The impact of land initialization on seasonal forecasts of surface air temperature: role of snow data assimilation in the Northern Hemisphere

Abstract:

Land initializations (i.e., snow, soil moisture, leaf area index) have been recognized as important sources of seasonal climate predictability besides ocean and atmosphere initializations. However, studies focusing on assessing how land data assimilation (DA) contributes to seasonal forecast skills are still lacking due to the limited number of large-scale land DA studies. In this study, taking advantage of the snow outputs from a multivariate global land DA system (i.e., DART/CLM), we systematically investigated the role of large-scale snow DA in influencing seasonal forecasts of surface air temperature. Three suites of ensemble seasonal forecast experiments were performed using the Community Earth System Model (CESM v1.2.1), in which three different snow initialization datasets were used. They are (1) CLM4 simulation without DA, (2) CLM4 simulation with MODIS snow cover DA, and (3) CLM4 simulation with joint GRACE and MODIS snow DA. Each suite of the experiment starts from multiple initialization dates of eight years from 2003 to 2010 and has three-month lead times. All experiments used the same atmosphere initializations from ERA-Interim (perturbed to get 8 ensembles) and the same prescribed SSTs. Our results show that snow DA plays an important role in surface air temperature predictions in regions such as Europe, western Canada, northern Alaska, Mongolia Plateau, Tibetan Plateau, and the Rocky Mountains. The analyses also account for multiple lead times as snow can influence the atmosphere through immediate snow-albedo effect and through delayed snow hydrological effect after snow melts and wets the soil. This is a first study to quantify the impacts of snow initializations on seasonal forecasts of surface air temperature with an emphasis on large-scale snow DA. The insights are helpful to both land DA studies as well as research on seasonal climate forecasts.