Random Forest Algorithm for High-Resolution Land Use Mapping in Google Earth Engine: A Case Study in Kota

Land cover is the most important indicator of anthropogenic footprint and land use is extremely useful for planning and development. Mapping land use for large areas requires multiple data tiles to download and process which poses significant difficulty in terms of time and computing power. Google Earth Engine (GEE) has provided a unique opportunity to process and classify large datasets using cloud computing technology. Modern sophisticated classification algorithms like Random Forest (RF) have the potential to accurately estimate land use classes from high-resolution Sentinel-2 multispectral data. This study utilizes RF and Sentinel-2 data in land use mapping using the GEE platform. The overall accuracy of 94% obtained in this study proves the potential of RF in high-resolution land use mapping. The results from the band variable importance suggest that the SWIR band played a crucial role in classification.