Wednesday, 14 August 2002: 4:30 PM
The effects of subgrid model mixing and numerical filtering in simulations of mesoscale convective systems
Many mesoscale cloud simulations of grid spacing O(1km) employ a
turbulent-closure model for subgrid-scale mixing as well as a numerical spatial
filter to avoid numerical errors. In this study, the effects of mixing by
subgrid models and numerical filtering in simulations of mesoscale cloud
systems are investigated using the newly developed Weather Research and
Forecasting (WRF) model. Because the WRF model's numerical scheme is free of
numerical filters, we have been able to examine the sensitivity of squall-line
simulations to the parameters in popular turbulence-closure schemes and to the
parameters in the numerical filters (which were added to the WRF model for this
purpose). The basic test case is the three-dimensional development of a line
thermal in a moist unstable, no-shear environment. In this case the simulated
flow is composed of thunderstorm cells forming and dissipating along a line.
This study was motivated by our initial attempt to simulate this flow using the
WRF model's 5th-order upstream differencing scheme for the advection terms
along with a widely used 1.5-order TKE scheme. Using this combination of
advection scheme and turbulence-closure model, the solution was characterized
by many poorly resolved grid-scale cells. However, by increasing the
proportionality constant C in the eddy viscosity coefficient, the cells
became well-resolved; but further increases in the constant overly smoothed the
cells. Such solution sensitivity was not noticed in previous cloud models as
they have typically used numerical filters that were responsible for most of
the needed smoothing at small scales. We have performed a series of simulations
with the WRF model using 3rd- and 5th-order advection schemes and, for
comparison with the original Klemp-Wilhelmson cloud model, a 4th-order centered
scheme with a 4th-order filter. We find that, as compared with the scheme
using a 4th-order filter, the upstream schemes using subgrid models with a
viscosity coefficients 1.5 to 2 times larger than those normally used, retain more
power in short scales, but without an unwanted buildup of energy.
Supplementary URL: http://www.mmm.ucar.edu/individual/skamarock/takemi.pdf