As another follow up to Jonathan’s post, David Glenn from NWS Morehead City provided me with Weatherflow data for Sandy (2012). The data included five minute wind speed observations for sixteen sites in the Outer Banks. Per David, none of these sites reported any outages during the storm, providing us with wind speed observations higher than in some of our previous gust factor analyses. The five minute temporal resolution of the data provides us with a more comprehensive wind data set. As in previous analyses, gust factors were only calculated for sustained winds winds of 10 mph or greater. A total of 8,800 gust factors were computed for Sandy from this WeatherFlow data.
The chart below shows a scatter plot of five minute sustained wind speeds in mph versus gust factors for the 8,800 available observations. A logarithmic fit was also added to the plot. As in previous analyses, the chart demonstrates an inverse relationship between the wind speed and gust factor as well as a decrease in the variability in observations as wind speeds decrease.
A few interesting observations:
- Previous regressions indicated an R^2 value ranging from .0005-.4229. The analysis presented here suggested a higher R^2 value of .4595, indicative of a stronger correlation between sustained wind and gust factor. I suggest this can be partially explained based on the locations of the WeatherFlow sites. These sites are mostly located over Cape Hatteras and Ocracoke in the Outer Banks, where flow is primarily maritime with little surface land roughness. As a result, the wind direction and surface roughness contributions to explain gust factor variance are not likely as important in the case of these Outer Banks sites.
- Previous analyses showed very few occurrences of gust factors below 1.2. This current analysis shows gust factor values below 1.2, particularly for stronger wind speeds. Hence, the previously considered asymptotic decay to 1.2 may need to be revised taking into account this data set with stronger wind speeds.
- The mean gust factor was calculated as 1.31. This is in line with what has been suggested in recent studies for coastal stations by the CSTAR TC Winds group.
- The equation for the regression curve (y=-.4595ln(x) + 2.928) is similar to the other storms previously examined, but it tends to be on the higher side of the curves for lower end sustained winds. This is consistent with the results presented in Jonathan’s previous post for Sandy.