Short-term prediction of downburst winds: A double-step modification enhanced approach

Abstract

The downburst wind is an extreme wind event with rapidly increasing and afterwards descending wind velocities in a short period. The prediction of downburst winds, especially for the peak wind velocity and its occurrence time, is a critical issue in engineering communities. In this paper, typical commonly used time-series models are firstly employed to predict the downburst winds based on measured data, and the measured results are found highly underestimated with a prominent time lag. According to the faced problem, a double-step modification enhanced approach is developed to improve the prediction accuracy of time-series models. Central to this approach is the dual modifications, in which the first modification remedies the underestimation by correcting the error linearly correlated to the measured wind velocity, while the second modification compensates the residual error negatively proportional to the wind velocity after the first modification. The developed approach is then compared with other time-varying and nonlinear forecasting models, and the dual modifications are found to be able to highly improve the prediction accuracies of these models. Finally, the efficacy of the developed approach is further verified via case studies concerning the one-step and multi-step ahead predictions. The satisfied predicted results prove the effectiveness of the developed approach in the short-term prediction of downburst winds. Thus, the double-step modification enhanced approach can be used for early-warning purposes.