An effort has been made at improving the currently used wind interpolation method for the TCMWindTool. Currently, the tool uses the modified Rankine vortex model to create a base sustained wind speed grid. After examining several storms and comparing to Hurricane Research Division HWind analyses, it was hypothesized that there are systematic errors in the interpolated wind speeds. In particular, there appeared to be a strong correlation between the interpolated wind speed error and the distance from storm center for each storm quadrant. The research team examined 271 available HWind analyses and calculated error as the difference between the interpolated wind speeds and the analyzed wind speeds. The error was normalized by the estimated maximum sustained wind speeds in the TCM product to allow for inter-storm comparison. For each analysis time and storm quadrant, error was binned in intervals of 5 km distances from storm center out to the 34 knot maximum wind radius and average error was calculated within the bins. Figure 1 shows the resulting average error for all storm analysis times and quadrants. It is clear from the figure that there are indeed systematic errors as a function of storm quadrant and distance from storm center. In particular, there are large positive values in the normalized error from the radius of maximum winds out to a distance approximately 100-150 km from the storm center. This is indicative of the TCM product and interpolation leading to wind speeds higher than analyzed within these regions. For radii outside 150 km from the storm center, the error is much more minimal, with normalized values hovering near zero.
In order to quantify the error as a function of distance from storm center and quadrant, a four degree polynomial fit was applied to the data. Due to the sensitivity in the interpolation to the radius of maximum winds and the likely errors in the HWind analyses near the storm center, data points outside the radius of maximum winds were excluded when developing the error function. The resulting R^2 values were well above .90, indicative of a good fit to the data. Figure 2 shows the averaged wind speed data along with the developed polynomial fits.
Outside of the radius of maximum winds, the error functions were subtracted from the modified Rankine vortex interpolated wind field. Figure 3 shows a sample comparison for Danielle (2008) of the HWind analysis, modified Rankine Vortex wind field, and the new interpolated wind field after the error function was applied to the data. The new wind field displays much more similarity to the HWind analysis. This is particurly true within a distance of 200 km from the storm center, where the wind field is reduced to values much more in line with the HWind analysis. Furthermore, the four quadrant asymmetries are more defined in the new interpolated wind fields. After examining all 271 new wind fields, the error function consistently reveals these features.
As a next step, the research team is currently collaborating with the TCMWindTool developers for potential additions to future versions of the tool. Since the new wind field is simply an additive to the existing interpolated wind field, the incorporation of the improved wind field into the TCMWindTool is not expected to be labor intensive. The goal is in future years to have the various Weather Forecasting Offices with the option to apply the error function to the interpolated wind fields prior to outputting the raw sustained wind speed grids prior to forecaster modification.