There are three recognized meteorological regimes that can produce strong downslope winds: (1) a strongly stable layer near mountain-top level below a layer of lower stability, (2) a large amplitude wave that breaks thereby self-inducing a critical level, and (3) the presence of a critical level in the mean flow. The sensitivity to initial conditions of windstorms in the first two regimes has been previously explored, with results suggesting low predictive skill at time scales of as little as 6-12 hours. However, the predictability of the third regime has not been explored using the same methodology, and there are some indications that windstorms in this regime may exhibit greater predictability. Here we compare the predictability of downslope winds developing in the presence and absence of mean-state critical levels by comparing the Santa Ana wind events in Southern California with chinook winds along the eastern slopes of the Front Range of the Colorado Rockies. Santa Ana winds are approximately northeasterly winds that often possess a mean-state critical level. In contrast, mean-state critical levels are seldom present during chinook winds.
The relative predictability of several Santa Ana and Chinook events is assessed by comparing the ensemble spread at different forecast lead times using the NCAR Ensemble Forecast dataset. The NCAR ensemble consisted of 10 WRF-ARW members. The outer nest of these simulations had a horizontal grid spacing of 15 km, and the inner nest, which covered the entire contiguous United States, had a horizontal grid spacing of 3 km with 40 vertical levels. The ensemble was run for a period of approximately two and a half years from 7 April 2015 to 30 December 2017. Several Santa Ana and Front Range windstorm events occurred during this timeframe. This includes a Santa Ana event that developed in conjunction with a deep upper-level trough, without a mean-state critical level.