Confirming successful eradication of rats from Isabel Island, Mexico
Looking across Isabel Island. Image - J.A. Soriano / GECI archive.
A central problem in the science of pest eradication is how to determine whether eradication has been successful. This is most problematic on offshore islands where access is often difficult and costly. One approach is to ‘wait and see’, in which a survey is conducted after enough time has elapsed for an easily detectable population to have grown from any survivors. Failure in this case means a renewed eradication operation to remove the recovered population. A better approach is to confirm success immediately following an operation, as failure in this case means that the few survivors are spatially localised and can be removed with low-cost mop-up operations. Here, Dean Anderson working with Araceli Herrera of the Conservación de Islas, México, use data from an eradication operation of black (ship) rats from Isabel Island, Mexico, to illustrate a spatial-survey model for confirming eradication. Isabel Island (82 ha) is located in the entrance of the Gulf of California, Mexico (Fig. 1). The island supports a rich vertebrate biodiversity and is an important breeding site for nine species of sea birds. On 1 and 7 May 2009 compressed-grain bait loaded with the toxin brodifacoum was distributed by helicopter across the interior of the island, and by hand along the coast. At 12, 19, 24 and 30 months following the last baiting, 17 wax tags were deployed at 200m spacings to detect surviving rats (Fig. 2) and checked after 5 nights.
The spatial-survey model uses the detection-survey data in which no pests are detected. Dean and Araceli adopted a worst-case scenario for detecting failed eradication by assuming that a single pregnant female survived. Clearly a single animal is harder to detect than multiple animals, but a pest population can recover from a single mother. They asked the model: if a single pregnant female remained on the island, what is the probability that it would be detected with the array of wax tags deployed?
The model randomly placed the home-range centre of a pregnant rat on the island (Fig. 2). The probability that the array of wax tags would detect this female was then calculated, based on the home-range size of rats and the probability of a rat chewing a tag. If some months have elapsed since the eradication operation, the model estimates the probability of detecting her progeny, based on population growth rates and individual dispersal. Bayesian logic is subsequently used to produce a probability of eradication success, given no chewed wax tags.
Dean and Araceli incorporate the uncertainty associated with the parameters and propagate this uncertainty through to the results, which influence interpretation and management decisions. To do this, they repeated the model 1000 times and each time chose a new random starting location, and new values from each of the parameter distributions.
The first confirmation survey was conducted 12 months after baiting. Given the negative survey results at that time, the median probability of success was 0.91 (1.0 = 100% success), and the lower confidence nterval 0.69, which indicated low confidence in success and high uncertainty (Table). All subsequent surveys indicated that the eradication had been successful. To achieve a satisfactory probability of success immediately following the initial operation, the wax tags needed to be spaced at 50m intervals (Fig. 3). This would require 16 times the actual number of wax tags deployed, or 272 wax tags.
Table. Median and 95% credible intervals (CI) of estimated probability of eradication following wax tag survey
| Probability of success after: | Median | Lower CI | Upper CI |
| 12 months | 0.91 | 0.69 | 0.98 |
| 19 months | 1.00 | 0.99 | 1.00 |
| 24 months | 1.00 | 0.99 | 1.00 |
| 30 months | 1.00 | 0.99 | 1.00 |
Survey data from Isabel Island illustrates how the spatial-survey model can be used to make eradication projects more cost effective. A survey based on wax tags at 50m spacings conducted on the island immediately following baiting would have provided strong evidence of success and avoided the need to conduct subsequent surveys. More importantly, the model can be used prior to future operations to determine the survey effort required to confirm success. The benefits of this are clear. Funders are supplied with accurate estimates of project costs and objective measures of success. Further, if survivors are detected immediately after the operation, the small number present is likely to be easier to remove than a fully recovered population detected at a later date. Lastly, if the management team is confident of eradication success, subsequent restoration plans, such as the reintroduction of endangered species, can be implemented sooner rather than later.
This project was led by Conservación de Islas with the support of Mexican Government agencies and funding from national and international foundations.
Dean Anderson
Araceli Samaniego Herrera (Conservación de Islas, Mexico)
Fig 1. Location of Isabel Island, Mexico.
Fig. 2. Isabel Island showing the locations of wax tags (small black circles). The red star indicates the single randomly located pregnant female and the red circles represent eight dispersed offspring modelled 1 year after baiting.
Fig. 3. The median and 95% credible intervals (CI) of the probability of operational success decrease with increasing spacing of wax tags (devices). The horizontal dashed line is the threshold above which the lower 95% CI should be above.
