Improving habitat models for rare and inconspicuous species: A multi-part conditional approach

Background/Question/Methods Determination of occurrence and/or habitat requirements for rare or cryptic species is difficult because low detection leads to limited predictive capabilities. Furthermore, low detection probabilities increase the costs of direct abundance estimation, necessitating surrogate measures for large-scale surveys. This problem is especially poignant because rare and inconspicuous species are often of management or conservation concern. The pygmy rabbit (Brachylagus idahoensis) is petitioned to be listed under the Endangered Species Act. Published habitat models for this is small, inconspicuous rabbit suffer from poor sensitivity, and improvements are necessary in order to evaluate their distribution and conservation status. Here we demonstrate the use different types of indirect observations (burrows and fecal pellets) and niche-based covariates in a multi-part conditional model of rabbit occurrence. This approach is advantageous because it allows for the fit of all process and observational uncertainties (i.e. that of both pellets and burrows) to be included. Burrows and rabbit pellets were sampled along five 200-m line transects at each of 38 6-ha sites within a 13 500-ha study area in Rich County, Utah. Models of rabbit sign conditioned on the presence of burrows were used to interpolate rabbit habitat suitability for the entire study area.

Results/Conclusions We improve on previous efforts by modeling the process and observation uncertainty for each of the sign types (burrows and fecal pellets). We show that observation uncertainty could account for performance issues encountered by habitat-only models. Because detection is imperfect, surveys conducted in a cursory manner could easily fail to detect pygmy rabbits in areas of low rabbit/burrow density, thus causing the problem of low sensitivity.