TY - JOUR
T1 - Functional responses in habitat selection
T2 - clarifying hypotheses and interpretations
AU - Holbrook, Joseph D.
AU - Olson, Lucretia E.
AU - DeCesare, Nicholas J.
AU - Hebblewhite, Mark
AU - Squires, John R.
AU - Steenweg, Robin
N1 - Publisher Copyright:
© 2019 by the Ecological Society of America
PY - 2019/4
Y1 - 2019/4
N2 - A fundamental challenge in habitat ecology and management is understanding the mechanisms generating animal distributions. Studies of habitat selection provide a lens into such mechanisms, but are often limited by unrealistic assumptions. For example, most studies assume that habitat selection is constant with respect to the availability of resources, such that habitat use remains proportional to availability. To the contrary, a growing body of work has shown the fallacy of this assumption, indicating that animals modify their behavior depending on the context at broader scales. This has been termed a functional response in habitat selection. Furthermore, a diversity of methods is employed to model functional responses in habitat selection, with little attention to how methodology might affect scientific and conservation conclusions. Here, we first review the conceptual and statistical foundations of methods currently used to model functional responses and clarify the ecological tests evaluated within each approach. We then use a combination of simulated and empirical data sets to evaluate the similarities and differences among approaches. Importantly, we identified multiple statistical issues with the most widely applied approaches to understand functional responses, including: (1) a complex and important role of random- or individual-level intercepts in adjusting individual-level regression coefficients as resource availability changes and (2) a sensitivity of results to poorly informed individual-level coefficients estimated for animals with low availability of a given resource. Consequently, we provide guidance on applying approaches that are insensitive to these issues with the goal of advancing our understanding of animal habitat ecology and management. Finally, we characterize the management implications of assuming similarity between the current approaches to model functional responses with two empirical data sets of federally threatened species: Canada lynx (Lynx canadensis) in the United States and woodland caribou (Rangifer tarandus caribou) in Canada. Collectively, our assessment helps clarify the similarities and differences among current approaches and, therefore, assists the integration of functional responses into the mainstream of habitat ecology and management.
AB - A fundamental challenge in habitat ecology and management is understanding the mechanisms generating animal distributions. Studies of habitat selection provide a lens into such mechanisms, but are often limited by unrealistic assumptions. For example, most studies assume that habitat selection is constant with respect to the availability of resources, such that habitat use remains proportional to availability. To the contrary, a growing body of work has shown the fallacy of this assumption, indicating that animals modify their behavior depending on the context at broader scales. This has been termed a functional response in habitat selection. Furthermore, a diversity of methods is employed to model functional responses in habitat selection, with little attention to how methodology might affect scientific and conservation conclusions. Here, we first review the conceptual and statistical foundations of methods currently used to model functional responses and clarify the ecological tests evaluated within each approach. We then use a combination of simulated and empirical data sets to evaluate the similarities and differences among approaches. Importantly, we identified multiple statistical issues with the most widely applied approaches to understand functional responses, including: (1) a complex and important role of random- or individual-level intercepts in adjusting individual-level regression coefficients as resource availability changes and (2) a sensitivity of results to poorly informed individual-level coefficients estimated for animals with low availability of a given resource. Consequently, we provide guidance on applying approaches that are insensitive to these issues with the goal of advancing our understanding of animal habitat ecology and management. Finally, we characterize the management implications of assuming similarity between the current approaches to model functional responses with two empirical data sets of federally threatened species: Canada lynx (Lynx canadensis) in the United States and woodland caribou (Rangifer tarandus caribou) in Canada. Collectively, our assessment helps clarify the similarities and differences among current approaches and, therefore, assists the integration of functional responses into the mainstream of habitat ecology and management.
KW - Lynx canadensis
KW - Rangifer tarandus caribou
KW - functional response
KW - generalized linear mixed models
KW - habitat availability
KW - habitat selection
KW - habitat use
KW - resource selection function
UR - http://www.scopus.com/inward/record.url?scp=85062713155&partnerID=8YFLogxK
U2 - 10.1002/eap.1852
DO - 10.1002/eap.1852
M3 - Article
C2 - 30653797
AN - SCOPUS:85062713155
SN - 1051-0761
VL - 29
JO - Ecological Applications
JF - Ecological Applications
IS - 3
M1 - e01852
ER -