关键词: 数据融合, 增强型时空自适应反射融合模型, 地表城市热岛效应, 地理探测器

Abstract: Objective　Nanchang, one of China's well-known "stove cities," has experienced a pronounced intensification of the surface urban heat island (SUHI) effect in recent decades, posing considerable risks to human health, including the increasing risks of cardiovascular and respiratory diseases, and also disrupting the urban thermal-ecological balance. Although the variation patterns of SUHI have also been widely explored, existing studies have two major limitations. First, long-term SUHI spatiotemporal evolution, interactions among multidimensional thermal driving factors, and future thermal environment predictions are seldom studied within a unified framework. Secondly, especially for Nanchang city, frequent summer cloud cover leads to a lack of high-resolution, temporally continuous land surface temperature (LST) datasets for Nanchang's main urban area. In this study, the Enhanced Spatial and Temporal Adaptive Reflectance Fusion Model (ESTARFM) was used to spatially and temporally fuse multisource LST data from Landsat and MODIS, generating a 22-year continuous summer LST dataset at a 5-year interval for Nanchang's main urban area. Then the spatiotemporal evolution patterns of SUHI in Nanchang were systematically explored, and the individual and interactive effects of multiple natural and socioeconomic factors on LST were quantified. Finally, the future spatial distribution of heat islands with moderately strong and above levels in Nanchang was simulated. The study is expected to provide a scientific basis for the formulation of urban planning strategies for thermal environment optimization and sustainable development in Nanchang. Methods　Firstly, consistent preprocessing, including reprojection and resampling, was performed on Landsat 5/8 C2 L2 and MOD11A1/A2 LST data from 2000, 2005, 2009 (used in place of the poor-quality 2010 data), 2016, and 2022. And pixel-level fusion using ESTARFM was conducted to generate cloud-free summer LST products with a spatial resolution of 30 m. Secondly, the surface urban heat island intensity (SUHII) and the Urban-Heat-Island Ratio Index (URI) were derived to classify the urban thermal environments into seven categories, ranging from strong cold island to strong heat island. The spatiotemporal variations in these categories and the transitions between different heat island grades were subsequently quantified. Thirdly, nine representative driving factors (land use/land cover, LULC; NDVI, NDBI, MNDWI, GDP, population density, PM2.5, DEM, and slope) covering terrain, ecology, building density, population, economic activity, and air pollution were incorporated into the Geodetector model to evaluate their individual explanatory power and pairwise interaction in explaining LST spatial variation. Finally, an ANN-CA-Markov coupling model was established to simulate the spatial evolution of heat islands, where the seven factors of significance test were divided into three factor combinations, namely the main factor group, the secondary factor group and the full factor group. The observed 2022 heat island distribution data were used to evaluate the performance of the model, and the full-factor scheme with the highest simulation accuracy was subsequently selected to predict the spatial distribution of moderately strong and strong heat islands in 2034. The correlation coefficient (CC) and determination coefficient (R²) were used to evaluate LST fusion accuracy, while the Kappa coefficient was employed to evaluate the reliability of the spatial prediction model. Results and Discussion　The URI of Nanchang's main urban area increased steadily from 0.167 in 2000 to 0.306 in 2022, and the annual growth rate significantly accelerated from 2016 to 2022. During the research period, the total proportion of heat island regions in Nanchang increased from 34.77% to 43.6%, accompanied by the continuous expansion of moderately strong and strong heat island coverage from the old urban districts of Donghu and Xihu to southern Xinjian District and southern Nanchang County. Geodetector analysis identified that LULC has the strongest explanatory power for LST variation (q = 0.51), followed by NDVI, NDBI, and MNDWI, whereas DEM and slope are not significant, which is consistent with the predominantly flat terrain in the study region. All pairwise factor interactions exhibited bivariate enhancement, with the interaction between LULC and NDVI producing the highest explanatory power (q = 0.68). Model validation based on the actual heat island distribution data in Nanchang in 2022 showed that the full-factor group achieved the highest Kappa coefficient (0.76), supporting the combination of these significant test factors. The prediction of moderately strong and strong heat islands in Nanchang City in 2034 shows that, under the current unregulated urban sprawl scenario, the proportion of moderately strong and strong heat islands is predicted to increase from 31.89% in 2022 to 39.09% in 2034, with the southern parts of Xinjian District and Nanchang County emerging as the primary expansion zones of highintensity heat islands. Conclusions　The SUHI in Nanchang's main urban area intensified markedly from 2000 to 2022, with high-grade heat islands progressively expanding outward from the central urban area to the suburban new towns. Among the representative driving factors, land use type was the dominant driver governing the spatial differentiation of LST, while vegetation, built-up impervious surfaces, and urban water bodies exhibited synergistic effects with land use patterns in regulating urban thermal environment. If the current extensive urban development pattern continues without effective intervention, the SUHI effect will significantly deteriorate by 2034. To systematically mitigate SUHI in Nanchang, targeted countermeasures for optimizing land use structure are proposed, including adopting a compact land use model, establishing continuous green corridors across suburban thermal hotspots to prevent heat island aggregation, increasing urban vegetation coverage, protecting river and wetland networks, and synchronously implementing air pollution control and urban thermal environment management. Future research should incorporate multi-season, long-term thermal monitoring and multi-scenario heat island simulations under different greening strategies to provide stronger theoretical support for refined urban thermal environment governance.

Key words: data fusion, enhanced spatial and temporal adaptive reflectance fusion model, urban heat island effect; geodetector