Using DART and State Augmentation to Update a Surface Layer Parameter in WRF

Uncertainty in empirical parameters is a major source of model error in numerical weather prediction (NWP) models.  In certain surface layer parameterization schemes in the Weather Research and Forecasting (WRF) model (and other NWP models), there is one such parameter, the Zilitinkevich constant C_zil, which governs the strength of the thermal coupling between the land surface and the surface layer.  This constant has a direct effect of the heat and moisture fluxes through the surface layer, and a secondary effect on the momentum flux.  C_zil therefore has a large effect on the growth and structure of the atmospheric boundary layer.  There are no known methods to directly measure the correct value of C_zil, so its value must be estimated.  By default WRF treats C_zil as a global constant, but recent research suggests it should not be a constant in space or in time.

To develop an estimate of C_zil that varies both in space and in time, we couple a WRF ensemble with the Data Assimilation Research Testbed (DART).  We use an ensemble adjustment Kalman filter in this WRF-DART ensemble prediction system to assimilate surface and upper-air observations and update the WRF state vector of temperature, wind, and moisture variables.  Using a process called state augmentation, we also append C_zil to the state vector and allow DART to update C_zil as well.

We demonstrate the benefit of estimating a variable C_zil on low-level temperature and wind forecasts.  To do this we compare the performance of a constant-C_zil and a variable-C_zil ensemble from late September to mid-October 2012.  This time period coincides with high-resolution meteorological observations made during the Mountain Terrain Atmospheric Modeling and Observations Program (MATERHORN) at Dugway Proving Ground in western Utah.