STNN-based weighted regression models for the prediction of urban land surface temperature-A study of spatiotemporal association

Read the full article See related articles

Listed in

This article is not in any list yet, why not save it to one of your lists.
Log in to save this article

Abstract

In view of rapid environmental degradation, the city of Delhi faces prolonged heat waves and rising temperatures. Policymakers can leverage AI-based prediction models by utilizing remote sensing techniques to predict urban heat patterns, which enables data-driven decision-making for climate resilience. Earlier research shows a wide interdependency between LST (land surface temperature) and other risk factors, e.g., air pollutants, climatic, and land use characteristics variables. In this study, a significant positive correlation is observed between ozone and LST, whereas other pollutants show an opposite correlation with LST. This inverse relationship occurs because pollutants absorb solar radiation, limiting the amount of sunlight reaching the Earth’s surface, which in turn reduces the urban LST. However, the western and northwest regions of Delhi show consistently higher LST throughout the year of the study due to dense residential and commercial hubs. In addition, the present study aims to develop three STNN (Spatiotemporal Neural Network)-based weighted regression models (viz., STNNWR-v1, STNNWR-v2, and STNNWR-v3). Attention is paid to lowering the model complexity by reducing the number of parameters. The proposed models are compared with other traditional spatiotemporal models, which proves the performance of STNNWR-v3 is significantly better in terms of evolution metrics and computation time. The performance metrics for STNNWR-v3 for 30 independent trials are as follows: R2 score 0.931±0.003 and MSE 2.886±0.118(mean ± standard deviation). Moreover, execution time analysis further highlights that the computation time required for each independent trial is significantly less due to the smaller number of parameters used.

Article activity feed