Scholarlink

<p>The changing local climate and human comfort in metropolitan cities like Mumbai due to the expansion of urban impervious surfaces emphasize the significance for researchers studying urban heat islands (UHIs). The present study aims to analyze and predict future land use and land cover (LULC) and land surface temperature (LST) of Mumbai city using satellite-derived datasets and machine learning methods. To achieve this objective, thirty years of data from 1994 to 2024 at an interval of ten years have been considered, and cellular automata-artificial neural network (CA-ANN), random forest (RF), and extreme gradient boosting (XGBoost) regression models have been applied to predict the LULC and LST for the year 2034. The results reveal that the temperature has rapidly increased (4&#xa0;°C) from 1994 to 2004, and thereafter it has increased (&lt; 0.5&#xa0;°C) at a minimal rate. The CA-ANN model utilizes the previous LULC maps and thematic layers to predict the LULC in 2034 with an accuracy of 79%, projecting the urban area of Mumbai city to reach around 377 km<sup>2</sup>. Based on the predicted LULC and other factors, the RF and XGBoost regression models predicted the 2034 LST map with an RMSE value of less than 1&#xa0;°C. Moreover, from 1994 to 2034, it has been observed that the UHI region of the study area varied between 29% and 30%, which is comparatively lower than the other metropolitan cities, while the area of urban heat spot (UHS) has continuously increased (5%) due to the increasing trend of built-up area. This study provides a critical resource for policymakers aiming to mitigate urban heat and to support urban planning and environmental management strategies for sustainable growth in Mumbai city. Furthermore, the methodology of this study can be helpful in other studies for predicting future LULC and LST.</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

Evaluation and prediction of LST and LULC changes for urban heat island analysis in Mumbai using remote sensing and machine learning techniques

  • 期刊论文

  • 发表日期：

    2026年3月23日

查看全文
  • Pankaj Prasad,
  • C. Jeganathan,
  • Pallab Kumar Das,
  • Indrajit Mukherjee,
  • Akhilesh Agrawal

摘要

The changing local climate and human comfort in metropolitan cities like Mumbai due to the expansion of urban impervious surfaces emphasize the significance for researchers studying urban heat islands (UHIs). The present study aims to analyze and predict future land use and land cover (LULC) and land surface temperature (LST) of Mumbai city using satellite-derived datasets and machine learning methods. To achieve this objective, thirty years of data from 1994 to 2024 at an interval of ten years have been considered, and cellular automata-artificial neural network (CA-ANN), random forest (RF), and extreme gradient boosting (XGBoost) regression models have been applied to predict the LULC and LST for the year 2034. The results reveal that the temperature has rapidly increased (4 °C) from 1994 to 2004, and thereafter it has increased (< 0.5 °C) at a minimal rate. The CA-ANN model utilizes the previous LULC maps and thematic layers to predict the LULC in 2034 with an accuracy of 79%, projecting the urban area of Mumbai city to reach around 377 km2. Based on the predicted LULC and other factors, the RF and XGBoost regression models predicted the 2034 LST map with an RMSE value of less than 1 °C. Moreover, from 1994 to 2034, it has been observed that the UHI region of the study area varied between 29% and 30%, which is comparatively lower than the other metropolitan cities, while the area of urban heat spot (UHS) has continuously increased (5%) due to the increasing trend of built-up area. This study provides a critical resource for policymakers aiming to mitigate urban heat and to support urban planning and environmental management strategies for sustainable growth in Mumbai city. Furthermore, the methodology of this study can be helpful in other studies for predicting future LULC and LST.