Snow Ensemble Uncertainty Project (SEUP): Exploring Variability and Uncertainty in Modeled SWE Estimates Using an Ensemble-based Approach
Abstract
Seasonal snow is critical to the Earth's climate and hydrological systems over a significant portion of the total land area, particularly in the northern hemisphere. Accurate real-time and long-term estimates of snow water equivalent (SWE) are required for snowmelt forecasting, water resource allocation, and for quantifying trends in snow mass associated with climate change. However, there is still considerable uncertainty across the suite of observational techniques and modeling approaches that provide global snow data. Additionally, data are generally not available with the latency, accuracy and spatiotemporal resolution required to address all needs.
In this study, an analysis of spatial and temporal variability in modeled snow water equivalent (SWE) and other land surface variables was conducted to better understand the factors that contribute to uncertainty and guide future research. An ensemble of distributed SWE estimates over North America was developed using different combinations of land surface models (LSMs) and forcing data. Four different land surface models (LSMs) of varying complexity were run using the NASA Land Information System (LIS) and three different forcing datasets were used to drive each of the LSMs. The 12-member ensemble was run on a 5-km grid over the time period 2000 - 2017, with the first nine years used as model spin-up and the remaining years (2009-2017) for analysis. The results were evaluated to assess the validity of the results, to identity spatial regions and temporal periods with high uncertainty, and to better understand the processes that drive model spread. We compared the model estimates to various ground-based and airborne snow estimates, independent modeled data, and satellite remote sensing products. We assess the results throughout the snow season and in different landscapes and snow climates. Finally, we evaluate the various components of the water and energy budget by comparison to other modeled data. The results show mountainous and forested areas have the highest SWE uncertainty, while tundra and northern plains may contribute the greatest uncertainty in continental scale total snow mass. Additionally, greater variability results from models than forcing data, which seem to be largely due to how models handle sublimation and snow-vegetation interaction.- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2019
- Bibcode:
- 2019AGUFM.C43C..06V
- Keywords:
-
- 0736 Snow;
- CRYOSPHERE;
- 0740 Snowmelt;
- CRYOSPHERE;
- 0798 Modeling;
- CRYOSPHERE