Spatial data mining technique to evaluate forest extent changes using GIS and remote sensing

Development of new computational, visual analytical and statistical methods to process, analyse, and understand complex and massive geospatial and temporal data is of vital importance at present in the world. Therefore, Spatial Data Mining (SPD) technique is very useful tool to access environmental phenomena. Spatial data mining is the process of discovering interesting and previously unknown, but potentially useful patterns from large spatial datasets. The main purpose of the present study was to identify forest extent changes during two decades using SPD techniques. For this study, multi-temporal satellite images (Land sat 5 TM 1992 and ASTER 2006) were used. Nuwaraeliya was selected as the study area for this research. Two thematic maps were derived from following tow approaches. The first and second approaches were consisted of unsupervised and supervised classification, respectively; derived thematic maps (unsupervised and supervised) were combined with Geographical Information System (GIS) overlay technique to generate a new map. These three maps were reclassified and converted to American Standard Code for Information Interchange (ASCII) format which is suitable formatting interface for SDM modelling. In order to carry out spatial data mining, Back-propagation algorithm was used. Overall accuracy of ASTER was 96.2 whereas that of land sat TM was 94. Results revealed that the extent of forest cover was lost by 5.28% in the present study area within in the period from 1992 to 2006. The results of this study are expected to be useful for researchers, managers and policy makers for updating existing forest maps, detecting forest changes and planning.