Forecast Sensitivity of Waves and Submeso Motions to Initialization Strategy and PBL Physics in the Stable Boundary Layer

The need for a combined deterministic and non-deterministic verification strategy for the study of the stable boundary layer and development of improved model physics for hazard predictions is demonstrated. Terrain-induced gravity waves can generate intermittent bursts of turbulence by enhancing submeso fluctuations (i.e., fluctuations on scales from minutes to tens of minutes), hypothesized to be important for hazard prediction in the stable boundary layer (SBL). The sensitivity of sub-kilometer (0.444-km horizontal grid spacing) and high vertical resolution Weather Research and Forecasting model (WRF) forecasts to the choice of three initialization strategies and four planetary boundary layer (PBL) turbulence-physics schemes is examined for six cases characterized by terrain-induced gravity waves and submeso motions over Central PA. Three initialization strategies are tested: a 12-h forecast cold-started from GFS analyses at 0000 UTC; a 24-h forecast initialized from GFS 12 h earlier; and a 12-h forecast preceded by a 12-h, pre-forecast, data-assimilation period that includes local observations. The pre-forecast data assimilation configuration is used to test four PBL parameterizations: local, TKE-based (MYJ, QNSE, and MYNN) and nonlocal, k-profile (YSU). Model forecasts are verified against high-frequency observations from a special data network located at Rock Springs, PA. The verification of deterministic (> 2 h) scales is conducted using standard statistical approaches. For the verification of non-deterministic, submeso motions, a new stochastic verification technique based on the wavelet transform and wavelet decomposition/reconstruction is developed. It uses the wavelet transform as a band-pass filter to obtain 1) the spectral contribution of each fluctuation and 2) the reconstructed time-series at each frequency interval. The number of fluctuations at each frequency interval is then used to quantify the variability within each time series, thus allowing a detailed quantification of the fluctuations in the observations versus in the modeled fields at multiple temporal scales. This new stochastic scale evaluation is shown to facilitate the verification of submeso variability independent of time and phase error while accounting for the scale and amplitude of the fluctuations. The use of data assimilation in a pre-forecast period is shown to have an advantage for deterministic temperature predictions, and for near-surface non-deterministic predictions with periods greater than 22 minutes for temperature and 43 minutes for wind. TKE-based local schemes have some advantage over the YSU nonlocal PBL scheme for near-surface deterministic winds and for the non-deterministic temperature modes. Overall, similar wave regimes (e.g., trapped waves) and underlying circulations (e.g., rotors) are resolved independently of initialization strategy and/or PBL physics. However, improvements in the forecast of wave-turbulence interactions in the SBL are noted when coupling data assimilation with TKE-based PBL physics.