Habitat-based cetacean density models for the U.S. Atlantic and Gulf of Mexico

The International Whaling Commission placed a moratorium on commercial whaling in 1986, curtailing the biggest direct anthropogenic threat to many cetacean populations. But other threats have persisted, such as bycatch in fisheries1, ship strikes2, oil spills3,4, and other pollutants5. New threats have been recognized, including naval active sonar6,7,8, other anthropogenic sources of noise9,10, and climate change11. In the United States, national laws protect cetaceans. The Marine Mammal Protection Act (MMPA) prohibits intentional or incidental killing, injuring, or harassment of cetaceans and specifies the circumstances and rules under which permits may be issued for such activities. The Endangered Species Act (ESA) prohibits harm to species threatened with extinction, including 16 cetacean species, and requires conservation of their habitat. The National Environmental Policy Act (NEPA) specifies the process by which U.S. national government agencies must evaluate the potential environmental effects of their actions, consider alternatives, and conduct public reviews. Agency actions that involve decisions to issue permits under the MMPA or ESA are usually subject to this process. To evaluate the potential effects of proposed activities on cetacean populations, interested parties require a detailed understanding of the spatiotemporal distributions of these populations. Recent developments have created an urgent need for this information in U.S. waters of the Atlantic and Gulf of Mexico, when the U.S. Bureau of Ocean Energy Management (BOEM) proposed to open a large portion of the Atlantic continental shelf to oil and natural gas development and to expand oil and gas leasing in the Gulf of Mexico. Concurrently, the U.S. Navy began development of a new Environmental Impact Statement assessing the effects of training activities proposed for a large portion of the western North Atlantic, while the National Marine Fisheries Service (NMFS) proposed to expand the geographic area designated as critical habitat for endangered North Atlantic right whales, and to reevaluate the status of regional populations of humpback and Bryde’s whales under the ESA. To estimate the abundance of cetacean species in U.S. waters and assess how they are distributed geographically and seasonally, NMFS and other U.S. government organizations have conducted visual line-transect surveys for over 35 years, yielding two parallel modeling efforts. One effort, prompted by the national regulatory framework, applied distance sampling methodology12 to estimate the abundance of cetacean species within large geographic strata13,14,15. The other, driven by a desire to describe cetacean habitats at a fine spatiotemporal scale, developed regression models that related the presence of cetacean species to environmental correlates such as sea surface temperature and then predicted the models across the seascape using gridded maps of the correlates, yielding fine-scale maps of habitat suitability16,17,18. Neither effort has proved entirely satisfactory for managing cetacean populations in the U.S. The regulatory framework requires an estimate of the number of affected individual animals in proposals for actions that could harm or disturb cetaceans. The broad-scale abundance studies estimated the number of individuals present in large geographic areas but these so-called “stratified models” did not show how they were distributed within each area. In contrast, the habitat suitability studies modeled spatial variability at fine resolutions, but produced estimates that used relative or unit-less scales (e.g. ranging from 0 to 1) that could not directly be used to estimate counts of affected individuals. The last decade has seen a unification of these two approaches into a two-stage method known as density surface modeling19,20, in which traditional distance sampling is coupled with multivariate regression modeling to produce density maps (individuals km−2) predicted from fine scale environmental covariates21. A challenge with density surface models (DSMs) is that a large number of sightings are needed to fit the regression model. Cetaceans are rare; often many surveys must be aggregated to obtain sufficient sightings. For example, a study of beaked whales in the eastern tropical Pacific aggregated 6 years of surveys to obtain just 90 sightings of Cuvier’s beaked whale and 106 of Mesoplodon beaked whales22. This problem is exacerbated if the modeler desires to fit different models for different regions or seasons under the presumption that different behaviors occur in those places and times, e.g. that baleen whales on summer feeding grounds exhibit different environmental preferences than those on winter calving grounds. Here, we present cetacean density models for the U.S. Atlantic and Gulf of Mexico. To maximize the number of taxa modeled and account for regional and seasonal variability, we aggregated nearly 1.1 million linear km of line-transect surveys conducted over 23 years by 5 institutions. We modeled density from 8 physiographic and 16 dynamic oceanographic and biological covariates, producing predicted density maps for 29 cetacean species and multi-species guilds, comprising 36 species in total. Results are freely available online at the Ocean Biogeographic Information System Spatial Ecological Analysis of Megavertebrate Populations (OBIS-SEAMAP) repository at http://seamap.env.duke.edu/models/Duke-EC-GOM-2015.