Wind and Wave Extremes over the World Oceans From Very Large Forecast Ensembles

TL;DR

This study uses large ensemble forecast data to improve estimates of extreme wind and wave conditions over global oceans, providing more accurate and confident risk assessments than previous methods.

Contribution

It introduces a novel approach utilizing very large ensemble forecasts to enhance the accuracy and confidence of extreme value estimates for wind and wave heights.

Findings

Ensemble forecasts match ENVISAT wind speeds better than ERA-Interim.

Return estimates are higher in extratropics/subtropics than ERA-I but lower than some previous studies.

Confidence intervals are significantly reduced, and non-parametric estimates align well with parametric methods.

Abstract

Global return value estimates of significant wave height and 10-m neutral wind speed are estimated from very large aggregations of archived ECMWF ensemble forecasts at +240-h lead time from the period 2003-2012. The upper percentiles are found to match ENVISAT wind speed better than ERA-Interim (ERA-I), which tends to be biased low. The return estimates are significantly higher for both wind speed and wave height in the extratropics and the subtropics than what is found from ERA-I, but lower than what is reported by Caires and Sterl (2005) and Vinoth and Young (2011). The highest discrepancies between ERA-I and ENS240 are found in the hurricane-prone areas, suggesting that the ensemble comes closer than ERA-I in capturing the intensity of tropical cyclones. The width of the confidence intervals are typically reduced by 70% due to the size of the data sets. Finally, non-parametric estimates of return values were computed from the tail of the distribution. These direct return estimates compare very well with Generalized Pareto estimates.