Analyzing Habitat Selection In Animals Without Well-Defined Home Ranges

Ecology, May, 2000 by Dag Oystein Hjermann

DAG OYSTEIN HJERMANN [1]

Abstract. S. M. Arthur and colleagues presented a statistical framework that allows habitat availability to change for each observation of an animal, thus making habitat-use analysis possible when the habitat changes or it is difficult to determine a home range for each individual. I here modify their method by letting habitat availability vary on a continuous scale, rather than defining the area within a circle as equally available. The modified method also makes it possible to deal with continuously varying time intervals between observations and the influence of environmental variables (e.g., temperature, time of day) on movement activity. As an example, I use this method to analyze habitat selection of a bush-cricket (the wart-biter, Decticus verrucivorus).

Key words: autocorrelation, avoiding; Decticus verrucivorus; diffusion (random walk); habitat availability; habitat selection; habitat-use analysis in free-ranging animals; habitat use, scale dependent; movement pattern; statistical method.

INTRODUCTION

In studies of habitat selection it is typically assumed that the entire study area is available to all individuals, or that all of each individual's home range is available at all times (Manly et al. 1993, Arthur et al. 1996). However, this is appropriate only when (1) the relative amounts of the different habitat types are stable throughout the study period, (2) a well-defined home range can be identified for each individual, and (3) each position of the animal can be regarded as one independent choice of habitat within the home range (Manly et al. 1993). Recently, Arthur et al. (1996) presented an elegant and very useful statistical technique that relaxes these assumptions by allowing habitat availability to change for each observation of the animal. It is assumed that an animal in a given time period selects from only the parts of the habitat relatively close to its current position, not from its entire home range. However, Arthur et al. (1996) assumed that all habitat within a certain radius of the anima l's position is equally available, while the rest of the habitat is unavailable (the "availability radius" approach, or AR). Although this may be an adequate approximation under some circumstances, the true availability is usually expected to vary continuously rather than dichotomously with distance. In this report I use a continuous availability function (CAF) that depends on the time interval between recordings and on other non-spatial variables. Thus, an inherent property of the modified method is that the length of the time interval between observations is allowed to vary continuously, and it can easily take into account that environmental variables (e.g., temperature, time of day) may influence movement speed. As an example, I analyze habitat selection of a bush-cricket (the Wart-biter, Decticus verrucivorus) within a semi-natural habitat island.

METHODS

As any analysis of habitat use involves comparing the availability of each habitat with the actual habitat use, the definition of "availability" is critical. To take into account the fact that only a limited portion of the habitat is available to the animal at a given time step, Arthur et al. (1996) specified available habitat as a moving circle centered at the previous location of the animal. For instance, in their polar bear (Ursus maritimus) example, they assumed that if there were 3 d to the next observation, all habitat within 200 km of the bear's position was equally available for its next move (300 km if the observations were 6 d apart). These radii corresponded approximately to the 99th percentile of observed movement lengths. However, a histogram of the distribution of movement lengths usually indicates that the true availability varies on a continuous scale. Commonly, an animal is most likely to move some short distance, while the probability decreases gradually for longer distances. Moreover, it m ay be difficult to choose one or a few radii that correspond to the 99th percentile when the time interval between observations varies erratically, especially if factors other than habitat (e.g., temperature) also influence an animal's mobility. An alternative to this approach is to use a continuous availability function (CAF) that depends both on the time interval between observations and on biotic and environmental conditions. In presenting this approach I assume that the animals have been tracked individually, and that the habitat is mapped as a grid where each square is classified according to habitat type. The first step of the CAF approach is to estimate the distribution of the step length r, the distance an individual moves between two observations, given biotic and environmental conditions. Then it is possible to estimate the probability of moving to a given grid cell in the habitat, assuming no habitat selection (the null probability). For a given position of the animal, the sum of the null probabili ties for all grid cells of a given habitat type is taken as the availability of this habitat type prior to the next observation of the animal.