Using Cloud Liquid Water Path to Improve Severe Weather Forecasts with WRF-DART

Assimilating retrieved cloud properties from satellite data into convective-scale models has to date received limited attention in the research community despite its potential to provide a wide array of information to a model analysis. Products currently available include cloud liquid water path (CLWP), which represents the amount of cloud water and cloud ice present in an integrated column and cloud top (base) pressures (CTP, CBP), which represent the top and bottom pressure levels of a cloud layer respectively. Retrieval algorithms have been developed to derive these products from operational GOES-13 Imager data with a spatial resolution of 4 km at 15 minute intervals. These data are assimilated into a convective scale (3 km) resolution WRF model using an Ensemble Kalman Filter (EnKF) approach through the Data Assimilation Research Testbed (DART) software. The default DART distribution does not have the capability to assimilate CLWP directly requiring the development of a new forward operator for this purpose. The new CLWP forward operator combines the satellite cloud height information with the cloud property variables generated by WRF to determine where in the atmospheric column and how much to adjust its CLWP to match the observation. Testing of this forward operator with a case study occurring on 10 May 2010 in Oklahoma and Kansas is conducted using two experiments. The first only assimilates conventional (CONV) observations (ASOS, ACARS, and RAOB) while the second assimilates the conventional observations along with GOES CLWP (PATH). Both experiments begin their assimilation cycle at 1800 UTC using hourly mesoscale analyses for initial and boundary conditions. Assimilation cycles every 15 minutes until reaching 2100 UTC with forecasts generated thereafter for a total of 40 ensemble members. Comparison of the GOES CLWP observations at 2045 UTC to CONV and PATH model analysis fields at the same time show that PATH has a much improved representation of both the magnitude and spatial orientation of CLWP compared to CONV. Assimilating CLWP acts to suppresses convection in the model where none is present in satellite data while adding it where it does exist. Validating downward shortwave flux from each experiment against nearby Oklahoma Mesonet observations at 2100 UTC shows that PATH significantly reduces RMSE (165.5 vs. 240.7 Wm-2). Reduction in model error is generally maximized during the first 30 minutes of the forecast after 2100 UTC. Thereafter, the impact of satellite observations decreases with CONV and PATH ensemble mean forecasts converging.