3.1. Data and methods The procedure followed in generation of the suitability map. The data sources were social survey, field measurement, Upazila administrative map, topographic maps of 1:10,000 scale and ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer) satellite image for land evaluation and suitability classification. Soil and water samples were collected from 14 locations under four strata of land use patterns (agriculture, aquaculture, grassland and mangrove forest) in rainy and dry seasons to determine soil pH, texture, nitriteN, phosphate-P, organic matter content, organic carbon content, water pH, temperature, dissolved oxygen, nitrite-N, phosphate-P, total suspended solid and total dissolved solid following standard methods (APHA, 1976). Adequate surface water of desirable quality, soil types to hold water, skilled labour, fry sources, transportation facilities, availability of electricity, and marketing facilities were considered as the important criteria for prawn aquaculture development. Land use pattern, roads, electricity supply line and market location were taken from 1:10,000 scale topographic maps and 1:50,000 scale administrative map and then updated with ASTER satellite image, intensive field survey and participatory interviews with local community. All these information were used to identify suitable areas for prawn aquaculture development. The GIS software used in this study was ArcView for windows (version 3.2) developed by Environmental Systems Research Institute Inc., USA. The spatial extension module was used for surface interpolation in ArcView. The values of the analyzed and collected type of data were expanded to the sites where no samples were available using interpolation methods. The interpolation gives us values in such points where we have no measurements. The goodness of interpolation can be characterized by the discrepancy of the interpolated value from the true value. Comparison of results among Spline, IDW and TIN interpolation techniques showed that Spline interpolation was best in the area due to little change in physiography (Priyakant et al., 2000) over the horizontal distance of the study area which was used as input for further analysis. On the other hand, IDW and TIN produced different results though IDW usually used for regional while TINs used for detailed and large-scale applications. All maps and images were transformed into Universal Transverse Mercator (UTM) projection. Meter was defined for the software system as the unit of the scale and the unit of the map (Biagi et al., 2002; Hossain et al., 2007, 2008). Remote sensing image analysis used ENVI (version 3.4) developed by Environmental Systems Research Institute Inc., USA. Topographic maps at 1:10,000 scale published by the Bangladesh Inland Water Transport Authority (BIWTA) were used for geometric correction. Geometric correction was performed with bilinear transformation (Research Systems Inc., 2000a) and the root mean square error was (0.46) controlled within one pixel (15 m) for 18 ground control points (GCPs).
3.2. Classification procedure Isodata unsupervised classification (use information from the image itself to identify spectral clusters, which are interpreted as classes) was performed considering minimum and maximum classes of 5–10, 10–15 and 15–20, where the 10–15 classes turned out to be useful. Supervised classification were carried out on the basis of Region of Interest (ROIs), where the ground truth or so-called training areas (collected during field investigation) were regions of terrain with known properties or characteristics (Research Systems Inc., 2000b). Maximum likelihood classification strategy was applied and found to be most useful for discriminating the category of interest. Water and soil quality parameters, road network, markets, fry sources and electricity have been weighted and scored in terms of significance for prawn farming. After image processing, reference points were chosen for ground verification. All the reference points were surveyed for the ground truthing of Companigonj Upazila and compared with the preliminary map to real position. ArcView (The Environmental systems Research Institute Inc., USA) and MS Excel (The Microsoft Corporation, USA) software were used