Nonlinear and Non-Gaussian

p2-16 January 22 14:40-15:40

Exploring Non-Gaussian Approaches for Deterministic Data Assimilation

M. Buehner (ECCC), D. Jacques (ECCC), A. Perez (McGill University) and I. Zawadzki (McGill University)

Abstract

The assimilation of weather radar data using the Latent Heat Nudging algorithm is being evaluated for application to regional deterministic prediction in Canada. This approach is relatively simple to implement and improves precipitation forecasts during the first few forecast hours. It also provides a benchmark against which more advanced non-Gaussian data assimilation (DA) methods that rely on an ensemble of short-term forecasts can be compared. The simplest such approach constructs an ensemble composite analysis state by selecting the ensemble member that is most consistent with the radar observations within a local region, thus avoiding the usual assumptions required by standard DA methods. More generally, the analysis state can be computed as the weighted ensemble mean using spatially varying weights obtained with an approach similar to the localized particle filter. Examples of analyses and forecasts generated with these different approaches will be compared. Strategies for combining such non-Gaussian DA approaches with the currently operational 4D-EnVar approach will also be discussed.

Contact information

Dr. Mark Buehner Environment and Climate Change Canada