Recouping Lost Information when Mark-Recapture Data are Pooled: A Case Study of Milksnakes (Lampropeltis triangulum) in the Upper Midwestern United States

There is a surprising lack of data available on snake demography. This is largely because snakes are cryptic and difficult to detect. To address this problem, researchers often use nonrandom sampling methods (which can result in sampling artifacts) or pool capture occasions (which results in a loss of information). We conducted two spatially and temporally overlapping mark–recapture studies on Milksnakes (Lampropeltis triangulum). Both studies used fixed trap locations (artificial cover objects) but varied in their sampling duration and intensity. We used multimodel inference within an information-theoretic approach to estimate adult annual survival (0.72 ± 0.160 standard error [SE]) and abundance (N = 85 ± 35.2 SE). To estimate density, we used a spatially explicit model (1.19 individuals/ha ± 0.331 SE). We applied a novel approach to address sampling bias that resulted from the use of fixed trap locations. This involved modeling maximum detection frequency as a covariate, which better accounted for individual heterogeneity in detection probabilities than did mixtures, sex, behavior, or month. This approach allowed for the recovery of information that was lost when capture occasions were pooled to account for low detection. We found support for this technique via reanalysis of datasets of known population size. This approach may be useful for mark–recapture studies that rely on fixed trap locations.