Habitat Features Predict Carrying Capacity of a Recovering Marine Carnivore

Abstract

The recovery of large carnivore species from over‐exploitation can have socioecological effects; thus, reliable estimates of potential abundance and distribution represent a valuable tool for developing management objectives and recovery criteria. For sea otters (Enhydra lutris), as with many apex predators, equilibrium abundance is not constant across space but rather varies as a function of local habitat quality and resource dynamics, thereby complicating the extrapolation of carrying capacity (K) from one location to another. To overcome this challenge, we developed a state‐space model of density‐dependent population dynamics in southern sea otters (E. l. nereis), in which K is estimated as a continuously varying function of a suite of physical, biotic, and oceanographic variables, all described at fine spatial scales. We used a theta‐logistic process model that included environmental stochasticity and allowed for density‐independent mortality associated with shark bites. We used Bayesian methods to fit the model to time series of survey data, augmented by auxiliary data on cause of death in stranded otters. Our model results showed that the expected density at K for a given area can be predicted based on local bathymetry (depth and distance from shore), benthic substrate composition (rocky vs. soft sediments), presence of kelp canopy, net primary productivity, and whether or not the area is inside an estuary. In addition to density‐dependent reductions in growth, increased levels of shark‐bite mortality over the last decade have also acted to limit population expansion. We used the functional relationships between habitat variables and equilibrium density to project estimated values of K for the entire historical range of southern sea otters in California, USA, accounting for spatial variation in habitat quality. Our results suggest that California could eventually support 17,226 otters (95% CrI = 9,739–30,087). We also used the fitted model to compute candidate values of optimal sustainable population abundance (OSP) for all of California and for regions within California. We employed a simulation‐based approach to determine the abundance associated with the maximum net productivity level (MNPL) and propose that the upper quartile of the distribution of MNPL estimates (accounting for parameter uncertainty) represents an appropriate threshold value for OSP. Based on this analysis, we suggest a candidate value for OSP (for all of California) of 10,236, which represents 59.4% of projected K. © 2021 The Authors. The Journal of Wildlife Management published by Wiley Periodicals LLC on behalf of The Wildlife Society.

Diffusion modeling reveals effects of multiple release sites and human activity on a recolonizing apex predator

“…Thus, we modeled motility δ ( s ) as a log-linear function of covariates that have been found to be important drivers of sea otter space use and behavior [ 20 , 21 ]. Based on previous studies, our covariates included depth, as a binary indicator (depth =1 where <40 m, and 0 otherwise), distance to shore, slope of the ocean floor, and shoreline complexity [ 20 , 38 – 41 ]. Shoreline complexity was calculated for each location by summing the number of locations within a 1,000 m neighborhood that contained shoreline [ 20 ].…”

Diffusion modeling reveals effects of multiple release sites and human activity on a recolonizing apex predator

“…Thus, we modeled motility δ(s) as a log-linear function of covariates that have been found to be important drivers of sea otter space use and behavior (Williams et al, 2019;Lu et al, 2020). Based on previous studies, our covariates included depth, as a binary indicator (depth = 1 where < 40 m, and 0 otherwise), distance to shore, slope of the ocean floor, and shoreline complexity (Williams et al, 2019;Tinker et al, 2021;Bodkin et al, 2004;Johnson et al, 2009;Gregr et al, 2008).…”

“…Other studies have included multiple fine scale habitat covariates in population models to explain carrying capacity of sea otters (Tinker et al, 2021). As more spatial data become available for Southeast Alaska, similar covariates could be included in the diffusion model.…”

“…We included two indicator covariates in our formulation of K(s), representing areas with different management regimes, but variation in motility over the region also explains spatial variation in carrying capacity in the model. Other covariates, such as kelp canopy cover and benthic substrate composition, which have been shown to be important drivers of sea otter carrying capacity elsewhere (Tinker et al, 2021), could be included in future models (i.e., as covariates affecting density dependence and/or motility).…”

“…While an extension to modeling abundance in their framework is certainly possible, ecological diffusion naturally models abundance as well as movement toward and concentration in favorable habitat Tinker et al (2019). andTinker et al (2021) modeled sea otter recolonization (of Southeast Alaska and the central coast of California, respectively) by parsing the study areas into distinct units and specifying immigration and emigration among them. While doing so offers computational advantages, inferences are restricted to those defined units.…”