Spatiotemporal data integration for marine megafauna SDMs in dynamic environments: A point process approach

Species distribution models (SDMs) for marine megafauna, such as cetaceans, traditionally use occurrence data from scientific surveys. These surveys follow standardised protocols and provide spatially structured data, making them a reliable source of information on distributions. However, they are costly and temporally limited, offering only a snapshot of distributions in time. Because cetaceans track dynamic oceanographic features, models built on survey data alone often lack temporal transferability. In contrast, platforms of opportunity (such as whale-watching vessels) allow obtaining occurrence data with high temporal resolution and near-daily sampling. These data are restricted to small coastal areas and subject to spatial bias, yet remain informative through time. Given the complementary nature of these sources and the absence of a general spatiotemporal framework for combining designed line-transect detections with opportunistic presence-only data, here we present an integrated modelling approach that leverages their strengths: broad spatial structure from surveys and high temporal resolution from whale watching. We combine these sources in a joint-likelihood log-Gaussian Cox process (LGCP), fitted in a Bayesian hierarchical framework using integrated nested Laplace approximations (INLA). The survey contributes a line-transect distance-sampling likelihood, and whale watching contributes an availability-restricted presence-only likelihood. Simulations show that integration lets the model absorb the temporal signal from whale watching without losing spatial structure from the survey. We apply the approach to short-beaked common dolphins (Delphinus delphis) off mainland Portugal. The integrated model improves spatial fit and combines heterogeneous data within a coherent framework, though dynamic covariate effects were weak, limiting temporal variation in spatial predictions. The results demonstrate potential and limitations of spatiotemporal integration. They clarify data requirements when combining surveys with opportunistic data and provide a general, extensible framework for integrating marine datasets that differ in design, extent, and temporal resolution.