Presentation on theme: "Toward Improving Representation of Model Microphysics Errors in a Convection-Allowing Ensemble: Evaluation and Diagnosis of mixed- Microphysics and Perturbed."— Presentation transcript:
Toward Improving Representation of Model Microphysics Errors in a Convection-Allowing Ensemble: Evaluation and Diagnosis of mixed- Microphysics and Perturbed Microphysics Parameter Ensembles in the 2011 HWT Spring Experiment Jeff Duda, Xuguang Wang, Fanyou Kong, Ming Xue School of Meteorology and Center for Analysis and prediction of storms University of Oklahoma, Norman, OK Acknowledgements: Dan Dawson (NSSL), Kevin Thomas (CAPS), Keith Brewster (CAPS), Yunheng Wang (CAPS) Warn-on-Forecast and High Impact Weather workshop, Norman, OK, Feb. 8-9, 2012
Introduction Sources of error in an NWP forecast: – IC/LBCs – Model error Dynamics Physics Methods to account for microphysics errors – Multiple microphysics – Perturbed parameter within a single microphysics scheme
Background NOAA HWT 2011 spring experiment 52 member storm-scale ensemble forecast system (SSEF) run by CAPS at OU – ∆x = 4 km (no convective parameterization) Once-daily forecasts out to 36 hours – Initialized at 0000 UTC Use of 3DVAR and cloud analysis to assimilate radar data at initialization 35 forecasts from 27 April to 10 June
Setup Focus on two sub-ensembles – Mixed microphysics (six members) Thompson (control) Ferrier+ (modified for NMMB) Milbrandt-Yau (new for WRF v. 3.2) Morrison WDM-6 WSM-6 – Perturbed parameter (five members) WSM-6 and four sets of perturbations N 0r, N 0g, and graupel/hail density perturbed Unperturbed values: N 0r =8.0 x 10 6 m -4, N 0g =4 x 10 6 m -4, graupel density = 500 kg m -3 NameN 0r N 0g ρ graupel (kg m -3 ) WSM6 (ctrl)8 x 10 6 4 x 10 6 500 WSM6-M18 x 10 6 4 x 10 4 913 WSM6-M28 x 10 7 4 x 10 6 500 WSM6-M38 x 10 5 4 x 10 2 913 WSM6-M48 x 10 5 4 x 10 3 913
Outline of results Measures of skill for the two sub ensembles and their combination using various metrics Understand the difference among double moment microphysics schemes and single moment scheme with perturbed parameters for a case study using equivalent intercept parameters.
Summary of measure of skill Perturbed parameter has larger spread, but performed worse than mixed-microphysics Both subensembles (and their pooled combination) under dispersive for precip; also have a slight positive bias Pooled ensemble better than both sub-ensembles for some metrics – Perhaps due only to larger ensemble size Appropriate choices of parameter values?
Equivalent intercept parameter for double moment schemes Double moment vs. single moment microphysics – Number concentration prognosed as well as mass mixing ratio – Given assumed particle size distribution (generally a gamma distribution), can diagnose the value of the intercept parameter N 0 Loss of physical meaning for N 0 for non-zero shape parameter distributions Testud et al. (2001): normalized intercept parameter – For a fixed water species mass in a grid box, higher N 0 smaller particles greater surface area more evaporation, stronger cold pools
Conclusions Some consistency between N 0r and cold pool strength, accumulated precip – More analysis needed on N 0g Selected N 0r values for perturbed parameter sub-ensemble seem appropriate Evidence of size-sorting of raindrops Future work – Evaluate other parameters Storm propagation speed – Evaluate other case studies – Include object-based evaluation of QPF (MODE)