Urban Geothermal Resource Potential Mapping Using Data-Driven Models—A Case Study of Zhuhai City

Abstract

Geothermal energy, with its promise of sustainability and a minimal environmental impact, offers a viable alternative to fossil fuels that can allow us to meet the increasing energy demands while mitigating concerns over climate change. Urban areas, with their large energy consumption, stand to benefit significantly from the integration of geothermal systems. With the growing need to harness renewable energy sources efficiently, the detection of urban subsurface resources represents a critical frontier in the pursuit of sustainability. The Guangdong Bay area, known for its abundant geothermal resources, stands at the forefront of this green energy revolution, so, in our study, we chose to evaluate Zhuhai City, which is a city representative of the resource-rich area of Guangdong. With the progress of geographic information system (GIS) technology, the land surface temperature (LST) has been used to monitor the spatial distribution characteristics of geothermal anomalies. However, relatively few studies have been conducted in the field of urban geothermal resources. In this study, we calculated the LST of Zhuhai City using Landsat 8 remote sensing data and then investigated the distributions of geothermal hot springs. Spatial data layers were constructed, including the geological structure, DEM and derivatives, lithology, and urban regions, and, based on technology with the integration of machine learning, their spatial correlations with geothermal anomalies were analyzed. The support vector machine (SVM) and the multilayer perceptron (MLP) were employed to produce maps of potential geothermal resources, and their susceptibility levels were divided into five classes: very low, low, moderate, high, and very high. Through model interpretation, we found the moderate-susceptibility class to dominate at 26.90% (SVM) and 46.27% (MLP) according to the two models. Considering the influence of artificial areas, we also corrected the original LST by identifying urban areas of thermal anomalies via the urban thermal anomaly leapfrog fusion extraction (UTALFE) method; following this augmentation, the results shifted to 24.16% (SVM) and 28.67% (MLP). Meanwhile, the area under the curve (AUC) values of all results were greater than 0.65, showing the superior performance and the high applicability of the chosen study area. This study demonstrates that data-driven models integrating thermal infrared remote sensing technology are a promising tool for the mapping of potential urban geothermal resources for further exploration. Moreover, after correction, the reclassified LST results of urban areas are more authentic and suitable for the mapping of potential geothermal resources. In the future, the method applied in this study may be considered in the exploration of more southeastern coastal cities in China.