Ensemble approaches to mesoscale predictability and dynamics

Ensemble forecasting is a well-established method for quantifying the loss of predictive skill with increasing lead time due to errors in the initial conditions and model. The emergence of ensemble-based data assimilation provides a way to populate these ensembles with initial conditions that reflect state-dependent analysis error, which is particularly challenging on the mesoscale using other methods. Moreover, the fact that the ensemble approach does not require adjoint models or variational minimization algorithms, promotes data assimilation as a tool used by individual investigators. In addition to reviewing the merits of ensemble-based data assimilation for mesoscale prediction, I will discuss methods employing ensembles for sensitivity analysis, observation impact studies, and dynamical diagnosis. Examples illustrating these methods will be given for tropical cyclones and extratropical transition.