Land use land cover change modeling by integrating artificial neural network with cellular Automata-Markov chain model in Gidabo river basin, main Ethiopian rift

Abstract

Modeling land use land cover (LULC) change is crucial to understand its spatiotemporal trends to protect the land resources sustainably. The appraisal of this study was to model LULC change from 1985 to 2050 owing to the business-as-usual scenario (BAU) in Gidabo River Basin (GRB) located in the Main Ethiopian Rift Valley. Different dependent and independent spatial datasets were used viz, 1985, 2003 and 2021 Landsat imagery; topography features, proximity variables, population density and evidence likelihood. Since the future projection requires the historical land use as a baseline, historical land use trends were detected using hybrid image classification procedure in ERDAS Imagine and nine major land cover classes were identified. Multi-Layer Perceptron Neural Network and Cellular Automata-Markov Chain model built-in TerrSet software were implemented to project the 2035 and 2050 LULC. The study depicts, GRB experienced significant LULC dynamics and will also be extended for the coming several years. Agriculture land, settlement and water body showed significant gains at the expense of forest, shrub and grasslands loss. Land use changes beyond land's capability played a significant role in triggering land degradation. To minimize these adverse consequences of land use change, environmentally-friendly management measures must be implemented. The outcome of this study will be helpful in providing the opportunity to develop adequate land and water resource conservation strategy plan for the future.