Natural tags based on DNA fingerprints or natural features of animals are now becoming very widely used in wildlife population biology. However, classic capture-recapture models do not allow for misidentification of animals which is a potentially very serious problem with natural tags. Statistical analysis of misidentification processes is extremely difficult using traditional likelihood methods but is easily handled using Bayesian methods. We present a general framework for Bayesian analysis of categorical data arising from a latent multinomial distribution. Although our work is motivated by a specific model for misidentification in closed population capture-recapture analyses, with crucial assumptions which may not always be appropriate, the methods we develop extend naturally to a variety of other models with similar structure. Suppose that observed frequencies f are a known linear transformation f=A'x of a latent multinomial variable x with cell probability vector pi=pi(theta). Given that full conditional distributions [theta | x] can be sampled, implementation of Gibbs sampling requires only that we can sample from the full conditional distribution [x | f, theta], which is made possible by knowledge of the null space of A'. We illustrate the approach using two data sets with individual misidentification, one simulated, the other summarizing recapture data for salamanders based on natural marks.

Abstract. Misidentification of animals is potentially important when naturally existing features (natural tags) are used to identify individual animals in a capture-recapture study. Photographic identification (photoID) typically uses photographic images of animals' naturally existing features as tags (photographic tags) and is subject to two main causes of identification errors: those related to quality of photographs (non-evolving natural tags) and those related to changes in natural marks (evolving natural tags). The conventional methods for analysis of capture-recapture data do not account for identification errors, and to do so requires a detailed understanding of the misidentification mechanism. Focusing on the situation where errors are due to evolving natural tags, we propose a misidentification mechanism and outline a framework for modeling the effect of misidentification in closed population studies. We introduce methods for estimating population size based on this model. Using a simulation study, we show that conventional estimators can seriously overestimate population size when errors due to misidentification are ignored, and that, in comparison, our new estimators have better properties except in cases with low capture probabilities (,0.2) or low misidentification rates (,2.5%).

Summary1. For many species, noninvasive photographic identification offers a powerful and cost-effective method for estimating demographic parameters and testing ecological hypotheses in large populations. However, this technique is prone to misidentification errors that can severely bias capturerecapture estimates.2. We present a simple ad hoc data conditioning technique that minimizes bias in survival estimates across all rates of misidentification. We use simulated data sets to characterize trade-offs in bias, precision and accuracy of survival estimators for a range of misidentification probabilities, sampling intensities, survival rates and population sizes using this conditional approach. 3. Misidentification errors resulted in mean survival estimates that were negatively biased by as much as )24AE9% when errors were ignored. Applying the conditional approach resulted in very low levels of bias across parameter space. However, the main cost of conditioning is a loss of precision, which was particularly severe at low sampling intensities. Overall, the conditional approach was superior to the nonconditional approach [in terms of root mean square error (RMSE) of survival estimates] in 51% of the parameter combinations that we explored. 4. We apply the data conditioning technique to a 3-sample capture-recapture data set compiled from 2551 images of a migratory wildebeest, Connochaetes taurinus, population in northern Tanzania. We estimate the false rejection rate (i.e., the probability of failing to match two photographs of the same individual) using a test set of 'known-identity' individuals. With this information, we compare survival estimates derived from conditioned data (û = 0AE698 ± 0AE176), unconditioned data (û = 0AE706 ± 0AE121) and simulated data to illustrate some of the key considerations for deciding whether to apply a conditional approach to a photographic data set.5. These analyses demonstrate that ignoring misidentification error can lead to substantial bias in survival estimates. When sampling intensity and misclassification error rates are both relatively high, use of our conditioned data approach is preferred and yields survival estimates with lower RMSE. However, when sampling intensity and misclassification error are both small, the standard approach using unconditioned data yields smaller RMSE.

Development and use of multistate mark-recapture models, which provide estimates of parameters of Markov processes in the face of imperfect detection, have become common over the last 20 years. Recently, estimating parameters of hidden Markov models, where the state of an individual can be uncertain even when it is detected, has received attention. Previous work has shown that ignoring state uncertainty biases estimates of survival and state transition probabilities, thereby reducing the power to detect effects. Efforts to adjust for state uncertainty have included special cases and a general framework for a single sample per period of interest. We provide a flexible framework for adjusting for state uncertainty in multistate models, while utilizing multiple sampling occasions per period of interest to increase precision and remove parameter redundancy. These models also produce direct estimates of state structure for each primary period, even for the case where there is just one sampling occasion. We apply our model to expected-value data, and to data from a study of Florida manatees, to provide examples of the improvement in precision due to secondary capture occasions. We have also implemented these models in program MARK. This general framework could also be used by practitioners to consider constrained models of particular interest, or to model the relationship between within-primary-period parameters (e.g., state structure) and between-primary-period parameters (e.g., state transition probabilities).

Misidentification of animals is potentially important when naturally existing features (natural tags) such as DNA fingerprints (genetic tags) are used to identify individual animals. For example, when misidentification leads to multiple identities being assigned to an animal, traditional estimators tend to overestimate population size. Accounting for misidentification in capture-recapture models requires detailed understanding of the mechanism. Using genetic tags as an example, we outline a framework for modeling the effect of misidentification in closed population studies when individual identification is based on natural tags that are consistent over time (non-evolving natural tags). We first assume a single sample is obtained per animal for each capture event, and then generalize to the case where multiple samples (such as hair or scat samples) are collected per animal per capture occasion. We introduce methods for estimating population size and, using a simulation study, we show that our new estimators perform well for cases with moderately high capture probabilities or high misidentification rates. In contrast, conventional estimators can seriously overestimate population size when errors due to misidentification are ignored.