**3. AOT retrieval using the image-based integrated method: the example of Limassol, Cyprus**

The example described in this chapter includes AOT and surface reflectance field measurements in Limassol Cyprus during Landsat-5 TM and Landsat-7 ETM+ satellite overpass. Only 11 Landsat satellite images with low cloud cover are used. These images were chosen due to their availability and clarity. The Landsat-5 TM images used are for 24/6/2010, 10/7/2010, 27/8/2010, and 28/9/2010, and the Landsat-7 ETM+ images used are for 13/4/2010, 29/4/2010, 31/5/2010, 16/6/2010, 7/11/2010, 9/12/2010, and 2/5/2011. During each satellite overpass, AOT measurements were taken using the Microtops II sunphotometer at five different locations in Limassol and the stationery Cimel sunphotometer, which is part of the AERONET network, located in the center of the study area. These AOT values are later compared to the AOT values derived from the algorithm using the satellite images. At the same time, the SVC HR-1024 spectroradiometer was used to measure the reflectance value of a large nonvariant target dark asphalt surface located near the center of the satellite image.

Geometric and radiometric correction is performed on the 11 Landsat-5 and Landsat-7 satellite images, as well as atmospheric correction using the modified DP method. The image-based integrated method is applied to the Landsat-5 and Landsat-7 images taken over specific days of 2010–2011. Eqs. (1–9) are used to retrieve AOT from bands 1–4 of the Landsat-5 and Landsat-7 satellite images, following the steps outlined. As the *in-situ* AOT measurements are at the 500 nm wavelength, Landsat band 1 is used to derive AOT levels in this method.

In this example, the model is run in the ERDAS Imagine and MATLAB software to generate an image that is consists of AOT values. The model is then applied to every surface using the Landsat band 1 satellite image, since the AOT required are in the 500 nm wavelength. The resulting image features the AOT values for every pixel in the image, as featured in **Figure 2**. Areas with no data, due to cloud presence or negative AOT values, appear in white. All AOT values can be exported to ASCII files to create a georeferenced AOT dataset, which is then imported into the ArcGIS software.

Using the ArcGIS software, each pixel is converted to a point and each point is associated with the calculated AOT value. The white area of the image contains values of "no data" (**Figure 3**, left). To create a GIS thematic map showing the AOT distribution in the urban area of Limassol, the interpolation method is used in order to estimate the values from the "no data" regions. The ordinary Kriging interpolation tool is applied, where the unknown values of the "no data" area are interpolated using the weighted average of neighboring samples. **Figure 3** (right) shows the results of the Kriging interpolation for the Limassol example, where the dots indicate the points with AOT values, while the colored sections are the resulting AOT values determined from Kriging interpolation.

**Figure 2.** AOT values derived from algorithm.

**Figure 3.** (Left) Points with AOT dataset. (Right) Interpolation with Kriging method.

Once the interpolation is completed, an AOT thematic map is created, which is classified according to AOT values. The AOT values are displayed in different colors according to the AOT concentration and a legend of the AOT values is created. This facilitates the ability to visualize the concentration and distribution of AOT values over an urban area. **Figure 4** features a map of Limassol following Kriging interpolation, which provides synoptic views of the AOT distribution. As is evident, high AOT levels are present throughout the city, especially in the industrial estates and on busy streets, with lower AOT levels present in parks, stadiums, and outside the city.

The Image-Based Integrated Method for Determining and Mapping Aerosol Optical Thickness http://dx.doi.org/10.5772/65279 149

**Figure 4.** Thematic map of Limassol (09/12/2010) after Kriging interpolation, with AOT values.

In the example, the thematic map is overlaid with GIS vector data from the Lands and Surveys Department for Limassol to facilitate the identification of sources of AOT values within the area of interest (**Figure 5**). Each polygon in the thematic map is connected to a GIS database, where data such as plot number and area information can be identified. In this way, the GIS database, combined with the AOT data, can create thematic maps to illustrate trends in different areas.

**Figure 5.** Vector overlay.

**Figure 2.** AOT values derived from algorithm.

148 Aerosols - Science and Case Studies

and outside the city.

**Figure 3.** (Left) Points with AOT dataset. (Right) Interpolation with Kriging method.

Once the interpolation is completed, an AOT thematic map is created, which is classified according to AOT values. The AOT values are displayed in different colors according to the AOT concentration and a legend of the AOT values is created. This facilitates the ability to visualize the concentration and distribution of AOT values over an urban area. **Figure 4** features a map of Limassol following Kriging interpolation, which provides synoptic views of the AOT distribution. As is evident, high AOT levels are present throughout the city, especially in the industrial estates and on busy streets, with lower AOT levels present in parks, stadiums,

In the example, the image-based integrated method is used to produce GIS maps indicating the AOT distribution over Limassol. In order to visualize high and low AOT levels, thematic maps are generated using colors ranging from blue to green to yellow to red for the specified AOT range. A separate thematic map is created for each of the 11 Landsat-5 and Landsat-7 satellite images. **Figure 6** features the GIS thematic map using Landsat-7 satellite imagery from 24/6/2010.

**Figure 6.** GIS thematic map, 24 June 2010, with vector overlay.

The AOT values derived from the algorithm are compared with the AOT values measured with the Microtops and Cimel sun photometers. **Figure 7** shows all the locations of the *in-situ* AOT measurements using sun photometers. Locations 1–5 are measured with the Microtops sunphotometer, while location 6 is measured with the Cimel sunphotometer. In **Figure 7**, location 7 refers to the Department of Labor Air Quality Monitoring Site, which was used only for reference. An accuracy assessment is done to compare the AOT from the GIS map with the *in-situ* AOT as measured from the Microtops and Cimel sunphotometers during satellite overpass. A linear regression is conducted of the AOT levels measured with the sunphotometers and the AOT values derived from the algorithm, with a coefficient of determination *R*2 of 0.977.

**Table 1** compares the AOT measurements taken on site using a sunphotometer against the AOT values derived from the GIS maps using the proposed algorithm by location and date, as well as the correlation coefficient (*r*) between the on-site AOT measurements and algorithmderived AOT values from GIS, for each location. In order to determine the accuracy of the GIS model, the root mean square deviation (RMSD) is also calculated between the on-site AOT measurements and algorithm-derived AOT values from GIS for each location. The results of the RMSD, correlation coefficients, and the coefficient of determination verify that the AOT derived from the GIS map correlate strongly with the on-site AOT values as measured by sunphotometers.

**Figure 7.** Location of *in-situ* AOT measurements.

AOT range. A separate thematic map is created for each of the 11 Landsat-5 and Landsat-7 satellite images. **Figure 6** features the GIS thematic map using Landsat-7 satellite imagery from

The AOT values derived from the algorithm are compared with the AOT values measured with the Microtops and Cimel sun photometers. **Figure 7** shows all the locations of the *in-situ* AOT measurements using sun photometers. Locations 1–5 are measured with the Microtops sunphotometer, while location 6 is measured with the Cimel sunphotometer. In **Figure 7**, location 7 refers to the Department of Labor Air Quality Monitoring Site, which was used only for reference. An accuracy assessment is done to compare the AOT from the GIS map with the *in-situ* AOT as measured from the Microtops and Cimel sunphotometers during satellite overpass. A linear regression is conducted of the AOT levels measured with the sunphotometers and the AOT values derived from the algorithm, with a coefficient of determination *R*2

**Table 1** compares the AOT measurements taken on site using a sunphotometer against the AOT values derived from the GIS maps using the proposed algorithm by location and date, as well as the correlation coefficient (*r*) between the on-site AOT measurements and algorithmderived AOT values from GIS, for each location. In order to determine the accuracy of the GIS model, the root mean square deviation (RMSD) is also calculated between the on-site AOT measurements and algorithm-derived AOT values from GIS for each location. The results of the RMSD, correlation coefficients, and the coefficient of determination verify that the AOT

24/6/2010.

150 Aerosols - Science and Case Studies

of 0.977.

**Figure 6.** GIS thematic map, 24 June 2010, with vector overlay.



**Table 1.** Comparison of on-site AOT and AOT from GIS algorithm (4/2010–5/2011).
