Multi-Sensor Spatiotemporal Fusion for 30-m Daily Gapless Snow Cover Mapping

Abstract

High spatiotemporal resolution remote sensing data is crucial for monitoring heterogeneous mountainous snow cover. Although spatiotemporal fusion presents a promising approach for high-resolution snow monitoring, cloud contamination and sparse observations remain a critical constraint on its large-scale and long-term implementation. To address this issue, we propose an adaptive time-series fusion framework to generate 30-m daily gapless snow cover data based on the Enhanced Spatial and Temporal Adaptive Reflectance Fusion Model (ESTARFM). We integrate multi-source coarse-resolution data and multi-source fine-resolution data to increase the number of valid pixels and enhance data density to capture the rapid spatiotemporal variations of snow cover. Additionally, we introduce time-series image pairs to adapt the ESTARFM method, which overcomes the spatial completeness limitation of fine-resolution data and dynamically selects the spatiotemporal information closest to the target time for each pixel. Comprehensive evaluations confirm the high accuracy of the fused results, as demonstrated by the consistency with reference data (R = 0.776–0.964). Furthermore, validation with ground-based snow observations shows that the fused 30-m daily snow cover data not only outperforms the widely used 500-m data in capturing the temporal dynamics of snow cover, as evidenced by its strong alignment with ground-based snow phenology metrics, but also provides new insights into the spatial distribution of mountainous snow cover. In areas with elevations below 3,500 m, slopes under 25°, or shaded slopes, the 30-m data captures small-scale, sparse, and fragmented snow cover, offering significant potential for hydrological research and practical applications that require accurate snow cover estimation.