A satellite radiance-bias correction scheme for data assimilation

Authors


Abstract

Recently, much progress has been made in the direct assimilation of satellite radiance measurements in numerical weather-prediction systems. In order to use radiances from the TIROS Operational Vertical Sounder (TOVS), biases between the observed radiances and those simulated from the model first guess must be corrected. The original scheme for TOVS radiance-bias correction at the European Centre for Medium-Range Weather Forecasts utilized a global scan correction, and a linear air-mass correction, with the observed radiances from the Microwave Sounding Unit channels 2, 3 and 4 as predictors. The new scheme differs in two fundamental ways. Analysis of radiance data shows a significant residual scan bias which depends strongly on latitude for some channels. The new scheme applies a latitudinally dependent scan correction to take this into account. The air-mass predictors are now computed from the background field, since the background field contains a more consistent representation of the air mass and surface characteristics than the observed microwave radiances. Four new predictors are used, 1000–300 hPa thickness, 200–50 hPa thickness, model surface skin temperature and total precipitable water. In particular, the skin-temperature predictor is able to differentiate between ocean and sea-ice, performing much better than the old scheme in the winter hemisphere. The use of model predictors is a change in philosophy away from correction of the observations to correction of the computed forward radiances. This leads to a natural extension where the gradient of the bias correction can be taken into account in variational retrieval schemes.

Ancillary