Tuesday, 25 January 2011
Numerical weather prediction (NWP) model predictions are used for a variety of forecasting applications, including driving atmospheric transport and dispersion (AT&D) models. . Ensembles of NWP models both provide a better deterministic forecast as well as quantify the uncertainty of the prediction because the spread in the ensemble predictions is correlated to forecast uncertainty. The NWP ensemble should represent good spread in low-level wind direction and atmospheric boundary layer (ABL) depth to obtain appropriate spread in concentration predictions from AT&D models. To adequately sample the probability distribution function of the forecast atmospheric state, it is necessary to account for several sources of uncertainty, including the initial conditions, lateral boundary conditions, and model physics parameterizations. Limited computational resources typically constrain the size of ensembles, so choices must be made about which members to include when configuring an ensemble.
This study examines an NWP ensemble using the WRF-ARW model. This ensemble varies physics parameterizations for six randomly selected forecast periods in each month of 2009. Various statistical guidance methods are investigated and employed to verify and calibrate the ensemble forecasts, and to down-select a small number of physics configurations. We compare the performance of a small, down-selected, calibrated ensemble with a larger ensemble. Verification focuses on meteorological parameters at the surface and in the atmospheric boundary layer (ABL) that are the most relevant to AT&D modeling. Some of these parameters include low-level winds and temperatures; temperature was chosen because there are not widespread observations of surface fluxes or ABL depth.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner