Evaluation of Contemporary Ocean Wave Models in Rare Extreme Events: The “Halloween Storm” of October 1991 and the “Storm of the Century” of March 1993

View More View Less
  • 1 Oceanweather Inc., Cos Cob, Connecticut
  • | 2 Coastal Engineering Research Center, USAE Waterways Experiment Station, Vicksburg, Mississippi
  • | 3 Florida Institute of Technology, Melbourne, Florida
  • | 4 Environment Canada, Downsview, Ontario, Canada
  • | 5 Oceanweather Inc., Cos Cob, Connecticut
© Get Permissions
Full access

Abstract

Two recent severe extratropical storms, the “Halloween storm” of October 26–November 2 1991 (HOS) and the “storm of the century” (SOC) of March 12–15 1993, are characterized by measurements of sea states of unprecedented magnitude off the east coast of North America. A Canadian buoy moored in deep water south of Nova Scotia recorded peak significant wave heights (HS) exceeding 16 m in both storms. In SOC, a NOAA buoy moored southeast of Cape Hatteras recorded a peak HS of 15.7 m, a record high for NOAA buoys. These extreme storm seas (ESS) exceed existing estimates of the 100-yr estimated design wave in these regions by about 50%. The extensive wave measurements made in both storms from buoys moored in deep water provide a rare opportunity to validate modern ocean wave models in wave regimes far more severe than those used for model tuning. In this study, four widely applied spectral wave models (OWI1G, Resio2G, WAM4, and OWI3G) are adapted to the western North Atlantic basin on fine mesh grids and are driven by common wind fields developed for each storm using careful manual kinematic reanalysis. The alternative wave hindcasts are evaluated against time series of measured HS and dominant wave period obtained at nine U.S. and Canadian buoys moored in deep water between offshore Georgia and Newfoundland. In general, it was found that despite the large differences in model formulation, the hindcasts were almost uniformly skillful in specification of the evolution of wave height and period in these two storms. The skill was much greater than achieved routinely in real time wave analyses provided by some of these same models operating at U.S., Canadian, and European centers, confirming that at least for these particular models, typically large errors in operational surface marine wind field analyses are the dominant source of errors in operational wave analyses and forecasts. However, all models were found to systematically underpredict the magnitude of the peak sea states in both storms at buoys that recorded peak HS in excess of about 12 m (ESS). This bias in ESS wave heights was 3.2 m for OWI1G, 1.9 m for Resio2G, 2.2 m for OWI3G, and 1.5 m for WAM4. These results provide an interesting assessment of the Progress made in the past decade in ocean wave modeling, both in terms of improvements of 1G and 2G models, and the introduction of 3G models. The 2G and 3G models show a slight advantage over the 1G model in simulating the most extreme wave regimes. These results suggest strongly that, for applications where supercomputers are not available, and especially for most operational applications where only integrated properties of the spectrum (e.g., HS) are required or where errors in forcing wind fields are typical of real time objective analyses and forecasts, highly developed and validated 1G and 2G wave models may continue to be used. However, accurate specification of ESS is especially critical for application of wave models to determine the extreme wave climate for ship, offshore, and coastal structure design. Therefore, further study is required to isolate the contribution of remaining wind field errors and model physics and numerics to the underprediction of ESS in extreme storms. The common phenomenological link between these two storms in the regions of ESS appears to be wave generation along a dynamic fetch associated with intense surface wind maxima or jet streaks (JS), which maintain high spatial coherency over at least 24 h and propagate at speeds of 15–20 m s−1. ESS were observed only at those buoys directly in the path of the core of such features. This finding suggests that high-resolution wave models are required to model ESS, but these are justified only if the small-scale JS phenomena can be resolved in operational analysis and forecast systems.

Abstract

Two recent severe extratropical storms, the “Halloween storm” of October 26–November 2 1991 (HOS) and the “storm of the century” (SOC) of March 12–15 1993, are characterized by measurements of sea states of unprecedented magnitude off the east coast of North America. A Canadian buoy moored in deep water south of Nova Scotia recorded peak significant wave heights (HS) exceeding 16 m in both storms. In SOC, a NOAA buoy moored southeast of Cape Hatteras recorded a peak HS of 15.7 m, a record high for NOAA buoys. These extreme storm seas (ESS) exceed existing estimates of the 100-yr estimated design wave in these regions by about 50%. The extensive wave measurements made in both storms from buoys moored in deep water provide a rare opportunity to validate modern ocean wave models in wave regimes far more severe than those used for model tuning. In this study, four widely applied spectral wave models (OWI1G, Resio2G, WAM4, and OWI3G) are adapted to the western North Atlantic basin on fine mesh grids and are driven by common wind fields developed for each storm using careful manual kinematic reanalysis. The alternative wave hindcasts are evaluated against time series of measured HS and dominant wave period obtained at nine U.S. and Canadian buoys moored in deep water between offshore Georgia and Newfoundland. In general, it was found that despite the large differences in model formulation, the hindcasts were almost uniformly skillful in specification of the evolution of wave height and period in these two storms. The skill was much greater than achieved routinely in real time wave analyses provided by some of these same models operating at U.S., Canadian, and European centers, confirming that at least for these particular models, typically large errors in operational surface marine wind field analyses are the dominant source of errors in operational wave analyses and forecasts. However, all models were found to systematically underpredict the magnitude of the peak sea states in both storms at buoys that recorded peak HS in excess of about 12 m (ESS). This bias in ESS wave heights was 3.2 m for OWI1G, 1.9 m for Resio2G, 2.2 m for OWI3G, and 1.5 m for WAM4. These results provide an interesting assessment of the Progress made in the past decade in ocean wave modeling, both in terms of improvements of 1G and 2G models, and the introduction of 3G models. The 2G and 3G models show a slight advantage over the 1G model in simulating the most extreme wave regimes. These results suggest strongly that, for applications where supercomputers are not available, and especially for most operational applications where only integrated properties of the spectrum (e.g., HS) are required or where errors in forcing wind fields are typical of real time objective analyses and forecasts, highly developed and validated 1G and 2G wave models may continue to be used. However, accurate specification of ESS is especially critical for application of wave models to determine the extreme wave climate for ship, offshore, and coastal structure design. Therefore, further study is required to isolate the contribution of remaining wind field errors and model physics and numerics to the underprediction of ESS in extreme storms. The common phenomenological link between these two storms in the regions of ESS appears to be wave generation along a dynamic fetch associated with intense surface wind maxima or jet streaks (JS), which maintain high spatial coherency over at least 24 h and propagate at speeds of 15–20 m s−1. ESS were observed only at those buoys directly in the path of the core of such features. This finding suggests that high-resolution wave models are required to model ESS, but these are justified only if the small-scale JS phenomena can be resolved in operational analysis and forecast systems.

Save