Predicting Urban Land Use and Mitigating Land Surface Temperature: Exploring the Role of Urban Configuration with Convolutional Neural Networks

Abstract

The objective of this research was to examine the influence of urban configuration on the mitigation of land surface temperature (LST) and the prediction of land use and land cover change through the utilization of convolutional neural network modeling. The results indicate that the formation of different urban heat island patterns is significantly influenced by both urban geometry and land use land cover (LULC) types. However, there is no significant correlation between these factors and LST across all configuration metrics. The associations between landscape configuration and land cover types exhibit variability contingent upon the particular forest cover categories under examination. Furthermore, the application of predictive LULC mapping reveals a divergent pattern, characterized by a rise in the overall extent of vegetation but a decline in the inner context of the Shiraz metropolitan area. The projected trajectory of built-up areas indicates a continued trend of urban expansion. The unique landscape patterns are a result of the distinct characteristics of each LULC. According to recommendations, to address the issue of mean LST, it is advisable for urban landscape planning to give priority to cohesion, density, and continuity while simultaneously minimizing fragmentation, variability, and complexity. This research provides valuable insights into the following aspects for urban planners, policymakers, and practitioners to address the following. (1) Selecting appropriate landscape metrics: this study identifies suitable landscape metrics to represent and interpret different landscape structures and land use land cover changes (LULCCs) over time. (2) Understanding regional variations: this research highlights that different landscape metrics have distinct effectiveness in different regions. This knowledge helps urban planners and policymakers to tailor their strategies and interventions based on specific regional characteristics, ensuring more effective mitigation of the urban heat island effect. (3) Different correlations with configuration metrics: this study reveals that the correlations between landscape configuration and LST differ in land cover types. (4) Anticipating future changes: this research utilizes machine learning models to predict future LULCC and landscape metrics. This information is valuable for urban planners and policymakers in anticipating and preparing for future urban expansion and changes in vegetation areas. It enables them to proactively design and implement strategies to manage the surface urban heat island effect. Urban planners and policymakers can utilize these insights to develop comprehensive strategies that integrate land-use design, landscape configuration, and urban form to mitigate the negative impacts of urbanization.