Leveraging machine learning algorithms in dynamic modeling of urban expansion, surface heat islands, and carbon storage for sustainable environmental management in coastal ecosystems.
Journal:
Journal of environmental management
PMID:
39305877
Abstract
Climate change and rapid urbanization are dramatically altering coastal ecosystems worldwide, with significant implications for land surface temperatures (LST) and carbon stock concentration (CSC). This study investigates the impacts of day and night time LST dynamics on CSC in Cox's Bazar, Bangladesh, from 1996 to 2021, with future projections to 2041. Using Landsat and MODIS imagery, we found that mean daytime LST increased by 3.57 °C over the 25-year period, while nighttime LST showed a slight decrease of 0.05 °C. Concurrently, areas with no carbon storage increased by 355.78%, while high and very high CSC zones declined by 14.15% and 47.78%, respectively. The Integrated Valuation of Ecosystem Services and Tradeoffs (InVEST) model estimated a 28.64 km reduction in high CSC areas from 1996 to 2021. Statistical analysis revealed strong negative correlations between LST and vegetation indices (R = -0.795 to -0.842, p < 0.001) and positive correlations with built-up indices (R = 0.812 to 0.893, p < 0.001). Cross-sectional analysis showed that areas within 2 km of the coastline experienced a lower rate of LST increase (0.03 °C/year) compared to inland areas (0.05 °C/year). A Cellular Automata-Artificial Neural Network model projected that by 2041, 22.51% of the study area may experience LST >32 °C, while areas with LST <24 °C may decrease to 1.68%. These observations underscore the pressing necessity for sustainable strategies in urban planning and conservation in swiftly evolving coastal areas, especially considering the challenges posed by climate change and population growth.