Unsupervised Learning for Land Cover Mapping of Casablanca Using Multispectral Imaging

Abstract

Precise and current land use data hold immense significance in facilitating efficient urban planning and appropriate environmental oversight. This paper proposes an approach to the unsupervised classification of Casablanca's land use using the Google Earth Engine (GEE) platform. The study relies on multispectral satellite imagery, in particular data from Landsat satellites, to extract meaningful land use categories without resorting to manual labeling. The operational process includes data collection, pre-processing, unsupervised clustering, and graphical display of results. By applying the k-means and Lvq clustering algorithms, the urban area is split into distinct groups, each representing a specific land use class. The resulting land use map provides valuable data on Casablanca's urban fabric, highlighting wooded areas, agricultural land, built infrastructure, water bodies, and barren land. This automated approach demonstrates GEE's potential as a powerful tool for analyzing land use, enabling informed, data-driven decisions on urban development and environmental monitoring. The methodology outlined can serve as a reference for similar research in other regions, helping to advance remote sensing and geospatial analysis techniques in urban and environmental studies. The effectiveness of these two algorithms is assessed in terms of overall accuracy and kappa coefficient. The k-means algorithm showed moderate accuracy, while the Lvq algorithm showed the least satisfactory results.