Estimates of short‐range forecast‐temperature error correlations and the implications for radiance‐data assimilation

The importance of an accurate specification of short‐range forecast‐error structures for the successful assimilation of satellite sounding data is discussed, with special emphasis on the role of vertical (inter‐level) error correlations. Two different statistical approaches for estimating these correlations are examined. The first is based on comparing radiosonde profiles with the short‐range forecast. The second compares forecasts of different lengths which verify at the same time. It will be shown that, in some aspects, the two approaches produce consistent results which suggests that significant geographic and seasonal variability exists in the vertical error correlations. The effect of not modelling this variability in the analysis of TOVS (TIROS‐N (Television Infra‐Red Observation Satellite) Operational Vertical Sounder) radiance data is simulated using a linear retrieval estimator, and it is shown that it may result in a serious misinterpretation of the information in the observations.