Quantifying forest land-use changes using remote-sensing and CA-ANN model of Madhupur Sal Forests, Bangladesh

Abstract

The conversion of forest cover due to anthropogenic activities is of great concern in the Madhupur Sal Forest in Bangladesh. This study explored the land use changes in the Sal Forest area from 1991 to 2020, with the prediction of 2030 and 2040. This study examined and analyzed the changes in five land use classes viz., waterbodies, settlement, Sal Forest, other vegetation, and bare land, and predict those classes using Cellular Automated Artificial Neural Network (CA-ANN) model. The Sankey diagram was employed to represent the change percentage of Land Use and Land Cover (LULC). The LULC for 1991, 2000, 2010, and 2020 derived from Landsat TM and Landsat OLI images, were used to predict the periods of 2030 and 2040. During the last 30 years, the Sal Forest area decreased by 23.35%, whereas the settlement and bare land area increased by 107.19% and 160.89%. The greatest loss of the Sal Forest was observed from 1991 to 2000 by 46.20%. At the same period of time the settlements were increased by 92.68% indicating the encroachment of settlement in the Sal Forest area. The Sankey diagram revealed a major conversion was found between other vegetation and the Sal Forest area. There was a vis-à-vis between other vegetation and the Sal Forest area from 1991 to 2000 and from 2000 to 2010. Interestingly, there was no conversation of the Sal Forest area to other land use from 2010 to 2020, and the prediction showed that the Sal Forest area will be increased by 52.02% in 2040. The preservation and increment of the Sal Forest area suggested strong governmental policy implementation to preserve the forest.