Abstract: Snow exhibits high reflectivity and significant reflective anisotropy, playing an important role in climate change and the global radiation energy balance. To improve the accuracy of snow cover information extraction and reflectance property inversion from remote sensing data, this study comprehensively assessed three typical BRDF models—ART, RTLSRS, and FASMAR—using multi-angle POLDER data and single-angle MODIS data under different snow conditions (stable and changing periods). The results show that when using POLDER data as the source, all three models can effectively fit the reflective characteristics of snow in different states. Among them, RTLSRS achieves the highest fitting accuracy, followed by FASMAR, while ART tends to underestimate reflectivity under large forward observation angles. When using MODIS data as the source, during the stable snow state period, the simulation accuracy of the RTLSRS and FASMAR models is better than that of the ART model. During the changing snow state period, the errors of the kernel-driven models RTLSRS and FASMAR increase, whereas the ART model, capable of performing inversions with single-day data, demonstrates better accuracy, though it exhibits larger fitting errors in the shortwave infrared band.

Key words: Snow, Reflectance, Bidirectional reflectance distribution function (BRDF), POLDER, MODIS