The Predictive Ocean Atmosphere Model for Australia (POAMA) is an operational seasonal dynamical forecast system based on a coupled ocean - atmosphere model. A key component of this system is ocean data assimilation which is used to produce ocean initial conditions for seasonal forecasts. PEODAS (POAMA Ensemble Ocean Data Assimilation System) is a new assimilation system for POAMA-2. A significant advantage of PEODAS is that it provides a natural way to estimate cross covariance between the fields of different physical variables. The ensemble is constructed with the Australian Community Ocean Model (ACOM2), where each member is forced by differently perturbed ERA40 forcing fields over sequential 3-day intervals. The perturbations are based on a multivariate EOF decomposition of differences between NCEP and ERA40 reanalysis products.