Impact assessment of land-use and land-cover on urban heat waves using remote sensing and machine learning algorithms

Understanding the response mechanisms of land-use and land-cover is essential for effectively implementing adaptation strategies for heat waves. This research employed a support vector machine model using LANDSAT imagery to classify land use and land cover and evaluate the alterations in these categories in Daegu, South Korea, between 2000 and 2020. The results indicated a significant expansion of built-up areas and bare land, which increased to 165.75 km² and 182.93 km², respectively, by 2020. Additionally, a more in-depth analysis was conducted to identify the primary factors influencing the substantial changes in urban thermal conditions that contribute to regional heat waves. The factors considered include land surface temperature, housing distribution, population density, normalized difference bareness index, normalized difference built-up index, urban heat island, and urban thermal field variance index. These factors were used to forecast land-use and land-cover maps using a gradient-boosting classifier as a machine-learning algorithm. The findings demonstrate that the gradient boosting classifier provides a more accurate prediction of land cover changes within the city, with accuracy and precision scores recorded at 0.81 and 0.82, respectively. Furthermore, the kappa statistics for support vector machine-classified maps and predicted maps, validated against ground truth data from Google Earth, were 92.5% and 95% for 2010 and 2020 and 86.7% and 86.8% for 2010 and 2020, respectively. This methodology can be replicated in other regions by using satellite remote sensing images to generate land-use and land-cover maps and other pertinent data for predictive analysis.