Working with Geospatial features, Sentinel-2 MSI and Machine Learning for predicting tree canopy level Landsat-8 derived LST at 30m

The study utilizes multi-spectral imagery from Landsat-8 and Sentinel-2, along with high-resolution land cover, canopy height, and vector data, to estimate Land Surface Temperature (LST) and predict extreme heat events in Delhi. Data from the summer of 2022 is analyzed, focusing on March to May for LST and NDVI estimation. The methodology includes calculating LST through a series of steps involving spectral radiance, brightness temperature, NDVI, and emissivity. Preprocessing steps prepare the data for gridbased statistical analysis, followed by machine learning techniques to predict LST and heat events. Results indicate that ensemble methods, particularly Random Forest and boosting algorithms, provide the most accurate predictions for both LST and extreme heat events.