Fisheries-wildlife interactions are a major barrier to ecological sustainability in marine wild-capture fisheries. Mitigation of incidental capture (bycatch) of threatened, endangered and protected species including sharks, turtles, cetaceans, pinnipeds, and seabirds is a central goal of sustainable fisheries management. Predicting where and when fisheries-wildlife interactions are likely to occur can inform targeted monitoring and mitigation strategies, such as gear modifications, and the design of protected areas or other effective area-based conservation measures (OECMs). Transformational approaches to mitigation, such as dynamic management, mobile protected areas and ecological forecasting are also evolving. Predictive modelling of the likelihood of interactions with respect to the spatiotemporal dynamics of fisheries, marine species distributions, and the physical dynamics of the ocean is often used to inform these approaches. However, the skill of predictive models for anticipating where, when and under what conditions fisheries-wildlife interactions might occur is highly variable, and often dependent upon system specifics. This talk will discuss case studies that use physical data from remote sensing and ocean models in data-driven predictive modelling of fisheries-wildlife interactions, with particular focus on the roles of meso- and submesoscale features such as fronts and eddies in the spatial structuring of interaction likelihood.