Using modelled prey to predict the distribution of a highly mobile marine mammal. Issue 11 (28th August 2020)

• Record Type: Journal Article
• Authors: Pendleton, Daniel E.; Holmes, Elizabeth E.; Redfern, Jessica; Zhang, Jinlun
• Editors: Maiorano, Luigi
• Is Part Of: Diversity & distributions. Volume 26:Issue 11(2020)
• Page Start: 1612
• Page End: 1626
• Publication Date: 2020-08-28
• Subjects: Arctic -- boosted regression tree -- bowhead whale -- forecast -- hindcast -- maxent -- ocean model -- species distribution model -- zooplankton; Biodiversity -- Periodicals; Biodiversity conservation -- Periodicals; 577
• DOI: 10.1111/ddi.13149
• Languages: English
• ISSNs: 1366-9516

Abstract

Abstract: Aim: Species distribution models (SDMs) are a widely used tool to estimate and map habitat suitability for wildlife populations. Most studies that model marine mammal density or distributions use oceanographic proxies for marine mammal prey. However, proxies could be a problem for forecasting because the relationships between the proxies and prey may change in a changing climate. We examined the use of model‐derived prey estimates in SDMs using an iconic species, the western Arctic bowhead whale ( Balaena mysticetus ). Location: Western Beaufort Sea, Alaska, USA. Methods: We used Biology Ice Ocean Modeling and Assimilation System (BIOMAS) to simulate ocean conditions important to western Arctic bowhead whales, including important prey species. Using both static and dynamic predictors, we applied Maxent and boosted regression tree (BRT) SDMs to predict bowhead whale habitat suitability on an 8‐day timescale. We compared results from models that used bathymetry with those that used only BIOMAS simulated variables. Results: The best model included bathymetry and BIOMAS variables. Inclusion of dynamic variables in SDMs produced predictions that reflected temporal dynamics evident from the survey data. Bathymetry was the most influential variable in models that included that variable. Zooplankton was the most important variable for models that did not include bathymetry. Models with bathymetry performed slightly better than models with only BIOMAS derived variables. Main conclusions: Bathymetry and modelled zooplankton were the most important predictor variables in bowhead whale distribution models. Our predictions reflected within‐year variability in bowhead whale habitat suitability. Using modelled prey availability, rather than oceanographic proxies, could be important for forecasting species distributions. Predictor variables used in our study were derived from a biophysical ocean model with demonstrated ability to project future ocean conditions. A natural next step is to use output from our biophysical ocean model to understand the effects of Arctic climate change.