Assessing spatial uncertainty associated with forest fire boundary delineation
Assessing spatial uncertainty associated with forest fire boundary delineation
复制标题
DOI：
10.1007/s10980-005-0071-7
复制
发表时间：
2005-09-01
影响因子：
5.2
通讯作者：
Lertzman, KP
中科院分区：
文献类型：
--
作者：
Jordan, GJ;Fortin, MJ;Lertzman, KP
管理森林景观的不确定性来自许多来源，包括森林生态系统及其干扰过程固有的复杂性。然而，在收集全面、有代表性和准确的数据集方面的局限性往往阻碍了在景观一级了解森林生态系统。关于过去动乱的历史参考数据集也大多缺乏。然而，在地面火灾的情况下，可以通过使用树木年代学分析火灾痕迹来获得过去火灾的记录。虽然可以确定干扰的时间序列，但个别森林表面火灾的空间界限仍然不确定。在这里，我们研究如何补丁为基础的方法（模糊集成员资格）和基于边界的不确定性方法（边界成员资格）可以帮助确定森林火灾事件及其边界位置相关的空间不确定性。我们比较这些方法使用火痕数据从黄松（黄松）和花旗松（花旗松）采样在33 1公顷的地块在1500公顷的研究领域内的斯坦河流域（不列颠哥伦比亚省）。基于补丁的火灾地图，使用多个约束，从1785年到1937年。我们将所得到的总火灾事件图与基于边界的方法进行了比较，发现根据基于块的方法所选择的值，存在负相关性（尽管非常适度：r = -0. 1，p
Uncertainty in managing forested landscapes arises from many sources, including complexities inherent in forest ecosystems and their disturbance processes. However, gaining knowledge about forested ecosystems at the landscape level is often impeded by limitations in collecting comprehensive, representative, as well as accurate data sets. Historical reference data sets about past disturbances are also mostly lacking. In the case of ground fires, however, records of past fires can be obtained by analyzing fire scars using dendrochronology. While the temporal series of disturbance can be determined, there is still uncertainty about the spatial limits of individual forest surface fires. Here, we investigate how a patch-based method (fuzzy set membership) and a boundary-based uncertainty method (boundary membership) can help determine the spatial uncertainty related to forest fire events and their boundary locations. We compare these methods using fire scar data from ponderosa pine (Pinus ponderosa) and Douglas-fir (Pseudotsuga menziesii) sampled at 33 1-ha plots in a 1500-ha study area within the Stein River watershed (British Columbia). Patch-based fire maps, using multiple constraints, were derived for years 1785-1937. We compared the resulting total fire event maps with the boundary-based method, finding that depending on values chosen for the patch-based method, negative correlation was present (though very modest: r = -0. 1, p