Digital elevation model from Cartosat-1 stereo image data

Cartosat-1 is a remote sensing satellite, which was launched in May 2005 by the Indian Space Research Organization (ISRO). It has fore (F) and aft (A) panchromatic cameras which are capable of acquiring stereoscopic data along the orbital track, with a tilt in flight direction of +26° and -5°, respectively. The cameras are mounted on the satellite in such a way that near simultaneous imaging of the same area from two different angles is possible; this facilitates the generation of accurate three-dimensional maps. Figure 4.1a illustrates the along-track viewing geometry, which enables Cartosat-1 to acquire a stereo pair in about 54 seconds. High-quality DEMs are possible to be produced using these data. The produced data has a horizontal spatial resolution of 2.5 meters and cover a 30-km swath. A detailed description of Cartosat-1 technical features and data products is given by Cartosat-1 Data User‟s Handbook (2006) and Krishnaswamy and Kalyanaraman (2009).

4.2.1.1 Cartosat-1 data accuracy

A study of accuracy assessment for DEMs prepared by the Cartosat-1 stereo image data has been done by Evans et al. (2008), who compared the vertical accuracy amongst the National Elevation Dataset (NED), Shuttle Radar Topography Mission (SRTM) DEM, and the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) DEM. They concluded that Cartosat-1 DEMs compare quite well with the most accurate reference DEMs. In another study of the Cartosat-1 Scientific Assessment Programme (C-SAP), Titarov (2008) evaluated the Cartosat-1 Geometric potential and found that the derived DEM accuracy has a 2 m Root Mean Square Error (RMSE) for flat areas and 7 m RMSE for mountainous areas (as compared to reference DEMs). In general, Cartosat-1 has a good potential for generating DEMs with a grid spacing of about 10 m and accuracy (RMSE) of about 3 m, but it has been reported that this can be achieved by only using GCPs (Kocaman et al., 2008). The DEM generated from Cartosat-1 has been also evaluated by Kumar (2006) in Dehradum, India. The study concluded that the generated DEM from Cartosat-1 data are very useful for topographic analysis in the field of water resources, agriculture, etc. All the above mentioned studies and references refer to the potential of Cartosat-1 stereo images in producing a high quality and reliable DEM. Therefore, Cartosat-1 images were chosen among other satellite images to be used in this study, taking in consideration the accuracy and the high resolution.

4.2.1.2 Data acquisition and correction

The Cartosat-1 imagery data used in this study consist of a single stereo pair acquired on April 12, 2007 with geographic extents of 3519442.5-3549592.5 Northing and 737292.5-771587.5 Easting and spatial reference WGS_1984_UTM_ZONE_36N.

Figure 4.1b shows the acquired image with some distinctive features (i.e. dams or rivers).

To gather useful information from satellite images in a way that they can be used in mapping and GISs, these images must be prepared with the aim of removing distortion. This process is called orthorectification. Additionally, for 2.5 m ground sampling distance (GSD), the data are not accurate enough and they require a

correction which can be made by GCPs - the so called bias correction (Jacobsen et al., 2008).

Within the evaluation program of Cartosat-1 images, Kocaman et al. (2008) found that the GCP distribution seems to have an influence on the accuracy of the generated DEM and recommended the use of about 6 GCPs as minimum. Same minimum number of GCPs was also recommended by PCI Geomatics (2006).

During a field campaign in October, 2007, a GCP set including 47 points was taken using Differential Corrected GPS measurements covering the study area. These GCPs were taken from easily identifiable points on the satellite image and on grounds such as road intersections, farm boarders, and tunnels/bridges (Fig. 4.2a). The taken GCP set covers the catchment area of Wadi Kafrein in a detailed way as shown in Fig. 4.2b.

Fig. 4.1b: The acquired Cartosat-1 stereo image data for Lower Jordan Valley.

Fig. 4.1a: Along track imaging geometry of the Cartosat-1 fore- and aft-viewing cameras (Evans et. al 2008).

Fig. 4.2a: Taking a GCP measurement in Jordan Valley using Differential Corrected GPS measurements.

Fig. 4.2b: The GCP set taken to correct the acquired Cartosat-1 stereo image data.

4.2.1.3 Methodology and results

For the aim of this study, a high accuracy DEM has been extracted from Cartosat-1 satellite images with a final cell size of 5m x 5m. Data correction and DEM extraction procedures are described in detail by PCI Geomatics (2006) and are available online under: http://www.pcigeomatics.com/support/tutorials/pdf/cartosat_tutorial.pdf.

Using the prepared DEM within GIS environment, the elevation characteristics have been calculated. The maximum elevation in the study area is 1,079 m on the upper northeastern side of the study area. Elevation then decreases westward to a 139 m bsl (below sea level) minimum near the Kafrein dam reservoir, with an average elevation for the whole catchment area of 605 m. The catchment has a width of around 15 km and a length of 17 km, while the greatest variations in elevation occur in the first 5 km moving from northeast toward south west. In the southern boundaries near Na‟ur and Aldassiya cities, the slope is steeper and more than half of the drop in elevation occurs in the first 2-3 km (Fig. 4.3).

The high resolution DEM has been used as a main input for the modelling process and was used to calculate the needed hydrological charectarestics of the study area. Using the prepared DEM the catchment watershed was calculated; stream network and steepness map were prepared.

Fig. 4.3: High accuracy DEM extracted from Cartosat-1 imagery data with final cell size of 5m x 5m.

Wadi Bahhath

The accuracy of the derived DEM has been evaluated by comparing the produced DEM with ground truthing GPS measured points. Several previous elevation models are available for the Wadi Kafrein catchment with resolutions which are not suitable for detailed rainfall-runoff investigations. Figure 4.4 shows 3 different DEM for Wadi Bahhath where big variations in resolution can be noticed. The width of all maps in Fig. 4.4 is 7 km.

Fig. 4.4a: 100 m cell size b: 20m cell size c: DEM with 5m cell size