Snow: Dataset development, NWS products evaluations, and its impact on CFS subseasonal to seasonal prediction

Xubin Zeng
University of Arizona
  2 March, 2:30, in 2890

Abstract:
Snow (water equivalent, depth, and fraction) has a major impact on the energy and water cycle and land-atmosphere interactions. Despite this importance, high-quality snow datasets are lacking and NWP and climate models have difficulty in snow data assimilation and in snow modeling. Through our recent progress documented in 6 papers, here I will discuss several snow issues that are highly relevant to the NWS weather, water, and climate prediction.
 
First, we found large snow depth errors over U.S. in snow initializations from NCEP global (GFS and CFS) and regional (NAM) models (Dawson et al. 2016; doi: 10.1175/JHM-D-15-0227.1). The snow water equivalent (SWE) errors are even larger due to deficiencies in snow density. Subsequently we developed a new snow density parameterization for land data assimilation (Dawson et al. 2017a; doi: 10.1175/JHM-D-16-0166.1) that is significantly better than those used in the above snow initialization or in the NCEP land model (Noah).

Second, we developed a new and innovative method to obtain daily 4 km SWE and snow depth data from 1981 to present over continental U.S. based on USDA SNOTEL point SWE and snow depth measurements, NWS COOP point snow depth measurements, and PRISM daily gridded precipitation and temperature datasets (Broxton et al. 2016a; doi: 10.1002/2016EA000174). The robustness of our method and our product has been demonstrated using three approaches. Using this dataset, we found large SWE errors in reanalyses (including CFSR) and Global Land Data Assimilation Systems (including GLDAS-Noah) (Broxton et al. 2016b; doi: 10.1175/JHM-D-16-0056.1) and from satellite remote sensing (Dawson et al., 2017b, under preparation). Furthermore, the primary reasons for these underestimates are identified.

Finally, we found major impacts of snow initialization on CFS subseasonal to seasonal forecasting over Northern Hemisphere mid- and high-latitudes in the transition season (Apr-Jun), which are even greater than SST effects (Broxton et al. 2017, submitted). Furthermore, snow initialization deficiencies are primarily compensated by CFS atmospheric model deficiencies (most probably those related to atmospheric radiative transfer).

These results suggest that, to improve short-term to seasonal forecasting in the spring and early summer, CFS (GFS, and NGGPS) should improve snow initialization first, followed by atmospheric radiative transfer improvement (e.g., clouds and aerosols), and then followed by land model improvement.