An Estimate of Structural Uncertainty in QuikSCAT Wind Vector Retrievals

Abstract

ifferences between Quick Scatterometer (QuikSCAT) level-2b wind vectors from the Jet Propulsion Laboratory [JPL; the Direction Interval Retrieval with Thresholded Nudging (DIRTH) product] and from the Remote Sensing Systems Co. (RSS; smoothed versions 3.0 and 4.0) for one sample month are presented. Each dataset is derived from the same observations, but processing methods result in differences between wind vectors. These differences originate from 1) uncertainty in the geophysical model functions that relate backscatter to wind, 2) noise in the backscatter measurements, and 3) spatial filtering. Statistics of wind vector differences from RSS and JPL are used as an indication of structural uncertainty in QuikSCAT wind retrievals. When grouped by 1 m s?1 bins, systematic differences are largest beyond 20 m s?1, where wind speeds from version 3.0 (version 4.0) of RSS can be more than 15 (10) m s?1 higher (lower) than JPL wind speeds. Below 20 m s?1, systematic differences on the order of tenths of a meter per second are attributed to differences in the retrieval methods, rain and ice contamination, and cross-swath position. Even once the recommended data flags are applied, differences in individual wind speed retrievals exceed 10 m s?1 in a few cases but are much smaller in regions of the swath for which the viewing geometry allows more reliable retrievals. In all parts of the swath, the standard deviations of the differences are smaller than 1.0 m s?1. The analyses provide a measure of the structural uncertainty in QuikSCAT wind velocity that is due to the retrieval process, although such comparisons are not able to determine which dataset is closest to the actual wind.