Spatiotemporal Monitoring of Land Use-Land Cover and Its Relationship with Land Surface Temperature Changes Based on Remote Sensing, GIS, and Deep Learning

• Published in: Journal of the Indian Society of Remote Sensing 2024-11, Vol.52 (11), p.2461-2481
• Main Authors: Karimian, Razieh, Rangzan, Kazem, Karimi, Danya, Einali, Golzar
• Format: Article
• Language: English
• Subjects: Earth and Environmental Science; Earth Sciences; Remote Sensing/Photogrammetry; Research Article
• ISSN: 0255-660X; 0974-3006
• DOI: 10.1007/s12524-024-01958-3

Over the past century, rapid population growth and continuous exploitation of natural resources have caused numerous changes. A notable transformation involves the modification of land surface temperature (LST), which is impacted by changes in Land Use-Land Cover (LULC). The most important approach in discovering changes is to increase the accuracy of classification methods. Deep learning techniques have been successfully used, and this improved performance has been carried over to image classification. This study aims to monitor and mapping the spatial and temporal changes of LULC and LST in Izeh city using remote sensing, GIS, and deep learning. LULC and LST maps for the years 2001 and 2021 were created by processing thermal and multispectral bands. Two methods were used to generate the land use-land cover map: pixel-based (Max Likelihood (ML)) and object-based (Fully Convolutional Network (FCN)). During this period, the percentages of changes in water, urban, and wasteland classes increased, whereas those for grassland, forest, and wetland classes decreased. The average LST changes followed this order: wasteland > urban > grassland > forest > wetland > water. The normalized differential vegetation index (NDVI), normalized differential water index (NDWI), and normalized differential build-up index (NDBI) were utilized to analyze the relationship between LST and LULC. A linear, positive relationship between LST and NDBI was observed, indicating the direct effect of urban development on the increase in LST in the study area. The overall accuracy for LULC maps using the ML method was over 80.74% in 2001 and over 90.76% in 2021. With the FCN method, the accuracy was over 93% in 2001 and over 98% in 2021. Finally, evaluation the spatiotemporal environmental effects of unchecked human activity on LULC and its relationship with LST can be achieved using remote sensing, GIS, and deep learning approaches.