Habitat-Based Density Models for Three Cetacean Species off Southern California Illustrate Pronounced Seasonal Differences

Abstract

Managing marine species effectively requires spatially and temporally explicit knowledge of their density and distribution. Habitat-based density models, a type of species distribution model (SDM) that uses habitat covariates to estimate species density and distribution patterns, are increasingly used for marine management and conservation because they provide a tool for assessing potential impacts (e.g., from fishery bycatch, ship strikes, anthropogenic sound) over a variety of spatial and temporal scales. The abundance and distribution of many pelagic species exhibit substantial seasonal variability, highlighting the importance of predicting density specific to the season of interest. This is particularly true in dynamic regions like the California Current, where significant seasonal shifts in cetacean distribution have been documented at coarse scales. Finer scale (10 km) habitat-based density models were previously developed for many cetacean species occurring in this region, but most models were limited to summer/fall. The objectives of our study were two-fold: (1) develop spatially-explicit density estimates for winter/spring to support management applications, and (2) compare model-predicted density and distribution patterns to previously developed summer/fall model results in the context of species ecology. We used a well-established Generalized Additive Modeling framework to develop cetacean SDMs based on 20 California Cooperative Oceanic Fisheries Investigations (CalCOFI) shipboard surveys conducted during winter and spring between 2005 and 2015. Models were fit for short-beaked common dolphin (Delphinus delphis delphis), Dall's porpoise (Phocoenoides dalli), and humpback whale (Megaptera novaeangliae). Model performance was evaluated based on a variety of established metrics, including the percentage of explained deviance, ratios of observed to predicted density, and visual inspection of predicted and observed distributions. Final models were used to produce spatial grids of average species density and spatially-explicit measures of uncertainty.Becker et al. Seasonal Differences in Cetacean DistributionsResults provide the first fine scale (10 km) density predictions for these species during the cool seasons and reveal distribution patterns that are markedly different from summer/fall, thus providing novel insights into species ecology and quantitative data for the seasonal assessment of potential anthropogenic impacts.

Evaluating stakeholder‐derived strategies to reduce the risk of ships striking whales

“…Speed reductions should reduce risk year‐round, according to the relationship between speed and the probability that a collision is lethal (Conn & Silber, ). The effect of expanding the ATBA between January and June should be evaluated because the line‐transect data used to derive the whale distributions were collected between July and December and studies have found seasonal changes in fin (Scales et al, ) and humpback (Becker et al, ) whale distributions off California. These evaluations should use data collected throughout the Bight to ensure they capture potential seasonal changes in whale distributions and include changes in ship traffic in the Channel and western approach.…”

Predicting cetacean abundance and distribution in a changing climate

“…These two systems differ in the specific data used and in assimilation details, but both provide data‐constrained state estimates for our study area. We limited the predictors to those consistent between the two sources (Becker et al, ): sea surface temperature (SST) and its standard deviation sd (SST), calculated for a 3 × 3‐pixel box around the modelling segment mid‐point), mixed layer depth (MLD, defined by a 0.5°C deviation from the SST), sea surface height (SSH) and sd(SSH). A simple offset (+0.035 m) was applied to the near‐real‐time SSH data to match the historical reanalysis dataset, which had a different reference level (Scales et al, ).…”

“…The first included the full set of dynamic and bathymetric variables (“Model 1”; Table ). These habitat covariates have been used successfully in previous SDMs in the study area (Becker et al, ).…”

Prediction of Large Whale Distributions: A Comparison of Presence–Absence and Presence-Only Modeling Techniques

“…These variables represent surface water mass identity, physical processes such as mixing and upwelling, and bathymetric features that influence prey availability. For each of these, the same variable or a related variable, e.g., mixed layer depth rather than thermocline depth, has been found to be important as a predictor variable in the eastern tropical Pacific or other regions (Forney et al, 2012;Becker et al, 2017).…”