An Evaluation of Paired Regional/Convection-Allowing Model-Forecast Vertical Profiles in Warm-Season, Thunderstorm-Supporting Environments

Storm Prediction Center (SPC) forecasters extensively use model-derived vertical profiles of temperature, moisture, and wind to help predict when, where, and whether thunderstorms will develop and, if thunderstorms develop, their potential severity and hazards (e.g., tornadoes, hail, and wind). Historically, output from regional models such as the North American Mesoscale (NAM) and Rapid Refresh (RAP) has been used to derive these forecast vertical profiles. Though these models are unable to explicitly represent thunderstorms, historically, their higher resolution relative to global models and timeliness relative to convection-allowing models has fostered their extensive use in operations starting in the 1990s.

In the last decade, convection-allowing models have proven capable of providing reliable forecasts of convective mode and potential storm severity in the Day 1 forecast period. The added detail that their finer horizontal grid spacing permits is a double-edged sword, however: they also can crudely predict horizontal convective roll circulations within the daytime planetary boundary layer, across which there exists significant variation in boundary layer thermodynamic profiles and thus derived stability parameters. Thus, local variability resolved by convection-allowing models may not be representative of the intrinsically more skillful mesoscale environment of relevance to SPC.

To date, no research has been conducted to isolate the influence of horizontal grid spacing (and accompanying shift from parameterized to explicit representation of deep, moist convection) on short-range (0-24 h) model forecast skill for vertical temperature, moisture, and wind profiles in warm-season, thunderstorm-supporting environments. The outcome of such research has great potential to impact SPC operations, particularly as NOAA moves toward model system unification with FV3: the need for separate regional model configurations is obviated if convection-allowing models are capable of short-range predictions of vertical temperature, moisture, and wind profiles with an identical or greater level of skill.

This motivates an investigation of vertical profiles from two sets of paired regional/convection-allowing models, the NAM/NAM-Nest and RAP/HRRR, focusing on a twenty-five–day period in May 2017 during the NOAA Hazardous Weather Testbed Spring Forecasting Experiment. The research is guided by the hypothesis that convection-allowing models are equally skillful to their regional parent in depicting vertical profiles and associated thermodynamic parameters in warm-season, thunderstorm-supporting environments over a large sample. To large extent, the analysis supports the hypothesis: convection-allowing model-derived soundings are equally skillful, with nearly identical bias and mean absolute error, as their regional parent. However, the analysis also documents shared regional/convection-allowing model biases, which vary as a function of analysis time (1200 UTC vs. 0000 UTC), dynamical core, geographic region, and mesoscale environment (e.g., instability magnitude, presence of a capping inversion, etc.).