The University of Queensland, School of Geography, Planning and Environmental Management, St. Lucia, Brisbane 4072, Australia.
Conserv Biol. 2010 Aug;24(4):974-83. doi: 10.1111/j.1523-1739.2010.01451.x. Epub 2010 Feb 19.
Adaptive management is an iterative process of gathering new knowledge regarding a system's behavior and monitoring the ecological consequences of management actions to improve management decisions. Although the concept originated in the 1970s, it is rarely actively incorporated into ecological restoration. Bayesian networks (BNs) are emerging as efficient ecological decision-support tools well suited to adaptive management, but examples of their application in this capacity are few. We developed a BN within an adaptive-management framework that focuses on managing the effects of feral grazing and prescribed burning regimes on avian diversity within woodlands of subtropical eastern Australia. We constructed the BN with baseline data to predict bird abundance as a function of habitat structure, grazing pressure, and prescribed burning. Results of sensitivity analyses suggested that grazing pressure increased the abundance of aggressive honeyeaters, which in turn had a strong negative effect on small passerines. Management interventions to reduce pressure of feral grazing and prescribed burning were then conducted, after which we collected a second set of field data to test the response of small passerines to these measures. We used these data, which incorporated ecological changes that may have resulted from the management interventions, to validate and update the BN. The network predictions of small passerine abundance under the new habitat and management conditions were very accurate. The updated BN concluded the first iteration of adaptive management and will be used in planning the next round of management interventions. The unique belief-updating feature of BNs provides land managers with the flexibility to predict outcomes and evaluate the effectiveness of management interventions.
自适应管理是一个迭代过程，旨在收集有关系统行为的新知识，并监测管理行动对生态后果的影响，以改进管理决策。尽管该概念起源于 20 世纪 70 年代，但它很少被积极纳入生态恢复中。贝叶斯网络（BNs）作为一种有效的生态决策支持工具正在兴起，非常适合自适应管理，但将其应用于这种能力的示例很少。我们在自适应管理框架内开发了一个 BN，该 BN 侧重于管理林地中野生放牧和规定燃烧制度对鸟类多样性的影响。我们根据基线数据构建 BN，以预测鸟类数量与栖息地结构、放牧压力和规定燃烧之间的关系。敏感性分析的结果表明，放牧压力增加了攻击性吸蜜鸟的数量，而吸蜜鸟反过来又对小型雀形目鸟类产生了强烈的负面影响。然后进行了减少野生放牧和规定燃烧压力的管理干预措施，之后我们收集了第二组实地数据，以测试这些措施对小型雀形目鸟类的反应。我们使用这些数据（其中包含可能因管理干预而导致的生态变化）来验证和更新 BN。在新的栖息地和管理条件下，小型雀形目鸟类数量的网络预测非常准确。更新后的 BN 完成了自适应管理的第一轮迭代，并将用于规划下一轮管理干预措施。BN 的独特信念更新功能为土地管理者提供了预测结果和评估管理干预措施有效性的灵活性。
