Downslope windstorms are of major concern to those living near the Boulder, Colorado area, often striking with little warning, bringing clear air wind gusts of 35-50 m/s or higher, and producing widespread damage across the city. Models used for forecasting these dangerous events are often not accurate. Hence, there is a need to apply different linear and non-linear statistical modeling techniques to a 10-year mountain-windstorm dataset. A set of eighteen predictors, based on a decade of data, are used in this study. Linear regression, neural networks and support vector models are employed to relate the predictors to windstorm events. For the linear model, stepwise linear regression is applied. It is difficult to determine which predictor is the most important, although significance testing indicates 700 hPa flow is highly significant. The nonlinear techniques employed, support vector regression and a feedforward neural network did not filter out any predictors. The study indicates that there is a potential for improvement in peak wind forecasting using different methods and predictors. The models are evaluated using RMSE and median residuals. The support vector regression model performed best. Stepwise linear regression yielded results that were accurate to within 8 m/s, whereas a neural network reduced errors of 6 to 7 m/s and support vector regression had errors of 4 to 6 m/s. 85% of these forecasts based on nonlinear techniques predicted maximum wind gusts with an RMSE of less than 6 m/s, and all forecasts predicted wind gusts with an RMSE of below 12 m/s. In comparison, a linear model forecast wind gusts better than 6 m/s 60% of the time, and better than 12 m/s 95% of the time. These results suggest that meaningful improvements to mountain wind forecasts are achieved by application of newer non-linear techniques, such as neural networks and support vector regression.
This publication was presented at the following:
Not available
Authors who have authored or contributed to this publication.