Addressing Numerical Challenges Associated with WRF LES Modeling: Shallow Clouds over the ARM SGP Site

Description

A number of LES models exist that could potentially be used routinely at the ARM SGP site. The Weather Research and Forecasting (WRF) model has the benefit of community development that has resulted in the availability of multiple options for individual physics parameterizations and their continual improvement. Multiple nested grids can easily be configured, enabling realistic temporally and spatially varying boundary conditions for the LES domain. Several data assimilation packages have already been developed that could be used in the future to merge ARM data with the WRF simulation to produce high-resolution analyses over the SGP site. In addition, treatments of aerosol chemistry and cloud-aerosol interactions have been implemented in the model that could be included for routine operations and/or research purposes. However, Yamaguchi and Feingold (2012) showed for one case that WRF LES simulations are potentially sensitive to convergence issues because of its treatment of acoustic waves in solving the fully compressible equations. The impact of this issue on the routine simulations of clouds for the wide range of realistic atmospheric conditions was not addressed. By performing a series of sensitivity simulations, we found that the convergence issues only occurs when there is a large specific humidity gradients across the inversion at the top of the boundary layer combined with sufficiently strong wind speed. This study and a companion poster (Endo et al. 2015) both found that the convergence issues are tied to numerical noise that is generated shortly after the initialization of the simulation, and can occur even before clouds form. We used the long-term measurements from the ARM Planetary Boundary Layer (PBL) Value Added Product to show that meteorological conditions predisposing WRF to exhibit convergence issues occur infrequently at the SGP site. WRF LES simulations that employ typical conditions associated with shallow clouds at the SGP site do not produce the numerical noise that leads to the convergence issues. In addition, when the Weighted Essentially Non-Oscillatory (WENO) advection scheme combined with increased divergence damping is used, the numerical noise at the beginning of the simulation can be effectively removed and the convergence issue is not present. We found that WENO must be applied to momentum, moisture, and scalars. When WENO is applied to only moisture and scalars the numerical noise is reduced but not removed entirely. The combination of an improved advection treatment and the climatological analyses shows that the numerical issues pointed out in Yamaguchi and Feingold (2012) should not be a factor in considering using WRF LES for routine use at the SGP site.