| Description |
This dissertation examines Weather Research and Forecasting (WRF) simulations of Great Salt Lake Effect (GSLE) precipitation. An evaluation of banded and nonbanded GSLE-event simulations shows that WRF has low skill predicting GSLE precipitation. An object-based verification method is used in this evaluation to quantify a precipitation bias that contributes to WRF models' low skill. We also analyze WRF simulations of the 27 October 2010 banded GSLE event to evaluate the sensitivity of precipitation prediction to the choice of microphysics parameterization (MP). WRF simulations of 11 banded and eight nonbanded GSLE events are evaluated with subjective, traditional, and object-based verification. Subjectively, a majority of simulations of banded GSLE events produce realistic precipitation features, whereas a majority of simulations of nonbanded GSLE events do not. Simulations of both banded and nonbanded GSLE events record low equitable threat scores, but simulations of banded GSLE events outperform simulations of nonbanded events. Verification using the Method for Object-based Diagnostic Evaluation (MODE) developed by Davis et al. shows that simulations of banded and nonbanded GSLE events exhibit a southward (rightward and downstream relative to the flow) bias in event total precipitation location that limits forecast skill. WRF simulations of the 27 October 2010 GSLE event are sensitive to the choice of MP. Precipitation simulated using the Thompson MP scheme (THOM) verifies best against radar-estimated precipitation and gauge observations. The Goddard, Morrison, and WRF double-moment 6-class (WDM6) schemes produce more precipitation than THOM, with WDM6 producing the most. Analyses of hydrometeor mass tendencies show that WDM6 creates more graupel and total precipitation than the other schemes and indicate that the rate of graupel and snow production can strongly influence the precipitation efficiency in simulations of lake-effect storms. These results show that significant improvements in deterministic model skill and/or the use of an ensemble approach are necessary to improve the reliability of GSLE simulations. Improved deterministic model skill will likely require observations of GSLE hydrometeor characteristics to improve MP, while rectifying the southward (rightward and downstream relative to the flow) precipitation location bias is crucial for deterministic and ensemble forecasting success. |
