Coping With Heterogeneity to Detect Species on a Large Scale: N-Mixture Modeling Applied to Red-Legged Partridge Abundance

The reliability of long-term population estimates is crucial for conservation and management purposes. Most game species population monitoring programs assume that count indices are proportionally related to abundance. However, this assumption is untenable when detection varies spatially and temporally. We assessed whether N-mixture models, which allow detection modeling using spatially and temporally repeated count data, were relevant for monitoring the population of red-legged partridge (Alectoris rufa). Unbiased estimates are needed for calculating hunting quotas for this game species. We used the simple aural playback point-countmethod and a dataset of 121 point-transects (n = 16,288 counts) collected from 1992 to 2010. Covariates such as date, hour, wind, rain, and vegetation density influenced detection probability. Estimated abundance ranged from 0 to 15 males per point and exhibited variable coefficients of variation depending on sites (range: 0.14-1.31, mean: 0.44). We found a positive and linear relationship between the abundance estimated by the N-mixture model and the densities provided by 2 other counting methods, square sampling and kilometric abundance index, but not with blank beat. We established a maximum detection radius for the playback surveys, which enabled conversion of abundance at the points into density. The N-mixture modeling approach is more cost-effective for game species monitoring than capture-recapture methods or square sampling, and more reliable than indices of relative abundances. (C) 2014 The Wildlife Society.