Adaptive Methods in Numerical Weather Prediction

Ensemble Kalman Filtering is a sequential Monte Carlo method commonly used in meteorology to track atmospheric states and make numerical weather predictions (NWP). With the intent to introduce statisticians to this important area of application we address some of the practical aspects of the ensemble Kalman Filter in dynamic systems. We focus on three topics related to NWP: extending the ensemble Kalman filter to handle non-gaussian distributions, using ensembles to reduce storage and to gain computational efficiency, and identifying sources of error due to using sample covariance matrices in the Kalman filter. Although our paper provides a brief overview, we suggest that future research is motivated by this survey.