Overcoming Optical Observation Limitations: Automatic Dense Time-Series Mapping of Impervious Surfaces in Cloudy and Snow-Covered Regions

Timely and accurate monitoring of impervious surfaces is essential for urbanization assessment and sustainable development. However, dense time-series impervious surface mapping is severely constrained by the poor effective coverage of optical observation, which is frequently disrupted by cloud cover, precipitation, and snow, especially in regions with challenging weather conditions. To address these limitations and achieve synchronized refinement in both spatial and temporal dimensions, we propose a novel automatic dense time-series impervious surface mapping method based on multi-source remote sensing data. By integrating Sentinel-1 SAR, Sentinel-2 optical imagery, NPP VIIRS night-time lights, and topographic data, the proposed framework ensures data continuity and enhances mapping robustness. A key innovation of our approach is the hierarchical iterative refinement strategy, which leverages reliable coarse time-series outputs to iteratively improve finer-scale mapping, enabling semi-monthly monitoring at a 10-meter resolution. Moreover, we introduce an automated training sample generation strategy by fusing three existing land cover products, substantially reducing manual annotation efforts while maintaining high sample quality. Empirical evaluations in Beijing (a snow-covered region) and Wuhan (a cloudy and rainy region) demonstrate the method's superior performance, with semi-monthly mapping accuracies of 97.65% and 96.96%, respectively, and overall accuracy improvements exceeding 6.5%. These results highlight the method's ability to overcome the limitations of optical data discontinuity and support high-frequency, high-accuracy impervious surface monitoring in diverse and weather-challenged urban environments.