Sampling Scale Mismatch Drives Discrepancies between Satellite and TAO Buoy Measurements of Ocean Surface Winds

Satellite (QuikSCAT and ASCAT) and buoy winds measurements at TAO mooring sites generally agree, with 80\% of vector-wind comparisons falling within $\pm 2~\mathrm{m/s}$ and $\pm 25\degree$. We investigate the environmental conditions for the discrepancies in the remaining 20\% of mismatches, identified using Density-Based Spatial Clustering of Applications (DBSCAN). These discrepancies are concentrated in the ITCZ and the western Pacific, regions of frequent precipitation, yet the mismatches are not caused by rain contamination. Instead, the discrepancies are associated with higher relative humidity and sea-air temperature differences, along with lower winds, reduced Stokes drift velocity, and lower wind-wave significant height. The discrepancies also exhibit greater hourly variability in vector-winds, relative humidity, and sea-air temperature difference. Using a random forest model, we demonstrate that these discrepancies are systematically predictable from buoy-measured meteorological variables and wave reanalysis data. The results indicate that differences in the spatial and temporal sampling scales between satellites and buoys primarily drive the discrepancies, which can be slightly reduced by temporal smoothing of buoy winds and spatial smoothing of satellite winds. This implies that most mismatches arise from sampling scale differences rather than deficiencies in either observing system.