Estimating impacts of poison operations using mark-recapture analysis and population viability analysis: an example with New Zealand robins (Petroica australis)

Several recent studies have used “roll calls” – searches for individually-marked birds – to assess impacts of aerial poison operations on non-target species. Roll calls have advantages over methods such as 5-minute bird counts, call counts, and dead body counts, but roll calls are based on the assumption that detection rates are 100%, or that detection rates are constant over time and space. They also require more than one group of birds, at a poisoned and unpoisoned site for example, for valid statistical comparisons. With minor adjustment of field methods, however, a series of surveys can be treated as a mark-recapture experiment, allowing powerful analysis without such restrictive assumptions. Survival and detection rates can be estimated independently for each time interval, and alternative models fitted to the data for factors affecting those rates. Using the software package MARK, we analyzed data for New Zealand robins (Petroica australis) on Tiritiri Matangi Island to estimate impact of an aerial brodifacoum operation conducted in September 1993 to eradicate kiore (Rattus exulans). The population was established by translocations in 1992 and 1993, and consisted of 40 birds at the time of the operation. Taking seasonal and yearly variation in detection and survival rates into account, we estimate that 11% of robins (4-5 birds) died as a result of the poison. There was strong evidence that survival probability was lower immediately after the operation, despite the small number of birds disappearing. The poison effect was confined to the interval from September-October 1993, including the first two weeks after the poison drop, and survival for subsequent monthly intervals was consistent with the normal summer rate. Using simulation modelling based on our data from the population, we predict that the operation set population growth back by about one year but had no longer-term impact. __________________________________________________________________________________________________________________________________