Developing a monitoring program for dusky grouse in Montana

Abstract

Rigorous monitoring programs for Dusky Grouse (Dendragapus obscurus), a game species, are lacking. Difficult-to-reach habitat and low probability of detection makes monitoring Dusky Grouse difficult. Our objectives were to 1) evaluate Dusky Grouse habitat associations and generate a state-wide map predicting Dusky Grouse habitat, 2) evaluate sampling methods and survey conditions for maximizing Dusky Grouse detection, and 3) evaluate protocols (i.e., number of sites and visits) and analytical methods for producing annual unbiased and precise indices of abundance. We created our habitat model using resource selection functions, random forest, and an ensemble approach. We compared spring v. summer sampling, use of electronic playback to increase detection, effect of route type (off-trail, trail, road) on point counts, and the effect of weather, background noise, day, and time on probability of detection. We evaluated and compared four analytical methods using simulations: time-to-detection model with hierarchical distance sampling, N-mixture model, raw count (naive) and hierarchical distance sampling model. Multiple habitat characteristics affected relative probability of Dusky Grouse use including tree height and conifer forest vegetation types. Both habitat modeling methods were highly predictive and therefore we used an ensemble (frequency histogram) approach to create a state-wide map of Dusky Grouse habitat that was used to identify appropriate sampling sites for population monitoring. Spring point-count surveys conducted with electronic playback were most effective. Surveys located along roads/trails best balanced the trade-offs between sampling effort and survey design requirements, despite limiting inferences to Dusky Grouse populations located in prime habitat along roads/trails. Detection of Dusky Grouse was highest on clear days, with little wind and background noise, with surveys occurring 9-162 minutes post-sunrise during 3-23 May. Simulation results indicated that N-mixture models where 80 sites visited four times resulted in unbiased estimates of population size with the highest precision. Transect-based distance sampling survey protocols during the spring also produced unbiased and acceptably precise (< or = 15% CV) estimates of grouse density when > or = 35 transects of > or = 2.6-km length were surveyed per area of inference (e.g., administrative region). Our results provide baseline information necessary for the development of a state-wide monitoring program for Montana.