Understanding land use land cover change dynamics using machine learning algorithms in the Abelti watershed, Omo-Gibe Basin, Ethiopia

Accurate and precise land cover information is essential to subsequent applications, though it is highly sought after. The purpose of this study is to select the better land use land cover (LULC) classifier and investigate change detection. Support vector machine (SVM) and random forest (RF) algorithms were applied using Google Earth Engine (GEE) platform to categorize LULC satellite data in the Abelti watershed. SVM and RF have overall classification accuracy of 87.46% and 91.19%, respectively and thus RF classifier was selected for LULC change detection analysis. Results show that agricultural land was grown by 8.53% between 1992 and 2002, 6.44% between 2002 and 2012, and 14.94% between 2012 and 2022. Between 1992 and 2002, the settlement area grew by 69.91%, between 2002 to 2012 by 72.17%, and between 2012 and 2022, it expanded by 21.44%. Shrub land was also decreased by 38.60% between 1992 and 2022. Additionally, there was a change in bare land between 1992 and 2012 which decreased by 31.97%, then increased by 74.05% between 2012 and 2022. Finally, Agriculture, waterbody, and settlement areas showed an increasing trend of 12.57, 0.27 and 8.91%, respectively, while forest, shrubland, and bareland showed a decreasing trend of 6.21, 10.97 and 3.23%, respectively during 1992–2022. Consequently, utilizing a RF algorithm is a crucial method for classifying multispectral satellite data and in detecting LULC changes. The study results provide useful information for policymakers and planners in the implementation of sustainable land resource planning and management in the context of environmental change.