WindRAD Scatterometer Quality Control in Rain

Rain backscatter corrupts Ku-band scatterometer wind retrieval by mixing with the signatures of the σ∘ (backscatter measurements) on the sea surface. The measurements are sensitive to rain clouds due to the short wavelength, and the rain-contaminated measurements in a wind vector cell (WVC) deviate from the simulated measurements using the wind geophysical model function (GMF). Therefore, quality control (QC) is essential to guarantee the retrieved winds’ quality and consistency. The normalized maximum likelihood estimator (MLE) residual (Rn) is a QC indicator representing the distance between the σ∘ measurements and the wind GMF; it works locally for one WVC. JOSS is another QC indicator. It is the speed component of the observation cost function, which is sensitive to spatial inconsistencies in the wind field. RnJ is a combined indicator, and it takes both local information (Rn) and spatial consistency (JOSS) into account. This paper focuses on the QC for WindRAD, a dual-frequency (C and Ku band) rotating-fan-beam scatterometer. The Rn and RnJ have been established and thoroughly investigated for Ku-band-only and combined C–Ku wind retrieval. An additional 0.4% of WVCs are rejected with RnJ, as compared to Rn for both Ku-band-only and combined C–Ku wind retrievals. The number of accepted WVCs with high rain rates (>7 mm/h) is reduced by half, and the wind verification with respect to ECMWF winds is generally improved. The C-band measurements are little influenced by rain, so the Ku-based Rn is more effective for the combined C–Ku wind retrieval than the total Rn from both the C and Ku bands. The rejection rate of the combined C–Ku retrievals reduces by about half compared to the Ku-band-only retrieval, with similar wind verification statistics. Therefore, adding the C band into the retrieval suppresses the rain effect, and acceptable QC capabilities can be achieved with fewer rejected winds.