Scholarlink

<p>Rapid urbanization and climate change are intensifying heat-related risks worldwide, especially in urban areas of arid regions. Despite the urgent need for timely heat stress assessment and prediction, this area remains underexplored in arid regions. This study utilizes 11 types of daily weather data from 1982 to 2023 to calculate the Heat Index (HI) and develop predictive models for Riyadh City, Saudi Arabia. Six machine learning models—Long Short-Term Memory (LSTM), Gradient Recurrent Unit (GRU), Convolutional Neural Networks, Random Forest (RF), Gradient Boosting Machines, and Extreme Gradient Boosting Machines (XGBoost)—were trained and evaluated using standard performance metrics. Results show that the HI averages 29.65&#xa0;°C in January, peaking at 47.78&#xa0;°C, with March showing the least variability. Summer months reveal consistent HI peaks of 31.73&#xa0;°C, 33.06&#xa0;°C, and 32.91&#xa0;°C, indicating periods of noticeable discomfort, while extreme values in July and August reflect dangerous conditions. December indicates a rebound in HI (28.25&#xa0;°C) with greater variability, signaling intermittent but significant stress events even during winter. Prediction models reveal that while LSTM and GRU effectively capture temporal dependencies (R² ≈ 0.80), they have limited accuracy (RMSE &gt; 1.0) in predicting high-heat stress. Conversely, RF, GBR, and XGBoost achieve higher accuracy, with RF demonstrating exceptional capability in modeling complex thermal dynamics (R² = 0.9950, RMSE = 0.2663). These findings provide valuable insights for developing adaptive urban planning and early warning systems to improve thermal resilience in arid and climate-vulnerable cities.</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

Advanced deep learning and ensemble methods for urban heat stress prediction in arid metropolitan environments of Riyadh

  • 期刊论文

  • 发表日期：

    2026年1月7日

查看全文
  • Hamad Ahmed Altuwaijri,
  • Abdulla Al Kafy,
  • Mahin Rahman

摘要

Rapid urbanization and climate change are intensifying heat-related risks worldwide, especially in urban areas of arid regions. Despite the urgent need for timely heat stress assessment and prediction, this area remains underexplored in arid regions. This study utilizes 11 types of daily weather data from 1982 to 2023 to calculate the Heat Index (HI) and develop predictive models for Riyadh City, Saudi Arabia. Six machine learning models—Long Short-Term Memory (LSTM), Gradient Recurrent Unit (GRU), Convolutional Neural Networks, Random Forest (RF), Gradient Boosting Machines, and Extreme Gradient Boosting Machines (XGBoost)—were trained and evaluated using standard performance metrics. Results show that the HI averages 29.65 °C in January, peaking at 47.78 °C, with March showing the least variability. Summer months reveal consistent HI peaks of 31.73 °C, 33.06 °C, and 32.91 °C, indicating periods of noticeable discomfort, while extreme values in July and August reflect dangerous conditions. December indicates a rebound in HI (28.25 °C) with greater variability, signaling intermittent but significant stress events even during winter. Prediction models reveal that while LSTM and GRU effectively capture temporal dependencies (R² ≈ 0.80), they have limited accuracy (RMSE > 1.0) in predicting high-heat stress. Conversely, RF, GBR, and XGBoost achieve higher accuracy, with RF demonstrating exceptional capability in modeling complex thermal dynamics (R² = 0.9950, RMSE = 0.2663). These findings provide valuable insights for developing adaptive urban planning and early warning systems to improve thermal resilience in arid and climate-vulnerable cities.