The implications of accounting for observation error correlations on network design.
Many observation types assimilated in NWP are known to have correlated errors. Recent advances in estimating observation error correlations (OECs) means that a number of operational NWP centers are now accounting explicitly for them during their assimilation. One consequence of this is that it reduces the need for excessive thinning of the observations, meaning that a much larger proportion of the observations can be assimilated. This may be particularly beneficial for high-resolution DA, as it is known that observations with positively correlated errors are more informative about small-scale structures than observations without correlated errors. Hence a higher density of observations with significant OECs are necessary to utilize the small-scale information. This talk will focus on the implications of accounting for OECs on the design of optimum observing networks. Particular attention will be given to the sensitivity of such optimum networks to the accuracy of the estimated observation error covariance matrix, and suggest ways to mitigate the negative effects of known inadequacies within this matrix.
Dr. Alison M. Fowler University of Reading