Diagnosing current land surface models' ability to represent snowpack evolution in complex terrain and forested Central Rockies

The timing and amount of spring snowmelt runoff in mountainous regions are critical for water resources management. This study developed an integrated data set including snow water equivalent (SWE) observations from 112 SNOTEL sites in the Colorado Headwaters region for the 2008 water year, 2004-2008 observations (surface heat fluxes, radiation budgets, soil temperature and moisture) from two AmeriFlux sites (Niwot Ridge and GLEES), MODIS snow cover, and river discharge. These data were used to evaluate the ability of six widely-used land-surface/snow models (Noah, Noah-MP, VIC, CLM, SAST, and LEAF-2) in simulating the seasonal evolution of snowpack in central Rockies.

All models captured the seasonality of SWE evolution fairly well, although they underestimated both early-spring (March-April) snow accumulation and late spring ablation. Underestimating snowmelt from mid-May to mid-June allowed models to compensate for lower SWE in spring, and consequently resulted in a prolonged snow season. For Noah, Noah-MP, SAST, and LEAF, 50% of their simulated sites had biases ranging from -200 mm to 100 mm in maximum SWE, while CLM and VIC models had larger variations among sites. Most of the 112 simulated sites in those models tended to reach the maximum SWE late. While Noah-MP and LEAF closely followed observed ablation, the other four models had slower spring-snow depletion. No single model performed the best (or the worst) to reproduce three important features of snow evolution: maximum SWE depth, and the timing of maximum and minimum SWE.

The models exhibited large disparities in simulating the surface energy partitioning, which is equally important for correctly representing snow-atmospheric interactions in weather and climate models. All models were able to simulate increased snow albedo following fresh snowfall and its reduction due to snow aging and compaction, but they differed substantially in the magnitude and diurnal cycle of albedo. Only CLM and Noah-MP reproduced the observed “w-shaped” diurnal cycle of surface albedo. All models, especially Noah and VIC, underestimated the solar energy absorbed at the forest-soil-snow interface from December to March. That resulted in too little outgoing long-wave radiation and sensible heating returned to the atmosphere, which could be a crucial deficiency for coupled weather and climate models. Those model disparities and deficiencies can be further traced down by examining the treatment (or lack of it) of turbulence and radiation processes within and under the vegetation canopy. Excessive shortwave and longwave radiation transfer from canopy to ground/snow often led to larger sensible heating from canopy to ground/snow surface, and resulted in larger, undesired weekly SWE change in both accumulation and ablation phases. Accurate radiation transfer between canopy and ground/snow surface appears to be essential for capturing both snowpack evolution and snow-vegetation-atmosphere interactions.