An Ensemble Ocean Data Assimilation System for Seasonal Prediction

Abstract

Abstract A new ensemble ocean data assimilation system, developed for the Predictive Ocean Atmosphere Model for Australia (POAMA), is described. The new system is called PEODAS, the POAMA Ensemble Ocean Data Assimilation System. PEODAS is an approximate form of an ensemble Kalman filter system. For a given assimilation cycle, a central forecast is integrated, along with a small ensemble of forecasts that are forced with perturbed surface fluxes. The small ensemble is augmented with multiple small ensembles from previous assimilation cycles, yielding a larger ensemble that consists of perturbed forecasts from the last month. This larger ensemble is used to represent the system’s time-dependent background error covariance. At each assimilation cycle, a central analysis is computed utilizing the ensemble-based covariance. Each of the perturbed ensemble members are nudged toward the central analysis to control the ensemble spread and mean. The ensemble-based covariances generated by PEODAS potentially yield dynamically balanced analysis increments. The time dependence of the ensemble-based covariance yields spatial structures that change for different dynamical regimes, for example during El Niño and La Niña conditions. These differences are explored in terms of the dominant dynamics and the system’s errors. The performance of PEODAS during a 27-yr reanalysis is evaluated through a series of comparisons with assimilated and independent observations. When compared to its predecessor, POAMA version 1, and a simulation with no assimilation of subsurface observations, PEODAS demonstrates a quantitative improvement in skill. PEODAS will form the basis of Australia’s next operational seasonal prediction system.