An objective approach to modelling biases in satellite radiances: Application to AIRS and AMSU-A

Bias correction is a practical necessity for the assimilation of satellite radiances in numerical weather-prediction systems. However, the parametric form used to represent the observation bias often relies on predictors that are chosen subjectively. This paper proposes two diagnostics to assess the relevance of every predictor, derived from comparison of two approaches to estimating corrections based on the same parametric form. These diagnostics are applied to AIRS and AMSU-A, and the predictors used in the ECMWF operational bias correction are studied. The relevance of a correction to the radiative-transfer-model absorption coefficient is also investigated. Finally, a method to objectively construct the bias parametric form, starting from a large set of potential predictors, is proposed.