A machine learning approach for the efficient estimation of ground-level air temperature in urban areas

Abstract

The increasingly populated cities of the 21st Century face the challenge of being sustainable and resilient spaces for their inhabitants. However, climate change, among other problems, makes these objectives difficult to achieve. The Urban Heat Island phenomenon that occurs in cities, increasing their thermal stress, is one of the stumbling blocks to achieve a more sustainable city. The ability to estimate temperatures with a high degree of accuracy allows for the identification of the highest priority areas in cities where urban improvements need to be made to reduce thermal discomfort. In this work we posit that image-to-image deep neural networks (DNNs) can effectively correlate spatial and meteorological variables of an urban area with street-level air temperature. To this end, we introduce a novel DNN-based model leveraging a U-Net architecture to tackle this modeling task. We evaluate the proposed model through experiments in a use case focused on the city of Bilbao, Spain. Our method achieves regression performance metrics comparable to those of the numerical model it was trained against, with mean absolute error values below 2°C and a Pearson correlation close to 1. Additionally, it demonstrates strong regression performance against true temperature values recorded by on-site weather stations, enhancing the precision of estimates produced by numerical models. These results confirm that DNNs offer a fast and computationally efficient alternative for the data-driven estimation of ground-level air temperature.