Analyzing historical land use land cover changes using machine learning approach integrated with geospatial techniques in Omo Gibe River Basin, Ethiopia

Changes in land use encompasses residential development, agricultural expansion, and exploration of terrestrial ecosystems. However, previous research often relied on crude sampling strategies for feature selection and conventional classification methods, which did not adequately convey the complexity and heterogeneity of the basin. This article focuses on building suitable sampling protocols and machine learning algorithms to detect LULC changes. The Google Earth Engine has been utilized to process and extract large datasets. The classification method was evaluated based on credible references. LULC changes were analyzed over the last 39 years: before significant human-induced changes, a pre-intervention period from 1986, dam construction along with socioeconomic changes during 2006, and 2024 post-dam construction periods to attain the difference. The Classification was performed in open-source QGIS 3.36.3 machine learning Dzetsaka Classification Plugins (DCP). The RF produced 91.7 % overall accuracy, surpassing SVM (88 %) and KNN (75 %). This indicates RF exhibits strong performance in classifying geographically intricate regions, outperforming other algorithms. Therefore, RF was adopted to classify all available datasets. The result shows significant increases in agricultural land rose from 0.86 % to 36.73 %, built-up area from 0.09 % to 6 %, and water bodies from 0.64 % to 1.32 % between 1986 and 2024. Conversely, a notable decline was attained in forest class, dropping from 51.57 % in 1986 to 26.66 % in 2024, while shrub/scrub land also shows a slight decrease from 37.54 % in 1986 to 28.85 % in 2024. The NDVI values present a declining trend in the gentle slope with the largest proportions of water, shrub, agriculture and bareland classes. This study suggests a cutting-edge algorithm for future research paths and sharp detection of different landscape features for sustainable environmental management. The outcomes of this research aims to enhance land policy, management, and public comprehension of land use dynamics in the region.