We investigated the potential for using mark–recapture models to estimate abundance of bottlenose dolphin populations in open systems (e.g., bays, estuaries). A major challenge in these systems is that immigration and emigration occur during sampling, thus violating one of the most basic assumptions of mark–recapture models. We assumed that dolphins using our study site were composed of both residents (those that used the study area almost exclusively during our study), and transients (those that passed through our study area but did not remain long), and examined several mark–recapture estimators for their ability to accurately and precisely estimate the abundance of residents and the superpopulation (i.e., residents + transients). Using simulated data, we found that a novel approach accounting for transients resulted in estimators with less bias, smaller absolute relative error, and confidence interval coverage closer to nominal than other approaches, but this novel approach required intensive sampling and that the “correct” transient pattern be specified. In contrast, classical mark–recapture estimators for closed populations often overestimated the number of residents and underestimated the superpopulation. Using photo-identification records, a model-averaged estimate of the superpopulation of bottlenose dolphins in and around Choctawhatchee Bay, Florida was 232 (SE = 13) animals. We estimated resident abundance at 179 (SE = 8), which was lower than the number of unique animals we encountered (188). Our results appear promising for developing monitoring programs for bottlenose dolphins and other taxa in open systems. Our estimators should prove useful to wildlife managers who wish to base conservation decisions on estimates of the number of animals that reside primarily in their study or management area. © 2011 The Wildlife Society.