Land Use Land Cover Classification using Machine Learning, the Case of Upper Awash Basin Ethiopia.

Land use and land cover (LULC) mapping is associated with ecological changes, socio-economic issues, hydrology, climate change, and ecosystem services. It has always been of interest to researchers, governments, and international organizations. However, due to its dynamic nature, LULC maps should be updated frequently.

The Awash Basin, nestled in the Horn of Africa, serves as a critical region marked by diverse topography. Home to the Awash River, this basin plays a pivotal role in influencing both natural ecosystems and human activities across Ethiopia. In this study, we undertake a comprehensive LULC analysis of 2020 aimed at generating a 30-m spatial resolution LULC map of the upper awash basin using Landsat 8 imagery with enhanced spatial resolution and a Random Forest (RF) algorithm on the Google Earth Engine (GEE) platform. In the present study, multi-temporal data is used, including the spectral bands, some spectral indices, and different topographic factors. The accuracy assessment result shows that overall accuracy and Kappa coefficients are 0.86 and 0.97 for the years 1990 and 2020 respectively, indicating a robust classification methodology. The gradual decline in agriculture and vegetation alongside an increase in urban and barren land suggests ongoing land use transformations in the Awash Basin. These trends align with patterns of urbanization and potential shifts in agricultural practices.