146 An Evaluation of Vertical Thermodynamic Profiles and Derived Stability Parameters from Parallel FV3- and Spectral-Model GFS Forecasts

Monday, 13 January 2020
Hall B (Boston Convention and Exhibition Center)
Dillon V. Blount, Univ. of Wisconsin–Milwaukee, Milwaukee, WI; and C. Evans, I. L. Jirak, and A. R. Dean

In June 2019, after several years of testing and development, NCEP upgraded its Global Forecasting System, or GFS, model with the finite-volume cubed-sphere FV3 dynamical core. Prior to this model upgrade, an extensive parallel evaluation of spectral- and FV3-based GFS forecasts was conducted. Similar to other NCEP operational-model evaluations, metrics such as 500 hPa anomaly correlation, fields such as accumulated precipitation, and features such as extratropical and tropical cyclones at short- to medium-range (e.g., 12-120 h) lead times were used to characterize model skill or performance. For such evaluation methods, retrospective and real-time testing have indicated that the FV3-based GFS is equally or slightly more skillful to the spectral version of the GFS it replaced.

Both the spectral- and finite-volume versions of the GFS parameterize turbulence, including mixing within the planetary boundary layer, using a hybrid eddy-diffusivity mass-flux formulation, as described by Han et al. (2016, Wea. Forecasting). In this formulation, local mixing by small eddies under stable conditions is represented via eddy diffusivity, mixing under weakly unstable conditions is represented via a nonlocal countergradient mixing term, and mixing in convective boundary layers is represented via a mass-flux formulation. In convective boundary layers, such as are common in the daytime over land during the local summer months, the mass-flux formulation generally results in deeper vertical mixing (through greater entrainment atop the boundary layer from the free atmosphere) as compared to the countergradient mixing formulation.

However, as highlighted by SPC forecaster evaluations of GFS output, a long-standing issue with both approaches to vertical mixing under unstable conditions in the GFS is the model's tendency to predict too deep of vertical mixing in the central and western United States. This results in forecast surface conditions that are too warm and dry, convective available potential energy for near-surface-based parcels that is too low, and drylines (which propagate due to turbulent vertical mixing) that propagate too far to the east during local daytime hours. Though the FV3- and spectral-based GFS performed similarly in pre-implementation parallel forecasts for forecast parameters of importance to SPC operations, the low dewpoint, low instability, and overmixing biases noted above were somewhat worse in the FV3-based GFS for the cases considered. More investigation is necessary, however, to better understand FV3 performance characteristics at short- to medium-range lead times in the severe- and fire-weather environments for which SPC has lead forecast responsibility.

In this study, model-forecast vertical thermodynamic profiles (soundings) and derived stability parameters (e.g., convective available potential energy, convective inhibition, etc.) from parallel 2019 spectral- and FV3-based GFS forecasts across the United States are verified against available rawinsonde observations in environments uninfluenced by active or recent precipitation. The goals of this study are two-fold: to better inform SPC forecasters' operational use of GFS guidance (e.g., are there certain environments or parameters for which it performs sufficiently well?) and to provide insight regarding specific model strengths and weaknesses to model and turbulence parameterization developers. Accordingly, the results are composited by environments and locations of interest to SPC operations; e.g., observed instability exceeding specified thresholds, observed surface relative humidity below specified thresholds, and for specific geographic regions. Full results will be presented during the meeting.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner