Sentinel-3A and -3B satellite SAR altimeter wind speeds are routinely validated against moored buoys but increasingly benchmarked against ERA5 reanalysis when buoys are unavailable. We assess both reference choices over the Chinese nearshore seas using 14 months of matchups between satellite passes, moored meteorological buoys, and ERA5 wind fields. Against buoys, both altimeters show near-zero mean bias and a roughly flat error level across the wind-speed range. Against ERA5, the discrepancy grows strongly and monotonically with wind speed: the mean satellite-minus-ERA5 bias shifts from approximately −1.5 m/s in light wind to +1.4 m/s above Beaufort 6. Sentinel-3A tracks the buoy more closely than Sentinel-3B (r = 0.71 vs. 0.58), a gap that persists in the month-to-month RMSE record. Because the altimeter–buoy residual is flat, the wind-speed structure in the altimeter–ERA5 residual can be attributed to ERA5 rather than to the satellite retrieval. A compact quadratic fit to the altimeter–ERA5 residual, inverted to adjust ERA5 wind speed, removes essentially all of this wind-speed-dependent bias (per-bin mean bias reduced to within ±0.3 m/s), although the remaining scatter is set by representativeness differences that no bias model can remove. These results caution against uncorrected use of ERA5 as a stand-alone reference for altimeter evaluation in light-wind and extreme-wind regimes—over the semi-enclosed Chinese seas studied here, and more generally wherever sparse buoy coverage makes reanalysis the fallback reference for satellite wind validation.
Tan et al. (Fri,) studied this question.