Modeling the relationship between land use/land cover and land surface temperature in Dhaka, Bangladesh using CA-ANN algorithm

Abstract

Rises in land surface temperature (LST) significantly impacted by land use/land cover (LULC) changes. The increase in LST massively impacted the urban biodiversity, ecosystem and population health. This study aims to estimate the changes in LULC classes and identify their impacts on LST in Dhaka city, Bangladesh using Landsat satellite images from 2000 to 2020. Based on the past estimated change maps of LULC and LST, the study finally predicted the future LULC and LST scenario for the year 2030. The support vector machine algorithm was applied to perform the LULC classification. Artificial neural network and cellular automata algorithms were used to predict the LST and LULC changes for 2030. Results suggested a significant reduction in vegetation cover (5%) and an increase in built-up area (14%) from 2000 to 2020. Due to this massive increase in built-up areas, the LST increment took place by 7.24 °C in the last two decades. The maximum temperature was recorded in built-up areas (34 °C), and water bodies (19 °C) exhibited minimum temperature. A strong positive correlation was found between LST and Normalized Difference Built-up Index (NDBI), where negative relation estimated between LST and Normalized Difference Vegetation Index (NDVI) and Normalized Difference Water Index (NDWI). The predicted results for 2030 also exhibit significant loss of urban green cover areas by 13% and rises in built-up areas by 21%. The maximum LST will likely be increased by 9.29 °C in the predicted year. For ensuring sustainable urban development and minimizing the urban heat island effects, this study will play a significant role by providing effective guidelines for urban planners, policymakers and respective authorities of Dhaka city.

Keywords

• Land use/land cover changes
• Land surface temperature
• Cellular automata
• Artificial neural network
• Prediction