Advances in molecular biology offer promise to the study of demographic characteristics of rare or hard-re-capture species, because individuals can now be identified through noninvasive sampling such as fecal collection or hair snags. However, individual genotyping using such methods currently leads to a novel problem that we call a "shadow effect," because some animals not captured previously are believed to be recaptures due to their DNA profile being an indistinguishable shadow of previously captured animals. We evaluate the impact of the shadow effect on the two methods most commonly used in applied population ecology to estimate the size of closed populations: Lincoln-Petersen and multiple-recapture estimators in program CAPTURE. We find that the shadow effect can cause a negative bias in the estimates of both the number of different animals and the number of different genotypes. Furthermore, with Lincoln-Petersen estimators, the shadow effect can cause estimated confidence intervals to decrease even as bias increases. Because the bias arises from heterogeneity in apparent "capture" probabilities for animals with genetic shadows vs. those without, a model in program CAPTURE that is robust to capture heterogeneity (Mh-jackknife) does not underestimate the number of genotypes in the population and only slightly underestimates the rotal number of individuals As the shadow effect increases, CAPTURE is better able to correctly identify heterogeneity in capture probability and to pick Mh-jackknife, so that the higher levels of shadow effect have less bias than medium levels. The shadow effect will occur in all estimates of demographic rates (including survival) that use DNA sampling to determine individual identity, but it can be minimized by increasing the number of individual loci sampled.
© 2000 by the Ecological Society of America. L. Scott Mills, John J. Citta, Kevin P. Lair, Michael K. Schwartz, and David A. Tallmon 2000. ESTIMATING ANIMAL ABUNDANCE USING NONINVASIVE DNA SAMPLING: PROMISE AND PITFALLS. Ecological Applications 10:283–294. http://dx.doi.org/10.1890/1051-0761(2000)010[0283:EAAUND]2.0.CO;2.