Looking beyond Day 2 Forecasts of Severe Convective Weather in the Central US: A Comparison Between the Unified Model Prediction Across Scales (MPAS) and Nested WRF-ARW Approaches for Convection Permitting Models

Over the past decade a number of studies using nested CONUS-scale convection permitting forecasts (CPM), derived from initial and boundary conditions generated from larger-scale hydrostatic models, have demonstrated skill in predicting day 1 (i.e., the 12 - 36 hour time period for 00 UTC forecasts, Sobash 2016) convective storm locations and characteristics (Clark et al. 2012; Schwartz et al. 2015a, 2015b; Sobash et al. 2016).  Until recently, the computational costs associated with CPM resolution has prevented these forecasts to extend beyond 36 or 48 hours on a regular basis.  Importantly, the next generation of global models and computer systems will be capable of regionally employing convection-permitting resolution for the first time operationally.  Investigation into the predictive skill of longer-term CPM forecasts beyond day 1 would determine whether the forecasts have sufficient value to warrant their cost.  Another important issue is to determine whether a systematic skill difference exists between nested-regional versus unified-global modeling approaches, i.e., to determine whether the use of one-way lateral boundary conditions in the nested regional model impacts the forecasts.

The research here seeks to address these two questions.  First, do nested CPMs have any skill in predicting convective activity out to five days?  The predicted parameters include storm locations, timing, and intensity.  Second, does the use of a one-way nested CPM prediction system versus a global multi-scale prediction system with regional CPM resolution alter the skill of these five day forecasts?  These questions will be addressed using NCAR’s WRF-ARW and the Model for Prediction Across Scales (MPAS).  During the past two years, 5-day forecasts using MPAS have been created daily during the peak central U.S. spring-convective season using 00 UTC GFS analysis initial conditions and a variable resolution mesh with a nominal mesh-spacing of 3-km over the CONUS.  To answer the questions of the present study, the WRF-ARW model is used to provide nested 5 day CPM forecasts daily for May 2016.  WRF-ARW is configured similarly to MPAS (i.e., both contain identical physics parameterizations configured to their respective model cores), and is initialized and driven by MPAS boundary conditions over the same CONUS region.  

Using the method of producing probabilistic convective forecasts using updraft helicity as detailed by Sobash et al. (2017) we first access the overall viability and skill from both forecast models for day 1 when compared to the NCAR ensemble CPM system.  We will the examine the decay of forecast skill beyond day 1 from both systems.  Initial comparisons between MPAS and deterministic forecasts from the CPM members of the NCAR ensemble from day 1 suggest that MPAS is competitive with current day 1