Key points are not available for this paper at this time.
The development and validation of altimeter wind speed algorithms is investigated following the collation of the largest dataset to-date of coincident altimeter/buoy open ocean measurements. Nonlinear relationships between buoy wind and Topex backscatter are fitted to the 4500 points dataset using least-squares (LSQ). The addition of altimeter significant wave height (SWH) information causes a small but significant reduction of about 10% in root-mean-square (rms) error. The new LSQ algorithms yield significant improvement of the global wind speed bias and rms error compared to earlier models, but describe the wind to backscatter relationship poorly at extreme wind speeds. Best results are obtained with the Gourrion et al. (2000) model, improving on the Witter and Chelton (WC91) (1991) model used operationally. A residual dependence on sea state persists in all wind algorithms, which underestimate winds in young sea conditions on average by 1-1.5 m/s. A case study confirms that ordinary LSQ attribute excessive weight to the peak of the wind speed histogram and yield algorithms with poor performance at extreme winds. Measurement errors are shown to greatly influence the fitted models performance, as accounting for normally distributed errors in both altimeter and buoy measurements with orthogonal distance regressions (ODRs) yields significant improvements.
Gommenginger et al. (Tue,) studied this question.