Abstract
Data assimilation provides the optimal combination of background and observations, where the increments to analysis are weighted by their error settings. Knowledge of the true errors is crucial to get the optimal initial conditions for the numerical models; however, error estimation in data assimilation is not a straightforward problem.
The aims of this study are threefold: to demonstrate the importance of the true observation error in the assimilation scheme, to summarise the main sources of errors and, finally, to estimate the observation error of the Advanced Microwave Sounding Unit-A (AMSU-A) and Microwave Humidity Sounder (MHS) instruments (used by the National Oceanic and Atmospheric Administration (NOAA) and MetOp polar satellites) in the Aladin/CZ model. As a practical matter, we first study the inconsistency between the observation times and the background valid at the analysis time.
As a result, an increase in representativeness error with an assimilation window length was detected. Specifically, the estimated optimal window lengths are two (MHS) and three (AMSU-A) hours long.
Subsequently, based on the method described by Desroziers et al. (2005), the observation errors were estimated. Following the diagnostics, the predefined error setting was found to be overestimated for the both instruments in the Aladin/CZ model.
The most significant overestimation (up to 300 %) was detected for the MHS.