Improving 3D Urban Thermal Models from UAV Imagery Using Emissivity Correction and Multispectral-Based Material Classification

Urban surface temperature modeling is essential for understanding the spatial variability of heat in built environments, especially as cities face rising temperatures due to climate change and the worsening urban heat island effect. Conventional approaches for drone-based temperature measurements use longwave infrared radiation as a proxy for temperature, but accurate estimates depend on knowing the emissivity of the material. When a single emissivity value (typically 0.95) is applied to all materials, temperature estimates can have consistent directional errors by type (e.g., concrete) of a magnitude of 5-10 C° according to pilot data collected. We introduce a novel methodology for enhancing the accuracy of UAV-derived 3D thermal models through surface emissivity corrections and NDVI calibration. Thermal infrared, RGB, and multispectral imagery are collected using UAVs across diverse study areas in Binghamton, NY, as well as ground-based surface temperatures using a thermocouple sensor and infrared thermometer. By incorporating NDVI values and classified surface types, this study aims to meaningfully reduce the error associated with drone-based surface temperature estimates. This work contributes to the development of improved 3D thermal models, providing urban planners with better tools with which to make informed decisions for heat mitigation in cities.