ABSTRACT This study embarks on an evaluation of the performances of the two machine learning algorithms in the classification of land cover and to investigate LULC change detection in the Abelti Watershed, Omo-Gibe Basin, Ethiopia. The support vector machine (SVM) and random forest (RF) algorithms were applied using the Google Earth Engine (GEE) platform to categorize LANDSAT satellite imagery. The main land cover contains six LULC classes, including agriculture, barelands, forest settlement, shrubland, and water bodies. Stratified random sampling with 80% training and 20% for testing was used for training data- sets and test datasets. According to the classification results, agriculture, shrubs, forests, settlements, bare land, and water bodies were ranked first to sixth regarding the total surface area they covered in the watershed. The OA and K̂ values achieved based on the stratified random sampling technique were 87.46% and 0.83 for SVM, 91.19% and 0.88 for RF. The results indicate that the RF exhibits superior performance and is selected for land-cover detection analysis. As a result, agriculture, water bodies, 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.
Sulamo et al. (Fri,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: