Exploration of the Urban Heat Island Phenomenon and Air Temperature Estimation Using Machine Learning in Bangkok, Thailand

Abstract

The urban heat island effect (UHI) results from extensive urban areas and is recognised to significantly influence climate change. This research focuses on the intricate dynamics of UHI in the Bangkok Metropolis. It complex a thorough survey that covered a period of 10 years, from 2013 to 2022. The main goal was pinpointing areas influenced by UHI and devise an Urban Thermal Field Variability Index (UTFVI) in the Bangkok Metropolis. This research uses land surface temperature (LST) data from the Terra/Aqua satellite (MODIS) to investigate the relationship between nighttime temperature fluctuations and light intensity. The findings are based on night light data (NTL) from the Suomi-NPP satellite (VIIRS). Machine learning (ML) techniques also play a main role in shaping prediction models. It helps estimate temperatures in regions lacking weather monitoring stations. The study results demonstrated that the Bang Khun Thian area predominantly participate in agriculture. It recorded the highest UHI temperature. The Terra satellite data indicated that UTFVI was considered the “strong” UHI phenomenon (UTFVI = 0.015) in comparison to the “bad” ecological evaluation index (EEI) because of a recorded maximum temperature of 27.3 °C. Simultaneously, the Aqua satellite detected a peak temperature of 26 °C, classified as the “strong” UHI phenomenon (UTFVI = 0.0103) when compared to the “poor” ecological evaluation index (EEI). Furthermore, the research demonstrates a connection between nighttime lights and nighttime temperature rises at a minimal level (0.27–0.34) by calculating air temperature in regions lacking weather monitoring stations. The created model demonstrates good reliability, achieving an R2 = 81.6% and RMSE = 0.87. The establishment of this model is advantageous for cities that do not have extensive weather station coverage. The advancement of this model can be used to provide information to regions that lack weather monitoring stations.