Comparison of wind speeds obtained using numerical weather prediction models and topographic exposure indices for predicting windthrow in mountainous terrain

Windthrow prediction models require data concerning stand characteristics and wind exposure. Geographic information system databases typically contain elevation, forest cover, and logging history layers, therefore attributes can be extracted for points distributed across a given landscape. Climate stations in forested areas are rare, but the wind regime at regularly spaced points can be estimated using mesoscale numerical weather prediction models such as MC2, MM5, and RAMS. More traditionally, wind exposure is estimated using topographic exposure indices. Using gridded and cutblock edge segment databases for areas of mountainous terrain in central British Columbia (McGregor) and on southwestern Vancouver Island (WIT), we examined the spatial variability of simulated wind speeds and topographic exposure indices, simple correlations between variables, and the utility of these variables in predicting clearcut edge windthrow. Approximately half of the spatial variability in topographic and wind variables occurred for points spaced within 4 km. After restricting the dataset to one point from every 16 km2 panel, mean wind speed was found to be correlated with elevation (0.48, 0.86), but less well with topographic exposure indices (0.17–0.72). Correlations between local winds predicted during strong wind events and topographic exposure indices varied depending on the model used, ranging from non-significant to moderate (0.58). Concordance values for logistic regression models for predicting cutblock edge windthrow improved from 65.0 and 63.8 for base models with height and stand variables, to 70.2 and 68.2 with the addition of topographic exposure indices and extreme wind measures, for McGregor and WIT, respectively.