Proper orthogonal decomposition modelling for intermittent flow

Proper orthogonal decomposition (POD) develops a low-dimensional model of a high-dimensional random field through the eigenstructure of the process covariance matrix. These reduced order models provide critical insight into the properties of the field, but they may require assumptions not met by random fluid flow. In flow over a low-rise building, discussed in this paper, the flow is intermittent. Some periods may exhibit vortices and some periods not, resulting in transient behavior that is poorly modeled by the standard POD approach. In this paper, we propose an intermittent POD model that partitions the flow into separate regimes. The paper demonstrates the effectiveness of this viewpoint in capturing the marginal distribution of the pressure field.