Bighorn sheep habitat and model extrapolation across remote landscapes

Abstract

Determining a species’ habitat use is an essential first step in any wildlife conservation action. We described habitat use, animal movements and probable lambing areas in a remote, restricted-access region of the Mojave Desert. Differences in habitat use between sexes was apparent, supporting the often-reported concept of risk-aversion by females. Animals exhibited low variability in distances travelled, although males travelled further and with more variability than females. All females demonstrated what we interpret as lambing behavior during the same 2 ½ month periods over the two years, strongly supporting our inference of lambing sites. Water appeared critical to animal long-range movements, with no animal moving beyond 8.5 km from known sources. Modeling habitat use across the landscape of concern is another necessary step for conservation of species, allowing managers to plan for and predict the outcomes of management actions. In ecology, models are often created within relatively small areas, then extrapolated across larger regions of concern. The ability to extrapolate ecological models may be especially useful across remote areas, where consistent access by wildlife managers may be highly restricted. These restrictions on access require that most data be collected remotely, necessitating the need for extrapolating models developed in other areas. We used data from GPS-collared desert bighorn sheep to describe and model habitat use across the Pintwater Range, located on the Nevada Test and Training Range of southern Nevada, a highly restricted military training ground. We tested the efficacy of habitat model extrapolation by comparing the performance of two models derived from adjacent but independent desert bighorn sheep populations. The predictive power of seasonal habitat models derived from adjacent mountain ranges was lower than those derived from the local population. However, the performance of the extrapolated model suggests it could still be a feasible alternative for estimating general habitat use.