Abstract
The background error covariance matrix represents a key component of a data assimilation system. It can be shown mathematically that the analysis increments are directly proportional to the covariance matrix. As a result it is correct to state that the performance of a data assimilation system is related to the characteristics of the covariance matrix, in terms of: horizontal and vertical length scales, standard deviations and variances. Considering the information that the data assimilation system uses the observations to correct the model forecasts weighting the model and observations errors, thus the use of an unadjusted covariance matrix can impact the resulting analysis at a great level. At CPTEC efforts has been made in order to adjust the covariance matrix for its application at the operations. In this work it is presented a conceptual overview on the subject, enlightening the importance of the background error covariance matrix and how it is treated in an operational data assimilation system. Furthermore, it is also made a quantitative and qualitative characterization of the background error covariance matrix calculated using the CPTEC global forecast model and the differences in the resulting analysis increments.
Keywords:
background error covariance matrix; data assimilation; numerical weather prediction
