Extended Abstract (940K)12.3
Evaluating NWP Ensemble Configurations for AT&D Applications
Jared Lee, Penn State University, State College, PA; and S. E. Haupt, D. R. Stauffer, A. Deng, L. J. Peltier, and J. C. Wyngaard
Atmospheric transport and dispersion (AT&D) models such as SCIPUFF are often driven by output from numerical weather prediction (NWP) models. Ensembles of NWP models are used to assess meteorological uncertainty due to the many sources of uncertainty in atmospheric modeling. The spread in the ensemble predictions can serve as a proxy for forecast uncertainty. Several sources of uncertainty must be represented in order to configure an ensemble that adequately samples the probability distribution function of the forecast atmospheric state, including initial conditions, lateral boundary conditions, and model physics parameterizations. No agreement currently exists on the best approach to varying or perturbing model inputs and options to configure an NWP ensemble. This is partially because what is “best” may be application dependent. For instance, an NWP ensemble configured to produce the best spread for precipitation forecasts is not necessarily the configuration that would provide the best spread for AT&D predictions. To obtain appropriate spread in concentration predictions from AT&D models, there should be good spread in predictions from the NWP ensemble of, in particular, low-level wind direction and atmospheric boundary layer depth.
This study establishes some initial guidelines to determine an NWP ensemble configuration for use with AT&D models. A large ensemble is run for week-long training periods periodically throughout the year for a range of synoptic regimes. Roughly 20-member subsets of this large ensemble are evaluated for their spread in low-level wind direction and temperature, to determine which subset of members should be used in a long-term ensemble for a future study.
Session 12, Mesoscale Air Pollution and Meteorology II
Thursday, 21 January 2010, 8:30 AM-9:45 AM, B308
Previous paper Next paper