URBAN HEAT ISLAND MAPPING USING THERMAL SENSING

The problem of surface urban heat islands (SUHIs), which has been exacerbated due to rapid urbanization, is linked to the rise of land surface temperatures and poses threats to the environmental sustainability, human well-being, and the overall urban life quality. In this study, an automatic approach for generating SUHI maps is introduced, combining Landsat 8 data from the Thermal Infrared Sensor (TIRS) with deep learning methods. It has been developed to be applicable in many urban areas and is being tested on a selected set of cities based on imagery captured in a specific observation period ranging from January to April. The approach involves radiometric calibration, atmospheric correction, land surface temperature (LST) retrieval, and calculation of spectral indices, which include NDVI (Normalized Difference Vegetation Index) and NDBI (Normalized Difference Built-up Index). All of the listed features will be used as inputs to the Convolutional Neural Network (CNN) which is trained to perform classification of urban areas into low-, medium-, and high-heat intensity zones. Contrary to the traditional SUHI research approaches, which are mostly limited to threshold-based LST classification, the described approach combines thermal sensing with the CNN-based classification process.The framework presented herein highlights the feasibility of integrating thermal remote sensing and deep learning for scaling up the monitoring of urban heat using these two technologies as building blocks for further improvements.