High-Resolution Thermal Mapping and Simulation Scenarios for Land Cover Intervention Planning

The UrbanSquare service within Destination Earth aims to deliver an operational digital twin of cities, enabling urban planners to explore and assess environmental processes and intervention scenarios. One key component of this digital twin is the representation of the urban heat island (UHI) effect. UrbanSquare relies on land surface temperature (LST) observations derived from thermal satellite imagery, primarily Landsat data resampled and distributed at 30 m spatial resolution.

However, actionable urban digital twins require both finer spatial detail and the ability to simulate “what-if” scenarios driven by land-cover change. In particular, UHI mitigation planning calls for high-resolution thermal information (well below 30 m) and dynamic coupling between land-cover configurations and surface temperature responses.

We present a three-stage framework for generating scenario-ready LST maps at 5-meter resolution.

In Stage 1, a Random Forest model upscales Landsat LST from 30 m to 5 m using Sentinel-2 spectral bands and indices (B2, B3, B4, B8, B11, B12, NDVI, albedo), solar geometry variables, and ERA5 meteorological predictors. Sentinel-2 data are harmonized to 30 m for model training, then applied at super-resolution (5 m) for inference. Cross-validation assesses predictive performance in the absence of in situ measurements.

Stage 2 applies pixel-wise linear regression between the super-resolved LST time series and synchronous ERA5 air temperature across a summer period. This normalization removes temporal and meteorological variability and enables LST generation for user-defined air-temperature scenarios, ensuring consistent thermal comparisons.

Stage 3 constructs a lookup table of thermal signatures for each land cover class. When users modify land cover, pixels are reassigned the corresponding thermal signature. A diffusion process accounts for lateral heat dispersion, producing delta-temperature maps, uncertainty layers, and decomposed contributions from different factors (vegetation, albedo increase, etc.).

Soon to be integrated into the UrbanSquare digital twin, this framework enables exploration, comparison, and quantification of UHI mitigation strategies, supporting evidence-based urban planning and climate adaptation decisions.