Ensemble variability and predictability of mountain waves and downslope winds

With the advancement of numerical weather prediction (NWP) models,

forecasting mesoscale phenomenon such as downslope wind storms and

mountain waves is becoming common place. While these models are

capable of producing realistic and plausible solutions, their ability

to produce valid a-priori forecasts of detailed mesoscale structures

is largely unknown. Cases from the TREX field campaign provide

several opportunities to investigate the predictability of such flows.

In this study we use an Ensemble Kalman Filter to generate a 70 member

ensemble at horizontal resolutions as fine as 1 km. Several

downslope-wind and mountain-wave events from the TREX campaign are

considered. The ensemble is used to demonstrate the variability

possible in a mesoscale forecast given typical synoptic-scale

uncertainty.

Despite relatively small synoptic-scale variability differences in the

downslope-wind response among the ensemble members can grow quite

large. For example, the difference in downslope-wind forecasts

between composites of the 10 weakest and 10 strongest ensemble members

exceed 30 m/s in a short 6-12 hour forecast. At the same time the

differences between the average crest-level wind speed in the strong

and weak composites is less than 5 m/s. Thus, at least in these

cases, the topography increases error growth and decreases

predictability. In addition to downslope-wind predictability, the

ability of models to accurately forecast mountain waves is explored.

Aircraft observations are used to verify individual ensemble members

for several events during TREX. While some members compare well with

observations, most of the members contain large errors in both

amplitude and phase. This suggests an inherent lack of predictability

in a-priori mountain-wave forecasts.