Urban Vegetation Cover Prediction Using Sentinel-2 NDVI and Random Forest: A Brief Narrative Review

Abstract

A predictive model of urban vegetation cover is developed by integrating remote sensing technology, cloud computing, and machine learning algorithms. The study used the Normalized Difference Vegetation Index (NDVI), calculated from Sentinel-2 satellite imagery and analyzed in Google Earth Engine (GEE), to monitor vegetation conditions at a wide spatial scale. The research approach uses quantitative methods, including spatial analysis based on satellite imagery and predictive modeling with the Random Forest algorithm. The research process includes acquiring Sentinel-2 Level-2A images, pre-processing them with cloud masking and atmospheric correction, calculating NDVI values, and developing vegetation prediction models using machine learning methods. The results showed that the Random Forest model predicted vegetation cover with high accuracy, as indicated by a Coefficient of Determination (R²) of 0.85 and a Root Mean Square Error (RMSE) of 0.045. The resulting vegetation distribution map shows significant variations in vegetation density between natural vegetation areas, agricultural land, and built-up areas. The findings of this study show that integrating NDVI from Sentinel-2, Google Earth Engine, and the Random Forest algorithm is an effective approach for monitoring and predicting urban vegetation cover. The results of this study make a methodological contribution to the development of remote sensing-based geospatial analysis and provide a scientific basis for sustainable urban planning and green open space management in urban areas.