**3. Methodology**

**Figure 5.** CORINE 2000 land use (Level 1) in the area of interest ("*Choirokoitia- Frenaros*" waterpipe)

presented in Table 1.

160 Remote Sensing of Environment: Integrated Approaches

have been reported.

During the period 2007 to 2010, three major leakages were observed along different sections of the pipe (Figure 6). The locations of these leakages were not detected until 2 months after the leakage occurred due to the difficulty of the local authorities in identifying the problematic areas. The leakages occurred during 2007; 2008 and 2010; further details for these events are

**Figure 6.** The "*Frenaros - Choirokoitia* " waterpipe (in blue). Points 1-3 indicate the three areas were water leakages

The detection of the footpirnt of the *"Southern Conveyor Project"* was made based on interpre‐ tation techniques. The interpretation was conducted using free data from Google Earth database and using high resolution satellite images. Several histogram enhancement techni‐ ques were applied along with filters in order to improve the interpretation. As well, Principal Component Analysis (PCA) and classification techniques were also conducted.

In order to explore the capabilities of remote sensing for the detection of water leakages, two different methodologies were followed. For the *"Lakatameia"*waterpipe pilot study, ground spectroradiometric measurements were taken using a handheld spectroradiometer. A leakage event was created by filling several sections of the pipeline with water so that ground spectral signatures could be taken before and after the leakage. Spectroradiometric data were also recorded from different heights using a low altitude system. In this way, spectral signatures were able to simulate variation in spatial resolution (pixel size) before any other further application.

For the "*Frenaros - Choirokoitia* " water pipe case study, three major leakages have been recorded (see Table 1). Several Landsat 7 ETM+ medium resolution images, showing each leakage before and after the day the leakage was repaired, were used. A geometric and radiometric calibration of the images was performed, following by a multi-temporal analysis of all dataset based on either false composites or vegetation indices.

## **4. Resources**

In this section, the resources and processing used for each case study are presented. The resources are grouped into three main categories: (a) high resolution satellite data used for the *"Southern Conveyor Project"* area; (b) spectroradiometric ground data used for the "*Lakata‐ meia*" pipeline and (c) medium resolution satellite data used for the "*Choirokoitia- Fre‐ naros*" pipeline.

#### **4.1. High resolution satellite data**

IKONOS high resolution satellite images were used for the detection of the buried water pipe. The IKONOS sensor, launched in 1999, was the first high-resolution satellite imagery with a spatial resolution of less than 4m. In addition, free RGB satellite images from the Google Earth databasewereexploredandanalyzed(23-10-2003;13-06-2004;29-05-2008;30-05-2009)(Figure7).

**Figure 7.** IKONOS satellite image used for the detection of the buried water pipe (left) and free Google Earth images of the area (right).

GER 1500 spectroradiometer

used for mapping the pipeline

("*Lakatameia*")

Water losses

Underground water pipe water pipe

(a) (b)

**Figure 8.** (a): The GER 1500 spectroradiometer used for the collection of ground measurements and (b): the GNSS

Detection of Water Pipes and Leakages in Rural Water Supply Networks Using Remote Sensing Techniques

**Figure 9.** The low altitude system deployed over the leakage in the "*Lakatameia*" waterpipe.

Hyperspectral measurements recorded from the GER 1500 instrument needed to be recalcu‐ lated according to the characteristics of a specific multispectral satellite sensor. The authors modified these data to mimic Landsat 7 ETM+ satellite imagery based on Relative Spectral Response (RSR) filters since such data are freely distributed from the USGS. This data were used for the second case study (*"Frenaros - Choirokoitia"* waterpipe). RSR filters describe the instrument relative sensitivity to radiance in various parts of the electromagnetic spectrum (Wu et al. 2010). These spectral responses have a value of 0 to 1 and have no units since they are relative to the peak response (Figure 10). Bandpass filters are used in the same way in

Low altitude system (helium air balloon)

Problematic spot of the

Safety ropes

http://dx.doi.org/10.5772/39309

163

#### **4.2. Spectroradiometric data**

Spectroradiometric hyperspectral measurements were carried out using the GER 1500 field spectroradiometer (Figure 8a). The GER 1500 spectroradiometer records electromagnetic radiation between 350 nm to 1050 nm (visible and near infrared part of the spectrum) A calibrated Spectralon panel, with ≈100% reflectance, was also used simultaneously to measure the incoming solar radiation. The spectralon panel measurement was used as a reference, while the measurement over the crops as a target. Therefore, reflectance for each measurement can be calculated using the following equation (1):

$$\text{Reflectance} = \left( \text{Target Radius} / \text{Panel Radius} \right) \times \text{Calibration of the panel} \tag{1}$$

In order to avoid any errors due to changes in the prevailing atmospheric conditions (Milton et al. 2009), the measurements over the panel and the target were taken within minutes of each other. The coordinates of the measurements were mapped using a Global Navigation Satellite Systems (GNSS) (Figure 8b).

In addition, spectroradiometric measurements were taken from a low altitude system (Figure 9). The spectroradiometer was attached to the air balloon and raised over the pilot study area. Measurements were taken at several heights in the pilot study area and also in the surrounding area in order to compare their spectral signature profiles. As the airborne system was raised, the pixel size in the ground increased. Table 2 presents some characteristic heights where the pixel size corresponds to known satellite sensors.

Detection of Water Pipes and Leakages in Rural Water Supply Networks Using Remote Sensing Techniques http://dx.doi.org/10.5772/39309 163

**4.1. High resolution satellite data**

162 Remote Sensing of Environment: Integrated Approaches

of the area (right).

**4.2. Spectroradiometric data**

Systems (GNSS) (Figure 8b).

be calculated using the following equation (1):

pixel size corresponds to known satellite sensors.

IKONOS high resolution satellite images were used for the detection of the buried water pipe. The IKONOS sensor, launched in 1999, was the first high-resolution satellite imagery with a spatial resolution of less than 4m. In addition, free RGB satellite images from the Google Earth databasewereexploredandanalyzed(23-10-2003;13-06-2004;29-05-2008;30-05-2009)(Figure7).

**Figure 7.** IKONOS satellite image used for the detection of the buried water pipe (left) and free Google Earth images

Spectroradiometric hyperspectral measurements were carried out using the GER 1500 field spectroradiometer (Figure 8a). The GER 1500 spectroradiometer records electromagnetic radiation between 350 nm to 1050 nm (visible and near infrared part of the spectrum) A calibrated Spectralon panel, with ≈100% reflectance, was also used simultaneously to measure the incoming solar radiation. The spectralon panel measurement was used as a reference, while the measurement over the crops as a target. Therefore, reflectance for each measurement can

Reflectance = Target Radiance / Panel Radiance x Calibration of the panel ( ) (1)

In order to avoid any errors due to changes in the prevailing atmospheric conditions (Milton et al. 2009), the measurements over the panel and the target were taken within minutes of each other. The coordinates of the measurements were mapped using a Global Navigation Satellite

In addition, spectroradiometric measurements were taken from a low altitude system (Figure 9). The spectroradiometer was attached to the air balloon and raised over the pilot study area. Measurements were taken at several heights in the pilot study area and also in the surrounding area in order to compare their spectral signature profiles. As the airborne system was raised, the pixel size in the ground increased. Table 2 presents some characteristic heights where the

**Figure 8.** (a): The GER 1500 spectroradiometer used for the collection of ground measurements and (b): the GNSS used for mapping the pipeline

**Figure 9.** The low altitude system deployed over the leakage in the "*Lakatameia*" waterpipe.

Hyperspectral measurements recorded from the GER 1500 instrument needed to be recalcu‐ lated according to the characteristics of a specific multispectral satellite sensor. The authors modified these data to mimic Landsat 7 ETM+ satellite imagery based on Relative Spectral Response (RSR) filters since such data are freely distributed from the USGS. This data were used for the second case study (*"Frenaros - Choirokoitia"* waterpipe). RSR filters describe the instrument relative sensitivity to radiance in various parts of the electromagnetic spectrum (Wu et al. 2010). These spectral responses have a value of 0 to 1 and have no units since they are relative to the peak response (Figure 10). Bandpass filters are used in the same way in


**Table 2.** Pixel size from different heights using the low altitude system. The right column presents active satellite sensors with similar spatial resolution. Two lens with different field of view (FOV) have been be used in the GER 1500 spectroradiometer

spectroradiometers in order to transmit a certain wavelength band and block others. The reflectance from the spectroradiometer was calculated based on the wavelength of each sensor and the RSR filter as follows:

$$\text{Rband} = \Sigma \text{(Ri \* RSRi)} / \Sigma \text{RSRi} \tag{2}$$

**Figure 11.** Landsat 7 ETM+ satellite image (28/07/2008) over the "*Choirokoitia- Frenaros*" water pipe.

**Figure 10.** Relative Response filters for Bands 1-4 of Landat TM sensor (Alexakis *et al.* 2012)

Detection of Water Pipes and Leakages in Rural Water Supply Networks Using Remote Sensing Techniques

http://dx.doi.org/10.5772/39309

165

After the necessary pre-processing steps, several vegetation indices were evaluated. False colour composites were also applied in order to detect the water leakages from the entire dataset. The evaluation was made not only in the three areas of interest (leakage problem) but

Where:

Rband = reflectance at a range of wavelength (e.g. Band 1)

Ri = reflectance at a specific wavelength (e.g R 450 nm)

RSRi = Relative Response value at the specific wavelength

#### **4.3. Medium resolution satellite data**

Twelve medium resolution Landsat 7 ETM+ satellite images were used, dated before and after the local authorities fixed the leaks on the "*Frenaros-Choirokoitia"* pipeline (Figure 11; Table 3). ERDAS Imagine v. 10 software was used for the pre- and post-processing of satellite imagery. Pre-processing included geometric and atmospheric correction correction of the satellite imagery. Geometric correction of the satellite images was conducted using ground control points (GCPs), which included environmental features and ground coordinates. The Darkest Pixel (DP) atmospheric correction method was used, which is the most widely applied method of atmospheric correction that provides reasonable correction (Hadjimitsis et al., 2004; Hadjimitsis et al., 2009).

Detection of Water Pipes and Leakages in Rural Water Supply Networks Using Remote Sensing Techniques http://dx.doi.org/10.5772/39309 165

**Figure 10.** Relative Response filters for Bands 1-4 of Landat TM sensor (Alexakis *et al.* 2012)

spectroradiometers in order to transmit a certain wavelength band and block others. The reflectance from the spectroradiometer was calculated based on the wavelength of each sensor

**8o FOV (ground pixel -m)**

5 0.3 0.7 GeoEye (pan); WorldView-1

10 0.7 1.4 IKONOS (pan)

20 1.4 2.8 ALOS (pan)

50 3.5 7.0 IKONOS (multi) 75 5.2 10.5 ALOS (multi)

150 10.5 21.0 Landsat (pan) 200 14.0 28.0 IKONOS (multi)

**Table 2.** Pixel size from different heights using the low altitude system. The right column presents active satellite sensors with similar spatial resolution. Two lens with different field of view (FOV) have been be used in the GER 1500

**Satellite sensor**

Twelve medium resolution Landsat 7 ETM+ satellite images were used, dated before and after the local authorities fixed the leaks on the "*Frenaros-Choirokoitia"* pipeline (Figure 11; Table 3). ERDAS Imagine v. 10 software was used for the pre- and post-processing of satellite imagery. Pre-processing included geometric and atmospheric correction correction of the satellite imagery. Geometric correction of the satellite images was conducted using ground control points (GCPs), which included environmental features and ground coordinates. The Darkest Pixel (DP) atmospheric correction method was used, which is the most widely applied method of atmospheric correction that provides reasonable correction (Hadjimitsis et al.,

Rband =Σ(Ri \* RSRi)/ΣRSRi (2)

and the RSR filter as follows:

**Height from the ground**

Rband = reflectance at a range of wavelength (e.g. Band 1)

RSRi = Relative Response value at the specific wavelength

Ri = reflectance at a specific wavelength (e.g R 450 nm)

**4 FOV (ground pixel - m)**

164 Remote Sensing of Environment: Integrated Approaches

15 1.0 2.1

25 1.7 3.5

100 7.0 14.0

**4.3. Medium resolution satellite data**

2004; Hadjimitsis et al., 2009).

Where:

spectroradiometer

**Figure 11.** Landsat 7 ETM+ satellite image (28/07/2008) over the "*Choirokoitia- Frenaros*" water pipe.

After the necessary pre-processing steps, several vegetation indices were evaluated. False colour composites were also applied in order to detect the water leakages from the entire dataset. The evaluation was made not only in the three areas of interest (leakage problem) but also along the entire length of the water pipe. The results were mapped and statistical analysis was performed.

easier in areas with no coverage (bare soil) while in cultivated areas the interpretation was a difficult task. In addition, images taken just after rainfall or after watering crops, tend to

Detection of Water Pipes and Leakages in Rural Water Supply Networks Using Remote Sensing Techniques

http://dx.doi.org/10.5772/39309

167

Moreover, the tree pattern could reveal the footprint of the water pipe (see Figure 13). This pattern can be used for the detection of buried water pipes or can be used for monitoring

The IKONOS image used for this case study was able to maximize the visible footprint of the water pipe. Indeed, using the VNIR part of the spectrum and false colour composites (Figure 14) made possible the detection of both soil and crop marks. The IKONOS multispectral image was able to detect other parts of the water pipe network of the area, as shown in Figure 14 (right arrow). Spatial filter and PCA analysis applied to the image data (Figure 15) were able

In an attempt to evaluate if an automatic detection of such crop marks could be performed (e.g. classification), spectral profiles were examined. Spectral signatures from the image were evaluated as shown in Figure 15, which features areas of crop marks and of healthy vegetation. Scatter plots from these two areas (Figure 16) indicate that a spectral difference exists between

provide better results since soil marks could be easily spotted.

**Figure 13.** The footprint of the water pipe as a result of the tree pattern.

these areas, especially in the VNIR part of the spectrum.

further to enhance the interpretation.

possible problems resulting from tree roots.


**Table 3.** Satellite images used for this study

**Figure 12.** Google Earth satellite image used for the detection of the buried water pipe during different periods: (a): 23-10-2003; (b): 13-06-2004; (c): 29-05-2008; and (d): 30-05-2009.
