A comparison of optical-band based snow extent products during spring over North America

We compare the performances of two widely used hemispheric scale snow products during April, May, and June over North America. The Interactive Multisensor Snow and Ice Mapping System (IMS), based primarily on optical-band remotely sensed images, is the latest incarnation of a product that dates back to the 1960s and has been used as input to operational weather forecasting models as well as for establishing the historical climatology of snow extent over land surfaces. NASA's Moderate Resolution Imaging Spectroradiometer (MODIS) has been used for numerous applications since it was launched aboard the Terra satellite platform in 1999. The MODIS snow product is based primarily on optical-band reflectances. We include in our analysis only observations that are largely unobstructed by clouds as determined using the MODIS cloud detection algorithm. Then, after removing the influences of terrain and projection errors, we identify regions and land surface types where discrepancies between these two products occur. We also compare IMS and MODIS to the snow reanalysis produced by the Canadian Meteorological Center (CMC).We find that on seasonal time scales, the most pronounced differences between the IMS and MODIS snow products occurs during the ablation season over North America. Our results corroborate earlier studies showing pronounced differences over the northern tundra in June, where MODIS appears to be in agreement with other observations; as well as differences in April and May in the boreal forest, where evidence suggests that both products may be biased (although MODIS biases may be smaller) in comparison with the CMC product (which is based on station observations). The influence of clouds may be a factor even though the analysis includes only clear days. Another possible explanation for these discrepancies involves the impact of numerous small lakes over the North American landscape on the interpretation of satellite retrievals in the visible band, although there are other potential sources of error in both products. For example, comparison to the CMC reanalysis suggests that MODIS may be overestimating snow during the ablation season in the boreal forest. The resolution of these discrepancies may affect our understanding of the seasonal snow cover cycle, the evaluation of and development of parameterization schemes for climate models, and the development of a climate data record for snow cover.