Other Model Simulations
National Ensemble
National WRF
Model Details and Status
Runtime Status
Ensemble Formulation
Ensemble Goals
Ensemble Examples
Model Domain
Acknowledgments
Cluster Machine Status
Newest Output Fields
Added: 10 November 2009
Multi-Panel Output Fields
Standard Deviation Fields
FSU Southeast US WRF Ensemble Simulation Goals
Overview
The twice-daily WRF Ensemble Suite is designed to test the short-term sensitivity of seabreeze, nocturnal, and continental convection to an array of physical parameters and initialization differences. A control run using the exact same setup as with our 0000 UTC 72 hr simulations is first performed, followed by a suite of seven additional model simulations. Differences from the control run are manifest in initial and boundary conditions; planetary boundary layer (PBL) parameterizations; cumulus convective parameterizations; and amalgams of the previous contributions. Lessons learned from the 2006 UCAR/ASP Summer Colloquium on "The Challenge of Convective Forecasting," where case studies using different physics packages were evaluated to understand how convective simulations differed between them, and over three years of real-time mesoscale modeling and analysis are meshed to provide the motivation behind such this study.
Differences Between Ensemble Members
The simulations performed here diverge from one another in time due to impacts of the initial/boundary conditions ingested into the model, physical considerations owing to the methods of mathematical closure methods in the planetary boundary layer, physical considerations owing to vertical moisture profile adjustments in the presence of shallow convection, or a non-linear combination of two or more of these factors. As compared to a more traditional ensemble formulation, where differences in each member arise due to mathematical initial condition perturbations, the majority of the differences in the forecasts presented here are related to physical considerations. As such, with a knowledge of the strengths and limitations of the parameterizations employed to obtain the simulations' physical fields, a forecaster can more completely interrogate and apply the output from these simulations to their short-term forecast products. The details behind said strengths and differences are presented below.
Convective Parmaterizations
Two of the chosen convective schemes, Kain-Fritsch and Betts-Miller-Janjic, use mass flux and static profile convective adjustment schemes respectively. These primarily impact shallow convection but have larger-scale impacts upon long-term convective evolution as well. The Kain-Fritsch scheme searches from the bottom-up to find a parcel's LFC and cloud top, assigning a mass flux profile from there, and compensating for this mass flux via external downdrafts. The Betts-Miller-Janjic scheme uses observed and/or easily definable profiles to prescribe a shallow convective state to the simulated atmosphere, often times involving mixed profiles. Both schemes redistribute heat and moisture to conserve mass. From a forecaster's perspective, the Betts-Miller-Janjic scheme often turns on too often in environments with ample CAPE but that cannot support deep convection, leading to a bias toward convective development that may result in excessive QPF and resultant impacts on forecast temperature, wind, instability, and moisture fields; the Kain-Fritsch scheme does not suffer from this shortcoming. Two of the members use the Grell-Devenyi ensemble convective parameterization; physical insight into its differences will be provided here soon.
Planetary Boundary Layer Parameterizations
The selected boundary layer schemes, Yonsei University and Mellor-Yamada-Janjic, differ in how they close the mathematical formulations of turbulence processes (eddies, diffusion, and the like) within the PBL. The Yonsei Univ. scheme uses non-local eddy mixing triggered by surface heating, while the Mellor-Yamada-Janjic scheme uses local mixing. From a forecast perspective, each has its advantages and disadvantages, leading to the Mellor-Yamada-Janjic scheme often resulting in too shallow of a boundary layer and too moist in the PBL -- e.g. most of its negative influences lie in the PBL itself -- and the Yonsei Univ. scheme often having too shallow of a convective inhibition layer and thus being too moist above the PBL -- e.g. most of its negative influences lie at the top of the PBL.
Initial Conditions
GFS and NAM initial and boundary conditions are used to initialize members 1-4 and 5-8 of the ensemble formulation, respectively. While the NAM initial and boundary conditions are natively available at higher resolution than such files for the GFS, they are generally considered to be of lower quality overall, particularly in terms of their impacts upon forecasts beyond 36-48 hr. This occurs despite both models using the same data interpolation scheme to provide initial conditions to each model. Thus, primary differences in these fields lie in the data that is assimilated into their analysis packages; the GFS generally has more satellite-based data and more data overall due to a longer assimilation window while the NAM generally has more surface-based data but less data overall due to a shorter assimilation window. Other differences arise due to the first guess used in the data assimilation and model initialization process. As a forecaster, how these translate into differences in the forecasts shown here is highly non-linear and will vary from simulation to simulation.
Microphysical Parameterizations
Experiences with the larger domain WRF simulations performed at 0000 UTC daily led us to test using the WSM-6 scheme in place of the Lin et al. microphysical parameterization in two of the ensemble members. Specifically, the combination of the Lin et al. scheme plus the BMJ and Kain-Fritsch convective schemes at 15-25 km horizontal grid spacing was found to lead to overestimates of light precipitation and overintensification of tropical cyclone-like features over warm waters. Preliminary tests using the WSM-5 scheme show much improvement upon the Lin et al. scheme in these areas from the 0000 UTC larger domain simulations. In the ensemble formulation, we use the WSM-6 scheme, equivalent to WSM-5 except for the allowance of graupel, to maintain better consistency with the Lin et al. parameterization.
Non-Linear Combinations, Limitations, and Future Plans
Tests are performed with these changes separate and together as we do not expect that they will act entirely linearly with or independent of one another. Do note that there still exist limitations in this ensemble modeling system, akin to those that exist with any numerical modeling system, and tie in here to the physical packages used in the WRF-ARW model and the horizontal and vertical resolution of the model. Though this modeling system is still in its infancy, future plans include analyzing the differences between the simulation output both on a holistic scale and with specific convective features. Keeping the differences between the schemes in mind, it is hoped that these studies -- whether over the Southeast US or elsewhere -- will provide for the better understanding of convective phenomena and their evolution.
Overarching Goals
Currently, few operational centers run a suite of mesoscale models -- the SREF by NCEP is one such example -- and even fewer non-operational centers run an ensemble suite of real-time mesoscale models for public consumption and private research. One of the goals of these simulations is to provide such a framework across the southeast United States -- particularly Mississippi, Alabama, Georgia, and Florida -- for both operational forecasting and research purposes. Doing so will allow local NWS offices, broadcast and private sector meteorologists, and the research community to better analyze and understand the factors influencing convection across the region.
One of the unique aspects about the southeast United States is that multiple modes of convection are found on a daily to annual basis. Cold season convection ranges from stratiform rains to disorganized multicell complexes to supercell thunderstorms and mesoscale convective systems. Warm season convection ranges from the ambiguous "air mass convection" to seabreeze and landbreeze circulations to tropical cyclones and their convection. No other region on this continent sees such a wide variety of convection on an annual basis. It is hoped that a better understanding and predictability of these elements may come from these forecasts.
An ensemble of mesoscale model simulations is chosen here for three reasons: one, to contrast to the daily 0000 UTC simulations performed here; two, to provide for a better understanding of the regional biases inherent to each set of parameters chosen; and three, to provide for the development of a means of estimating uncertainty in short-term numerical weather prediction of convection and elements that lend themselves to convective development or supression. The hope behind starting these simulations is to encourage collaboration between researchers and the operational community. To that end, we openly solicit and request recommendations for model parameters -- both raw output and ensembled (mean/standard deviation) output -- and additional ensemble members. Bringing these communities together will hopefully provide for better collaboration on this and other future endeavors between research and operational forecasters.
© 2006-2009, Clark Evans. Disclaimer: These forecasts are experimental and NOT official forecasts. As with any model, these forecasts are prone to large forecast error. Please refer to official NWS forecasts for the latest weather information.