Predicting urban air temperature using UVA Infrared thermography and an artificial neural network

This study introduces a new approach to predicting ambient temperature profiles in urban environments, focusing on spatial thermal characteristics. Traditional methods for predicting microclimates in urban areas typically rely on either field measurements or simulation models, both of which have notable limitations. Field measurements often involve using mobile weather stations installed on carts, bicycles, or vehicles. While this method is widely adopted, it has drawbacks such as high costs for experimental equipment, significant labor demands, time consumption, low resolution, and potential reliability issues compared to fixed measurements. Similarly, simulation models, although validated and widely used, require specialized experience and knowledge, are limited in scale, and involve high computational costs. These challenges highlight the need for a more efficient and accessible approach.

To address these challenges, this study employs an artificial neural network (ANN) to predict ambient temperature profiles in diverse urban spatial characteristics using thermal images captured by Unmanned Aerial Vehicles (UAVs). Eight fixed weather stations are installed on a university campus in South Korea, each placed in areas with different spatial characteristics to measure ambient temperature. In addition to temperature, other parameters such as sky view factor and sun path are measured at these eight locations. The UAV captures thermal images every 30 minutes, covering the experimental area of the campus. These data are then used in ANN to predict trends in ambient temperature profiles.

The measurement period spans one week during the summer of 2024. The validation of the method will be presented, along with the establishment of the overall workflow for the prediction process. The main contribution of this study is the introduction of a new method that simplifies the prediction of outdoor ambient temperatures compared to traditional methods, addressing and overcoming their limitations. This approach has the potential to be used for predicting microclimate data in urban environments using typical aerial images and weather information. Such data can aid architects, landscape architects, urban planners, and policymakers in designing and analyzing urban spaces, ultimately supporting the creation of more sustainable built environments.