High-resolution Forcing Driven Offline ELM Snow Processing and Comparing to Observations in Two Alaska Tundra Regions

Land surface snow processes are of critical importance in northern high-latitude regions, however remaining one of great challenges to Earth System models. Its mostly due to highly heterogeneous surface across spatial-temporal scales and lacking of reliable data in those remote and harsh regions. Here we present snow process simulations using offline Energy Exascale Earth System Models (E3SM) Land Model (ELM) in Alaskan Coastal Tundra and Seward Peninsulas Tundra transition regions. Offline ELM is driven by 1km resolution climate forcing of DAYMET which sub-daily downscaled by GSWP3 v2 (1980-2014). The high resolution simulations are also using newly developed 0.01x0.01 degree northern American vegetation map. Specifically in this study, for assessing snow processes and its consequence on plant phenology, we derived DOYs (day of year) of snow melting and ground-covering at 24 and 4 km resolution from 1999-2014, from northern hemispheric daily snow coverage products of the US National Ice Centers Interactive Multi-sensor Snow and Ice Mapping System (USNIC-IMS). Compared to USNIC-IMS observations, ELM simulated yearly-averaged a couple of days earlier fully-snow-covering but a week or so earlier snow-melting, and thus about 5 days longer snow-free season. But model showed yearly trends of snow-free seasons reasonably well. Additionally simulations indicated that earlier snow melting could have been started since 1980s when observable warming appeared in the studied regions. The slightly earlier snow fully covering probably implied model driving forcing data discrepancy at 1 km resolution scale. While model-data discrepancy snow melting might be caused by either forcing or snow algorithms or both. Some preliminary analysis indicated that its partially due to modeled soil temperature bias under grass or shrub when snow covering. For an example, in Alaskan coastal tundra snow melting date data-model discrepancy appeared more severer than in more shrubby tundra of Seward Peninsulas. Our study demonstrated that data integration and model development and fidelity assessments are critical to further improve ELM performance in pan-Arctic.