Real-time species distribution models for conservation and management of natural resources in marine environments

Abstract

Species distribution models in marine environments typically use static variables, partly due to the limited availability of fine-resolution dynamic predictor variables and sufficiently detailed species abundance data. Our aim was to describe and quantify the dynamic coupling between the distribution of marine species (seabirds) and the natural variability of their habitat in real time through the combination of a high-resolution hydrodynamic model, aerial digital surveys and real-time species distribution modelling. We used a 2-step (delta) generalized additive model at 500 m spatial resolution for assessment and prediction of the changing patterns of wintering red-throated divers (RTDs) Gavia stellata in the outer Thames estuary, United Kingdom. Our dynamic species distribution models successfully resolved the major oscillations in the distribution of RTDs and confirmed their tight association with frontal zones where the probability of prey encounter was higher. The relative model standard errors (%) were generally below 30% in the high-density areas. Area under the curve (AUC) values indicated that the models were capable of distinguishing presence from absence about 75% of the time. The predictive power of the achieved distribution models made it possible to accurately identify areas where RTDs were con- centrated. Comparisons between visual aerial and digital stills aerial surveys documented that, in spite of similar patterns, the aerial digital surveys generally recorded significantly higher densities of RTDs than the visual aerial surveys. This study demonstrates how marine distribution models with assimilation of habitat variables from a well-calibrated fine-resolution hydrodynamic model coupled with the use of digital aerial surveys can facilitate the capture of detailed associations between seabirds and their dynamic habitats.