Scholarlink

<p>Amid the effects of climate change and rising urbanization, the interaction between urban heat islands (UHIs) and background climate factors has become critical to study. This study investigates the spatiotemporal variation of atmospheric urban heat island intensity (AUHII) in Guangdong Province, China, and evaluates the long-term influence of key climate variables: precipitation, relative humidity, and wind speed on AUHI. An integrated modeling approach was used, combining econometric techniques (Fully Modified Ordinary Least Squares and Dynamic Ordinary Least Squares) with machine learning and deep learning methods. The Random Forest (RF) model served as an initial benchmark, followed by a Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) framework to improve predictive accuracy. Results showed significant spatial and seasonal variations, with AUHII ranging from − 2.6 to 2.3&#xa0;°C for daytime, nighttime, and mean values. Seasonal extremes were observed in winter (-4.1 to 3.9&#xa0;°C) and summer (-1.8 to 1.4&#xa0;°C), with nighttime and winter exhibiting the strongest AUHI effects, particularly in western and southern cities. Relative humidity was the most influential factor, followed by precipitation. While the RF model identified key predictors, the CNN-LSTM model demonstrated stronger generalization, achieving testing R² values above 0.75 across most cities. Our findings enhance the understanding of the linkages between background climate variables and the AUHI effect, providing insight that can help urban planners and policymakers develop strategies to mitigate the effects of atmospheric urban heat islands.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Spatiotemporal assessment and background climate drivers of atmospheric urban heat island in Guangdong province, China

  • 期刊论文

  • 发表日期：

    2025年9月25日

查看全文
  • Abubakar Sabo Ahmad,
  • Li Yi,
  • Asim Biswas,
  • Ji Chen

摘要

Amid the effects of climate change and rising urbanization, the interaction between urban heat islands (UHIs) and background climate factors has become critical to study. This study investigates the spatiotemporal variation of atmospheric urban heat island intensity (AUHII) in Guangdong Province, China, and evaluates the long-term influence of key climate variables: precipitation, relative humidity, and wind speed on AUHI. An integrated modeling approach was used, combining econometric techniques (Fully Modified Ordinary Least Squares and Dynamic Ordinary Least Squares) with machine learning and deep learning methods. The Random Forest (RF) model served as an initial benchmark, followed by a Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) framework to improve predictive accuracy. Results showed significant spatial and seasonal variations, with AUHII ranging from − 2.6 to 2.3 °C for daytime, nighttime, and mean values. Seasonal extremes were observed in winter (-4.1 to 3.9 °C) and summer (-1.8 to 1.4 °C), with nighttime and winter exhibiting the strongest AUHI effects, particularly in western and southern cities. Relative humidity was the most influential factor, followed by precipitation. While the RF model identified key predictors, the CNN-LSTM model demonstrated stronger generalization, achieving testing R² values above 0.75 across most cities. Our findings enhance the understanding of the linkages between background climate variables and the AUHI effect, providing insight that can help urban planners and policymakers develop strategies to mitigate the effects of atmospheric urban heat islands.