Using a random forest model and public data to predict the distribution of prey for marine wildlife management

Abstract

Modern wildlife management relies on studies investigating the distribution patterns and habitat selection of wildlife at appropriately large scales for decision-making. One important aspect to consider in the assessment of habitat suitability and the underlying mechanism of animal distribution is the spatial distribution of their food resources. In marine areas, where many mammal and bird species occur over large spatial scales, the analysis of habitat use and distribution patterns requires information on the distribution of food resources at appropriately large scales (Huettmann and Diamond 2006). An important food resource for several species of marine birds and mammals are invertebrate organisms that live on the bottom of the sea and are collectively described as the benthos, a community that is especially productive and diverse at high latitudes (Carey 1991; Piepenburg 2005; Starmans et al. 1999). The distribution and productivity of benthic foragers such as ice seals, walrus (Odobenus rosmarus), sea ducks (Somateria spp., Melanitta spp.), and gray whales (Eschrichtius robustus) is influenced by the distribution of accessible benthic prey resources (Kaiser et al. 2006; Lovvorn et al. 2003; Moore et al. 2003). Benthic invertebrates thus form a key component in the trophic structure of marine ecosystems, and the distribution of marine benthic invertebrates is of major interest to wildlife managers (Solan et al. 2004). Most of the species mentioned above are of management concern, and either are (Spectacled Eider, Somateria fischeri; Steller's Eider, Polysticta stelleri) or have been proposed (walrus, ice seals) to be listed as 'threatened'. Therefore, the identification and delineation of critical habitat providing sufficient food resources for these species will become extremely important in the near future.