ISA-1: AI-Empowered Super-Resolution Mapping 1-m Impervious Surface Area in the Yangtze River Economic Belt

<i>Currently, only the results of a few example cities have been released. All results will be made available once the paper is accepted.</i><br>The Yangtze River Economic Belt (YREB) is a critical region for the ecological and economic sustainability in China, which faces significant challenges from rapid urbanization, particularly the expansion of artificial impervious surfaces area (ISA). While existing many well-known ISA products (e.g., ESA-10m) offer foundational insights, their coarse spatial resolution limits accuracy, especially in fragmented landscapes and complex terrains. To this end, we map a 1-m very-high-resolution ISA product by means of AI techniques, called ISA-1, covering 36 cities in YREB, contributing to the community. ISA-1 is generated by integrating 10-m Sentinel-2 images with a generative adversarial network (GAN)-based super-resolution segmentation framework. Our approach overcomes the limitations of traditional methods, achieving a 10× spatial enhancement to resolve fine-grained urban structures and dispersed rural ISA features. Validation across 36 municipal-level administrative regions along the main stream of the Yangtze River shows that ISA-1 reduces ISA overestimation in densely urbanized areas (e.g., Shanghai, Suzhou) by 0.04%–36.15% compared with ESA-10m, while detecting 5.00% –81.74% more ISA in mountainous regions (e.g., Garzê, Liangshan), demonstrating superior capability in capturing small-scale artificial impervious surfaces. The model achieved an overall accuracy of 94.42%, with F1 scores of 96.64% (non-ISA) and 83.61% (ISA), outperforming existing mid-resolution products in both geometric precision and semantic consistency. This advancement not only enhances the monitoring of urbanization’s ecological impacts, such as altered hydrological cycles and biodiversity loss, but also provides a scalable, cost-effective solution for high-resolution ISA mapping