Tropical Cyclone-Affected Ocean Surface Winds: A Comparison of Atmospheric Reanalyses With Altimeter, Scatterometer, Radiometer, and Buoy Measurements

Abstract

This study evaluates 1985–2017 ocean surface winds from atmospheric reanalysis products, particularly ERA5-12.5, focusing on 673 tropical cyclones (TCs) in the North Atlantic. Utilizing measurements from ten altimeters, seven scatterometers, 12 radiometers, and NDBC buoys, we perform a comprehensive error analysis through a suite of verification diagnostics involving normalization and successive decomposition. The analysis features detailed comparisons with individual satellite datasets and investigates four factors influencing reanalysis wind speed errors. Our results indicate that traditional interpolation significantly degrades the accuracy of NCEP/NCAR TC wind speeds, while the interpolation of ERA5 (resulting in ERA5-12.5) minimally affects error metrics and increases matched observational data to more than three times. In ERA5-12.5 diagnostics, bias contributes less to overall error than pattern errors, which are mainly driven by phase discrepancies. Additionally, ERA5-12.5 tends to underestimate high wind speeds, with significant underestimation occurring beyond certain wind speed thresholds. Random errors in ERA5-12.5 show minimal variation beyond a threshold distance of approximately 160 km from the TC center; however, they increase sharply as proximity to the TC center decreases. Furthermore, random errors and bias in ERA5-12.5 reveal no clear trends with increasing distance from the coast or over the years.