MEPS prepress abstract   -  DOI: https://doi.org/10.3354/meps14836

ABSTRACT: Geographic distributions are shaped by how species interact with environmental and climatic conditions. Species distribution models use these relationships to understand and predict how and where species are distributed; however, a single model type may not fully capture these dynamics. Using a multi-model approach, we performed species distribution modeling to elucidate the current geographic distribution of American (nèe West Indian) manatees, Trichechus manatus in the continental United States (USA) and predicted future changes to their distribution under different climatic scenarios. We used generalized linear and additive models and machine learning models like maximum entropy, random forests, and boosted regression trees. Although all models had high predictive capacity, the boosted regression tree model and the random forest had the highest validation scores. Important environmental variables included distance to shore, sea surface salinity (SSS) variance, sea surface temperature (SST) range, SST of June, and bathymetry. All models showed that areas farther north in the manatee range, like the northern Gulf of Mexico (nGOM) and the Atlantic coast north of Florida, have become increasingly important parts of their primary habitat in the USA. This northward trend is predicted to continue, with manatees having a higher probability of occurrence in areas farther north and at more future time periods (2050 vs 2100), concomitant with greater greenhouse gas emissions and associated warming that extends favorable habitat for manatees. Our results forecast widespread tropicalization of the nGOM and the southeast US Atlantic coast under future climate scenarios of which manatees are likely to be a sentinel.