Note this is a multi-part review of Arthur, with the focus in this post on the comparison of forecasts with HWind data.
On 3 July through 4 July, 2014 Hurricane Arthur impacted portions of eastern North Carolina with strong winds and very heavy rain as it moved northeast along the Carolina coast. Arthur’s track took the storm northeast, largely parallel to the South Carolina coast and then across the far eastern portion of North Carolina including much of the Outer Banks. Given this path, the strongest winds from Arthur were confined to the eastern Coastal Plain and especially the coastal region of North Carolina (Fig. 1). To evaluate the performance of CSTAR related research to operations activities, we desired to objectively verify the NDFD forecasts with a reliable surface wind analysis product.
Bryce Tyner from NC State obtained Real-time Hurricane Wind Analysis System (HWind) surface wind analyses from Dr. Mark Powell of Hwind Scientific for Hurricane Arthur. You can learn more about HWind product at this URL: http://www.hwind.co/About/Basis-of-Hwind. Bryce compared the HWind data to the NWS NDFD wind forecasts and the National Hurricane Center (NHC) TCM forecasts. Images consisting of various comparisons of the three products were produced for the synoptic times between 12 UTC on 3 July 2014 and 12 UTC 4 July 2014. The images were arranged in a 6-panel display with the HWind analysis in the top left, the NDFD 7-hour forecast in the top center, and the NHC TCM product in the top right. Difference images were constructed on the bottom row with the TCM product – HWind difference in the bottom left, the NDFD – HWind difference in the bottom center, and the NDFD – TCM product difference in the lower right. Five, 6-panel images are shown below, valid every 6 hours.
The comparative imagery below is somewhat noisy leading to a more difficult interpretation, but in general, both the NDFD and TCM bias become more positive and increases with time. Across eastern North Carolina, the HWind – NDFD difference product shows a general over prediction of the sustained winds in the NDFD, especially at some distance from the storm center with the bias typically in the 8-16 kt range (Figs. 4e and 5e). These errors appear most commonly near the outer edge of the 34 kt wind radii provided in the TCM guidance. The HWind – NDFD difference product show a complex pattern near the storm center (Figs. 4e and 5e) which is not surprising as differences in the forecast vs observed center location can lead to considerable errors. Verification of the 1-hr NDFD forecasts (not shown) showed smaller and less complex error near the circulation center. The TCM wind product shows an even larger over prediction over land, particularly in the western semi-circle (Fig. 5d). While the over prediction is not desirable, it should reduce some of the concern that the new CSTAR methodology might be too aggressive and result in wind speeds that are too low and conflict with tropical cyclone watch/warning products.
There is a lot of information to digest here and we would welcome any comments or interpretations on the analysis presented here.