Modeled distribution and abundance of a pelagic seabird reveal trends in relation to fisheries

Abstract

The northern fulmar Fulmarus glacialis is one of the most visible and widespread seabirds in the eastern Bering Sea and Aleutian Islands. However, relatively little is known about its abundance, trends, or the factors that shape its distribution. We used a long-term pelagic dataset to model changes in fulmar at-sea distribution and abundance since the mid-1970s. We used an ensemble model, based on a weighted average of generalized additive model (GAM), multivariate adaptive regression splines (MARS), and random forest models to estimate the pelagic distribution and density of fulmars in the waters of the Aleutian Archipelago and Bering Sea. The most important predictor variables were colony effect, sea surface temperature, distribution of fisheries, location, and primary productivity. We calculated a time series from the ratio of observed to predicted values and found that fulmar at-sea abundance declined from the 1970s to the 2000s at a rate of 0.83% (± 0.39% SE) per annum. Interpolating fulmar densities on a spatial grid through time, we found that the center of fulmar distribution in the Bering Sea has shifted north, coinciding with a northward shift in fish catches and a warming ocean. Our study shows that fisheries are an important, but not the only factor, shaping fulmar distribution and abundance trends in the eastern Bering Sea and Aleutian Islands.

Temporal shifts in seabird populations and spatial coherence with prey in the southeastern Bering Sea

“…Likewise, long-term, northward shifts in the distribution of seabirds have also been observed in the Bering Sea (Renner et al 2013, as well as the occurrence of new species farther north into the Chukchi Sea ). Additionally, long-term studies of seabird colonies in the Bering Sea indicate that food web changes may be more pronounced for near-surface foraging guilds compared to those foraging at depth .…”

“…An examination of 4 decades of at-sea surveys showed that for northern fulmars, colony effect was a key factor in modeling the distribution of this highly abundant species in the SEBS (Renner et al 2013). Colony effect can vary among seabirds, however; diving seabirds, such as thickbilled murre, show a stronger colony effect than surface-feeding black-legged kittiwakes (Sigler et al 2012).…”

Statistical integration of tracking and vessel survey data to incorporate life history differences in habitat models

“…Although tracking data is presence-only data, this analysis enables us to calculate a continuous density probability gradient (including near 0 values) spatially for use in our models. Bird density close to a breeding colony is stochastically high due to geometric spreading as birds commute to and from the island (i.e., colony effect; Ashmole and Ashmole 1967, Kinder et al 1983, Decker and Hunt 1996, Renner et al 2013, especially for central-place foragers, such as Streaked Shearwaters. The high density around the colony may not be a function of oceanic variables but merely a function of distance from the colony (Gre´millet and Boulinier 2009).…”

“…For the tracking data, the normalized monthly data sets (April-July) were pooled into a single data set to fit and refine the model across seasons (Louzao et al 2009), as pooled data may be more representative of habitat use for a given species compared to monthly data alone. The performance of species distribution models may differ among modeling techniques used (e.g., Yen et al 2004, Oppel et al 2012, Renner et al 2013. In order to assess the predictive capacity of the various modeling techniques in estimating shearwater distribution, we compared shearwater tracking models from 2007 using four different modeling techniques: generalized linear model (GLM), generalized additive model (GAM), random forest, and an ensemble of all models (ensemble model).…”

“…Species distribution data of marine vertebrates is traditionally obtained by direct sightings of non-marked individuals via shipboard, aerial, or other survey methodology. These methods provide useful information on overall distribution ranges of species and critical estimates of density and abundance, including diversity of different species in space and time (Ballance et al 1997, Renner et al 2013, Wong et al 2014. However, as ecological information (e.g., sex, breeding colony, and current reproductive condition) of observed individuals cannot be detected during surveys, especially for species that do not exhibit sexual dimorphism including many seabird species, few studies have been able to examine distributions of a given species as a function of ecological or life history state (e.g., colony of origin, sex, maturity, and breeding condition).…”