Maximum Temperature Prediction Using Remote Sensing Data Via Convolutional Neural Network

The study introduces a machine-learning model utilizing Convolutional Neural Network (CNN) to predict maximum temperatures in urban areas, leveraging satellite data, meteorological predictions, and additional remote sensing inputs, with the aim of providing high-resolution temperature prediction maps for Turin's urban region.

Abstract

Urban heat islands, defined as specific zones exhibiting substantially higher temperatures than their immediate environs, pose significant threats to environmental sustainability and public health. This study introduces a novel machine-learning model that amalgamates data from the Sentinel-3 satellite, meteorological predictions, and additional remote sensing inputs. The primary aim is to generate detailed spatiotemporal maps that forecast the peak temperatures within a 24-hour period in Turin. Experimental results validate the model's proficiency in predicting temperature patterns, achieving a Mean Absolute Error (MAE) of 2.09 degrees Celsius for the year 2023 at a resolution of 20 meters per pixel, thereby enriching our knowledge of urban climatic behavior. This investigation enhances the understanding of urban microclimates, emphasizing the importance of cross-disciplinary data integration, and laying the groundwork for informed policy-making aimed at alleviating the negative impacts of extreme urban temperatures.

Method

The authors employed a machine-learning model utilizing a Convolutional Neural Network (CNN) for high-resolution maximum temperature prediction in Turin. The model's input data is a 3D matrix comprising various channels, including spectral channels from Sentinel-3, DEM data, land cover information, weather components, and spatial coordinates. The UHI detection task is defined as an image-to-image regression task, aiming to find a function that represents the estimated maximum temperature for each pixel on the corresponding day. The dataset undergoes preprocessing steps to enhance model robustness, including normalization and geometric augmentations. The model's performance was evaluated using the mean absolute error (MAE) metric, and three separate models were trained at resolutions of 100, 50, and 20 meters per pixel. The results indicated that the ConvNext model outperformed the other models, achieving the lowest MAE values at all resolutions.

Main Finding

The authors discovered that by utilizing a machine-learning model incorporating Convolutional Neural Network (CNN) and integrating satellite data, meteorological predictions, and additional remote sensing inputs, they were able to generate high-resolution temperature prediction maps for Turin's urban region. The model demonstrated proficiency in predicting temperature patterns, achieving a Mean Absolute Error (MAE) of 2.09°C for the year 2023 at a resolution of 20 meters per pixel. This research enriched the understanding of urban climatic behavior and emphasized the significance of cross-disciplinary data integration for informed policy-making aimed at mitigating the negative impacts of extreme urban temperatures. Additionally, the study introduced a novel approach to high-resolution spatiotemporal prediction of maximum temperatures, providing comprehensive insights into Urban Heat Islands (UHIs) and their impact on urban environments.

Conclusion

The conclusion of the study is that the machine-learning model, utilizing a Convolutional Neural Network (CNN) and integrating satellite data, meteorological predictions, and additional remote sensing inputs, effectively predicts high-resolution maximum daily temperatures in Turin. The model achieved a Mean Absolute Error (MAE) of 2.09°C for the year 2023 at a resolution of 20 meters per pixel, providing comprehensive insights into urban heat islands and their impact on urban environments. The study also highlights the importance of cross-disciplinary data integration and suggests potential enhancements, such as integrating multistep predictions and supplementary input data, to further improve the model's predictive capabilities. These proposed enhancements could significantly augment the model's predictive capabilities and offer a more detailed understanding of the factors impacting the predictions.

Keywords

AI alignment, paradox, language models, adversarial attacks, model tinkering, input tinkering, output tinkering, rogue actors, AI research, human values, beneficial use of AI