Gravity wave phase discrepancies in WRF

The operational NCEP North American Mesoscale (NAM) WRF model, employing the NCEP dynamic core and NCEP physics package, has been consistently producing robust gravity wave signatures containing descent on the upwind side of terrain and ascent on the downwind side. Observed upslope precipitation often occurs where the model indicates strong descent. These waves occur under routine conditions and relatively mild topography, not requiring mountains or strong downslope winds, and seem to reflect real phenomena except that the phase and wavelength appears to be inconsistent with observations and previous modeling studies. The “high-resolution window” runs at NCEP, using an earlier version of the same model run at 5-km grid spacing with explicit convection, produces a similar result. However, NCEP also runs a version of the WRF using the NCAR dynamic core and NCAR physics over the same regional domain with the same input for initial and boundary conditions, at around 6-km grid spacing, and this produces a different, consistent, robust gravity wave signature with a much shorter wavelength and low-level descent beginning at the ridge crest.

This paper will document the wave structures, along with available observations, for several typical cases in different geographic areas.

The cause of these discrepancies is not known. The waves generated by the model should be related to the spectra of the model terrain while observed waves should be related to the spectra of actual terrain. Thus, it might even be possible that the surprising aspects of the solutions from WRF with the NCEP dynamic core could be correct for its topography, though that is hard to reconcile with the same wave structures appearing in 12-km and 5-km runs. The NCAR dynamic core seems to give solutions which might be more useful to forecasters, since the human forecaster's problem is to predict what nature will do regardless of what is or is not a mathematically correct solution to the numerically-posed problem.