**6. Monitoring urban land cover with the use of satellite remote sensing techniques as a means of flood risk assessment in Cyprus.**

#### **6.1. Introduction**

In order to improve image interpretation for water affected areas, several RGB composites were constructed, including microwave and textural bands. The optimum ones improved remarkably the final RGB composites and contributed to the delineation of the moisture af‐ fected areas, as shown in Fig. 13. Specifically, in Fig. 13a, the moisture affected areas are in‐ dicated in green tones. In Fig. 13b where only texture indicators were used the moisture affected areas are in light cyan color. On the one hand, the combination of speckle reducing Lee filter band and texture indicators in Fig. 13c, resulted in whitish color for flood prone areas. On the other hand, concerning the composite Fig. 13d, the combination of Mean, Var‐ iance and Homogeneity bands resulted in a light yellowish color for the moisture affected

(a)

tors RGB composite (R: Mean, G: Variance, B: Homogeneity)

**5.6. Conclusions**

c(b)

(c) (d)

**Figure 13.** a) RGB composite of the catchment area with the ALOS images before and after the precipitation event (R: Filtered image before precipitation, G: Filtered image after precipitation, B: Filtered image before precipitation - with green colors the areas where backscattering values were increased due to moisture effect are indicated). (b) Texture indicators RGB composite (R: Homogeneity, G: Contrast, B: Dissimilarity) (c). Combination of microwave bands and textural bands (R: Filtered image before precipitation, G: Filtered image after precipitation, B: Mean). (d) Texture indica-

In this study, ALOS PALSAR imagery data (acquired before and after a certain precipitation event) proved to be useful for evaluating their potential to detect increased land moisture

areas.

( b )

116 Remote Sensing of Environment: Integrated Approaches

This study uses an integrated approach that combines record of urban sprawl, land use and landscape metrics. Specifically, a remote sensing approach is applied to Aster satellite im‐ ages to analyze and identify patterns of urban changes within the spatial limits of Yialias watershed basin in the island of Cyprus. Moreover, there is an effort to optimize the classifi‐ cation products by combining spectral and texture data to the final.

#### **6.2. Data and methodology**

#### *6.2.1. Methodology*

Αn innovative methodology was developed for improving the classification accuracy of As‐ ter images concerning multi-temporal (2000 – 2010) record of urban land cover within the spatial limits of Yialias watershed basin in Cyprus. The phenomenon of spectral similarity of the spectral signatures of urban and marl/chalk formations, identified in the study area, stimulated the calculation of texture measurements in order to improve the traditional clas‐ sification products derived from spectral bands. Thus, with the use of ENVI 4.7 software 7 indicators of texture information were extracted for the images of 2000 and 2010. These indi‐ cators were evaluated for their separability concerning urban and marl / chalk and the opti‐ mum ones were used either individually or in combination with spectral bands in order to improve the land use / land cover (LULC) classification accuracy. The Kappa coefficient was used in order to evaluate the reliability of the classified products. In the final stage, the opti‐ mum LULC products were incorporated in Fragstats tool in order to record the changes in urban cover structures during the last decade with the use of sophisticated spatial metrics.

The major issue that this study had to deal with was the similarity of spectral signature re‐ sponse mainly between urban, marl/ chalk and soil features in the Aster images of 2000 and 2010. This problem is clearly denoted in Fig. 14. For this reason different kind of classifica‐ tion methods were used in order to optimize the final results and provide an alternative

Integrated Remote Sensing and GIS Applications for Sustainable Watershed Management: A Case Study from Cyprus

http://dx.doi.org/10.5772/39307

119

The pixel-based classification is considered to be the most classic way of classifying satellite imagery. For this reason, the first three bands of Aster image were used covering a spectral range from visible to near infrared part of spectrum. This process was accomplished in or‐ der to form a standard of comparison with the other classification products such as those of texture or combination of texture and spectral bands. After proceeding with evaluation ac‐ curacy, it was resulted that the Kappa coefficient for image acquired for 2000 was 0.684 and for 2010 was 0.695. These accuracies can be described as moderate and were ascribed to ur‐

According to Zhang & Zhu (2011), texture is defined as the spatial variation in gray value and is independent of color or luminance. Texture measures smoothness, coarseness and regularity of a region in an image (Gonzalez & Woods, 1992). Concerning satellite digital imagery texture quantifies the way two neighboring pixels relate each other within a small window centered on one of the pixels. It is generally used to describe the visual homogenei‐

way of creating efficient LULC cover maps.

**Figure 14.** Spectral response curve of typical ground objects

*6.4.1. Multispectral image classification*

ban and marl/chalk spectral conflict.

*6.4.2. Texture classification*

#### *6.2.2. Data*

For the purposes of the study, the following satellite and digital spatial data were incorporated:


The acquired ASTER images have a 10 year time interval in order the multi-temporal moni‐ toring of urban sprawl to be guaranteed. For this study, the first three spectral bands were used (VNIR and SWIR) with spatial resolution of 15 m. The exact acquisition dates of the images were: 12 May 2000 and 06 April 2010.

#### **6.3. Pre-processing techniques**

Geometric corrections were carried out using standard techniques with ground control points and a first order polynomial fit. For this purpose, topographical maps were used to track the position of ground control points in conjunction with the digital shoreline of Cy‐ prus extracted from the provided DEM. in the following, the DN values were converted to radiance values. For both images, the at-satellite radiance values were converted to at–satel‐ lite reflectance values. Finally, the darkest pixel atmospheric correction method was applied to every image (Hadjimitisis et al., 2004b). It has been found that atmospheric effects con‐ tribute significantly to the classification technique.

#### **6.4. Image classification**

In this study, the Iterative Self-Organizing Data Analysis Technique (ISODATA) method was used. The ISODATA algorithm operates as k-means clustering algorithm by merging the clusters if the separation distance in a multispectral feature is less than a value specified by the user and certain rules for splitting a certain cluster into two clusters. Accuracy assess‐ ment, which is an integral part of any image classification process, was calculated to esti‐ mate the accuracy of different methodologies of land cover classifications. An important statistic generated from the error matrix is the Kappa coefficient that is well suited for accu‐ racy assessment of LULC maps (Vliet, 2009). This statistic takes into account all the values in the matrix and produces an index that indicates the rate of improvement compared to ran‐ domly allocating pixels to different classes (Congalton & Green, 2008).

The major issue that this study had to deal with was the similarity of spectral signature re‐ sponse mainly between urban, marl/ chalk and soil features in the Aster images of 2000 and 2010. This problem is clearly denoted in Fig. 14. For this reason different kind of classifica‐ tion methods were used in order to optimize the final results and provide an alternative way of creating efficient LULC cover maps.

**Figure 14.** Spectral response curve of typical ground objects

#### *6.4.1. Multispectral image classification*

cators were evaluated for their separability concerning urban and marl / chalk and the opti‐ mum ones were used either individually or in combination with spectral bands in order to improve the land use / land cover (LULC) classification accuracy. The Kappa coefficient was used in order to evaluate the reliability of the classified products. In the final stage, the opti‐ mum LULC products were incorporated in Fragstats tool in order to record the changes in urban cover structures during the last decade with the use of sophisticated spatial metrics.

For the purposes of the study, the following satellite and digital spatial data were incorporated:

**•** A Digital Elevation Model (DEM) of 25m pixel size provided by the Department of Land and Surveys of Cyprus and created with the use of orthorectified stereopairs of airphotos

The acquired ASTER images have a 10 year time interval in order the multi-temporal moni‐ toring of urban sprawl to be guaranteed. For this study, the first three spectral bands were used (VNIR and SWIR) with spatial resolution of 15 m. The exact acquisition dates of the

Geometric corrections were carried out using standard techniques with ground control points and a first order polynomial fit. For this purpose, topographical maps were used to track the position of ground control points in conjunction with the digital shoreline of Cy‐ prus extracted from the provided DEM. in the following, the DN values were converted to radiance values. For both images, the at-satellite radiance values were converted to at–satel‐ lite reflectance values. Finally, the darkest pixel atmospheric correction method was applied to every image (Hadjimitisis et al., 2004b). It has been found that atmospheric effects con‐

In this study, the Iterative Self-Organizing Data Analysis Technique (ISODATA) method was used. The ISODATA algorithm operates as k-means clustering algorithm by merging the clusters if the separation distance in a multispectral feature is less than a value specified by the user and certain rules for splitting a certain cluster into two clusters. Accuracy assess‐ ment, which is an integral part of any image classification process, was calculated to esti‐ mate the accuracy of different methodologies of land cover classifications. An important statistic generated from the error matrix is the Kappa coefficient that is well suited for accu‐ racy assessment of LULC maps (Vliet, 2009). This statistic takes into account all the values in the matrix and produces an index that indicates the rate of improvement compared to ran‐

domly allocating pixels to different classes (Congalton & Green, 2008).

*6.2.2. Data*

**•** 2 ASTER Images

covering the study area.

118 Remote Sensing of Environment: Integrated Approaches

**6.3. Pre-processing techniques**

**6.4. Image classification**

images were: 12 May 2000 and 06 April 2010.

tribute significantly to the classification technique.

The pixel-based classification is considered to be the most classic way of classifying satellite imagery. For this reason, the first three bands of Aster image were used covering a spectral range from visible to near infrared part of spectrum. This process was accomplished in or‐ der to form a standard of comparison with the other classification products such as those of texture or combination of texture and spectral bands. After proceeding with evaluation ac‐ curacy, it was resulted that the Kappa coefficient for image acquired for 2000 was 0.684 and for 2010 was 0.695. These accuracies can be described as moderate and were ascribed to ur‐ ban and marl/chalk spectral conflict.

#### *6.4.2. Texture classification*

According to Zhang & Zhu (2011), texture is defined as the spatial variation in gray value and is independent of color or luminance. Texture measures smoothness, coarseness and regularity of a region in an image (Gonzalez & Woods, 1992). Concerning satellite digital imagery texture quantifies the way two neighboring pixels relate each other within a small window centered on one of the pixels. It is generally used to describe the visual homogenei‐ ty of images and is considered to be a common intrinsic property of all ground objects. For the description of texture histograms, gray level co-occurrence matrix (GLCM), local statis‐ tics and characteristics of the frequency spectrum are used. The GCLM mainly operates by calculating a matrix that is based on quantifying the difference between the grey levels of neighboring pixels in an image window. The main aim of this matrix is the quantification of the spatial pixel structure within this window. It was initially suggested as a mechanism for extracting texture measures (Haralick et al., 1973).

ment, homogeneity, entropy, dissimilarity, mean and variance were used for carrying out

Integrated Remote Sensing and GIS Applications for Sustainable Watershed Management: A Case Study from Cyprus

**1 Urban** 12.186 0.2841 2.082 0.129 2.779 35.591 4.662 **2 Vegetation 1** 0.5181 0.778 1.138 0.398 0.455 24.828 0 **3 Vegetation 2** 2.568 0.560 1.538 0.241 1.123 30.604 0.778 **4 Forest** 1.083 0.694 1.303 0.318 0.690 19.236 0.220 **5 Marl/Chalk** 24.808 0.198 2.049 0.137 3.882 49.939 3.759 **6 Bare Soil** 1.139 0.6605 1.269 0.344 0.755 25.799 0.316

**1 Urban** 6.856 0.422 1.948 0.151 1.875 28.594 2.606 **2 Vegetation 1** 3.319 0.533 1.528 0.254 1.284 17.77 0.906 **3 Vegetation 2** 1.867 0.688 1.398 0.299 0.801 26.178 0.948 **4 Forest** 0.612 0.723 1.223 0.328 0.562 13.248 0.199 **5 Marl/Chalk** 7.540 0.380 2.114 0.125 2.057 41.463 4.367 **6 Bare Soil** 0.337 0.831 0.892 0.485 0.337 18.45 0.160

It is clearly shown in Table 5 that marl formations and urban classes which cannot be differ‐ entiated (based on spectral features) vary in the means of contrast, homogeneity, dissimilari‐ ty and mean texture regarding the image corresponding to 2010 (Fig. 14). Concerning the texture bands of 2000 (Table 6) the greatest differences in values between marl and urban

Texture-based classification methodologies give the opportunity to end users to extend the traditional-based classifiers by incorporating the texture bands into the multispectral bands, in order to coalesce the spectral and spatial information in the final product. The ISODATA algorithm was applied to different texture products. Specifically, the algorithm was applied to the multiband texture images of 2000 and 2010 and to the PCA products (three first com‐ ponents) of 2000 and 2010 with corresponding Kappa coefficients of 0.694, 0.685, 0.710, 0.715

**Angular Second Moment**

**Angular Second Moment** **Dissimilarity Mean Variance**

http://dx.doi.org/10.5772/39307

121

**Dissimilarity Mean Variance**

the statistical texture analysis of all the typical ground objects (Tables 4, 5 and 6).

**Contrast Homogeneity Entropy**

**Table 5.** Analysis of texture features of basic objects for satellite image corresponding to 2010

**Contrast Homogeneity Entropy**

**Table 6.** Analysis of texture features of basic objects for satellite image corresponding to 2000

classes are indicated at mean and variance texture classes.

and 0.723.


#### **Table 4.** Description of the texture parameters

Initially, principal component analysis was applied to both satellite images in order to ex‐ tract the first principal component from each image which would subsequently be used for texture analysis. Thus, the first component of the two images was imported in ENVI 4.7 soft‐ ware and 7 different statistical indicators of texture such as contrast, angular second mo‐ ment, homogeneity, entropy, dissimilarity, mean and variance were used for carrying out the statistical texture analysis of all the typical ground objects (Tables 4, 5 and 6).


**Table 5.** Analysis of texture features of basic objects for satellite image corresponding to 2010

ty of images and is considered to be a common intrinsic property of all ground objects. For the description of texture histograms, gray level co-occurrence matrix (GLCM), local statis‐ tics and characteristics of the frequency spectrum are used. The GCLM mainly operates by calculating a matrix that is based on quantifying the difference between the grey levels of neighboring pixels in an image window. The main aim of this matrix is the quantification of the spatial pixel structure within this window. It was initially suggested as a mechanism for

**Texture Descriptor Equation Description**

Contrast measures the difference between the highest and lowest values of a contiguous set of pixels. Thus, low contrast image features means low

Image homogeneity is sensitive to the presence if near diagonal elements in

Calculates the disorder of an image and gives high values when an image is not

ASM measures texture uniformity. High ASM values occur when the distribution of gray levels values is constant.

Dissimilarity is similar to Contrast. However it weights increase linearly rather than weighting the diagonal

Measure of similarity in pixel values (mean pixel value) of the neighborhood resolution cells in an image block.

Variance measures homogeneity and increases when the grey level values

differ from their mean.

spatial frequencies.

texturally uniform

exponentially.

GLCM.

(i-j)2 g2 (i,j)

1+(*<sup>i</sup>* <sup>+</sup> *<sup>j</sup>*)2 g(i,j)

*g2(i, j)*log*(g(i,j))*

extracting texture measures (Haralick et al., 1973).

120 Remote Sensing of Environment: Integrated Approaches

*i*=0 *Ng*−1

*i*=0 *Ng*−1

*i=0 Ng*−*1*

*i*=0 *Ng*−1

*i*=0 *Ng*−1

*i*=0 *Ng*−1

*i*=0 *Ng*−1

Ng is the number of gray levels, entry (i, j) in the GLCM and u = ∑

∑ *j*=0 *Ng*−1

∑ *j*=0 *Ng*−<sup>1</sup> <sup>1</sup>

∑ *j=0 Ng*−*1*

∑ *j*=0 *Ng*−1

∑ *j*=0 *Ng*−1

∑ *j*=0 *Ng*−1

∑ *j*=0 *Ng*−1 g (i, j)2

g (i, j)

(i-u)2g (*i, j*)

*i*=0 *Ng*−1

Initially, principal component analysis was applied to both satellite images in order to ex‐ tract the first principal component from each image which would subsequently be used for texture analysis. Thus, the first component of the two images was imported in ENVI 4.7 soft‐ ware and 7 different statistical indicators of texture such as contrast, angular second mo‐

∑ *j*=0 *Ng*−1

g(i, j)

g (*i, j*) | *i* − *j* |

Contrast ∑

Homogeneity ∑

Entropy ∑

Angular Second Moment (ASM) ∑

Dissimilarity ∑

Mean ∑

Variance ∑

**Table 4.** Description of the texture parameters


**Table 6.** Analysis of texture features of basic objects for satellite image corresponding to 2000

It is clearly shown in Table 5 that marl formations and urban classes which cannot be differ‐ entiated (based on spectral features) vary in the means of contrast, homogeneity, dissimilari‐ ty and mean texture regarding the image corresponding to 2010 (Fig. 14). Concerning the texture bands of 2000 (Table 6) the greatest differences in values between marl and urban classes are indicated at mean and variance texture classes.

Texture-based classification methodologies give the opportunity to end users to extend the traditional-based classifiers by incorporating the texture bands into the multispectral bands, in order to coalesce the spectral and spatial information in the final product. The ISODATA algorithm was applied to different texture products. Specifically, the algorithm was applied to the multiband texture images of 2000 and 2010 and to the PCA products (three first com‐ ponents) of 2000 and 2010 with corresponding Kappa coefficients of 0.694, 0.685, 0.710, 0.715 and 0.723.

As investigated by O'Neil et al. (1988) due to correlation and overlap between landscape metrics, it is not necessary to calculate all landscape metrics. The specific metrics were se‐ lected because of their simplicity and effectiveness in depicting urban forms evolution (Al‐ berti & Waddel, 2002; Herold et al., 2002). It was found that there was an increase in built up areas during the period 2000 to 2010. The number of patches used in landscape analysis in‐ dicate the aggregation or disaggregation in the landscape. The considerable increase of the specific index during the time span 2000 - 2010 suggests urbanization in the study area char‐ acterized by dispersion. Moreover, a development of a number of isolated and fragmented built up areas occurred at the end of this period. Regarding largest patch index, the small increase between 2000 and 2010 indicates a corresponding small urban core increase. The in‐ creased urbanization rate is characterized by the appearance of new, dispersed settlements.

Integrated Remote Sensing and GIS Applications for Sustainable Watershed Management: A Case Study from Cyprus

**Metrics Description Comments**

It is an absolute measure of total edge length on a per unit area bases that facilitates comparison

http://dx.doi.org/10.5772/39307

123

Quantifies the percentage of total landscape area

Is a measure of landscape composition and calculates how much of the landscape is comprised

Measurement of the extent of subdivision or

Simple measure of patch context. It is extensively used for quantification of patch isolation

Measures the extent to which landscapes are

It reflects shape complexity across a range of spatial

among landscapes of different sizes

comprised by the largest patch

of a particular landscape.

aggregated or clumped

scales

Thus, the increase of edge density value by indicates an increase in the total length of the edge of the urban patches due to urban land use fragmentation. This finding is also en‐ hanced by the increase in weighted mean patch fractal dimension value indicating the urban

fragmentation of the patch type.

Equals the sum of the lengths of all edge segments divided by total landscape

Equals the area of the largest patch of the corresponding patch type divided by total landscape area and multiplied by 100.

Equals the sum of the areas of all patches of the corresponding patch type

Equals the number of patches of the corresponding

Equals the distance to the nearest neighboring patch of

Area weighted mean value of the fractal dimension values of all the patches

area

class

6 Contagion Describes the heterogeneity

**Table 7.** Properties of spatial metrics used in this study

the same type

of a landscape

**No Landscape**

1 Edge Density (ED)

2 Largest Patch Index

4 Number of Patches

Distance

<sup>5</sup> Euclidean Nearest Neighbor

<sup>7</sup> Area weighted mean patch fractal dimension

3 Class Area

**Figure 15.** Urban and marl/chalk features as indicated in Angular Second Moment texture indicator (left). Urban and marl/chalk features as indicated in Contrast texture indicator (right)

#### *6.4.3. Combined spectral and texture methodology*

The combined use of spectral and texture methodology function by combining spectral and texture bands (either original bands or PCA components) and creating a final integrated im‐ age. For this study, the following two combinations were accomplished and the ISODATA classifier was applied to them:


The overall accuracy of the methodology was considered as promising compared to the re‐ sults of the previous classification products derived from individual either spectral or tex‐ ture bands. Specifically, the Kappa coefficient values for the 1st category of combined classification for 2000 and 2010 was 0.702 and 0.732, respectively. In addition, the Kappa co‐ efficient values for the second category were 0.765 and 0.775 concerning 2000 and 2010 im‐ ages. These results led the research team to select these two final LULC cover maps concerning the period 2000 and 2010 for applying spatial landscape metrics.

#### **6.5. Landscape metrics**

Spatial landscape metrics are used in sustainable landscape planning and analysis of urban land use change (Botequilha et al., 2002). These metrics typically measure spatial configuration of landscapes, and can be used to enhance the understanding of relationships between spatial pat‐ terns and spatial processes (Herold et al., 2005). In this study, the FRAGSTATS tool was used in order to measure and analyze the diachronic changes of LULC regime of the study area and re‐ cord the urban sprawl phenomenon within the watershed. Specifically, seven spatial individu‐ al metrics were used for analyzing urban land cover changes and these were (Edge Density, Largest Patch Index, Class Area, Number of Patches, Area weighted mean patch fractal dimen‐ sion, Euclidean nearest neighbor distance and Contagion) (Table 7).

As investigated by O'Neil et al. (1988) due to correlation and overlap between landscape metrics, it is not necessary to calculate all landscape metrics. The specific metrics were se‐ lected because of their simplicity and effectiveness in depicting urban forms evolution (Al‐ berti & Waddel, 2002; Herold et al., 2002). It was found that there was an increase in built up areas during the period 2000 to 2010. The number of patches used in landscape analysis in‐ dicate the aggregation or disaggregation in the landscape. The considerable increase of the specific index during the time span 2000 - 2010 suggests urbanization in the study area char‐ acterized by dispersion. Moreover, a development of a number of isolated and fragmented built up areas occurred at the end of this period. Regarding largest patch index, the small increase between 2000 and 2010 indicates a corresponding small urban core increase. The in‐ creased urbanization rate is characterized by the appearance of new, dispersed settlements.


**Table 7.** Properties of spatial metrics used in this study

**Figure 15.** Urban and marl/chalk features as indicated in Angular Second Moment texture indicator (left). Urban and

The combined use of spectral and texture methodology function by combining spectral and texture bands (either original bands or PCA components) and creating a final integrated im‐ age. For this study, the following two combinations were accomplished and the ISODATA

**•** Use of all multispectral bands and the first three components after the application of PCA

The overall accuracy of the methodology was considered as promising compared to the re‐ sults of the previous classification products derived from individual either spectral or tex‐ ture bands. Specifically, the Kappa coefficient values for the 1st category of combined classification for 2000 and 2010 was 0.702 and 0.732, respectively. In addition, the Kappa co‐ efficient values for the second category were 0.765 and 0.775 concerning 2000 and 2010 im‐ ages. These results led the research team to select these two final LULC cover maps

Spatial landscape metrics are used in sustainable landscape planning and analysis of urban land use change (Botequilha et al., 2002). These metrics typically measure spatial configuration of landscapes, and can be used to enhance the understanding of relationships between spatial pat‐ terns and spatial processes (Herold et al., 2005). In this study, the FRAGSTATS tool was used in order to measure and analyze the diachronic changes of LULC regime of the study area and re‐ cord the urban sprawl phenomenon within the watershed. Specifically, seven spatial individu‐ al metrics were used for analyzing urban land cover changes and these were (Edge Density, Largest Patch Index, Class Area, Number of Patches, Area weighted mean patch fractal dimen‐

concerning the period 2000 and 2010 for applying spatial landscape metrics.

sion, Euclidean nearest neighbor distance and Contagion) (Table 7).

marl/chalk features as indicated in Contrast texture indicator (right)

*6.4.3. Combined spectral and texture methodology*

122 Remote Sensing of Environment: Integrated Approaches

**•** Use of all multispectral and texture bands

classifier was applied to them:

to texture bands.

**6.5. Landscape metrics**

Thus, the increase of edge density value by indicates an increase in the total length of the edge of the urban patches due to urban land use fragmentation. This finding is also en‐ hanced by the increase in weighted mean patch fractal dimension value indicating the urban sprawl phenomenon in the study area. Moreover, the fractal shape dimension value was al‐ ways slightly higher than 1, indicating a moderate shape complexity. In addition, the de‐ crease in Euclidean Nearest Neighbor Distance metric between 2000 and 2010 denoted a reduction in the distance between the built-up patches, suggesting coalescence (Table 8).

tential of use of texture bands in combination with multispectral imagery. Moreover, the ur‐ ban sprawl phenomenon was recorded in detail with the use of landscape metrics emphasizing to the flood inundation danger in an already flood prone watershed basin such as Yialias. The research team will continue the specific research by incorporating images of

Integrated Remote Sensing and GIS Applications for Sustainable Watershed Management: A Case Study from Cyprus

http://dx.doi.org/10.5772/39307

125

This study revealed that the integrated use of satellite remote sensing and GIS technology can contribute substantially to the sustainable management of a watershed basin. Interpreta‐ tion of multi-spectral satellite sensor data proved to be of great help in the development of updated LULC maps and record of the LULC regime and urban sprawl phenomenon in a catchment area. Moreover, a soil erosion model such as RUSLE was found to be efficiently applied at basin scale with quite modest data requirements in a Mediterranean environ‐ ment. The RUSLE model provides the end users with reliable quantitative and spatial infor‐ mation concerning soil erosion and erosion risk in general. Following, the results denoted the potential of Radar imagery in recording soil moisture regime of an inundated area as

The overall results pointed out the substantial contribution of satellite remote sensing to the

The project results reported here reports are based on findings of the SATFLOOD project (ΠΡΟΣΕΛΚΥΣΗ/ΝΕΟΣ/0609) that has been funded by the Cyprus Research Promotion Foundation. Thanks are also given to the Remote Sensing and Geo-Environment Laboratory of the Department of Civil Engineering & Geomatics at the Cyprus University of Technolo‐

1 Cyprus University of Technology, Faculty of Engineering and Technology, Department of Civil Engineering and Geomatics, Remote Sensing and Geo-Environment Lab, Cyprus

, Athos Agapiou1

and Adrianos Retalis3

,

higher spatial resolution to the classification model.

well its potential to improve classification accuracy.

gy for its continuous support (http://www.cut.ac.cy).

2 Meteorological Service of Cyprus, Cyprus

3 National Observatory of Athens, Greece

, Dimitrios D. Alexakis1

, Silas Michaelides2

sustainable management of a catchment area.

**7. Overall conclusions**

**Acknowledgements**

**Author details**

Diofantos G. Hadjimitsis1

Kyriacos Themistocleous1


**Table 8.** Landscape indices

However, it is important to mention that the landscape metrics results can be used as gener‐ al indicators and do not provide the users with absolute answers.

#### **6.6. Results**

The impacts of changes in land use patterns on hydrology due to extensive urbanization in the spatial limits of watershed is a critical issue in water resource management and water‐ shed land use planning. Land use and land cover maps of the study area for the years 2000 and 2010 were obtained using spectral bands, texture bands or combination of both of them. The major motivation for the use of alternative classification methodologies was the exis‐ tence of similar spectral signatures for urban and marl/chalk geologic formations located in the study area. These methodologies were evaluated for their accuracy and the optimum classification products were selected in order to be used to the research of urban land use regime evolution during the last decade. In both cases (2000 and 2010) the combination of three spectral bands with the first three principal components extracted from texture bands led to more accurate and reliable results. In the next stage, landscape spatial metrics were used to measure the urban sprawl phenomenon in the study area and its changes through time. Specifically, seven metrics were applied to the two final classified images. The results from the vast majority of the metrics, besides Euclidean distance measurement, denoted a steady dispersion of urban settlements within the area of watershed. Although there was not a significant total urban area increase during this period, a considerable urban sprawl phenomenon was recorded.

This study denoted that spatial measures, such as texture, can play an important role in the analysis of satellite imagery. The overall improvement of classification accuracy products derived from images of medium spatial resolution such as those of Aster highlights the po‐ tential of use of texture bands in combination with multispectral imagery. Moreover, the ur‐ ban sprawl phenomenon was recorded in detail with the use of landscape metrics emphasizing to the flood inundation danger in an already flood prone watershed basin such as Yialias. The research team will continue the specific research by incorporating images of higher spatial resolution to the classification model.
