Evaluating the Predictability of Gusts Using Durst Curves: A Historical Statistical Approach with Modern Observations

Gusts represent the maximum wind within a period and can pose a significant hazard to wind energy, fire weather, and engineering. Unfortunately, gusts are not resolvable in current operational weather models due to their limited spatial and temporal resolutions, resulting in the development of a wide range of gust parameterizations to estimate wind threat. Each method has its benefits and drawbacks, but many are computationally expensive and unable to diagnose gusts of differing duration which can be critical for applications such as engineering. The Durst Curve, or more colloquially gust curves, are a simple historic tool for estimating gusts of various durations based on empirical observations. Gust curves relate the ratio of gusts for a range of durations to the hourly mean wind. Assuming the mean wind can be estimated from an available forecast the maximum gust of the desired duration within that hour can be determined. However, the Durst Curve is based on incredibly limited data from the 1930's, relies on multiple statistical assumptions, and many not be representative of most environments.

The goal of this study is to leverage a large collection of modern wind observations to create a variety of gust curves, assess their variability, and their potential use as a wind threat estimation tool. We use multiple years of wind data from observational networks across the CONUS that report data at a sufficiently high frequency: Automated Surface Observing System (ASOS), New York State Mesonet (NYSM), and High Performance Wireless Research and Education Network (HPWREN). Collectively these constitute over 1000 surface stations across the CONUS allowing us to evaluate a wide variety of weather conditions and local environments. Using the high frequency wind data, we were able to super sample observations to calculate gusts of differing durations and the hourly mean wind to recreate gust curves every hour over several years. Our results show that gust curves vary widely not only between locations but overtime at a single location. This variability is frequently not distributed normally about the mean making utilization of an average or central tendency curve difficult. Some locations also exhibit a bimodal distribution in gust curves depending on the predominant wind direction and the roughness or fetch upstream. We also observe the spread of gust curves is greatly reduced during observed high wind conditions, increasing the reliability of gust estimates when estimation of wind threats are most desirable. However, while gust curves cluster more tightly under these conditions their average values still vary greatly between locations within the same network. Additionally, this convergence is reliant on restricting the variance of wind direction during high wind events to ensure gusts and mean wind originate from similar upstream directions. Thus, gust curves are seen as unreliable for periods of rapidly changing wind direction or for intense gusts during periods of relatively calm wind.