Putting a cart before the search: Successful habitat prediction for a rare forest herb
Putting a cart before the search: Successful habitat prediction for a rare forest herb
复制标题
DOI：
10.1890/04-1666
复制
发表时间：
2005-10-01
期刊：
影响因子：
4.8
通讯作者：
Gill, DE
中科院分区：
文献类型：
--
作者：
Bourg, NA;McShea, WJ;Gill, DE
稀有物种保护和超种群生物学理论的领域往往是相互关联的，因此共享几个基本的挑战。其中最重要的两项任务是先验地区分栖息地和非栖息地，然后在研究区域划定合适的栖息地斑块。我们结合分类树分析，分类和回归树（CART）模型的一个子集，以及地理信息系统（GIS）中环境变量的数字数据层，预测了土耳其胡（Xerophyllum asphodeloides）的适宜栖息地和潜在的新种群发生，土耳其胡是一种罕见的与南阿帕拉契亚松栎（Pinus-Quercus）森林相关的萱草类林下草本植物。在建模过程中，利用8个环境数据层的样本值和人口调查数据，生成了一个交叉验证的分类树，预测了研究区域的适宜栖息地。高程、坡度、森林类型和火灾频率是模型中的四个主要解释变量。约4%的研究区被划分为5个适宜的生境类，误分类错误率为4.74%。最终的13叶树正确分类了74%的已知存在区和90%的已知缺失区，地面调查结果发现了8个新的栖息地斑块。本研究结果对于阿巴拉契亚森林中疏叶松的保护和管理，以及栖息地模拟技术在加强对疏叶松超种群和干扰机制研究中的适用性具有重要意义。此外，它们还证实了基于CART和gis的物种分布建模方法的潜力和价值。我们的模型成功地定义了合适的栖息地，并在景观尺度上发现了一种稀有物种的新种群。对其他稀有物种的类似应用可能对解决这些和其他生态和保护问题非常有用，例如规划移植或重新引入实验，确定超种群碎片化阈值，以及制定保护策略。
The realms of rare species conservation and metapopulation biology theory are often interrelated, and hence share several basic challenges. Two of the most important are the critical and frequently difficult tasks of distinguishing a priori between habitat and nonhabitat, and then delimiting suitable habitat patches in a study area. We combined classification tree analysis, a subset of classification and regression tree (CART) modeling, with digital data layers of environmental variables in a geographic information system (GIS) to predict suitable habitat and potential new population occurrences for turkeybeard (Xerophyllum asphodeloides), a rare liliaceous understory herb associated with southern Appalachian pine-oak (Pinus-Quercus) forests, in northwestern Virginia. Sample values from eight environmental data layers and population survey data were used in the modeling process to produce a cross-validated classification tree that predicted suitable habitat in the study area. Elevation, slope, forest type, and fire frequency were the four main explanatory variables in the model. Approximately 4% of the study area was classified into five suitable habitat classes, with a misclassification error rate of 4.74%. The final 13-leaf tree correctly classified 74% of the known presence areas and 90% of the known absence areas, and ground-truthing surveys resulted in the discovery of eight new occupied habitat patches. Results of this study are important for conservation and management of X. asphodeloides, as well as for the applicability of the habitat modeling techniques to enhancing the study of metapopulations and disturbance regimes in Appalachian forests. In addition, they confirm the potential and value of CART and GIS-based modeling approaches to species distribution problems. Our model was successful at defining suitable habitat and discovering new populations of a rare species at the landscape scale. Similar application to other rare species could prove very useful for addressing these and other ecological and conservation issues, such as planning transplantation or reintroduction experiments, identifying metapopulation fragmentation thresholds, and formulating conservation strategies.