Hourly Urban Air Temperature Forecasting with Graph Machine Learning

The Urban Heat Island (UHI) effect, where urban areas exhibit higher temperatures than their rural surroundings, is a growing subject of concern due to its implications for human health, energy demand, and anthropogenic emissions. Accurate high‐resolution forecasts of the UHI intensity and, more generally, of urban air temperatures are therefore crucial for guiding mitigation and adaptation strategies, especially for real-time heat warning systems and reliable power load forecasting. We present a spatiotemporal graph-based machine-learning framework for hourly urban air-temperature forecasting that couples a Diffusion Convolutional Recurrent Neural Network (DCRNN) encoder with a multi-horizon Multi-Layer Perceptron (MLP) decoder. The model is trained and evaluated using a dense network of 113 low-cost temperature sensors in Bern, Switzerland. Spatial dependencies are learned on a directed, weighted sensor graph built from geographic proximity and environmental similarity, while temporal dynamics are modeled with gated recurrence. We forecast temperatures at each sensor up to 24 hours ahead and compare two settings: conditioning the decoder on past rural-station meteorological observations and future meteorological forecasts. At 24-hour lead time, the proposed model achieves an average RMSE of 2.99~K with past meteorology and 1.68~K with future meteorology. Relative to a per-sensor RNN baseline, it improves performance by an average of 13\% when only past meteorological data are available, but underperforms when future observations are provided, motivating further work on how to incorporate future exogenous variables. Finally, we demonstrate the practical value of the approach by combining the sensor forecasts with regression-kriging to produce hourly 50 m resolution temperature and UHI-intensity maps over Bern. Overall, the results show the promise of graph-based learning for city-scale, high-resolution temperature forecasting to support heat-risk management and urban planning.