Predicting Land Use and Land Cover Changes in the Chindwin River Watershed of Myanmar Using Multilayer Perceptron-Artificial Neural Networks

Abstract

This study investigates the potential anthropogenic land use activities in the 114,000-km2 Chindwin River Watershed (CRW) in northwestern Myanmar, a biodiversity hotspot. This research evaluates current and future land use scenarios, particularly focusing on areas that provide ecosystem services for local communities and those essential for biodiversity conservation. Remote sensing and geographical information systems were employed to evaluate land use changes in the CRW. We used a supervised classification approach with a random tree to generate land use and land cover (LULC) classifications. We calculated the percentage of change in LULC from 2010 to 2020 and projected future LULC change scenarios for approximately 2030 and 2050. The accuracy of the LULC maps was validated using Cohen’s Kappa statistics. The multilayer perceptron artificial neural network (MLP-ANN) algorithm was utilized to predict future LULC. Our study found that human settlements, wetlands, and bare land areas have increased while forest land has declined. The area covered by human settlements (0.36% of the total in 2000) is projected to increase from 264 km2 in 2000 to 424 km2 by 2050. The study also revealed that forest land has connections to other land categories, indicating a transformation of forest land into other types. The predicted future land use until 2050 reflects the potential impacts of urbanization, population growth, and infrastructure development in the CRW.