Assessing Machine Learning Algorithms for Land Use and Land Cover Classification in Morocco Using Google Earth Engine

Google Earth Engine constitutes a cloud-based geospatial data processing platform. It grants free access to vast volumes of satellite data along with unlimited computational power, enabling the monitoring, visualization, and analysis of environmental features on a petabyte scale. The platform's capacity to accommodate various land use and land cover (LULC) classification approaches, utilizing both pixel-based and object-oriented methods, has been facilitated by providing an array of machine learning algorithms. Earth observation data has emerged as a valuable resource, offering temporally and spatially consistent quantitative information compared to traditional ground surveys. It presents numerous opportunities for urban mapping, monitoring, and a wide array of physical, climatic, and socio-economic data to support urban planning and decision-making. In this study, Landsat 8 satellite data was harnessed for supervised classification. Three advanced machine learning techniques—Support Vector Machine (SVM), Random Forest (RF), and Minimum Distance (MD)—were employed to categorize areas within Morocco, encompassing water bodies, built-up regions, cultivated land, sandy areas, barren zones, and forests. The classification outcomes are presented using a set of accuracy indicators, including Overall Accuracy (OA) and the Kappa coefficient.