Mike Brennan directed me to a paper, “Uncertainty and Intercalibration Analysis of H*Wind” (DiNapoli et al, 2012). The paper provides uncertainty estimates of H*Wind analyses, using five Atlantic/Gulf of Mexico hurricanes. After various assumptions are made and some data points and analysis times are removed, the authors describe, “the results of the study are representative of well-sampled hurricanes over open water, which are not undergoing eyewall replacements or other major structural changes. Analyses of landfalling systems, understampled systems, and systems with concentric eyewalls ave additional variability that is beyond the scope of this study.”
Based on the storms analyzed, the results suggest a general positive bias to the wind speeds of approximately 6% near the storm center, increasing to approximately 13% near the region of tropical storm force winds.
A few key points that I think are important as far as our study is concern and use in our NDFD verification:
-Our NDFD verification has systematically shown an overprediction of wind speeds for almost all locations and all storms analyzed when comparing to H*Wind analyses. As presented in this paper, the H*Wind analyses tend to overdo the wind speeds. Hence, the NDFD over prediction may actually be worse than the H*Wind comparison results suggest.
-The H*Wind analyses are typically better in cases for stronger storms, with non concentric eye walls and away from the storm center
-The percentages presented in this paper cannot be assumed in our analyzed storms, since the storms we examined were weaker and near landfall. The errors in the H*Wind analyses for our storms are likely higher.
Overall, I still think H*Wind analyses provide a good method of NDFD verification, although there are certiantly some limitations, as this paper suggests. Since the paper suggests H*Wind analyses are most limited for weaker tropical storms and depresions, perhaps the verification with the CRONOS/ASOS data is more effective for these cases.