RawImagesAndOutput
Description
This dataset contains high-resolution (10 m) spatial layers developed to map and assess neighborhood-scale urban thermal risks across Dhaka City, Bangladesh. Developed within the cloud-based Google Earth Engine (GEE) platform, the dataset bridges the spatial limitations of conventional thermal sensors by downscaling Landsat 8 Land Surface Temperature (LST) data utilizing 10-meter Sentinel-2 multi-spectral indices (NDVI, NDBI, and MNDWI). The data utilizes an ensemble-based Random Forest regression model integrated with a bicubic residual correction approach to preserve localized sub-pixel thermal heterogeneity and ensure strict radiometric consistency. Validated with a root mean square error (RMSE) of 0.96°C and a high coefficient of determination ($R^2 = 0.83$), the repository provides the original 30 m LST, the downscaled 10 m LST, the source biophysical indices, and a classified Urban Thermal Zoning (UTZ) map. This dataset serves as an open-access resource for micro-scale urban climate modeling, heat island mitigation, and sustainable urban planning in dense, data-constrained environments.
Files
Institutions
- Rajshahi University of Engineering and TechnologyRajshahi Division, Rajshahi