Integrating Spatial Data Mining Technique to Identify Potential Areas for Forest Production Using GIS and Remote Sensing

Abstract

The purpose of the present study was to find potential areas for forest production using Spatial Data Mining (SDM) technique, GIS, and remote sensing. Artificial Neural Network (ANN) was the data mining method used. Current land use, average annual rainfall, soil erosion, slope, and fire hazards were selected as the criteria for the forest production. For this study, an ASTER-VNIR (Level 1B) satellite image acquired on 11 July 2006 was used. The image was classified into eight land use classes. Digital Elevation Model (DEM) was used to derive the slope map (20 m). The normalized Difference Vegetation Index (NDVI) was applied to detect areas of vegetation cover. A model was developed to evaluate areas vulnerable to soil erosion. Thematic paper maps (rainfall and soil) of the study area were screen digitized and converted to the raster format. The Alyuda NeuroIntelligence 2.2 application was used to implement the standard Back-Propagation (BP) algorithm for ANN modeling. Training stage of ANN helped to identify the optimal neural network structure. The final result was saved in spreadsheet format and converted to the GIS format for the output process, which was evaluated to produce the final map. This map was used to separate the suitable areas for the forest production into three suitability classes. The results showed that 19.65%, 28.43%, and 51.93% land areas were most suitable, suitable and less suitable, respectively, for the forest production. The SDM is more powerful than conventional cartography modeling technique for land suitability analysis.