Automated Detection of Lee Troughs Generated by Alberta Rockies

Lee troughs are important factors in many types of weather phenomena: gap winds, chinooks, severe thunderstorms, and freezing rain, to name a few. A long-term climatology of lee troughs could be very useful in determining how certain properties of lee troughs interact with these phenomena. However, to our knowledge, all papers on lee troughs involve only a small number of cases (usually less than 10, never more than 70) rather than a long-term climatology. This is probably because unlike other features, such as closed low-pressure systems, no reliable algorithm has been developed to automatically detect lee troughs. This motivated us to develop our own algorithm for doing so.

The input data for our algorithm consist of gridded pressure and height fields from the NARR (North American Regional Reanalysis). Currently, our analysis is two-dimensional, which means that it can handle only one input field at a time. So far, we have chosen either MSLP (mean-sea-level pressure) or 850-mb heights. In principle, though, one could choose any input field below about 3 km (~700 mb, above which the signature of lee trough becomes very weak).

For whichever input field is chosen, our algorithm consists of six steps:

1. Smooth the input field, to remove small-scale disturbances. (We are looking for synoptic-scale features, with lengths on the order of 100 to 1000 km.) This is done using a Cressman filter. 2. Find curvatures throughout the smoothed field, using a Lagrangian (or potentially higher-order) operator. 3. Find regions of maximum (positive) curvature in the smoothed field. Each of these regions is stored as a polygon. 4. Find the main skeleton line passing through each region of maximum curvature. Basically, this is a way of projecting each polygon down to a line. The main skeleton line is considered a trough axis (and potentially lee-trough axis). 5. Consider each pair of trough axes in the map. If the two axes have a pair of nearby endpoints which could be joined with a line having a similar angle to the two axes, these endpoints are joined. 6. For each trough axis, certain criteria are applied to determine whether or not it should be considered a lee trough. (First, the “main ridge” of the Alberta Rockies is defined.)

a. Distance between the trough axis and main ridge must fall below a certain threshold. b. Angle of the trough axis must be similar to the angle of the main ridge (approximately NNW to SSE). c. Values of the input field (either pressures or heights) must increase consistently “down” the main ridge, from NNW to SSE. d. The trough axis must exceed a certain length (again, we are looking for synoptic-scale phenomena).

So far, this algorithm has been very successful in our limited domain, downstream of the Alberta Rockies. However, the algorithm is highly parametric, so even in this domain its performance could probably be improved by more optimization of parameters. (We plan to use a genetic algorithm for this purpose.) Furthermore, if time allows, we would like to generalize this method so that it can be used in other domains. We hope that researchers will be able to use this algorithm to perform long-term climatologies of lee troughs.