[6-3]

Ensemble inflation in hybrid data assimilation

A. El Akkraoui (Science Systems and Applications Inc SSAI) , R. Todling (GMAO Global Modeling and Assimilation office GMAO/NASA)

The Global Modeling and Assimilation Office (GMAO) has recently implemented a 3D-Hybrid data assimilation system based on the square root Ensemble Kalman filter of Whitaker and Hamill (2002). To prevent the ensemble from collapsing after repetitive cycles of analysis updates, inflation is used as a way to boost ensemble spread, and to account for some system uncertainties. GMAO currently uses a combination of a relaxation to prior type of inflation (multiplicative), and an additive inflation that uses random samples of forecast error perturbations (drawn similarly to the NMC method). While the former accounts for sampling errors, the latter is supposed to represent model uncertainty in the ensemble. However, generating additive samples from a climatological distribution has no bearing on actual model errors or errors of the day. A consequence of such inadequacy is reflected in the ensemble spread being over or under represented in some regions and/or vertical levels. Besides, these perturbations are also shown not to grow during the assimilation window. This work will examine some alternative options to induce and sustain growth and discuss preliminary results of utilizing Incremental Analysis Update IAU-type forcing. This effort ultimately aims at replacing the additive inflation all together with stochastically perturbed tendencies.