1B.6 Assimilation and Evaluation of Air Force 557th Weather Wing SNODEP Products into NCEP Operational FV3GFS System

Monday, 7 January 2019: 9:45 AM
North 126BC (Phoenix Convention Center - West and North Buildings)
Jiarui Dong, NOAA, College Park, MD; and H. Wei, J. Meng, W. Zheng, J. S. Kain, S. V. Kumar, D. Mocko, R. Reichle, and C. D. Peters-Lidard

NOAA/NCEP Gridpoint Statistical Interpolation (GSI) is a unified variational data assimilation system utilized in the operational Global Forecast System (GFS). However, GSI lacks a land component and cannot assimilate land surface products such as soil moisture and snow. Thus, the land analysis in the GFS is conducted using NCEP’s Global Land Data Assimilation System (GLDAS) running under the NASA Land Information System (LIS), using the Noah LSM (Land Surface Model) to evolve land states and compute surface fluxes. The land states are updated using a “semi-coupled” approach, in which these states are generated from a parallel GLDAS driven by observed precipitation (a blend of satellite retrievals, gauge-based, and model-based at higher latitudes) and with near-surface forcing from the parent atmospheric data analysis system. However, assimilation of remotely-sensed estimates of land-surface states such as soil moisture and snowpack are not supported yet in GLDAS. To overcome this limitation, we are testing the latest version of LIS (NASA LIS version 7) in NCEP operational systems. This version of LIS includes NOAA operational land surface and hydrological models (NCEP’s Noah, versions from 2.7.1 to 3.6 and the future Noah-MP), high-resolution satellite and observational operators, and land DA tools. The land DA capabilities in this version of LIS have been transitioned to a developmental version of GLDAS to support NCEP’s land surface assimilation of satellite-based soil moisture and snow observations.

A set of offline numerical experiments driven by forcing from GFS forecasts have been conducted to evaluate the impact of assimilating snow with daily Global Historical Climatology Network (GHCN). An EnKF based approach in LIS has been used to evaluate the errors of snow modeling due to the inaccurate forcing, including temperature and precipitation biases. The statistics from LIS EnKF DA results with 20 members are better than all the other methods including AFWA SNODEP, operational GFS/GDAS product, LIS control run, and LIS DA with direct replacement. The land states from the snow DA are utilized to initialize the FV3GFS (a new version of the GFS that uses the FV3 dynamic core) daily for December 2017 experimental forecasts, and the 2-meter air temperature and specific humidity forecasts show improvements compared to FV3GFS without snow DA forecasts. The results will be discussed at the conference.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner