Estimating impacts of poison operations using mark-recapture analysis: hihi (Notiomystis cincta) on Mokoia Island

Hihi (Notiomystis cincta) were reintroduced to Mokoia Island, Lake Rotorua, New Zealand, in September 1994, and two years later there was an aerial drop of brodifacoum cereal pellets aimed to eradicate mice (Mus musculus). Using Program MARK, we analyzed data from resighting surveys to assess whether hihi had lower than normal survival in the 6-week interval following the drop. The resighting data were collected on a regular basis over a 3-year period, from 1994-97, allowing us to control for yearly and seasonal variation in resighting and survival probabilities. We initially established that the Cormack-Jolly- Seber model had a good fit to the data and could therefore be used as the global model for our analysis. We then compared a range of simpler candidate models, some of which included a poison effect (an unusual survival rate for the interval after the poison drop). Under the best model (that with the lowest AIC), the survival probability was constant over time and there was no poison effect. The estimated survival probability for the 6 weeks after the poison drop was 0.95, which is slightly higher than the value of 0.89 expected based on pre- and post- poison intervals. The approximate 95% confidence interval for the probability of a bird dying due to poison ranged from -0.17 (i.e., a decrease in mortality rate due to the poison) to +0.04. We therefore concluded that the poison caused at most a negligible increase in mortality, and that mark-recapture analysis on resighting data provided a powerful method for assessing the impact of the poison drop. We discuss the relative costs and benefits of radio tagging versus resighting surveys of banded birds for estimating impacts of poison operations. For species with relatively high resighting rates, such as hihi, analysis of resighting surveys is a much more reliable and cost-efficient methodology.