Ramsey David S L, Parkes John, Morrison Scott A
Landcare Research, Private Bag 11052, Palmerston North, New Zealand.
Conserv Biol. 2009 Apr;23(2):449-59. doi: 10.1111/j.1523-1739.2008.01119.x. Epub 2008 Nov 17.
A major challenge facing pest-eradication efforts is determining when eradication has been achieved. When the pest can no longer be detected, managers have to decide whether the pest has actually been eliminated and hence to decide when to terminate the eradication program. For most eradication programs, this decision entails considerable risk and is the largest single issue facing managers of such programs. We addressed this issue for an eradication program of feral pigs (Sus scrofa) from Santa Cruz Island, California. Using a Bayesian approach, we estimated the degree of confidence in the success of the eradication program at the point when monitoring failed to detect any more pigs. Catch-effort modeling of the hunting effort required to dispatch pigs during the eradication program was used to determine the relationship between detection probability and searching effort for different hunting methods. We then used these relationships to estimate the amount of monitoring effort required to declare eradication successful with criteria that either set a threshold for the probability that pigs remained undetected (type I error) or minimized the net expected costs of the eradication program (cost of type I and II errors). For aerial and ground-based monitoring techniques, the amount of search effort required to declare eradication successful on the basis of either criterion was highly dependent on the prior belief in the success of the program unless monitoring intensities exceeded 30 km of searching effort per square kilometer of search area for aerial monitoring and, equivalently, 38 km for ground monitoring. Calculation of these criteria to gauge the success of eradication should form an essential component of any eradication program as it allows for a transparent assessment of the risks inherent in the decision to terminate the program.
根除害虫工作面临的一个主要挑战是确定何时实现了根除。当再也检测不到害虫时，管理人员必须决定害虫是否真的已被消灭，从而决定何时终止根除计划。对于大多数根除计划而言，这一决策存在相当大的风险，并且是此类计划管理人员面临的最大单一问题。我们针对加利福尼亚州圣克鲁斯岛野猪（Sus scrofa）的根除计划解决了这个问题。我们采用贝叶斯方法，在监测未能再检测到任何野猪时，估计根除计划成功的置信度。利用根除计划期间捕杀野猪所需的捕猎努力的捕获努力建模，来确定不同捕猎方法下检测概率与搜索努力之间的关系。然后，我们利用这些关系，根据设定野猪未被检测到的概率阈值（I型错误）或最小化根除计划的净预期成本（I型和II型错误的成本）的标准，来估计宣布根除成功所需的监测努力量。对于空中和地面监测技术，基于任何一个标准宣布根除成功所需的搜索努力量高度依赖于对该计划成功的先验信念，除非监测强度超过每平方公里搜索区域30公里的空中监测搜索努力量，以及同等的地面监测38公里的搜索努力量。计算这些评估根除成功的标准应成为任何根除计划的重要组成部分，因为它允许对终止计划决策中固有的风险进行透明评估。
