Fecal genotyping to estimate small mammal population size, with a comparison to live mark-recapture estimates

Live capture-recapture is often considered the gold standard for estimating wildlife population size or density, but the approach can be limited by permitting requirements, required labor, welfare concerns, and biased estimates resulting from heterogeneity in individual behavior. Noninvasive genetic sampling (e.g., from fecal pellets) offers a powerful alternative approach, but this method's success varies among taxa, with little research available on its use in rodents. Here, we addressed a series of questions to develop a noninvasive genetic sampling approach for the endangered giant kangaroo rat ( Dipodomys ingens ): (1) how quickly does DNA degrade in natural conditions, (2) how many pellets are required to recover a genotype, (3) how often do multiple individuals contaminate a pooled sample from a single sampling location, and (4) how do variable and parameter estimates from noninvasive genetic sampling compare to live-trapping mark-recapture estimates? We found that fecal pellets were successfully genotyped up to 9 days (estimated probability of recovery = 0.78) after exposure to hot, arid conditions, but that rate fell precipitously soon after. Although giant kangaroo rats are territorial, multiple individuals deposited fecal pellets at the same sampling locations; however, single pellets contained sufficient DNA to recover genotypes and to identify individuals, so contamination was not a problem for this approach. Capture probabilities were lower using noninvasive genetic sampling ( = 0.26, SE = 0.01) than live trapping ( = 0.40, SE = 0.06). Population estimates were generally similar using noninvasive genetic sampling, although they were quite a bit higher ( = 64 , = 38) on one grid. Noninvasive genetic sampling can overcome many of the limitations of live-trapping for small mammals, but the approach should be tested in additional taxa and systems to provide more generalizable recommendations for sampling schemes.