Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Detection and Simulation of Urban Heat Islands Using a Fine-Tuned Geospatial Foundation Model for Microclimate Impact Prediction

Jannis Fleckenstein, David Kreismann, Tamara Rosemary Govindasamy, Thomas Brunschwiler, Etienne Vos, Mattia Rigotti

TL;DR

This work tackles the challenge of forecasting urban heat island dynamics in data-scarce cities by fine-tuning a geospatial foundation model (Granite-GFM) to predict high-resolution land surface temperatures and simulate greening interventions. The authors establish empirical ground-truth cooling patterns from green spaces, test extrapolation to future climates in an unseen city, and demonstrate greening inpainting to quantify mitigation effects. The approach yields improved predictive accuracy, demonstrates extrapolation capability to extreme heat conditions, and provides a practical simulation workflow for planning climate-resilient urban interventions across diverse hydro-climatic settings. The study highlights the potential of foundation-model-based workflows to support urban heat mitigation decisions when local data are sparse, by enabling forecasting, scenario analysis, and intervention evaluation from a unified framework.

Abstract

As urbanization and climate change progress, urban heat island effects are becoming more frequent and severe. To formulate effective mitigation plans, cities require detailed air temperature data, yet conventional machine learning models with limited data often produce inaccurate predictions, particularly in underserved areas. Geospatial foundation models trained on global unstructured data offer a promising alternative by demonstrating strong generalization and requiring only minimal fine-tuning. In this study, an empirical ground truth of urban heat patterns is established by quantifying cooling effects from green spaces and benchmarking them against model predictions to evaluate the model's accuracy. The foundation model is subsequently fine-tuned to predict land surface temperatures under future climate scenarios, and its practical value is demonstrated through a simulated inpainting that highlights its role for mitigation support. The results indicate that foundation models offer a powerful way for evaluating urban heat island mitigation strategies in data-scarce regions to support more climate-resilient cities.

Detection and Simulation of Urban Heat Islands Using a Fine-Tuned Geospatial Foundation Model for Microclimate Impact Prediction

TL;DR

This work tackles the challenge of forecasting urban heat island dynamics in data-scarce cities by fine-tuning a geospatial foundation model (Granite-GFM) to predict high-resolution land surface temperatures and simulate greening interventions. The authors establish empirical ground-truth cooling patterns from green spaces, test extrapolation to future climates in an unseen city, and demonstrate greening inpainting to quantify mitigation effects. The approach yields improved predictive accuracy, demonstrates extrapolation capability to extreme heat conditions, and provides a practical simulation workflow for planning climate-resilient urban interventions across diverse hydro-climatic settings. The study highlights the potential of foundation-model-based workflows to support urban heat mitigation decisions when local data are sparse, by enabling forecasting, scenario analysis, and intervention evaluation from a unified framework.

Abstract

As urbanization and climate change progress, urban heat island effects are becoming more frequent and severe. To formulate effective mitigation plans, cities require detailed air temperature data, yet conventional machine learning models with limited data often produce inaccurate predictions, particularly in underserved areas. Geospatial foundation models trained on global unstructured data offer a promising alternative by demonstrating strong generalization and requiring only minimal fine-tuning. In this study, an empirical ground truth of urban heat patterns is established by quantifying cooling effects from green spaces and benchmarking them against model predictions to evaluate the model's accuracy. The foundation model is subsequently fine-tuned to predict land surface temperatures under future climate scenarios, and its practical value is demonstrated through a simulated inpainting that highlights its role for mitigation support. The results indicate that foundation models offer a powerful way for evaluating urban heat island mitigation strategies in data-scarce regions to support more climate-resilient cities.
Paper Structure (14 sections, 6 figures, 2 tables)

This paper contains 14 sections, 6 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Training procedure and geospatial foundation model
  4. Study regions and data
  5. Experiments and results
  6. Ground truth internal and spillover cooling patterns
  7. Model performance evaluation and future climate forecasting
  8. Urban greening simulation through inpainting
  9. Conclusion
  10. Appendix
  11. Overview of selected study regions
  12. Evaluation of urban green area cooling effects using a geospatial foundation model
  13. Analysis of cooling gradients for urban green areas using a geospatial foundation model
  14. Comparison of model variants for extrapolation capability

Figures (6)

  • Figure 1: Baseline UHI characterization for Prague based on gradient and source–sink experiments.
  • Figure 2: Comparison of ground truth cooling phenomena with model predictions.
  • Figure 3: Projected UHI extent under RCP 2.6, 4.5, and 8.5 for 2030, 2050, and 2100 in Brașov.
  • Figure 4: Comparison of inpainting results against ground truth data in Prague.
  • Figure 5: Map of the twelve European study cities (red dots), selected for their diverse urban characteristics, hydro-climatic settings, and suitability within the HLS tiling system, which were further analyzed for their internal and spillover cooling effects. Brașov (blue dot) was chosen as an extrapolation site to test model generalization.
  • ...and 1 more figures