decision Support Model for Gediz Basin, Turkey 53

technological scenarios. On the other hand, effects of real-time irrigation programming at network level were not evaluated on water and agriculture policy scenarios in this investigation. Jalal et al. (2007) developed a model for optimal multi-crop irrigation areas associated with reservoir operation policies in an irrigation system. The objectives were to maximize the annual benefit of the system by supplying irrigation water for a proposed multi-crop pattern over the planning period. An irrigation program wasn't developed under

In addition, it is complicated to analyze the management of deficit resources from the points of view of social, economics and politics, which constitute the various dimensions of management planning. Farmers decide on which crops to grow and on the associated use of resources such as land, labor, water and capital. Governments, on the other hand, develop policies (e.g., subsidies, taxation, and infrastructural developments) that are targeted at influencing decisions made at the farm level in order to achieve aggregated changes which are deemed desirable on a municipal, provincial or national scale. At national level, overall policies and decisions are formulated on sectoral allocations of resources and economic activities. Strategies, policies and programs for sectoral development are included in sector plans. At sub-national level, potentials, constraints and objectives for agricultural development are identified. In this multi-level planning approach, the plans at different levels have to be consistent and interlinked (Acs et al., 2007; Laborte et al., 2007; Mousavi & Ramamurthy, 2000;). Clemmens (2006) carried out research on improving irrigated agriculture performance through the water delivery process. Reasons for poor performance of the schemes were discussed and a method was proposed to improve its performance. According to this research, the main problem was that operation of the irrigation systems was not tied to productivity. As a result, the dispersive nature of the large open canal distribution systems causes extreme variability in water delivery service to users. Diaz et al. (2007) developed a model using data from an on-demand pressurized water distribution network located in Sector VIII of the Genil-Cabra irrigation district of Santaella, Cordoba, Spain to simulate an irrigation season, and calculate the flows that circulate in the system at any given time during the irrigation day. Water demand frequencies were estimated by using the results from model solution. Statistical distribution approach was used in this process. In addition, the most appropriate periods were studied for determining peak demand. The results showed that the statistical methods slightly underestimated demand. It was concluded that a better fit is

achieved when a more flexible distribution such as Gamma Distribution is used.

Haie & Keller (2008) proposed two efficiency models: one is based on water quantity, and the other on quantity and quality, with the possibility of considering water reuse in both. These models were developed for two scales: the first was called Project Effective Efficiency, and the second Basin Effective Efficiency. The latter gives the influence of project on water resources systems of the basin while the former does not make such connection to the whole basin. The concept of equity in water allocation between large numbers of users in temporal and spatial dimensions weren't taken into consideration under the real-time programming conditions. Du et al. (2009) evaluated the Soil and Water Assessment Tool (SWAT) model for estimation of continuous daily flow based on limited flow measurements in the Upper Oyster Creek (UOC) watershed. Among the five main stem stations, four stations were statistically shown to have good agreement between predicted and measured flows. SWAT underestimated the flow of the fifth main stem station possibly because of the existence of

modeling outputs, alternative future of the irrigation practice in the region were explored and it was found that to maintain current irrigation practices would lead to worsening environmental and economic consequences. Investments in infrastructure improvements (about annualized US \$ 299 million) and crop pattern change would be necessary to sustain the irrigated agriculture and the associated environment in the region. Evans et al. (2003) carried out an investigation on efficiency and equity in irrigation management. The objective of this study was to address the problems of inefficiency and inequity in water allocation in the El Angel watershed, located in Ecuador's Sierra region. Water is captured in a high-altitude region of the watershed and distributed downstream to producers in four elevation-defined zones via a system of canals. Upstream and downstream producers face different conditions with respect to climate and terrain. A mathematical programming model was created to study the consequences of addressing chronic water scarcity problems in the watershed by shifting water resources between the four zones. The objective function of the model maximizes producer welfare as measured by aggregate gross margin, subject to limited supplies of land, labor and water. Five water allocation scenarios ere evaluated with respect to efficiency in land and water use and equity in income distribution. Results revealed that although water was the primary constrained resource downstream, in the upstream zones, land was far more scarce. The current distribution of water rights did not consider these differences and therefore was neither efficient nor equitable. Improvements in efficiency and equity were associated with 1) a shift of water to the lower zone, and 2) the use of lower levels of irrigation intensity upstream. A linear optimization model was used in this investigation instead of real-time water allocation programming for different growing stages of crops.

Generally, optimal multi-cropping patterns and irrigation areas associated with appropriate reservoir operation and irrigation scheduling are essential for increasing the overall efficiency of reservoir-irrigation systems. Speelman et al. (2008) analyzed the efficiency with which water was used in small scale irrigation schemes in North-West Province in South Africa and studied its determinants. In the study area, small-scale irrigation schemes play an important role in rural development, but the increasing pressure on water resources and the approaching introduction of water charges raise the concern for more efficient water use. The Data Envelopment Analysis (DEA) techniques and sub-vector efficiencies were used in the study. This process was carried out under constant returns to scale (CRS) and variable returns to scale (VRS) conditions. The most important aspect of operation is distribution of the right quantity of water to the crops at the right time. An optimal multi-cropping pattern is important, since it provides better opportunities for water conservation and reduces the impact of water constraint on the system (Georgiou & Papamichail, 2008; Hsiao et al., 2007;). Bartoloni et al. (2007) carried out an investigation in order to evaluate the impacts of agriculture and water policy scenarios on the sustainability of selected irrigated farming systems in Italy. Five main scenarios were developed reflecting aspects of agricultural policy, markets and technologies: Agenda 2000, world market, global sustainability, provincial agriculture and local community. These were combined with two water price levels, representing stylized scenarios for water policy. The effects of the scenarios on irrigated systems were simulated using multi-attribute linear programming models representing the reactions of the farms to external variables defined by each scenario. In this study, five Italian irrigated farming systems were considered: cereal, rice, fruit, vegetables and citrus. The results showed the diversity of irrigated systems and the different effects that water pricing policy might produce depending on the agricultural policy, market and

modeling outputs, alternative future of the irrigation practice in the region were explored and it was found that to maintain current irrigation practices would lead to worsening environmental and economic consequences. Investments in infrastructure improvements (about annualized US \$ 299 million) and crop pattern change would be necessary to sustain the irrigated agriculture and the associated environment in the region. Evans et al. (2003) carried out an investigation on efficiency and equity in irrigation management. The objective of this study was to address the problems of inefficiency and inequity in water allocation in the El Angel watershed, located in Ecuador's Sierra region. Water is captured in a high-altitude region of the watershed and distributed downstream to producers in four elevation-defined zones via a system of canals. Upstream and downstream producers face different conditions with respect to climate and terrain. A mathematical programming model was created to study the consequences of addressing chronic water scarcity problems in the watershed by shifting water resources between the four zones. The objective function of the model maximizes producer welfare as measured by aggregate gross margin, subject to limited supplies of land, labor and water. Five water allocation scenarios ere evaluated with respect to efficiency in land and water use and equity in income distribution. Results revealed that although water was the primary constrained resource downstream, in the upstream zones, land was far more scarce. The current distribution of water rights did not consider these differences and therefore was neither efficient nor equitable. Improvements in efficiency and equity were associated with 1) a shift of water to the lower zone, and 2) the use of lower levels of irrigation intensity upstream. A linear optimization model was used in this investigation instead of real-time water allocation programming for different growing stages of crops.

Generally, optimal multi-cropping patterns and irrigation areas associated with appropriate reservoir operation and irrigation scheduling are essential for increasing the overall efficiency of reservoir-irrigation systems. Speelman et al. (2008) analyzed the efficiency with which water was used in small scale irrigation schemes in North-West Province in South Africa and studied its determinants. In the study area, small-scale irrigation schemes play an important role in rural development, but the increasing pressure on water resources and the approaching introduction of water charges raise the concern for more efficient water use. The Data Envelopment Analysis (DEA) techniques and sub-vector efficiencies were used in the study. This process was carried out under constant returns to scale (CRS) and variable returns to scale (VRS) conditions. The most important aspect of operation is distribution of the right quantity of water to the crops at the right time. An optimal multi-cropping pattern is important, since it provides better opportunities for water conservation and reduces the impact of water constraint on the system (Georgiou & Papamichail, 2008; Hsiao et al., 2007;). Bartoloni et al. (2007) carried out an investigation in order to evaluate the impacts of agriculture and water policy scenarios on the sustainability of selected irrigated farming systems in Italy. Five main scenarios were developed reflecting aspects of agricultural policy, markets and technologies: Agenda 2000, world market, global sustainability, provincial agriculture and local community. These were combined with two water price levels, representing stylized scenarios for water policy. The effects of the scenarios on irrigated systems were simulated using multi-attribute linear programming models representing the reactions of the farms to external variables defined by each scenario. In this study, five Italian irrigated farming systems were considered: cereal, rice, fruit, vegetables and citrus. The results showed the diversity of irrigated systems and the different effects that water pricing policy might produce depending on the agricultural policy, market and technological scenarios. On the other hand, effects of real-time irrigation programming at network level were not evaluated on water and agriculture policy scenarios in this investigation. Jalal et al. (2007) developed a model for optimal multi-crop irrigation areas associated with reservoir operation policies in an irrigation system. The objectives were to maximize the annual benefit of the system by supplying irrigation water for a proposed multi-crop pattern over the planning period. An irrigation program wasn't developed under real-time conditions at the system level.

In addition, it is complicated to analyze the management of deficit resources from the points of view of social, economics and politics, which constitute the various dimensions of management planning. Farmers decide on which crops to grow and on the associated use of resources such as land, labor, water and capital. Governments, on the other hand, develop policies (e.g., subsidies, taxation, and infrastructural developments) that are targeted at influencing decisions made at the farm level in order to achieve aggregated changes which are deemed desirable on a municipal, provincial or national scale. At national level, overall policies and decisions are formulated on sectoral allocations of resources and economic activities. Strategies, policies and programs for sectoral development are included in sector plans. At sub-national level, potentials, constraints and objectives for agricultural development are identified. In this multi-level planning approach, the plans at different levels have to be consistent and interlinked (Acs et al., 2007; Laborte et al., 2007; Mousavi & Ramamurthy, 2000;). Clemmens (2006) carried out research on improving irrigated agriculture performance through the water delivery process. Reasons for poor performance of the schemes were discussed and a method was proposed to improve its performance. According to this research, the main problem was that operation of the irrigation systems was not tied to productivity. As a result, the dispersive nature of the large open canal distribution systems causes extreme variability in water delivery service to users. Diaz et al. (2007) developed a model using data from an on-demand pressurized water distribution network located in Sector VIII of the Genil-Cabra irrigation district of Santaella, Cordoba, Spain to simulate an irrigation season, and calculate the flows that circulate in the system at any given time during the irrigation day. Water demand frequencies were estimated by using the results from model solution. Statistical distribution approach was used in this process. In addition, the most appropriate periods were studied for determining peak demand. The results showed that the statistical methods slightly underestimated demand. It was concluded that a better fit is achieved when a more flexible distribution such as Gamma Distribution is used.

Haie & Keller (2008) proposed two efficiency models: one is based on water quantity, and the other on quantity and quality, with the possibility of considering water reuse in both. These models were developed for two scales: the first was called Project Effective Efficiency, and the second Basin Effective Efficiency. The latter gives the influence of project on water resources systems of the basin while the former does not make such connection to the whole basin. The concept of equity in water allocation between large numbers of users in temporal and spatial dimensions weren't taken into consideration under the real-time programming conditions. Du et al. (2009) evaluated the Soil and Water Assessment Tool (SWAT) model for estimation of continuous daily flow based on limited flow measurements in the Upper Oyster Creek (UOC) watershed. Among the five main stem stations, four stations were statistically shown to have good agreement between predicted and measured flows. SWAT underestimated the flow of the fifth main stem station possibly because of the existence of

Sustainable Management of Large Scale

1999; Kilic, 2004; Topraksu, 1971, 1974; Yonter, 2010).

Fig. 1. General plan of the Gediz Basin in Turkey.

main and secondary canals are under upstream control.

(Kilic & Tuylu, 2010).

Irrigation Systems: A Decision Support Model for Gediz Basin, Turkey 55

groundwater in the basin, especially near the sea where the slope is minimal (Girgin et al.,

The Demirköprü Dam was constructed on the Gediz River in 1960 for irrigation, energy and flood control. Total water storage in the dam reservoir determines the volume and duration of irrigation water supplies to Gediz Basin System. Roughly 751 million cubic meters of water per year is released to the Lower Gediz Irrigation System by means of three regulators constructed on the river: from upstream to downstream, Adala, Ahmetli and Emiralem

For the past decade, there has been a scarcity of water in the Lower Gediz Basin because of the increase in urban and industrial demands (Svendsen et al., 2001). Unplanned production patterns, inadequate system capacity, poor distribution and management of water, large numbers of divided and small sized plots for cropping, and uncontrolled and inappropriate use of water by the farmers are the major factors giving rise to low efficiency in the Gediz Basin Irrigation System. Water level or flow can be controlled from three points in the system: I- main regulator at the head of the main canal; II- offtake regulators at the heads of the secondary canals; and III- constant-head orifices at the turnout to each tertiary canal. The

The National Water Works (DSI) operates the major water control infrastructures, such as river regulators and dams. Also, water allocation to main canals is fixed by the DSI according to the size of command and cropping pattern. Irrigation associations are responsible for water delivery from the main canal to secondary canals. Water delivery to tertiary canals and plots is arranged by Village Irrigation Groups (VIGs) which are

complex flood control measures near to the station. SWAT estimated the daily flow at one tributary station well, but with relatively large errors for the other two tributaries. Any water allocation plan wasn't prepared for the district. Varis & Abu-Zaid (2009) carried out an investigation on socio-economic and environmental aspects of water management in the 21st century: trends, challenges and prospects for the Middle East and North Africa (MENA) region. Garizabal et al. (2009) carried out an investigation in order to analyze the evolution of the agro-environmental impact in a traditional irrigation land of the middle Ebro Valley (Spain) which was experienced changes in its management. It was determined that the drought of 2005 caused more intensive water use (86%), increasing in 33% the irrigation efficiency when compared to 2001 (53%), even though a high hydric deficit (24%) was caused. Ryu et al. (2009) developed a decision support system for sustainable water resources management in a water conflict resolution framework to identify and evaluate a range of alternatives for the Geum River Basin in Korea. Working with stakeholders in a "shared vision modeling" framework, management strategies were created to illustrate system tradeoffs as well as long term system planning. A multi-criterion decision making approach using subjective scales is utilized to evaluate the water resource allocation and management tradeoffs between stakeholders and system objectives. The real-time programming wasn't carried out in this process, and changing efficiency values for the systems in temporal and spatial dimensions weren't taken into consideration. Sheild et al. (2009) carried out an investigation to identify and quantify stakeholder references pertaining to water management programs in order to improve water policy design. The relative importance of water management attributes was evaluated and willingness-to-pay values were estimated. Results showed that the majority of respondents weighed preserving stream health and Hawaiian cultural practices in water allocation decisions and were willing to pay \$4.53 per month per household to improve stream health to an excellent condition. These results highlight the need to strongly align watershed-level preferences to better balance in-stream and offstream demands to help guide water managers to make more effective water allocation decisions.

In this investigation, the real-time irrigation programming model MONES 4.1 developed by Kilic (2010) was applied to the irrigation system known as Sector VII which is served by 28 tertiary canals in the Right Bank Irrigation System of Ahmetli Regulator in the Lower Gediz Basin, Turkey. Irrigation programs from the model for different periods were analyzed, and the results were compared with the actual irrigation applications in the system.

### **2. Description of the study area**

This investigation was carried out on the commands of 28 tertiary canals in Irrigation District of Sector VII in Ahmetli Right Bank Irrigation Network in Lower Gediz Basin Irrigation System in Turkey. The Basin is located within the Aegean Region of western Turkey at latitude 380 04' - 390 13' N, and longitude 260 42' - 290 45' E. The main water source for the Lower Gediz Irrigation System is the Gediz River, which is 275 km in length. Drainage area of the basin is roughly 17219 km2 (Figure 1). The Gediz Basin is a river deposit basin formed with the alluvium transported by the Gediz River and its tributaries. The basin's topography is characterized by hills and rolling country. The tributaries of the Gediz River have been filled with eroded silt and sediment by erosion. For this reason, flood flows can easily overtop the river banks. These conditions create a problem of high

complex flood control measures near to the station. SWAT estimated the daily flow at one tributary station well, but with relatively large errors for the other two tributaries. Any water allocation plan wasn't prepared for the district. Varis & Abu-Zaid (2009) carried out an investigation on socio-economic and environmental aspects of water management in the 21st century: trends, challenges and prospects for the Middle East and North Africa (MENA) region. Garizabal et al. (2009) carried out an investigation in order to analyze the evolution of the agro-environmental impact in a traditional irrigation land of the middle Ebro Valley (Spain) which was experienced changes in its management. It was determined that the drought of 2005 caused more intensive water use (86%), increasing in 33% the irrigation efficiency when compared to 2001 (53%), even though a high hydric deficit (24%) was caused. Ryu et al. (2009) developed a decision support system for sustainable water resources management in a water conflict resolution framework to identify and evaluate a range of alternatives for the Geum River Basin in Korea. Working with stakeholders in a "shared vision modeling" framework, management strategies were created to illustrate system tradeoffs as well as long term system planning. A multi-criterion decision making approach using subjective scales is utilized to evaluate the water resource allocation and management tradeoffs between stakeholders and system objectives. The real-time programming wasn't carried out in this process, and changing efficiency values for the systems in temporal and spatial dimensions weren't taken into consideration. Sheild et al. (2009) carried out an investigation to identify and quantify stakeholder references pertaining to water management programs in order to improve water policy design. The relative importance of water management attributes was evaluated and willingness-to-pay values were estimated. Results showed that the majority of respondents weighed preserving stream health and Hawaiian cultural practices in water allocation decisions and were willing to pay \$4.53 per month per household to improve stream health to an excellent condition. These results highlight the need to strongly align watershed-level preferences to better balance in-stream and offstream demands to help guide water managers to make

In this investigation, the real-time irrigation programming model MONES 4.1 developed by Kilic (2010) was applied to the irrigation system known as Sector VII which is served by 28 tertiary canals in the Right Bank Irrigation System of Ahmetli Regulator in the Lower Gediz Basin, Turkey. Irrigation programs from the model for different periods were analyzed, and

This investigation was carried out on the commands of 28 tertiary canals in Irrigation District of Sector VII in Ahmetli Right Bank Irrigation Network in Lower Gediz Basin Irrigation System in Turkey. The Basin is located within the Aegean Region of western Turkey at latitude 380 04' - 390 13' N, and longitude 260 42' - 290 45' E. The main water source for the Lower Gediz Irrigation System is the Gediz River, which is 275 km in length. Drainage area of the basin is roughly 17219 km2 (Figure 1). The Gediz Basin is a river deposit basin formed with the alluvium transported by the Gediz River and its tributaries. The basin's topography is characterized by hills and rolling country. The tributaries of the Gediz River have been filled with eroded silt and sediment by erosion. For this reason, flood flows can easily overtop the river banks. These conditions create a problem of high

the results were compared with the actual irrigation applications in the system.

more effective water allocation decisions.

**2. Description of the study area** 

groundwater in the basin, especially near the sea where the slope is minimal (Girgin et al., 1999; Kilic, 2004; Topraksu, 1971, 1974; Yonter, 2010).

Fig. 1. General plan of the Gediz Basin in Turkey.

The Demirköprü Dam was constructed on the Gediz River in 1960 for irrigation, energy and flood control. Total water storage in the dam reservoir determines the volume and duration of irrigation water supplies to Gediz Basin System. Roughly 751 million cubic meters of water per year is released to the Lower Gediz Irrigation System by means of three regulators constructed on the river: from upstream to downstream, Adala, Ahmetli and Emiralem (Kilic & Tuylu, 2010).

For the past decade, there has been a scarcity of water in the Lower Gediz Basin because of the increase in urban and industrial demands (Svendsen et al., 2001). Unplanned production patterns, inadequate system capacity, poor distribution and management of water, large numbers of divided and small sized plots for cropping, and uncontrolled and inappropriate use of water by the farmers are the major factors giving rise to low efficiency in the Gediz Basin Irrigation System. Water level or flow can be controlled from three points in the system: I- main regulator at the head of the main canal; II- offtake regulators at the heads of the secondary canals; and III- constant-head orifices at the turnout to each tertiary canal. The main and secondary canals are under upstream control.

The National Water Works (DSI) operates the major water control infrastructures, such as river regulators and dams. Also, water allocation to main canals is fixed by the DSI according to the size of command and cropping pattern. Irrigation associations are responsible for water delivery from the main canal to secondary canals. Water delivery to tertiary canals and plots is arranged by Village Irrigation Groups (VIGs) which are

Sustainable Management of Large Scale

applications is not important in pricing the water.

and 704.6 mm respectively (DMI Reports, 2008).

**3.1 Water allocation stages at network level** 

secondary to tertiary canals, and 3) allocation to plots.

performed interactively in order for each level of segment.

other levels are also carried out in similar ways.

program.

**3. Description of the irrigation programming model** 

Irrigation Systems: A Decision Support Model for Gediz Basin, Turkey 57

In the research area, water charges are collected annually by the Gediz Irrigation Association according to the crop type and size of the area. In other words, water from the open canal irrigation system is priced in TL/ha, and is paid for as a single payment for the whole season. Thus, the number of irrigations and the amount of water used in irrigation

Apart from this, producers form crop patterns according to tradition and their own preferences. This has an adverse effect on the efficient management of these systems. In addition, there is a loss of productivity because of the great age of systems like the one under study here. Size of the area irrigated by the canals and the crop patterns are shown in Table 1. In the research area, cotton, maize, tomatoes, watermelons and grapes are grown in ratios of

The district has a continental climate. Rain falls mostly in the winter months, while summers are dry. The effect of Aegean Sea is felt inland because the mountains run perpendicular to the sea. The land is irrigated in the period from May to September when rainfall is insufficient. Annual average temperature and rainfall (1975-2006) in the district are 16.9 ºC

The program performs the real-time allocation of water at network level in three main stages: 1) allocation of water from the main canal to the secondaries, 2) allocation from

The entire network is divided into different segments in the program. This means that the main canal cross-section between the points where two consecutive secondaries receive water is the primary level segment; a secondary canal cross-section between the points where two consecutive tertiaries receive the water is the secondary level segment, and a tertiary canal cross-section between the points where two consecutive plots receive water constitutes the third level segment. Each different level of segment takes an increasing consecutive index value from head of the network to the end. Therefore, the spatial description of each segment is carried out in the system, and the operation of the program is

Four main components are described for each segment in the program: 1) inflow discharge to head of the segment, 2) water conveyance loss through the segment, 3) amount of water received for irrigation from the segment, and 4) outflow discharge from the end of the segment. These data constitute one of the main components of real-time water allocation

Water distribution stages in the program are performed by running the seven modules interactively in order. Water is received by the plots from the tertiaries. For this reason, the planning process for tertiaries is described in detail, so as to show the effects of water allocation programs at the level of secondaries and the plots. The planning processes for

37.69%, 46.89%, 3.45%, 2.09% and 9.88% (Gediz Irrigation Association Reports, 2007).

responsible to the irrigation association. VIGs are the lowest unit of the irrigation associations, and are responsible for collecting and submitting farmers' water demand forms and managing water distribution at tertiary canal level. Farmers report their water requirements to the VIGs one or two days before the desired irrigation date, and VIGs decide the allocation of water to the plots according to the reports from the farmers. During a fixed length of system rotation period, farmers receive water from the canals to their plots according to this plan. Especially in peak irrigation period and under water scarcity conditions, farmers in the tail end of the network cannot use the system equally and cannot receive an adequate amount of water on schedule. Because, the farmers especially in the head of the canals continue receiving water from the system and decide for themselves whether an adequate amount of water has been received. Disagreements between the farmers are handled by the VIGs or irrigation associations.


Table 1. Crop pattern and size of the area irrigated by the canals.

responsible to the irrigation association. VIGs are the lowest unit of the irrigation associations, and are responsible for collecting and submitting farmers' water demand forms and managing water distribution at tertiary canal level. Farmers report their water requirements to the VIGs one or two days before the desired irrigation date, and VIGs decide the allocation of water to the plots according to the reports from the farmers. During a fixed length of system rotation period, farmers receive water from the canals to their plots according to this plan. Especially in peak irrigation period and under water scarcity conditions, farmers in the tail end of the network cannot use the system equally and cannot receive an adequate amount of water on schedule. Because, the farmers especially in the head of the canals continue receiving water from the system and decide for themselves whether an adequate amount of water has been received. Disagreements between the

(ha) Maize (ha) Watermelons

P.3 0.45 0.00 2.36 0.00 0.50 P.4 7.99 0.00 6.36 0.00 6.86 P.5 5.29 0.95 26.03 0.45 7.43 P.6 1.50 0.80 0.39 0.00 0.30 P.7 5.87 6.20 19.39 0.00 2.55 P.8 2.88 2.92 0.45 0.00 0.57 P.9 12.37 0.00 10.42 1.10 0.25 P.10 1.48 0.00 10.95 0.00 0.00 P.11 4.36 0.96 12.34 0.00 0.86 P.12 6.13 0.00 2.57 0.00 0.00 P.13 8.09 4.02 3.08 0.20 1.10 P.14 13.12 0.46 18.54 0.08 0.00 P.15 8.57 1.41 2.98 0.00 0.00 P.16 15.23 0.00 24.30 0.00 0.00 P.17 4.00 0.00 0.00 1.00 0.00 P.18 15.00 0.00 32.99 0.00 0.00 P.19 11.20 0.00 43.70 2.10 0.10 P.20 41.05 7.13 15.86 5.00 0.30 P.21 12.54 4.00 3.42 0.00 0.00 P.22 11.14 10.76 10.90 0.00 0.00 P.23 19.42 4.64 4.43 1.00 0.00 P.24 9.97 1.25 16.15 1.49 0.00 P.25 2.57 4.14 0.00 0.00 0.00 P.26 19.02 4.37 26.59 0.00 0.00 P.27 0.00 2.81 0.00 0.00 0.00 P.28 0.61 0.00 0.00 0.00 0.00 P.29 0.00 1.02 0.00 0.00 0.00 P.30 1.02 5.31 5.53 0.95 1.21

(ha)

Tomatoes (ha)

farmers are handled by the VIGs or irrigation associations.

Grapes

Table 1. Crop pattern and size of the area irrigated by the canals.

Cotton (ha)

Tertiary name

In the research area, water charges are collected annually by the Gediz Irrigation Association according to the crop type and size of the area. In other words, water from the open canal irrigation system is priced in TL/ha, and is paid for as a single payment for the whole season. Thus, the number of irrigations and the amount of water used in irrigation applications is not important in pricing the water.

Apart from this, producers form crop patterns according to tradition and their own preferences. This has an adverse effect on the efficient management of these systems. In addition, there is a loss of productivity because of the great age of systems like the one under study here. Size of the area irrigated by the canals and the crop patterns are shown in Table 1. In the research area, cotton, maize, tomatoes, watermelons and grapes are grown in ratios of 37.69%, 46.89%, 3.45%, 2.09% and 9.88% (Gediz Irrigation Association Reports, 2007).

The district has a continental climate. Rain falls mostly in the winter months, while summers are dry. The effect of Aegean Sea is felt inland because the mountains run perpendicular to the sea. The land is irrigated in the period from May to September when rainfall is insufficient. Annual average temperature and rainfall (1975-2006) in the district are 16.9 ºC and 704.6 mm respectively (DMI Reports, 2008).

### **3. Description of the irrigation programming model**

### **3.1 Water allocation stages at network level**

The program performs the real-time allocation of water at network level in three main stages: 1) allocation of water from the main canal to the secondaries, 2) allocation from secondary to tertiary canals, and 3) allocation to plots.

The entire network is divided into different segments in the program. This means that the main canal cross-section between the points where two consecutive secondaries receive water is the primary level segment; a secondary canal cross-section between the points where two consecutive tertiaries receive the water is the secondary level segment, and a tertiary canal cross-section between the points where two consecutive plots receive water constitutes the third level segment. Each different level of segment takes an increasing consecutive index value from head of the network to the end. Therefore, the spatial description of each segment is carried out in the system, and the operation of the program is performed interactively in order for each level of segment.

Four main components are described for each segment in the program: 1) inflow discharge to head of the segment, 2) water conveyance loss through the segment, 3) amount of water received for irrigation from the segment, and 4) outflow discharge from the end of the segment. These data constitute one of the main components of real-time water allocation program.

Water distribution stages in the program are performed by running the seven modules interactively in order. Water is received by the plots from the tertiaries. For this reason, the planning process for tertiaries is described in detail, so as to show the effects of water allocation programs at the level of secondaries and the plots. The planning processes for other levels are also carried out in similar ways.

Sustainable Management of Large Scale

Irrigation Systems: A Decision Support Model for Gediz Basin, Turkey 59

Fig. 3. The interface form containing some of the parameters of the model.

process was carried out by use of the formulas given below (Kilic & Anac, 2010).

The second main module determines the canal rotation groups at system level. This process is based on determining the canal groups which cannot receive water at the same time. In order to obtain the highest benefit from the system, the planning process was carried out in accordance with the operation of canals at maximum capacity. In other words, it was ensured that canals received water from the network at maximum capacity. This application also constitutes one of the main principles of water allocation by the rotation method. This

Water conveyance efficiency for a segment

\* ( / 100) max ( 1)

( 1) *<sup>f</sup> l l m r mr*

where, m = indices of secondary canals from head to the end of the network; u = indices of segments in secondary m from head to the end; r = indices of tertiary canals from head of the secondary to the end; QSmur = discharge remaining in the secondary after water is

*<sup>f</sup> QS ESC mur mu QT mu r* (1)

( ) 0 ( 1) *l l m r mr* (2)

Water allocation level

Characteristics of a canal segment

### **3.2 Description of the main modules in the program**

In this section, the modules for preparing real-time irrigation programs for tertiary levels will be described. One of the modules is the structural module. This component contains all the structural and hydraulic features of the network. The main parameters in this module are water carrying capacity and length of each secondary segment; the inflow discharge to head of the secondary cross-section; water conveyance efficiency and maximum water carrying capacity and size of the command of each tertiary. In this stage, it must be taken into consideration that water is delivered from secondary canals to the tertiaries, and these two allocation levels are described interactively in this module. Some parameters in the program are shown schematically together with the layout of the canals in Figure 2.

Fig. 2. Schematic description of parameters of the real-time programming model of an open canal system.

As seen on Figure 2, tertiaries T1-T5 receive water from secondary 1. The lengths of the secondary segments between these tertiaries are shown as Segment 1, Segment 2,.... Segment 5. Average flow velocities in the vertical cross-section of the canals, which are necessary in the description of hydraulic features of the system, were determined by the Velocity-Cross Section method as explained by Mays (1996). Water conveyance losses occurring in the canals were determined by Kilic & Tuylu (2010) according to the Inflow-Outflow method (ANCID, 2003). One of the interface forms containing the parameters explained above is shown in Figure 3.

#### Sustainable Management of Large Scale Irrigation Systems: A Decision Support Model for Gediz Basin, Turkey 59

58 Sustainable Natural Resources Management

In this section, the modules for preparing real-time irrigation programs for tertiary levels will be described. One of the modules is the structural module. This component contains all the structural and hydraulic features of the network. The main parameters in this module are water carrying capacity and length of each secondary segment; the inflow discharge to head of the secondary cross-section; water conveyance efficiency and maximum water carrying capacity and size of the command of each tertiary. In this stage, it must be taken into consideration that water is delivered from secondary canals to the tertiaries, and these two allocation levels are described interactively in this module. Some parameters in the

program are shown schematically together with the layout of the canals in Figure 2.

Fig. 2. Schematic description of parameters of the real-time programming model of an open

As seen on Figure 2, tertiaries T1-T5 receive water from secondary 1. The lengths of the secondary segments between these tertiaries are shown as Segment 1, Segment 2,.... Segment 5. Average flow velocities in the vertical cross-section of the canals, which are necessary in the description of hydraulic features of the system, were determined by the Velocity-Cross Section method as explained by Mays (1996). Water conveyance losses occurring in the canals were determined by Kilic & Tuylu (2010) according to the Inflow-Outflow method (ANCID, 2003). One of the interface forms containing the parameters explained above is

canal system.

shown in Figure 3.

**3.2 Description of the main modules in the program** 

Fig. 3. The interface form containing some of the parameters of the model.

The second main module determines the canal rotation groups at system level. This process is based on determining the canal groups which cannot receive water at the same time. In order to obtain the highest benefit from the system, the planning process was carried out in accordance with the operation of canals at maximum capacity. In other words, it was ensured that canals received water from the network at maximum capacity. This application also constitutes one of the main principles of water allocation by the rotation method. This process was carried out by use of the formulas given below (Kilic & Anac, 2010).

\* ( / 100) max ( 1) *<sup>f</sup> QS ESC mur mu QT mu r* (1)

$$f = l\_{m\{r+1\}} - l\_{mr}$$
 \\

$$(l\_{m\{r+1\}} - l\_{mr}) \ge 0 \tag{2}$$

where, m = indices of secondary canals from head to the end of the network; u = indices of segments in secondary m from head to the end; r = indices of tertiary canals from head of the secondary to the end; QSmur = discharge remaining in the secondary after water is

Sustainable Management of Large Scale

allocation zone a for rotation period i.

rotation period i.

definite rotation period.

the canals in the entire network.

Irrigation Systems: A Decision Support Model for Gediz Basin, Turkey 61

where, a = indices of the allocation zones in the system (in order from the head of the network to the end); na = the total number of allocation zones in the system; i = indices of rotation periods (in order from the beginning of the irrigation season to the end); ni = total number of rotation periods during the entire irrigation season; k = indices of tertiary canals delivering water simultaneously to allocation zone a in rotation period i (in order from the head of the secondary to the end); nkia = the number of tertiary canals delivering water simultaneously to allocation zone a in rotation period i; QTmaxkia = maximum water carrying capacity of tertiary k delivering water to allocation zone a in rotation period i (m3 sec-1); Qmaxia = sum of the maximum water carrying capacities of the tertiary canals delivering water simultaneously to allocation zone a in rotation period i (m3 sec-1); Aa = total size of the irrigated area in allocation zone a in rotation period i (ha); tia = ratio coefficient of

In the second stage, the system factor was determined for a definite rotation period.

1 *na ia a SF R t <sup>i</sup>*

where, R = length of the rotation period for the system (hours); SFi = system factor for the

In the third stage, length of irrigation time was determined for each allocation zone during a

where, IRTia = length of irrigation time for allocation zone a in rotation period i (hours).

processes in the previous stages were converted to a general equation.

max \* 1 1 *nkia na IRT A QT R t ia <sup>a</sup> kia ia k a*

constraint levels occurring in tertiary commands is performed in the next module.

In the fourth stage, the formula shown below was obtained when all the calculation

As can be understood from the calculation process explained above, the planning is carried out in accordance with the operation of the system at maximum capacity. In addition however, plant pattern and soil features of the allocation zones may be different from each other, which means that the irrigation water requirements of the zones will be different from each other too. Determination of irrigation water requirements of allocation zones and analysis of water

In the fifth module, irrigation water requirements of the crops grown in the command of each tertiary are determined as volume for a given period. The value of this parameter is used as a transition stage in determining the amount of water to be allocated from the resource to a given allocation zone. It is also used in determining the length of irrigation time necessary to meet the water requirements of the crops, and water deficits occurring in

*t AQ*max *ia a ia* For a 1, na ; For i 1, ni (5)

For i 1, ni (6)

For a 1, na ; For i 1, ni (8)

*IRT t SF* \* *ia ia i* For a 1, na ; For i 1, ni (7)

received by tertiary r from segment u of secondary m (m3 sec-1); ESCmu = water conveyance efficiency for segment u of secondary m (% 1km-1); QTmaxmu(r+1) = maximum carrying capacity of the consecutive tertiary (r+1), receiving water from segment u of secondary m (m3 sec-1); lmr = the distance from the point where tertiary r receives water to the head of secondary m (km); lm(r+1) = distance from the point where consecutive tertiary (r+1) receives water to the head of secondary m (km); f = Length of a secondary canal segment between the consecutive tertiaries r and (r+1) receiving water from secondary m (km).

Each tertiary canal validating the conditions indicated in formulas (1) and (2) will be in the same rotation group and can receive water in maximum capacity from the secondary at the same time. On the other hand, if the conditions are not validated by the tertiary, this canal will be in the consecutive rotation zone together with the tertiaries validating the conditions. The canal rotation groups were formed by carrying out the process repetitively for the entire network. Thus, the system is divided into different allocation zones to ensure efficient usage of resources and operation of the network.

The third module determines borders, sizes and numbers of the allocation zones devised by the model in the system. Total sizes of the commands irrigated by the canal rotation groups serving each allocation zone also describe the borders and sizes of these zones. Indices are given to the allocation zones in an increasing order from head of the network to the end. In the program, borders of the allocation zones are represented by the names and indices of the head and end segments of the canals irrigating the area. These borders are described by the formulas given below in the model.

$$AR = \sum\_{z=1}^{nz\_a} A r\_{ax} \text{ For } \mathbf{a} = \mathbf{1}\_\prime \text{ na} \tag{3}$$

where, a = indices of allocation zones in the system (in order from head of the network to the end); na = total number of allocation zones in the system; z = indices of tertiary canals delivering water simultaneously to allocation zone a (in order from head to the end of the secondary); nza = total number of tertiary canals delivering the water simultaneously to allocation zone a; ATaz = size of the area irrigated by tertiary z in allocation zone a (ha); AR = total size of the irrigated area in allocation zone a (ha).

In the fourth module, the lengths of irrigation times to be allocated to the zones during the system rotation period are determined in accordance with the principle of delivering equal amounts of water per unit of area in each allocation zone. In other words, whatever the location of a plot in the network, the capacity of the canal receiving water from the system, or the water conveyance efficiency, this plot will benefit from the water resource and system equally in temporal and spatial dimensions. This process is described for each canal rotation group which cannot receive water at the same time. The lengths of water allocation periods for different allocation zones in a given irrigation period were determined in four main stages as explained below.

In the first stage**,** ratio coefficient values were determined for each allocation zone in a definite rotation period. These calculations are formulated below.

$$Q \max\_{i \neq \dots \neq i} \frac{n k\_{ia}}{k = 1} QT \max\_{k \neq i} \tag{4}$$

received by tertiary r from segment u of secondary m (m3 sec-1); ESCmu = water conveyance efficiency for segment u of secondary m (% 1km-1); QTmaxmu(r+1) = maximum carrying capacity of the consecutive tertiary (r+1), receiving water from segment u of secondary m (m3 sec-1); lmr = the distance from the point where tertiary r receives water to the head of secondary m (km); lm(r+1) = distance from the point where consecutive tertiary (r+1) receives water to the head of secondary m (km); f = Length of a secondary canal segment between

Each tertiary canal validating the conditions indicated in formulas (1) and (2) will be in the same rotation group and can receive water in maximum capacity from the secondary at the same time. On the other hand, if the conditions are not validated by the tertiary, this canal will be in the consecutive rotation zone together with the tertiaries validating the conditions. The canal rotation groups were formed by carrying out the process repetitively for the entire network. Thus, the system is divided into different allocation zones to ensure efficient usage

The third module determines borders, sizes and numbers of the allocation zones devised by the model in the system. Total sizes of the commands irrigated by the canal rotation groups serving each allocation zone also describe the borders and sizes of these zones. Indices are given to the allocation zones in an increasing order from head of the network to the end. In the program, borders of the allocation zones are represented by the names and indices of the head and end segments of the canals irrigating the area. These borders are described by the

where, a = indices of allocation zones in the system (in order from head of the network to the end); na = total number of allocation zones in the system; z = indices of tertiary canals delivering water simultaneously to allocation zone a (in order from head to the end of the secondary); nza = total number of tertiary canals delivering the water simultaneously to allocation zone a; ATaz = size of the area irrigated by tertiary z in allocation zone a (ha); AR

In the fourth module, the lengths of irrigation times to be allocated to the zones during the system rotation period are determined in accordance with the principle of delivering equal amounts of water per unit of area in each allocation zone. In other words, whatever the location of a plot in the network, the capacity of the canal receiving water from the system, or the water conveyance efficiency, this plot will benefit from the water resource and system equally in temporal and spatial dimensions. This process is described for each canal rotation group which cannot receive water at the same time. The lengths of water allocation periods for different allocation zones in a given irrigation period were determined in four main

In the first stage**,** ratio coefficient values were determined for each allocation zone in a

max max 1 *nkia <sup>Q</sup> QT ia kia <sup>k</sup>* 

For a 1, na (3)

(4)

1 *nza AR ATaz <sup>z</sup>* 

the consecutive tertiaries r and (r+1) receiving water from secondary m (km).

of resources and operation of the network.

formulas given below in the model.

stages as explained below.

= total size of the irrigated area in allocation zone a (ha).

definite rotation period. These calculations are formulated below.

$$
\omega\_{ia} = A\_a \not\!Q \max\_{ia} \text{ For } \mathbf{a} = \mathbf{1}, \text{ ra } \text{\textquotedblleft}, \text{For } \mathbf{i} = \mathbf{1}, \text{\textquotedblright} \end{bmatrix} \tag{5}
$$

where, a = indices of the allocation zones in the system (in order from the head of the network to the end); na = the total number of allocation zones in the system; i = indices of rotation periods (in order from the beginning of the irrigation season to the end); ni = total number of rotation periods during the entire irrigation season; k = indices of tertiary canals delivering water simultaneously to allocation zone a in rotation period i (in order from the head of the secondary to the end); nkia = the number of tertiary canals delivering water simultaneously to allocation zone a in rotation period i; QTmaxkia = maximum water carrying capacity of tertiary k delivering water to allocation zone a in rotation period i (m3 sec-1); Qmaxia = sum of the maximum water carrying capacities of the tertiary canals delivering water simultaneously to allocation zone a in rotation period i (m3 sec-1); Aa = total size of the irrigated area in allocation zone a in rotation period i (ha); tia = ratio coefficient of allocation zone a for rotation period i.

In the second stage, the system factor was determined for a definite rotation period.

$$S\_{\hat{i}}^{\varepsilon} = \mathcal{R} \Big/ \sum\_{u=1}^{m} t\_{iu} \text{ For } \mathbf{i} = \mathbf{1}\_{\prime} \text{ ni} \tag{6}$$

where, R = length of the rotation period for the system (hours); SFi = system factor for the rotation period i.

In the third stage, length of irrigation time was determined for each allocation zone during a definite rotation period.

$$\text{R}\mathbb{R}T\_{\text{int}} = t\_{\text{int}} \text{ \* s}\mathbb{F}\_{\text{i}} \text{ For } \mathbf{a} = \mathbf{1} \text{ \* } \text{na} \text{ \* For } \mathbf{i} = \mathbf{1} \text{ \* ni} \tag{7}$$

where, IRTia = length of irrigation time for allocation zone a in rotation period i (hours).

In the fourth stage, the formula shown below was obtained when all the calculation processes in the previous stages were converted to a general equation.

$$\mathbf{a}\_{\text{RRT}\_{\text{lat}}} - \left[A\_a \int\_{k=1}^{nk\_{\text{lat}}} \varrho \mathbf{r} \, \max\_{\text{fia}} \mathbf{a}\_{\text{fia}} \right] \bullet \left[\bigwedge\_{a=1}^{na} \frac{a}{\sum\_{a=1}^{I\_{\text{fia}}}} \right] \mathbf{F} \mathbf{or} \; \mathbf{a} = \mathbf{1}, \; \mathbf{na} \; \text{;} \mathbf{For } \mathbf{i} = \mathbf{1}, \; \mathbf{ni} \tag{8}$$

As can be understood from the calculation process explained above, the planning is carried out in accordance with the operation of the system at maximum capacity. In addition however, plant pattern and soil features of the allocation zones may be different from each other, which means that the irrigation water requirements of the zones will be different from each other too. Determination of irrigation water requirements of allocation zones and analysis of water constraint levels occurring in tertiary commands is performed in the next module.

In the fifth module, irrigation water requirements of the crops grown in the command of each tertiary are determined as volume for a given period. The value of this parameter is used as a transition stage in determining the amount of water to be allocated from the resource to a given allocation zone. It is also used in determining the length of irrigation time necessary to meet the water requirements of the crops, and water deficits occurring in the canals in the entire network.

Sustainable Management of Large Scale

Irrigation Systems: A Decision Support Model for Gediz Basin, Turkey 63

Crop type, irrigation water requirement, size of the area and irrig. efficiency for a plot

Spatial description of plots in a definite allocation zone

Some of the interactive solution reports for a tertiary command

Fig. 4. A sample interface form used for description of the plant pattern, irrigation water

In the sixth module, alternative irrigation programs are prepared by changing the values of parameters in the program as desired. For example, water deficiency levels occurring in each tertiary canal can be determined by changing the length of system rotation period. Thus, the most suitable length of rotation period can be decided by taking into consideration the deficit levels occurring in the entire network. This module can derive alternative solutions for desired numbers of irrigation periods. Apart from this, adequate carrying capacity of the canals needed to meet the water requirements of the crops can be determined by running this module. Also, optimum size of command which can be irrigated by the infrastructure of the network in reality can be determined. In addition, the priority and degree of maintenance and renovation works of the system can be decided by this module. Thus, insight is provided to the decision maker into the use of limited labor and financial resources at an optimum level. In this process, the results of possible operation plans from the model solution can be analyzed before making a final decision on operation strategies of

In the seventh module, detailed report files are prepared for each alternative solution, and these are presented to the decision maker as tables. Therefore, an evaluation can be achieved for the entire system. A sample report file interface form for this process is shown in Figure 5. The flow chart of the MONES 4.1 model devising irrigation programs at network level is

requirements of the crops, and levels of water deficit in the system.

the system.

given in Figure 6.

Irrigation water requirements and constraint levels occurring in the command of each tertiary canal were determined using the formulas given below.

$$\text{WVD}\,V\_{\text{akc\'i}} = \Sigma \left(D\_{\text{akc\'i}}\,^\*A\_{\text{akc\'}}\,^\*\mathbf{10}\right) \tag{9}$$

where WDVakci = total amount of irrigation water requirement as volume of crop c grown in the command of tertiary k, in allocation zone a, for the rotation period i (m3); Dakci = total amount of irrigation water requirement of crop c, irrigated by tertiary k, in allocation zone a, in rotation period i (mm). The value of this parameter was determined using a well known package, Cropwat (FAO, 1992). About 30-45% of available moisture between the permanent wilting point and field capacity was allowed to be depleted, and the soil moisture was refilled to field capacity at each irrigation. Exceeding the soil moisture depletion level by over 50% was not allowed, as explained by Doorenbos & Kassam (1979). Aakc = size of the growing area of crop c irrigated by tertiary k, in allocation zone a (ha).

Amount of irrigation water allocated to the crops grown in the command of a tertiary canal were determined by the formula given below.

$$A\mathcal{W}\_{kin} = QT \max\_{kia} \ast\_{kia} \ast\_{iia} \ast\_{\mathcal{B}600} \tag{10}$$

where AWkia = amount of irrigation water allocated to the crops grown in the command of tertiary k, in allocation zone a, during the rotation period i (m3); QTmaxkia = maximum water carrying capacity of tertiary k delivering water to allocation zone a in rotation period i (m3 sec-1); IRTia = length of irrigation time for allocation zone a in rotation period i (hours); 3600 = the coefficient converting hour to second.

Water constraint levels occurring in the command of each tertiary were determined by the formula given below.

$$\text{VDL}\_{\text{kin}} = ((\text{VDV}\_{\text{akci}} - \text{AV}\_{\text{kin}}) / \text{VDV}\_{\text{akci}}) \, \text{\*} 100 \, \left(\text{if } \text{VDV} \, \text{V}\_{\text{akci}} \ge \text{AW}\_{\text{kia}}\right) \tag{11}$$

where WDLkia = water constraint level occurring in the command of tertiary k, in allocation zone a, in rotation period i (%).

In the fourth module, the length of irrigation time allocated for each zone during the system rotation period was determined in accordance with the operation of the network at maximum capacity. In the fifth stage on the other hand, the level is determined at which the irrigation water requirements of the crops can be met by the actual infrastructure of the system. For this purpose, water constraint levels occurring in each tertiary command are determined. One of the interface forms carrying out these processes in the model is shown in Figure 4.

Irrigation water requirements of the crops vary at different growing stages. There is not a linear relationship between the amount of water given to the system and yield of crops. Thus, the Yield Response Factor (ky) takes different values at each growing stage (Kilic, 2004). Numbers and borders of allocation zones, canal rotation groups, and length of irrigation times for the zones may change depending on the conditions in different periods. The MONES 4.1 (Kilic, 2010) package provides the real-time irrigation programs by taking into consideration of varying conditions in the system.

#### Sustainable Management of Large Scale Irrigation Systems: A Decision Support Model for Gediz Basin, Turkey 63

62 Sustainable Natural Resources Management

Irrigation water requirements and constraint levels occurring in the command of each

where WDVakci = total amount of irrigation water requirement as volume of crop c grown in the command of tertiary k, in allocation zone a, for the rotation period i (m3); Dakci = total amount of irrigation water requirement of crop c, irrigated by tertiary k, in allocation zone a, in rotation period i (mm). The value of this parameter was determined using a well known package, Cropwat (FAO, 1992). About 30-45% of available moisture between the permanent wilting point and field capacity was allowed to be depleted, and the soil moisture was refilled to field capacity at each irrigation. Exceeding the soil moisture depletion level by over 50% was not allowed, as explained by Doorenbos & Kassam (1979). Aakc = size of the

Amount of irrigation water allocated to the crops grown in the command of a tertiary canal

where AWkia = amount of irrigation water allocated to the crops grown in the command of tertiary k, in allocation zone a, during the rotation period i (m3); QTmaxkia = maximum water carrying capacity of tertiary k delivering water to allocation zone a in rotation period i (m3 sec-1); IRTia = length of irrigation time for allocation zone a in rotation period i (hours);

Water constraint levels occurring in the command of each tertiary were determined by the

where WDLkia = water constraint level occurring in the command of tertiary k, in allocation

In the fourth module, the length of irrigation time allocated for each zone during the system rotation period was determined in accordance with the operation of the network at maximum capacity. In the fifth stage on the other hand, the level is determined at which the irrigation water requirements of the crops can be met by the actual infrastructure of the system. For this purpose, water constraint levels occurring in each tertiary command are determined. One of the interface forms carrying out these processes in the model is shown

Irrigation water requirements of the crops vary at different growing stages. There is not a linear relationship between the amount of water given to the system and yield of crops. Thus, the Yield Response Factor (ky) takes different values at each growing stage (Kilic, 2004). Numbers and borders of allocation zones, canal rotation groups, and length of irrigation times for the zones may change depending on the conditions in different periods. The MONES 4.1 (Kilic, 2010) package provides the real-time irrigation programs by taking

*WDL WDV AW WDV* (( ) / ) \* 100 *kia akci kia akci* if WDV AW akci kia (11)

*WDV D A* ( \* \*10) *akci akci akc* (9)

*AW QT IRT* max \* \* 3600 *kia kia ia* (10)

tertiary canal were determined using the formulas given below.

growing area of crop c irrigated by tertiary k, in allocation zone a (ha).

were determined by the formula given below.

3600 = the coefficient converting hour to second.

into consideration of varying conditions in the system.

formula given below.

in Figure 4.

zone a, in rotation period i (%).

Fig. 4. A sample interface form used for description of the plant pattern, irrigation water requirements of the crops, and levels of water deficit in the system.

In the sixth module, alternative irrigation programs are prepared by changing the values of parameters in the program as desired. For example, water deficiency levels occurring in each tertiary canal can be determined by changing the length of system rotation period. Thus, the most suitable length of rotation period can be decided by taking into consideration the deficit levels occurring in the entire network. This module can derive alternative solutions for desired numbers of irrigation periods. Apart from this, adequate carrying capacity of the canals needed to meet the water requirements of the crops can be determined by running this module. Also, optimum size of command which can be irrigated by the infrastructure of the network in reality can be determined. In addition, the priority and degree of maintenance and renovation works of the system can be decided by this module. Thus, insight is provided to the decision maker into the use of limited labor and financial resources at an optimum level. In this process, the results of possible operation plans from the model solution can be analyzed before making a final decision on operation strategies of the system.

In the seventh module, detailed report files are prepared for each alternative solution, and these are presented to the decision maker as tables. Therefore, an evaluation can be achieved for the entire system. A sample report file interface form for this process is shown in Figure 5.

The flow chart of the MONES 4.1 model devising irrigation programs at network level is given in Figure 6.

Sustainable Management of Large Scale

to desired parameters in the model.

alternative solutions of the model.

from those applied in reality in the system.

plan applied in the district in reality.

given in Table 2.

**4. Results and discussion** 

Irrigation Systems: A Decision Support Model for Gediz Basin, Turkey 65

Firstly, data input process to the related modules was carried out in the program (Figure 6). This step inquired whether data input to the system was completed or not. Another stage in running the model is derivation of alternative irrigation programs. If decision maker decides that alternative programs must be devised for a given irrigation period, the model is run again by the second conditional return, shown in the flow chart in Figure 6. By running this module, it is possible to derive alternative irrigation programs by making necessary changes

At the end of this process, optimum operation strategies for the system can be decided by analyzing the water constraint levels occurring in canals, the length of irrigation periods necessary for different allocation zones, the amount of water used in the network, and the allocation of deficit resources to different irrigation periods. Consequently, before deciding on an operation strategy for the system, the optimum program can be selected by analyzing

Significant levels of differences occurred between the water allocation plan applied in the research area in reality and the model solution. It was determined that canal rotation groups, allocation zones and irrigation programs from the model solution were different

In the research area, a 10 day system rotation period is applied by the current water allocation program in the network. Tertiaries are held open during this period by the Gediz Irrigation Association, and water is received in the plots by the farmers without any planning when it is released to the canals (Gediz Irrigation Association Reports, 2007). In other words, in the water allocation plan applied in reality, irrigation water is given to all the tertiary canals at the same time, and an attempt is made to irrigate the entire command in 10 days. This application prevents the operation of the system at maximum capacity and does not enable the optimum benefit to be obtained from production. Such a practice causes the irrigation water to be taken from the canals immediately, especially by the farmers whose plots are in the head of the network. Uncontrolled water allocation prevents it from being received in the desired amount and at the desired time by the farmers whose plots are in the tail end of the system. Consequently, it is not possible to provide social equity in temporal and spatial dimensions in use of the system and allocation of deficit resources to large numbers of users by means of the water allocation

In addition, no scientific plant patterning has been carried out in the research area or elsewhere in Turkey. Thus, producers choose crop patterns according to tradition and market conditions. In this state, it is necessary to prepare real-time irrigation programs at network level, and the parameters necessary for optimum growing conditions must be taken into consideration. In this way, optimum crop yield and benefit can be obtained by using the system capacity at maximum level. Allocation Zones (AZ), canal rotation groups and size of irrigated areas obtained by running the irrigation programming model MONES 4.1 are


Fig. 5. A sample interface form for the preparation of alternative solution reports.

Fig. 6. Model flowchart for real-time irrigation programs.

Firstly, data input process to the related modules was carried out in the program (Figure 6). This step inquired whether data input to the system was completed or not. Another stage in running the model is derivation of alternative irrigation programs. If decision maker decides that alternative programs must be devised for a given irrigation period, the model is run again by the second conditional return, shown in the flow chart in Figure 6. By running this module, it is possible to derive alternative irrigation programs by making necessary changes to desired parameters in the model.

At the end of this process, optimum operation strategies for the system can be decided by analyzing the water constraint levels occurring in canals, the length of irrigation periods necessary for different allocation zones, the amount of water used in the network, and the allocation of deficit resources to different irrigation periods. Consequently, before deciding on an operation strategy for the system, the optimum program can be selected by analyzing alternative solutions of the model.

### **4. Results and discussion**

64 Sustainable Natural Resources Management

Length of system rotation period (day)

Fig. 5. A sample interface form for the preparation of alternative solution reports.

Irrigation programming report for a definite period

Fig. 6. Model flowchart for real-time irrigation programs.

Significant levels of differences occurred between the water allocation plan applied in the research area in reality and the model solution. It was determined that canal rotation groups, allocation zones and irrigation programs from the model solution were different from those applied in reality in the system.

In the research area, a 10 day system rotation period is applied by the current water allocation program in the network. Tertiaries are held open during this period by the Gediz Irrigation Association, and water is received in the plots by the farmers without any planning when it is released to the canals (Gediz Irrigation Association Reports, 2007). In other words, in the water allocation plan applied in reality, irrigation water is given to all the tertiary canals at the same time, and an attempt is made to irrigate the entire command in 10 days. This application prevents the operation of the system at maximum capacity and does not enable the optimum benefit to be obtained from production. Such a practice causes the irrigation water to be taken from the canals immediately, especially by the farmers whose plots are in the head of the network. Uncontrolled water allocation prevents it from being received in the desired amount and at the desired time by the farmers whose plots are in the tail end of the system. Consequently, it is not possible to provide social equity in temporal and spatial dimensions in use of the system and allocation of deficit resources to large numbers of users by means of the water allocation plan applied in the district in reality.

In addition, no scientific plant patterning has been carried out in the research area or elsewhere in Turkey. Thus, producers choose crop patterns according to tradition and market conditions. In this state, it is necessary to prepare real-time irrigation programs at network level, and the parameters necessary for optimum growing conditions must be taken into consideration. In this way, optimum crop yield and benefit can be obtained by using the system capacity at maximum level. Allocation Zones (AZ), canal rotation groups and size of irrigated areas obtained by running the irrigation programming model MONES 4.1 are given in Table 2.

Sustainable Management of Large Scale

time programming.

peak irrigation period, July.

package are given in Tables 4-6.

met completely during the 12 day system rotation period.

Irrigation Systems: A Decision Support Model for Gediz Basin, Turkey 67

determined in accordance with the irrigation water requirements of crops at different growing stages, size of area, location and soil features of plots receiving water from canals, infrastructure of the network, length and layout of canals, water carrying capacity of different canal segments, conveyance efficiencies and hydraulic features of the network. In this process, the most suitable length of system rotation period for the parameters stated above constitutes one of the main components in determining the irrigation times in real-

The MONES 4.1 package (Kilic, 2010) was run for the entire irrigation season in the research area, and irrigation programs were obtained. In order to bring out some of the main features of the model, three different irrigation periods were handled, beginning on 6th June, 10th July and 18th August, respectively. One of the main reasons for selecting these periods is to evaluate irrigation programs for different growing stages of the crops. Thus, the program was run for irrigation water requirements of the crops, which vary during the growing period, and the results were evaluated. The second main reason for selecting these periods was to investigate the irrigation programs of June and August, together with the program of

In irrigation programming in the model, depletion of 30-45% of the available moisture between permanent wilting point and field capacity was allowed, and the soil moisture was refilled to field capacity at each irrigation. Exceeding the soil moisture depletion level by

The extent to which irrigation water requirements of the crops could be met on 6 June, 10 July and 18 August for system rotation periods of 7, 10, 11 and 12 days was determined by running the model. In other words, it was shown by alternative solutions to what level the irrigation water requirements of crops in given periods could be met by system capacity. The length of alternative rotation periods stated above did not cause any problem from the point of view of minimum irrigation intervals of the crops (Kilic & Ozgurel, 2005; Kodal et al., 1997; Sagardoy et al., 1982). However, it was determined that water deficit levels in some canals exceeded 45%, especially in rotation periods which were shorter than necessary. These canals serve a larger area than they can irrigate, because of the unplanned production pattern. In this state, irrigation water requirements of the crops cannot be met completely because of the inadequate carrying capacity of some canals and a shorter length of rotation period than necessary. Summarized results of the MONES 4.1

As seen in Table 4, water deficits reaching 38.15% (P.7 tertiary) occurred for the 10 day system rotation period in the irrigation applications beginning on 6th June. The maximum levels of water deficit occurring during the 7, 11 and 12 day system rotation periods were 56.31%, 32.10% and 26.05%, respectively. An increase in the length of rotation period diminished the levels of water constraints occurring in the canals. The lengths of these periods were also suitable for minimum irrigation interval of the crops in the research area. However, it is clear that 56.31% water deficiency level occurring in the 7 day system rotation period is not suitable for optimum irrigation and growing conditions of these crops. On the other hand, irrigation water requirements of all 25 tertiaries, except P.7, P.9 and P.26, were

over 50% was not allowed, as explained by Doorenbos & Kassam (1979).


Table 2. Allocation zones, canal rotation groups and total size of the commands in the MONES 4.1 model.

Three different allocation zones were determined in the research area, as maximum carrying capacity of the secondary canal is not adequate for delivering water to all tertiaries at the same time. The number of tertiary canals allocating water to AZ I, AZ II and AZ III also decreased progressively along the length of the secondary, as the capacity of the secondary canal diminishes from the head of the network to the end. In other words, fewer tertiary canals could deliver water at the same time to a given allocation zone at the end of the network than at the head, because of the progressive reduction of the secondary canal's capacity. For example, at the head of the network, 20 tertiary canals (P.3-P.22) can irrigate the command of AZ I at the same time, while only 4 tertiaries (P.27-P.30) can irrigate the command of AZ III simultaneously, as it is at the end of the network. In this way, all the canals in the network were operated in maximum level.

Maximum lengths of irrigation times (IRT) allocated for the zones (AZ) during the 7, 10, 11 and 12-day alternative system rotation periods (R) for the district are given in Table 3.

As can be understood from the Table 3, the maximum lengths of irrigation times allocated for different zones vary depending on the length of the alternative system rotation periods. These allocation times were planned in order to give equal amounts of water to the unit areas of different allocation zones.


Table 3. Lengths of irrigation times allocated for the zones during the alternative system rotation periods.

Irrigation water requirements of the allocation zones in different periods are different from each other. The whole irrigation water requirement of the allocation zones could not be met completely in the maximum length of irrigation time allocated for the zones during the system rotation period. This causes deficit irrigation applications. For this purpose, irrigation programs developed for each tertiary canal are analyzed in detail for each irrigation period. In this way it was possible to analyze the effects of the lengths of irrigation times on water deficits occurring in different periods.

In addition, irrigation times for the whole season are not determined as fixed time points at the beginning of the irrigation season in real-time programming. Irrigation times are

AZI P.3-P.22. 505.65 AZII P.23-P.26. 115.04 AZIII P.27-P.30. 18.46

canals in the network were operated in maximum level.

7 days (168 hours)

times on water deficits occurring in different periods.

areas of different allocation zones.

Allocation Zones

rotation periods.

Table 2. Allocation zones, canal rotation groups and total size of the commands in the

Three different allocation zones were determined in the research area, as maximum carrying capacity of the secondary canal is not adequate for delivering water to all tertiaries at the same time. The number of tertiary canals allocating water to AZ I, AZ II and AZ III also decreased progressively along the length of the secondary, as the capacity of the secondary canal diminishes from the head of the network to the end. In other words, fewer tertiary canals could deliver water at the same time to a given allocation zone at the end of the network than at the head, because of the progressive reduction of the secondary canal's capacity. For example, at the head of the network, 20 tertiary canals (P.3-P.22) can irrigate the command of AZ I at the same time, while only 4 tertiaries (P.27-P.30) can irrigate the command of AZ III simultaneously, as it is at the end of the network. In this way, all the

Maximum lengths of irrigation times (IRT) allocated for the zones (AZ) during the 7, 10, 11 and 12-day alternative system rotation periods (R) for the district are given in Table 3.

As can be understood from the Table 3, the maximum lengths of irrigation times allocated for different zones vary depending on the length of the alternative system rotation periods. These allocation times were planned in order to give equal amounts of water to the unit

> 10 days (240 hours)

AZI 95.06 135.80 149.38 162.96 AZII 51.93 74.18 81.60 89.02 AZIII 21.01 30.02 33.02 36.02 Table 3. Lengths of irrigation times allocated for the zones during the alternative system

Irrigation water requirements of the allocation zones in different periods are different from each other. The whole irrigation water requirement of the allocation zones could not be met completely in the maximum length of irrigation time allocated for the zones during the system rotation period. This causes deficit irrigation applications. For this purpose, irrigation programs developed for each tertiary canal are analyzed in detail for each irrigation period. In this way it was possible to analyze the effects of the lengths of irrigation

In addition, irrigation times for the whole season are not determined as fixed time points at the beginning of the irrigation season in real-time programming. Irrigation times are

Alternative system rotation periods

11 days (264 hours)

12 days (288 hours)

Tertiary canal rotation groups Total size of the irrigated area (ha)

Allocation Zone

MONES 4.1 model.

determined in accordance with the irrigation water requirements of crops at different growing stages, size of area, location and soil features of plots receiving water from canals, infrastructure of the network, length and layout of canals, water carrying capacity of different canal segments, conveyance efficiencies and hydraulic features of the network. In this process, the most suitable length of system rotation period for the parameters stated above constitutes one of the main components in determining the irrigation times in realtime programming.

The MONES 4.1 package (Kilic, 2010) was run for the entire irrigation season in the research area, and irrigation programs were obtained. In order to bring out some of the main features of the model, three different irrigation periods were handled, beginning on 6th June, 10th July and 18th August, respectively. One of the main reasons for selecting these periods is to evaluate irrigation programs for different growing stages of the crops. Thus, the program was run for irrigation water requirements of the crops, which vary during the growing period, and the results were evaluated. The second main reason for selecting these periods was to investigate the irrigation programs of June and August, together with the program of peak irrigation period, July.

In irrigation programming in the model, depletion of 30-45% of the available moisture between permanent wilting point and field capacity was allowed, and the soil moisture was refilled to field capacity at each irrigation. Exceeding the soil moisture depletion level by over 50% was not allowed, as explained by Doorenbos & Kassam (1979).

The extent to which irrigation water requirements of the crops could be met on 6 June, 10 July and 18 August for system rotation periods of 7, 10, 11 and 12 days was determined by running the model. In other words, it was shown by alternative solutions to what level the irrigation water requirements of crops in given periods could be met by system capacity. The length of alternative rotation periods stated above did not cause any problem from the point of view of minimum irrigation intervals of the crops (Kilic & Ozgurel, 2005; Kodal et al., 1997; Sagardoy et al., 1982). However, it was determined that water deficit levels in some canals exceeded 45%, especially in rotation periods which were shorter than necessary. These canals serve a larger area than they can irrigate, because of the unplanned production pattern. In this state, irrigation water requirements of the crops cannot be met completely because of the inadequate carrying capacity of some canals and a shorter length of rotation period than necessary. Summarized results of the MONES 4.1 package are given in Tables 4-6.

As seen in Table 4, water deficits reaching 38.15% (P.7 tertiary) occurred for the 10 day system rotation period in the irrigation applications beginning on 6th June. The maximum levels of water deficit occurring during the 7, 11 and 12 day system rotation periods were 56.31%, 32.10% and 26.05%, respectively. An increase in the length of rotation period diminished the levels of water constraints occurring in the canals. The lengths of these periods were also suitable for minimum irrigation interval of the crops in the research area. However, it is clear that 56.31% water deficiency level occurring in the 7 day system rotation period is not suitable for optimum irrigation and growing conditions of these crops. On the other hand, irrigation water requirements of all 25 tertiaries, except P.7, P.9 and P.26, were met completely during the 12 day system rotation period.

Sustainable Management of Large Scale

Total irrigation water requirement (m3)

applications beginning on 10th of July.

given in Table 6.

period.

Tertiary name

Irrigation Systems: A Decision Support Model for Gediz Basin, Turkey 69

tertiaries except P.7, P.9 and P.26 were met completely during the 12 day system rotation

P.3 4908.777 0 0 0 0 P.4 29787.900 0 0 0 0 P.5 58798.374 0 0 0 0 P.6 4147.970 0 0 0 0 P.7 49443.972 64.91 50.43 45.61 40.78 P.8 9229.546 0 0 0 0 P.9 35387.586 51.49 31.27 24.53 17.79 P.10 19003.545 9.05 0 0 0 P.11 27568.383 38.11 12.15 3.49 0 P.12 12696.326 0 0 0 0 P.13 23099.029 26.39 0 0 0 P.14 47988.167 29.08 0 0 0 P.15 18626.505 9.03 0 0 0 P.16 59186.227 0 0 0 0 P.17 6949.991 0 0 0 0 P.18 72267.113 0 0 0 0 P.19 86136.363 21.13 0 0 0 P.20 98793.526 0 0 0 0 P.21 28307.164 39.69 14.41 5.98 0 P.22 46591.310 0 0 0 0 P.23 41744.648 0 0 0 0 P.24 42601.463 34.44 6.91 0 0 P.25 9010.986 0 0 0 0 P.26 73765.516 49.22 28.02 20.95 13.88 P.27 3624.255 0 0 0 0 P.28 868.814 0 0 0 0 P.29 1315.566 35.12 7.88 0 0 P.30 19554.267 0 0 0 0

Table 5. Water deficit levels occurring in the alternative rotation periods for the irrigation

Results obtained by running the model for irrigation period beginning on 18th August are

Water deficit levels occurring in the length of alternative system rotation periods (%) 7 days 10 days 11 days 12 days


Table 4. Water deficit levels occurring in the alternative rotation periods for the irrigation applications beginning on 6th of June.

July is the peak irrigation period in the district. The results obtained from model solution for the irrigation applications beginning on 10th July are given in Table 5.

As seen in Table 5, 64.91%, 50.43%, 45.61% and 40.78% of maximum water constraints occurred respectively for system rotation periods of 7, 10, 11 and 12 days. It is not suitable for optimum irrigation conditions that water constraint levels occurring in the rotation periods of 7 and 10 days be over 50% (Doorenbos & Kassam, 1979). In other words, high water requirements in some canals cannot be met completely because of the inadequate canal carrying capacity and shorter than necessary length of rotation period. It is clear that a 10 day system rotation period applied in reality in the network caused a yield loss, especially in the peak irrigation period. However, irrigation water requirements of all 25

P.3 4767.272 0 0 0 0 P.4 28708.209 0 0 0 0 P.5 60016.997 0 0 0 0 P.6 4015.468 0 0 0 0 P.7 39412.259 56.31 38.15 32.10 26.05 P.8 9088.454 0 0 0 0 P.9 30811.573 44.48 21.26 13.51 5.77 P.10 18433.697 6.28 0 0 0 P.11 26643.547 36.01 9.14 0.19 0 P.12 12028.759 0 0 0 0 P.13 22117.774 23.18 0 0 0 P.14 46031.517 26.13 0 0 0 P.15 17683.992 4.25 0 0 0 P.16 57319.330 0 0 0 0 P.17 6567.543 0 0 0 0 P.18 70256.602 0 0 0 0 P.19 83344.510 18.54 0 0 0 P.20 94137.366 0 0 0 0 P.21 26919.283 36.65 10.06 1.20 0 P.22 44876.254 0 0 0 0 P.23 39653.301 0 0 0 0 P.24 40992.809 31.92 3.31 0 0 P.25 8675.461 0 0 0 0 P.26 70853.951 47.18 25.11 17.76 10.40 P.27 3564.963 0 0 0 0 P.28 812.503 0 0 0 0 P.29 1294.043 34.06 6.37 0 0 P.30 19086.784 0 0 0 0 Table 4. Water deficit levels occurring in the alternative rotation periods for the irrigation

July is the peak irrigation period in the district. The results obtained from model solution for

As seen in Table 5, 64.91%, 50.43%, 45.61% and 40.78% of maximum water constraints occurred respectively for system rotation periods of 7, 10, 11 and 12 days. It is not suitable for optimum irrigation conditions that water constraint levels occurring in the rotation periods of 7 and 10 days be over 50% (Doorenbos & Kassam, 1979). In other words, high water requirements in some canals cannot be met completely because of the inadequate canal carrying capacity and shorter than necessary length of rotation period. It is clear that a 10 day system rotation period applied in reality in the network caused a yield loss, especially in the peak irrigation period. However, irrigation water requirements of all 25

the irrigation applications beginning on 10th July are given in Table 5.

Water deficit levels occurring in the length of alternative system rotation periods (%) 7 days 10 days 11 days 12 days

Tertiary name

Total irrigation water requirement (m3)

applications beginning on 6th of June.

tertiaries except P.7, P.9 and P.26 were met completely during the 12 day system rotation period.


Table 5. Water deficit levels occurring in the alternative rotation periods for the irrigation applications beginning on 10th of July.

Results obtained by running the model for irrigation period beginning on 18th August are given in Table 6.

Sustainable Management of Large Scale

June are shown in Figure 7.

Amount of irrigation water (m3

)

560000.00 580000.00 600000.00 620000.00 640000.00 660000.00 680000.00 700000.00 720000.00

tertiaries P.7, P.9, P.11 and P.21 of AZ I respectively (Table 4).

**to them in alternative system rotation periods** 

of view of water use effectiveness at the network level.

Irrigation Systems: A Decision Support Model for Gediz Basin, Turkey 71

**4.1 Irrigation water requirements of allocation zones and amounts of water allocated** 

Allocation of irrigation water in the research area was evaluated at the level of zones. Three different irrigation applications, started on 6 June, 10 July and 18 August, were taken into consideration for the allocation zones, which were served by different canal rotation groups. Irrigation water requirements in tertiaries, length of irrigation times for the canals and water deficit levels occurring in these areas were analyzed for irrigation programs devised for 7, 10, 11 and 12 day alternative system rotation periods. Results were evaluated from the point

The irrigation water requirement of AZ I and the amount of water allocated to this zone during the alternative system rotation periods for the irrigation applications started on 6

Amount of water allocated in alternative system rotation periods

Irrigation water requirement in AZ I

7 days 10 days 11 days 12 days

Alternative system rotation periods (Days)

Fig. 7. Irrigation water requirement of AZ I and the amount of water allocated to this zone during the alternative system rotation periods in irrigation applications started on 6 June.

As seen in Figure 7, as the length of alternative system rotation periods increased, the amount of water delivered to AZ I increased and deficit levels diminished. In the 7 day system rotation period, 95.06 hours of irrigation time were allocated to AZ I for the irrigation application started on 6 June (Table 3). In this process, a 12.78% water deficit occurred in this zone (Figure 7). In contrast, it was seen that 56.31%, 44.48%, 36.01% and 36.65% water deficits occurred in

When the length of the system rotation period was increased to 10 days, 135.80 hours of irrigation time was allocated to AZ I. In this state, the water deficit level in tertiary P.7 took the value of 38.15% (Table 4). In the 11 day system rotation period, the water deficit level decreased to 32.10% in tertiary P.7 (Table 4) for 149.38 hours of irrigation time (Table 3) and


Table 6. Water deficit levels occurring in the alternative rotation periods for the irrigation applications beginning on 18th of August.

Maximum levels of water constraints occurring in the system rotation periods of 7, 10, 11 and 12 days were 61.30%, 45.28%, 39.94% and 34.60% respectively in the P.7 tertiary (Table 6). These ratios were lower than the maximum deficiency levels which occurred in the irrigation period beginning on 10th July. However, the water deficiency level (45.28%) which occurred in the 10 day system rotation period which was applied in reality in the research area was quite high. On the other hand, irrigation water requirements of all 25 tertiaries, except P.7, P.9 and P.26 were met completely during the 12 day system rotation period.

P.3 4367.812 0 0 0 0 P.4 27970.355 0 0 0 0 P.5 52577.673 0 0 0 0 P.6 3981.409 0 0 0 0 P.7 44672.084 61.30 45.28 39.94 34.60 P.8 8976.829 0 0 0 0 P.9 32689.617 47.60 25.70 18.41 11.12 P.10 16509.593 1.81 0 0 0 P.11 24626.182 30.87 0 0 0 P.12 11958.468 0 0 0 0 P.13 22001.799 22.78 0 0 0 P.14 43465.261 21.84 0 0 0 P.15 17676.586 4.21 0 0 0 P.16 53336.634 0 0 0 0 P.17 6818.113 0 0 0 0 P.18 64475.342 0 0 0 0 P.19 75985.677 10.77 0 0 0 P.20 93733.214 0 0 0 0 P.21 27051.418 36.95 10.49 1.68 0 P.22 43428.077 0 0 0 0 P.23 40029.311 0 0 0 0 P.24 38629.018 27.84 0 0 0 P.25 8779.407 0 0 0 0 P.26 67129.658 44.33 21.03 13.27 5.50 P.27 3520.494 0 0 0 0 P.28 852.726 0 0 0 0 P.29 1277.901 33.25 5.21 0 0 P.30 18020.521 0 0 0 0 Table 6. Water deficit levels occurring in the alternative rotation periods for the irrigation

Maximum levels of water constraints occurring in the system rotation periods of 7, 10, 11 and 12 days were 61.30%, 45.28%, 39.94% and 34.60% respectively in the P.7 tertiary (Table 6). These ratios were lower than the maximum deficiency levels which occurred in the irrigation period beginning on 10th July. However, the water deficiency level (45.28%) which occurred in the 10 day system rotation period which was applied in reality in the research area was quite high. On the other hand, irrigation water requirements of all 25 tertiaries, except P.7, P.9 and P.26 were met completely during the 12 day system rotation period.

Water deficit levels occurring in the length of alternative system rotation periods (%) 7 days 10 days 11 days 12 days

Tertiary name

Total irrigation water requirement (m3)

applications beginning on 18th of August.

### **4.1 Irrigation water requirements of allocation zones and amounts of water allocated to them in alternative system rotation periods**

Allocation of irrigation water in the research area was evaluated at the level of zones. Three different irrigation applications, started on 6 June, 10 July and 18 August, were taken into consideration for the allocation zones, which were served by different canal rotation groups. Irrigation water requirements in tertiaries, length of irrigation times for the canals and water deficit levels occurring in these areas were analyzed for irrigation programs devised for 7, 10, 11 and 12 day alternative system rotation periods. Results were evaluated from the point of view of water use effectiveness at the network level.

The irrigation water requirement of AZ I and the amount of water allocated to this zone during the alternative system rotation periods for the irrigation applications started on 6 June are shown in Figure 7.

Fig. 7. Irrigation water requirement of AZ I and the amount of water allocated to this zone during the alternative system rotation periods in irrigation applications started on 6 June.

As seen in Figure 7, as the length of alternative system rotation periods increased, the amount of water delivered to AZ I increased and deficit levels diminished. In the 7 day system rotation period, 95.06 hours of irrigation time were allocated to AZ I for the irrigation application started on 6 June (Table 3). In this process, a 12.78% water deficit occurred in this zone (Figure 7). In contrast, it was seen that 56.31%, 44.48%, 36.01% and 36.65% water deficits occurred in tertiaries P.7, P.9, P.11 and P.21 of AZ I respectively (Table 4).

When the length of the system rotation period was increased to 10 days, 135.80 hours of irrigation time was allocated to AZ I. In this state, the water deficit level in tertiary P.7 took the value of 38.15% (Table 4). In the 11 day system rotation period, the water deficit level decreased to 32.10% in tertiary P.7 (Table 4) for 149.38 hours of irrigation time (Table 3) and

Sustainable Management of Large Scale

24000.00 24100.00 24200.00 24300.00 24400.00 24500.00 24600.00 24700.00 24800.00

Amount of irrigation water (m3

found to be suitable.

are shown in Figure 10.

)

Irrigation Systems: A Decision Support Model for Gediz Basin, Turkey 73

Amount of water allocated in alternative system rotation periods

Irrigation water requirement in AZ III

7 days 10 days 11 days 12 days Alternative system rotation periods (Days)

Fig. 9. Irrigation water requirement of AZ III and amount of water allocated to this zone during the alternative system rotation periods in irrigation applications started on 6 June.

In irrigation applications started on 6 June, 21.01 hours and 30.02 hours of irrigation time were allocated to AZ III in the 7 and 10 day alternative system rotation periods (Table 3). As the hydraulic features of the canals in this zone were suitable for water requirements of the crop pattern and size of the irrigated area, most of the water requirements were met in this command. In the 7 and 10 day system rotation periods, 1.78% and 0.33% water deficits occurred respectively in AZ III. No water constraint occurred in the 33.02 hours of irrigation time allocated for the 11 day system rotation period in the same zone. In addition, it was a desired condition from the point of view of irrigation programming that the 11 day system rotation period was also suitable for AZ I and AZ II, and that no water constraint occurred in AZ III for this period. As a result, an 11 day system rotation period and 149.38 hours, 81.60 hours and 33.02 hours maximum irrigation times allocated to AZ I, AZ II and AZ III respectively for the irrigation applications started on 6 June were

The irrigation water requirement of AZ I and the amount of water allocated to this zone during the alternative system rotation periods in irrigation applications started on 10 July

In irrigation applications started on 10 July, which is in the peak period in the research area, a 15.39% water deficit occurred in AZ I during the 7 day system rotation period (Figure 10). In contrast, the water deficit at tertiary level in this zone increased depending on the rising irrigation water requirements in the peak period. High levels of water constraint (64.91%, 51.49%, 38.11% and 39.69%) occurred in the commands of tertiaries P.7, P.9, P.11 and P.21

respectively during the 7 day system rotation period (Table 5).

a 2.44% water constraint occurred in AZ I (Figure 7). As a result, most of the irrigation water requirement of AZ I was met in the 11 day system rotation period. Therefore, it was decided that the 11 day rotation period was suitable for AZ I in the irrigation applications started on 6 June.

The irrigation water requirement of AZ II and the amount of water allocated to this zone during the alternative system rotation periods in irrigation applications started on 6 June are shown in Figure 8.

Fig. 8. Irrigation water requirement of AZ II and amount of water allocated to this zone during the alternative system rotation periods in irrigation applications started on 6 June.

In this period, 51.93 hours, 74.18 hours, 81.60 hours and 89.02 hours of irrigation time were allocated to AZ II in 7, 10, 11 and 12 day alternative system rotation periods respectively (Table 3). A 29.04% water constraint occurred in AZ II for the 7 day system rotation period (Figure 8). However, if water deficits are analyzed at tertiary level for this period, it is seen that 31.92% and 47.18% water deficits occur in canals P.24 and P.26 respectively (Table 4). In other words, water deficits occurring in these tertiaries were higher than the deficit level of AZ II. The reason for this is that some of the canals serve larger areas than they should.

As seen in Figure 8, water deficits in AZ II showed a diminishing trend in the 10, 11 and 12 day system rotation periods, taking the values of 11.95%, 7.86% and 4.60% respectively. Although the water constraint level was 11.95% in AZ II in the 10 day system rotation period for the irrigation applications started on 6 June, a high (38.15%) level of deficit occurred in tertiary P.7 in AZ I in the same period. This made the 11 day system rotation period suitable for AZ II also.

The irrigation water requirement of AZ III and the amount of water allocated to this zone during the alternative system rotation periods in irrigation applications started on 6 June are shown in Figure 9.

a 2.44% water constraint occurred in AZ I (Figure 7). As a result, most of the irrigation water requirement of AZ I was met in the 11 day system rotation period. Therefore, it was decided that the 11 day rotation period was suitable for AZ I in the irrigation applications started on

The irrigation water requirement of AZ II and the amount of water allocated to this zone during the alternative system rotation periods in irrigation applications started on 6 June are

Amount of water allocated in alternative system rotation periods

Irrigation water requirement in AZ II

7 days 10 days 11 days 12 days Alternative system rotation periods (Days)

Fig. 8. Irrigation water requirement of AZ II and amount of water allocated to this zone during the alternative system rotation periods in irrigation applications started on 6 June.

In this period, 51.93 hours, 74.18 hours, 81.60 hours and 89.02 hours of irrigation time were allocated to AZ II in 7, 10, 11 and 12 day alternative system rotation periods respectively (Table 3). A 29.04% water constraint occurred in AZ II for the 7 day system rotation period (Figure 8). However, if water deficits are analyzed at tertiary level for this period, it is seen that 31.92% and 47.18% water deficits occur in canals P.24 and P.26 respectively (Table 4). In other words, water deficits occurring in these tertiaries were higher than the deficit level of AZ II. The reason for this is that some of the canals serve larger areas than they should.

As seen in Figure 8, water deficits in AZ II showed a diminishing trend in the 10, 11 and 12 day system rotation periods, taking the values of 11.95%, 7.86% and 4.60% respectively. Although the water constraint level was 11.95% in AZ II in the 10 day system rotation period for the irrigation applications started on 6 June, a high (38.15%) level of deficit occurred in tertiary P.7 in AZ I in the same period. This made the 11 day system rotation period suitable for AZ II also. The irrigation water requirement of AZ III and the amount of water allocated to this zone during the alternative system rotation periods in irrigation applications started on 6 June are

6 June.

shown in Figure 8.

Amount of irrigation water (m3

shown in Figure 9.

)

0.00 20000.00 40000.00 60000.00 80000.00 100000.00 120000.00 140000.00 160000.00 180000.00

Fig. 9. Irrigation water requirement of AZ III and amount of water allocated to this zone during the alternative system rotation periods in irrigation applications started on 6 June.

In irrigation applications started on 6 June, 21.01 hours and 30.02 hours of irrigation time were allocated to AZ III in the 7 and 10 day alternative system rotation periods (Table 3). As the hydraulic features of the canals in this zone were suitable for water requirements of the crop pattern and size of the irrigated area, most of the water requirements were met in this command. In the 7 and 10 day system rotation periods, 1.78% and 0.33% water deficits occurred respectively in AZ III. No water constraint occurred in the 33.02 hours of irrigation time allocated for the 11 day system rotation period in the same zone. In addition, it was a desired condition from the point of view of irrigation programming that the 11 day system rotation period was also suitable for AZ I and AZ II, and that no water constraint occurred in AZ III for this period. As a result, an 11 day system rotation period and 149.38 hours, 81.60 hours and 33.02 hours maximum irrigation times allocated to AZ I, AZ II and AZ III respectively for the irrigation applications started on 6 June were found to be suitable.

The irrigation water requirement of AZ I and the amount of water allocated to this zone during the alternative system rotation periods in irrigation applications started on 10 July are shown in Figure 10.

In irrigation applications started on 10 July, which is in the peak period in the research area, a 15.39% water deficit occurred in AZ I during the 7 day system rotation period (Figure 10). In contrast, the water deficit at tertiary level in this zone increased depending on the rising irrigation water requirements in the peak period. High levels of water constraint (64.91%, 51.49%, 38.11% and 39.69%) occurred in the commands of tertiaries P.7, P.9, P.11 and P.21 respectively during the 7 day system rotation period (Table 5).

Sustainable Management of Large Scale

the irrigation applications started on 10 July.

are shown in Figure 11.

6.13% in this period.

are shown in Figure 12.

Amount of irrigation water (m3

)

24600.00 24700.00 24800.00 24900.00 25000.00 25100.00 25200.00 25300.00 25400.00

Irrigation Systems: A Decision Support Model for Gediz Basin, Turkey 75

however, a 45.61% water constraint continued its effect in tertiary P.7. As these irrigation applications were in the peak period, the water requirements of the crops increased, and water constraint levels in the tertiaries also rose. As the water constraint decreased to 3.58% in AZ I for the 12 day system rotation period, the deficit level in the command of tertiary P.7 also diminished to 40.78%. Thus, the 12 day system rotation period was suitable in AZ I for

The irrigation water requirement of AZ II and the amount of water allocated to this zone during the alternative system rotation periods in irrigation applications started on 10 July

In irrigation applications started on 10 July, 51.93 hours of irrigation time was allocated to AZ II for the 7 day system rotation period (Table 3), and a 30.50% water deficit occurred in this zone (Figure 11). In this period, a high (49.22%) level of water deficit, which occurred in tertiary P.26 in AZ II, caused an increment of the water constraint level for the entire zone. 14.13%, 9.25% and 6.13% water constraints occurred in the 10, 11 and 12 day alternative system rotation periods respectively in AZ II. As seen on Figure 11, as the length of the alternative system rotation periods increased, water constraint levels showed a decreasing trend in this zone. Thus, the 12 day system rotation period was found to be suitable in AZ II for irrigation applications started on 10 July, owing to the fact that this rotation period was also suitable for AZ I, and that the lowest water constraint occurred in AZ II with a ratio of

The irrigation water requirement of AZ III and the amount of water allocated to this zone during the alternative system rotation periods in irrigation applications started on 10 July

Amount of water allocated in alternative system rotation periods

Irrigation water requirement in AZ III

7 days 10 days 11 days 12 days Alternative system rotation periods (Days)

Fig. 12. Irrigation water requirement of AZ III and amount of water allocated to this zone during the alternative system rotation periods in irrigation applications started on 10 July.

Fig. 10. Irrigation water requirement of AZ I and amount of water allocated to this zone during the alternative system rotation periods in irrigation applications started on 10 July.

Fig. 11. Irrigation water requirement of AZ II and amount of water allocated to this zone during the alternative system rotation periods in irrigation applications started on 10 July.

Although a low water deficit of 5.88% occurred in AZ I in the 10 day system rotation period, a fairly high level of water constraint (50.43%) occurred in tertiary P.7 in the same zone. Also, a water deficit of 4.59% occurred in AZ I in the 11 day system rotation period;

Amount of water allocated in alternative system rotation periods

Irrigation water requirement in AZ I

7 days 10 days 11 days 12 days Alternative system rotation periods (Days)

7 days 10 days 11 days 12 days

Alternative system rotation periods (Days)

Fig. 11. Irrigation water requirement of AZ II and amount of water allocated to this zone during the alternative system rotation periods in irrigation applications started on 10 July.

Although a low water deficit of 5.88% occurred in AZ I in the 10 day system rotation period, a fairly high level of water constraint (50.43%) occurred in tertiary P.7 in the same zone. Also, a water deficit of 4.59% occurred in AZ I in the 11 day system rotation period;

Fig. 10. Irrigation water requirement of AZ I and amount of water allocated to this zone during the alternative system rotation periods in irrigation applications started on 10 July.

Amount of water allocated in alternative system rotation periods

Irrigation water requirement in AZ II

550000.00 575000.00 600000.00 625000.00 650000.00 675000.00 700000.00 725000.00 750000.00

0.00 25000.00 50000.00 75000.00 100000.00 125000.00 150000.00 175000.00 200000.00

Amount of irrigation water (m3

Amount of irrigation water (m3

)

)

however, a 45.61% water constraint continued its effect in tertiary P.7. As these irrigation applications were in the peak period, the water requirements of the crops increased, and water constraint levels in the tertiaries also rose. As the water constraint decreased to 3.58% in AZ I for the 12 day system rotation period, the deficit level in the command of tertiary P.7 also diminished to 40.78%. Thus, the 12 day system rotation period was suitable in AZ I for the irrigation applications started on 10 July.

The irrigation water requirement of AZ II and the amount of water allocated to this zone during the alternative system rotation periods in irrigation applications started on 10 July are shown in Figure 11.

In irrigation applications started on 10 July, 51.93 hours of irrigation time was allocated to AZ II for the 7 day system rotation period (Table 3), and a 30.50% water deficit occurred in this zone (Figure 11). In this period, a high (49.22%) level of water deficit, which occurred in tertiary P.26 in AZ II, caused an increment of the water constraint level for the entire zone. 14.13%, 9.25% and 6.13% water constraints occurred in the 10, 11 and 12 day alternative system rotation periods respectively in AZ II. As seen on Figure 11, as the length of the alternative system rotation periods increased, water constraint levels showed a decreasing trend in this zone. Thus, the 12 day system rotation period was found to be suitable in AZ II for irrigation applications started on 10 July, owing to the fact that this rotation period was also suitable for AZ I, and that the lowest water constraint occurred in AZ II with a ratio of 6.13% in this period.

The irrigation water requirement of AZ III and the amount of water allocated to this zone during the alternative system rotation periods in irrigation applications started on 10 July are shown in Figure 12.

Fig. 12. Irrigation water requirement of AZ III and amount of water allocated to this zone during the alternative system rotation periods in irrigation applications started on 10 July.

Sustainable Management of Large Scale

August are shown in Figure 14.

0.00 25000.00 50000.00 75000.00 100000.00 125000.00 150000.00 175000.00

> 23000.00 23100.00 23200.00 23300.00 23400.00 23500.00 23600.00 23700.00 23800.00

Amount of irrigation water (m3

)

Amount of irrigation water (m3

)

Irrigation Systems: A Decision Support Model for Gediz Basin, Turkey 77

The irrigation water requirements of AZ II and the amount of water allocated to this zone during the alternative system rotation periods for irrigation applications started on 18

Amount of water allocated in alternative system rotation periods

Irrigation water requirement in AZ II

7 days 10 days 11 days 12 days Alternative system rotation periods (Days)

7 days 10 days 11 days 12 days Alternative system rotation periods (Days)

Fig. 15. Irrigation water requirements of AZ III and amount of water allocated to this zone during the alternative system rotation periods for irrigation applications started on 18 August.

Fig. 14. Irrigation water requirements of AZ II and amount of water allocated to this zone during the alternative system rotation periods for irrigation applications started on 18 August.

Amount of water allocated in alternative system rotation periods

Irrigation water requirement in AZ III

In the 7 and 10 day system rotation periods, 1.81% and 0.41% water deficits occurred in AZ III. No constraint occurred in AZ III for the 11 and 12 day rotation periods (Figure 12). Since the water carrying capacities of the canals were adequate for the size of the irrigated area and the water requirements of the crops in this zone, most of the requirements were met in that district. Thus, 12 day system rotation period was found to be suitable for the entire district containing three of the zones for the irrigation applications started on 10 July, which was in the peak period.

The irrigation water requirement of AZ I and the amount of water allocated to this zone during the alternative system rotation periods for the irrigation applications started on 18 August are given in Figure 13.

Fig. 13. Irrigation water requirement of AZ I and amount of water allocated to this zone during the alternative system rotation periods for the irrigation applications started on 18 August.

In irrigation applications started on 18 August in the research area, 12.42%, 4.70%, 3.60% and 2.82% water deficits occurred in AZ I for the 7, 10, 11 and 12 day alternative system rotation periods respectively (Figure 13). In addition , a high level of water constraint occurred in some tertiaries in AZ I.

In the 7 day system rotation period, 95.06 hours of irrigation time were allocated to AZ I (Table 3). During this process, 61.30%, 47.60% and 36.95% water deficits occurred in tertiaries P.7, P.9 and P.21 respectively (Table 6). For the 10 day system rotation period, 135.80 hours of irrigation time was allocated to AZ I, and a 45.28% water deficit occurred in tertiary P.7. When 149.38 hours of irrigation time was allocated to this zone in the 11 day system rotation period in order to reduce the water constraint in this canal (Table 3), the deficit level diminished to 39.94% in tertiary P.7 (Table 6). Since no constraint occurred in most of the canals in AZ I, the 11 day system rotation period was found to be suitable for this zone in this period.

In the 7 and 10 day system rotation periods, 1.81% and 0.41% water deficits occurred in AZ III. No constraint occurred in AZ III for the 11 and 12 day rotation periods (Figure 12). Since the water carrying capacities of the canals were adequate for the size of the irrigated area and the water requirements of the crops in this zone, most of the requirements were met in that district. Thus, 12 day system rotation period was found to be suitable for the entire district containing three of the zones for the irrigation applications started on 10 July, which

The irrigation water requirement of AZ I and the amount of water allocated to this zone during the alternative system rotation periods for the irrigation applications started on 18

Amount of water allocated in alternative system rotation periods

Irrigation water requirement in AZ I

7 days 10 days 11 days 12 days

Alternative system rotation periods (Days)

Fig. 13. Irrigation water requirement of AZ I and amount of water allocated to this zone during the alternative system rotation periods for the irrigation applications started on 18 August.

In irrigation applications started on 18 August in the research area, 12.42%, 4.70%, 3.60% and 2.82% water deficits occurred in AZ I for the 7, 10, 11 and 12 day alternative system rotation periods respectively (Figure 13). In addition , a high level of water constraint

In the 7 day system rotation period, 95.06 hours of irrigation time were allocated to AZ I (Table 3). During this process, 61.30%, 47.60% and 36.95% water deficits occurred in tertiaries P.7, P.9 and P.21 respectively (Table 6). For the 10 day system rotation period, 135.80 hours of irrigation time was allocated to AZ I, and a 45.28% water deficit occurred in tertiary P.7. When 149.38 hours of irrigation time was allocated to this zone in the 11 day system rotation period in order to reduce the water constraint in this canal (Table 3), the deficit level diminished to 39.94% in tertiary P.7 (Table 6). Since no constraint occurred in most of the canals in AZ I, the 11 day system rotation period was found to be suitable for

was in the peak period.

August are given in Figure 13.

540000.00 560000.00 580000.00 600000.00 620000.00 640000.00 660000.00 680000.00 700000.00

occurred in some tertiaries in AZ I.

this zone in this period.

Amount of irrigation water (m3

)

The irrigation water requirements of AZ II and the amount of water allocated to this zone during the alternative system rotation periods for irrigation applications started on 18 August are shown in Figure 14.

Fig. 14. Irrigation water requirements of AZ II and amount of water allocated to this zone during the alternative system rotation periods for irrigation applications started on 18 August.

Fig. 15. Irrigation water requirements of AZ III and amount of water allocated to this zone during the alternative system rotation periods for irrigation applications started on 18 August.

Sustainable Management of Large Scale

can be determined accurately.

real time conditions.

unit amount of deficit resources.

that district in different irrigation periods.

sustainability of deficit resources.

Irrigation Systems: A Decision Support Model for Gediz Basin, Turkey 79

In addition, irrigation water requirements of the crops grown in the district must be estimated in a scientific way for different growing stages. In this way, the amount of irrigation water to be allocated from the resource to the allocation zones in different periods

Whatever the location of a plot in the network, the capacity of the canal receiving water from the system, or the water conveyance efficiency, this plot must benefit from the water resource and system equally in temporal and spatial dimensions. In this process, the optimum length of irrigation time must be determined for each allocation zone by taking into consideration such parameters as the infrastructure of the network, the hydraulic features of the canals, the water conveyance efficiency, the soil features of the district, the location and size of the plots, the plant pattern, and the irrigation water requirements of the crops. Since there are a large number of water users in the system, irrigation applications must be monitored continuously by the technical personnel of the association. All these processes should be carried out serially with the aid of computers in

Apart from this, maintenance, repair, renovation and cleaning activities in the network must be performed regularly by the association, because these processes have a direct effect on to

The most important point is that decision support systems enabling real time irrigation programming at network level should be used in order to obtain the optimum benefit per

The MONES 4.1 model enabled operation of the system at maximum capacity by dividing the research area into three different allocation zones by taking into consideration the parameters stated above. Also, the most suitable length of system rotation periods according to the model solution was determined to be 11 days for June and August, and 12 days for the peak irrigation period in July. For the irrigation applications started on 6 June and 18 August, 149.38 hours, 81.60 hours and 33.02 hours maximum irrigation times allocated to AZ I, AZ II and AZ III respectively were found to be suitable. In irrigation applications started on 6 June, 2.44% and 7.86% water deficits occurred in AZ I and AZ II respectively. No water constraint occurred in AZ III in this period. In addition, 162.96 hours, 89.02 hours and 36.02 hours maximum irrigation times were allocated to AZ I, AZ II and AZ III respectively for the irrigation applications started on 10 July in the peak period. Also, while 3.58% and 6.13% water deficits occurred in AZ I and AZ II respectively, no water constraint occurred in AZ III in this period. For the irrigation applications started on 18 August, 3.60% and 5.76% water constraints occurred in AZ I and AZ II. On the other hand, no water deficit occurred in AZ III in this period. Since the water-carrying capacities of the canals were adequate for the size of the irrigated area and the water requirements of the crops in AZ III, most of the requirements could be met in

As a result, it can be seen that the application of irrigation programming techniques to such systems has a vital importance both for optimum operation of the system and for the

the irrigation programming and allocation of water at network level.

In the irrigation applications started on 18 August, 51.93 hours and 74.18 hours of irrigation time were allocated to AZ II in the 7 and 10 day system rotation periods (Table 3), also 26.21% and 9.13% water constraints occurred in AZ II in these rotation periods respectively (Figure 14). A 44.33% water constraint occurred in tertiary P.26 in the 7 day system rotation period. This deficit level fell to 21.03% in the 10 day rotation period (Table 6). On the other hand, since the 11 day system rotation period was suitable for AZ I and nearly 71% of the tertiaries serving the entire district were in AZ I, these conditions affected the rotation period of AZ II, and 11 days were found suitable for this zone in this irrigation period.

Irrigation water requirements of AZ III and the amount of water allocated to this zone during the alternative system rotation periods for irrigation applications started on 18 August are shown in Figure 15.

In this period, 1.79% and 0.28% water deficits occurred in the 7 and 10 day system rotation periods respectively in AZ III. No water constraint occurred in the 11 day rotation period in this zone (Figure 15).

As a result, the 11 day system rotation period was found to be suitable for irrigation applications started on 18 August, and 149.38 hours, 81.60 hours and 33.02 hours maximum irrigation times were allocated to AZ I, AZ II and AZ III respectively.

### **5. Conclusions**

In this investigation, irrigation programming model MONES 4.1 (Kilic, 2010) was applied to Sector VII in the Right Bank Irrigation System of Ahmetli Regulator in the Lower Gediz Basin, Turkey. Irrigation programs were devised for different growing stages and irrigation periods of the crops in the research area for 7, 10, 11 and 12 day system rotation periods.

Considerable differences occurred between the water allocation plan applied in the research area in reality and the model results, from the points of view of irrigation programs, canal rotation system, water allocation zones and deficit levels occurring in canals with different lengths of rotation periods. During the 10 day system rotation period which was applied in reality in the research area, all the tertiaries were kept open, no canal rotation program was applied on the network, and the irrigation area was not divided into different allocation zones in reality by the irrigation association. However, maximum water carrying capacity of the secondary canal serving Sector VII was inadequate for distribution of water to all tertiaries at the same time (Kilic, 2004; Kilic & Tuylu, 2010). Because of this, the tertiary canals in the network could not be operated at their maximum capacities according to the water allocation plan in reality.

In addition, for the 7 and 10 day system rotation periods, deficit levels exceeded 45% in some canals, because the lengths of these periods were not suitable for the infrastructure of the system, the hydraulic features of the canals, and the actual production pattern. As a result, it was not possible to irrigate the whole area during the 7 and 10 day system rotation periods.

In order to operate the system at the optimum level, the research area must be divided into allocation zones by running the entire network at maximum capacity. In order to achieve this, the canal rotation groups which are most suitable for the system must be determined.

In the irrigation applications started on 18 August, 51.93 hours and 74.18 hours of irrigation time were allocated to AZ II in the 7 and 10 day system rotation periods (Table 3), also 26.21% and 9.13% water constraints occurred in AZ II in these rotation periods respectively (Figure 14). A 44.33% water constraint occurred in tertiary P.26 in the 7 day system rotation period. This deficit level fell to 21.03% in the 10 day rotation period (Table 6). On the other hand, since the 11 day system rotation period was suitable for AZ I and nearly 71% of the tertiaries serving the entire district were in AZ I, these conditions affected the rotation period of AZ II, and 11 days were found suitable for this zone in this irrigation period.

Irrigation water requirements of AZ III and the amount of water allocated to this zone during the alternative system rotation periods for irrigation applications started on 18

In this period, 1.79% and 0.28% water deficits occurred in the 7 and 10 day system rotation periods respectively in AZ III. No water constraint occurred in the 11 day rotation period in

As a result, the 11 day system rotation period was found to be suitable for irrigation applications started on 18 August, and 149.38 hours, 81.60 hours and 33.02 hours maximum

In this investigation, irrigation programming model MONES 4.1 (Kilic, 2010) was applied to Sector VII in the Right Bank Irrigation System of Ahmetli Regulator in the Lower Gediz Basin, Turkey. Irrigation programs were devised for different growing stages and irrigation periods of the crops in the research area for 7, 10, 11 and 12 day system rotation periods.

Considerable differences occurred between the water allocation plan applied in the research area in reality and the model results, from the points of view of irrigation programs, canal rotation system, water allocation zones and deficit levels occurring in canals with different lengths of rotation periods. During the 10 day system rotation period which was applied in reality in the research area, all the tertiaries were kept open, no canal rotation program was applied on the network, and the irrigation area was not divided into different allocation zones in reality by the irrigation association. However, maximum water carrying capacity of the secondary canal serving Sector VII was inadequate for distribution of water to all tertiaries at the same time (Kilic, 2004; Kilic & Tuylu, 2010). Because of this, the tertiary canals in the network could not be operated at their maximum capacities according to the

In addition, for the 7 and 10 day system rotation periods, deficit levels exceeded 45% in some canals, because the lengths of these periods were not suitable for the infrastructure of the system, the hydraulic features of the canals, and the actual production pattern. As a result, it was not possible to irrigate the whole area during the 7 and 10 day system rotation

In order to operate the system at the optimum level, the research area must be divided into allocation zones by running the entire network at maximum capacity. In order to achieve this, the canal rotation groups which are most suitable for the system must be determined.

irrigation times were allocated to AZ I, AZ II and AZ III respectively.

August are shown in Figure 15.

water allocation plan in reality.

periods.

this zone (Figure 15).

**5. Conclusions** 

In addition, irrigation water requirements of the crops grown in the district must be estimated in a scientific way for different growing stages. In this way, the amount of irrigation water to be allocated from the resource to the allocation zones in different periods can be determined accurately.

Whatever the location of a plot in the network, the capacity of the canal receiving water from the system, or the water conveyance efficiency, this plot must benefit from the water resource and system equally in temporal and spatial dimensions. In this process, the optimum length of irrigation time must be determined for each allocation zone by taking into consideration such parameters as the infrastructure of the network, the hydraulic features of the canals, the water conveyance efficiency, the soil features of the district, the location and size of the plots, the plant pattern, and the irrigation water requirements of the crops. Since there are a large number of water users in the system, irrigation applications must be monitored continuously by the technical personnel of the association. All these processes should be carried out serially with the aid of computers in real time conditions.

Apart from this, maintenance, repair, renovation and cleaning activities in the network must be performed regularly by the association, because these processes have a direct effect on to the irrigation programming and allocation of water at network level.

The most important point is that decision support systems enabling real time irrigation programming at network level should be used in order to obtain the optimum benefit per unit amount of deficit resources.

The MONES 4.1 model enabled operation of the system at maximum capacity by dividing the research area into three different allocation zones by taking into consideration the parameters stated above. Also, the most suitable length of system rotation periods according to the model solution was determined to be 11 days for June and August, and 12 days for the peak irrigation period in July. For the irrigation applications started on 6 June and 18 August, 149.38 hours, 81.60 hours and 33.02 hours maximum irrigation times allocated to AZ I, AZ II and AZ III respectively were found to be suitable. In irrigation applications started on 6 June, 2.44% and 7.86% water deficits occurred in AZ I and AZ II respectively. No water constraint occurred in AZ III in this period. In addition, 162.96 hours, 89.02 hours and 36.02 hours maximum irrigation times were allocated to AZ I, AZ II and AZ III respectively for the irrigation applications started on 10 July in the peak period. Also, while 3.58% and 6.13% water deficits occurred in AZ I and AZ II respectively, no water constraint occurred in AZ III in this period. For the irrigation applications started on 18 August, 3.60% and 5.76% water constraints occurred in AZ I and AZ II. On the other hand, no water deficit occurred in AZ III in this period. Since the water-carrying capacities of the canals were adequate for the size of the irrigated area and the water requirements of the crops in AZ III, most of the requirements could be met in that district in different irrigation periods.

As a result, it can be seen that the application of irrigation programming techniques to such systems has a vital importance both for optimum operation of the system and for the sustainability of deficit resources.

Sustainable Management of Large Scale

10.1007/s11269-010-9601-4.

10.1002/ird.602.

English abstract).

*Agric. Sys*. 94: 862-873.

Turkiye, s. 137. (In Turkish with English abstract)

Turkey. (In Turkish with English abstract)

*Large Scale Irrigation Systems*. (with Turkish interface)

961-968.

Irrigation Systems: A Decision Support Model for Gediz Basin, Turkey 81

Haie, N. & Keller, A.A. (2008). Effective Efficiency as a Tool for Sustainable Water Resources

Hsiao, T.C., Steduto, P. & Fereres, E. (2007). A Systematic and Quantitative Approach to

Hussain I., Turral H., Molden D. & Din Ahmad U.M. (2007). Measuring and Enhancing the Value of Agricultural Water in Irrigated River Basins. *Irrig. Sci*. 25: 263-282. Jalal, M.M., Haddad, O.B., Karney, W.B. & Marino, A.M. (2007). Reservoir Operation in Assigning Optimal Multi-Crop Irrigation Areas. *Agric. Water Manage*. 90: 149-159. Kilic M. & Anac, S. (2010). Multi-Objective Planning Model for Large Scale Irrigation

Kilic, M. (2004). *Determining the Optimum Water Allocation in Irrigation Networks by Nonlinear* 

Kilic, M. (2010). MONES 4.1, *A Computer Program for Management, Operation and Evaluation of* 

Kilic, M. & Ozgurel, M. (2005). Resource Leveling and Optimization of Irrigation in a

Kilic, M. & Tuylu, G.I. (2010). Determination of water conveyance loss in the Ahmetli

Kodal, S., Aküzüm, T., Çakmak, B. & Kendirli, B. (1997). Irrigation Schedules with Adequate

Laborte, A.G., Van Ittersum, M.K. & Van Den Berg, M.M. (2007). Multi-Scale Analysis of

Lermontov, A., Yokoyama, L., Lermontov, M., & Machado. M.A.S. (2011). *A Fuzzy Water* 

Lorite, I.J., Mateos, L., Orgaz, F. & Fereres, E. (2007). Assessing Deficit Irrigation Strategies at

Mousavi, H. & Ramamurthy, A.S. (2000). Optimal Design of Multi-Reservoir Systems for

Ryu, J.H., Palmer, R.N., Jeong, S., Lee, J.H. & Kim, Y.O. (2009). Sustainable Water Resources

Sagardoy, J.A.I., Bottrall, A. & Uittenbogaard, G.O. (1982). *Organization, Operation and* 

Management in a Conflict Resolution Framework. *Journal of the American Water* 

*Maintenance of Irrigation Schemes*. FAO Irrigation and Drainage Paper No. 40, p. 166.

quality-index-for-watershed-quality-analysis-and-management

the Level of an Irrigation District. *Agric. Water Manage*. 91: 51-60.

Mays, L.M. (1996). *Water Resources Handbook*. McGraw-Hill, N.Y.

*Resources Association*, 45(2): 485-499.

Water Supply. *Advances in Water Resources*, 23: 613-624.

Improve Water Use Efficiency in Agriculture. *Irrig. Sci.* 25: 209-231.

Management. *Journal of the American Water Resources Association* (JAWRA) 44(4):

Systems: Method and Application. *Water Resources Management*, DOI:

*Programming Method*. Ph.D. Thesis, Ege Üniversitesi Fen Bil Enst, Bornova, Izmir,

Tertiary Canal Irrigation Unit. *Ege Univ. J. Agric. Fac.* 42(2): 97–108, Bornova, İzmir,

Regulator Irrigation System in the Lower Gediz Basin, Turkey. *Irrigation and Drainage*, Published online in Wiley Online Library (wileyonlinelibrary.com). DOI:

and Limited Water of Some Field Crops Grown in Urfa Region. *VI. National Kültürteknik Congress*, 5-8 June, Kirazlyayla, Bursa, p. 354-362. (In Turkish with

Agricultural Development: A Modeling Approach for Ilocos Norte, Philippines.

*Quality Index for Watershed Quality Analysis and Management, Environmental Management in Practice*, Elzbieta Broniewicz (Ed.), ISBN: 978-953-307-358-3, InTech, Available from: http://www.intechopen.com/articles/show/title/a-fuzzy-water-

### **6. References**


Acs, S., Berentsen, P.B.M. & Huirne, R.B.M. (2007). Conversion to organic arable farming in

ANCID. (2003). Open Channel Seepage and Control V (1.4) – *Best Practice Guideline for* 

Bartolini, F., Bazzani, G.M., Gallerani, V., Raggi, M. & Viaggi, D. (2007). The impact of water

Brumbelow, K. & Georgakakos, A. (2007). Optimization and Assessment of Agricultural

Cai, X., McKinney, D.C. & Rosegrant, M.W. (2003). Sustainability analysis for irrigation water management in the Aral Sea region. *Agricultural Systems*, vol. 76, 1043–1066. Chambers, R. (1988). *Managing Canal Irrigation: Practical Analysis from South Asia*. Cambridge

Clemmens, A.J. (2006). Improving Irrigated Agriculture Performance Through an Understanding of the Water Delivery Process. *Irrigation and Drainage* 55: 223-234. Diaz, J.A.R., Poyato, E.C. & Luque, R.L. (2007). Model to Forecast Maximum Flows in On-

DMI (National Meteorology Works) Reports. **(**2008). *Manisa and Akhisar Meteorology Reports*.

Doorenbos, J. & Kassam, A.H. (1979). *Yield Response to Water*. FAO Irrigation and Drainage

Du, B., Ji, X., Harmel, R.D. & Hauck, L.M. (2009). Evaluation of a Watershed Model for

Evans, E.M., Lee, D.R., Boisvert, R.N., Arce, B., Steenhuis, T.S., Prano, M. & Poats, S.V. (2003).

the El Angel watershed, Carchi, Ecuador. *Agricultural Systems* vol. 77, 1–22. FAO. (1992). *Cropwat; A Computer Program for Irrigation Planning and Management*. Irrigation and Drainage Paper No. 46, Food and Agriculture Organization, Rome, Italy. Garizabal, G.I., Valenzuela, C.J. & Abrahao, R. (2009). Evolution of the efficiency and agro-

Georgiou, P.E. & Papamichail, D.M. (2008). Optimization Model of an Irrigation Reservoir

Girgin, A., Gencel, G. & Gul, S. (1999). *Gediz Havzasndaki Sulamalarn Su Yönetimi Acsndan Başar Durumlar*. Izmir Su Kongresi, Izmir- Turkey, 4-5 Haziran, s. 317-334.

*Resources Planning and Management*, 133(3): 264-274.

*Water Resources Association* (JAWRA) 45(2): 475-484.

(2001-2007). *Spanish J. of Agricultural Research*, 7(2): 465-473.

Gediz Irrigation Association Reports. (2007). Manisa, Turkey.

The Netherlands: A dynamic linear programming analysis. *Agricultural Systems*,

*Channel Seepage Identification and Measurement Australian National Committee on Irrigation and Drainage (ANCID)*, p. 95 (http:/www.ancid.org.au/pdf/seepreports

and agriculture policy scenarios on irrigated farming systems in Italy: An analysis based on farm level multi-attribute linear programming models. *Agricultural* 

Water Sharing Scenarios under Multiple Socioeconomic Objectives. *Journal of Water* 

Demand Irrigation Distribution Networks. J. *Irrigation and Drainage Engineering*

Estimating Daily Flow Using Limited Flow Measurements. *Journal of the American* 

Achieving efficiency and equity in irrigation management: an optimization model of

environmental impact of a traditional irrigation land in the middle Ebro Valley

for Water Allocation and Crop Planning Under Various Weather Conditions. *Irrig.* 

**6. References** 

94(2): 405-415.

141003/Vol1.4\_Guidelines.pdf)

University Press, Cambridge, U.K.

Ankara, Turkey. (In Turkish)

*Systems*, vol. 93, 90–114.

133(3): 222–231.

*Sci*. 26: 487-504.

Paper No. 33, p. 193.


**Part 2** 

**Climate Change Mitigation and Adaptation** 

