A multi-sensor, multi-scale approach to mapping tree mortality in woodland ecosystems
A multi-sensor, multi-scale approach to mapping tree mortality in woodland ecosystems
复制标题
DOI：
10.1016/j.rse.2020.111853
复制
发表时间：
2020-08-01
影响因子：
13.5
通讯作者：
Anderegg, William R. L.
中科院分区：
文献类型：
--
作者：
Campbell, Michael J.;Dennison, Philip E.;Anderegg, William R. L.
占全球陆地面积近4%的林地生态系统正面临着各种威胁，包括持续时间越来越长的严重干旱、入侵昆虫的爆发以及病原体的快速传播。虽然已经开发了许多遥感方法来检测和量化森林环境中的死亡率，林地生态系统提出了独特的挑战，准确地映射树木死亡，由于相对较低的树冠覆盖，较小的和不规则形状的树冠，以及林下植被和土壤覆盖的反射率的影响更大。为了应对这些挑战，我们开发了一种多传感器，多尺度的方法，结合厘米分辨率无人机系统图像的分析优势，用于解释个体树级死亡率，机载激光雷达用于树冠测绘和量化树冠死亡率百分比，以及Landsat图像用于将死亡率估计提升到区域范围。这种方法利用了一种新的算法来描绘小，不规则的林地树冠使用激光雷达的形状。为了证明这种方法的应用，我们映射的范围和严重程度最近的树木死亡事件在松树，杜松（PJ）林地东南部犹他州。我们的研究结果表明，39%的PJ在该地区经历了一定程度的死亡率，与补丁超过50%的死亡率。对潜在死亡率驱动因素的分析显示，冠层覆盖、地形和近期冬季降水条件与死亡率最直接相关，尽管死亡率驱动因素模型的解释力较低。我们的方法表明，可用于树木死亡率映射和缩放在各种林地生态系统的方法，并可以提供一个强有力的基础，进一步的生态生理，生态和碳循环研究，涉及林地树木死亡率。
Woodland ecosystems, dominant on nearly 4% of all terrestrial land globally, are faced with a variety of threats, including increasingly prolonged and severe droughts, invasive insect outbreaks, and the rapid spread of pathogens. While many remote sensing methods have been developed for the detection and quantification of mortality in forested environments, woodland ecosystems present unique challenges to accurately mapping tree die-off due to relatively lower canopy covers, smaller and irregularly-shaped tree crowns, and greater influence of understory vegetation and soil cover on reflectance. To address these challenges, we developed a multi-sensor, multi-scale approach combining the analytical strengths of centimeter-resolution unmanned aerial system imagery for interpreting individual tree-level mortality, airborne lidar for crown mapping and quantifying percent canopy mortality, and Landsat imagery for upscaling mortality estimates to a regional scale. This approach utilizes a new algorithm for delineating the shapes of small, irregular woodland tree crowns using lidar. To demonstrate the application of this method, we map the extent and severity of a recent tree mortality event in pinon-juniper (PJ) woodlands of southeastern Utah. Our results suggest that 39% of PJ in this region has experienced some level of mortality, with patches exceeding 50% mortality. An analysis of potential mortality drivers revealed that canopy cover, terrain, and recent winter precipitation conditions are most directly linked with mortality, although the explanatory power of the mortality driver model was low. Our approach demonstrates a methodology that could be used for tree mortality mapping and scaling in a variety woodland ecosystems, and can provide a strong basis for further ecophysiological, ecological, and carbon cycle studies involving woodland tree mortality.