Autonomous mobile robots often require informa-tion about the environment beyond merely the shape of thework-space. In this work we present a probabilistic method formappingdynamics, in the sense of learning and representingstatistics about the flow of discrete objects (e.g., vehicles, people)as well as continuous media (e.g., air flow). We also demonstratethe capabilities of the proposed method with two use cases. Onerelates to motion planning in populated environments, whereinformation about the flow of people can help robots to followsocial norms and to learn implicit traffic rules by observingthe movements of other agents. The second use case relates toMobile Robot Olfaction (MRO), where information about windflow is crucial for most tasks, including e.g. gas detection, gasdistribution mapping and gas source localisation. We representthe underlying velocity field as a set of Semi-Wrapped GaussianMixture Models (SWGMM) representing the learnt local PDF ofvelocities. To estimate the parameters of the PDF we employ aformulation of Expectation Maximisation (EM) algorithm specificfor SWGMM. We also describe a data augmentation methodwhich allows to build a dense dynamic map based on a sparseset of measurements. In case only a small set of observations isavailable we employ a hierarchical sampling method to generatevirtual observations from existing mixtures.