Storm-Centered Assimilation using an Ensemble Kalman Filter

A significant challenge to storm-scale ensemble data assimilation is that the observations tend to make analyses of hurricanes more asymmetric than the background forecasts. Non-physical results such as double vortices and displacements far from the mean are apparent in analyses, even with accurate position observations and frequent assimilation. Dynamics often evolve these states toward axisymmetry, so it is difficult to distinguish between real and artificial asymmetries. We propose here a new framework for addressing this problematic signature of the impact of observations in order to investigate the dynamics and predictability of storm asymmetries.

The new framework is tested using an ensemble of cyclonic vortices in a shallow water model. Data assimilation is performed using an Ensemble Kalman Filter (EnKF) in a novel storm-centered framework. Solutions are compared against a control based on a standard EnKF scheme, revealing that vortices in the new framework are more symmetric and exhibit finer inner-core structure. Errors are reduced on average by 50% percent from the control case. Properties of the filtered simulation will be presented, as well as results from analysis of sensitivity to ensemble spread and observation network.