Abstract In the Pacific Northwest of the United States, 20th century timber harvesting resulted in major declines in area of forests supporting large live and dead trees (i.e., snags), that are not only key habitat elements for many wildlife species but also critical components of ecosystem function. Regional forest management guidance, such as the Northwest Forest Plan (1994) and Eastside Screens (1995), may aim to conserve and foster the development of late-successional old-growth forests, characterized by large live and dead trees. Satellite remote sensing supports regional monitoring efforts of these habitat characteristics, but managers may require additional guidance in order to leverage these data for large landscape or regional assessments. In this study, our objectives were to assess long-term (historical vs. contemporary) and short-term (1993–2017) changes in lands supporting large live trees and snags across 10 wildlife habitat types (WHTs) – disparate vegetation conditions that support significantly different wildlife communities – in Oregon and Washington, USA. We generated 30-m, annual maps of large live trees (>50 cm, >75 cm, and >100 cm diameter) and snags (>25 cm and > 50 cm diameter) based on the gradient nearest neighbor (GNN) imputation method. GNN integrates Landsat satellite imagery, geospatial climatic and topographic data, and USDA Forest Service Forest Inventory and Analysis data to predict forest attributes for all forested lands in the study area. GNN classification accuracy was poor to good for large live trees (Cohen’s kappa = 0.2 – 0.6) and poor to fair for snags (Cohen’s kappa = 0.1 – 0.3) in most WHTs, though performance was substantially lower in drier WHTs where large live trees and snags were rare. Our results highlighted long-term reductions in forest supporting large live trees and snags from historical to contemporary times, especially in wetter, more productive WHTs. In contrast, we observed short-term (1993–2017) increases in areas supporting large live trees and snags. Federal forests were both more similar to reference conditions and exhibited greater recent increases in areas supporting large live trees compared to nonfederal lands. Thus, Oregon and Washington have lost a substantial proportion of forests containing large live trees and snags and recent recruitment of these trees at regional scales is a slow process primarily occurring on federal lands. However, detecting such changes through current Landsat satellite mapping technologies remains challenging, highlighting the need for new mapping methods to aid in future management.

中文翻译：

---

量化大型活树的区域趋势和支持森林管理的障碍物的可用性

摘要 在美国太平洋西北部，20 世纪的木材采伐导致支持大型活树和死树（即树桩）的森林面积大幅减少，这些树木不仅是许多野生动物物种的关键栖息地要素，也是生态系统功能。区域森林管理指南，如西北森林计划（1994 年）和东边屏风（1995 年），可能旨在保护和促进后期连续生长的老林的发展，其特征是大型活树和死树。卫星遥感支持这些栖息地特征的区域监测工作，但管理人员可能需要额外的指导，以便利用这些数据进行大型景观或区域评估。在这项研究中，我们的目标是评估长期（历史与历史）。当代）和短期（1993 年至 2017 年）在美国俄勒冈州和华盛顿州 10 种野生动物栖息地类型 (WHT) 中支持大型活树和障碍物的土地变化 - 支持显着不同野生动物群落的不同植被条件。我们基于梯度最近邻 (GNN) 插补方法生成了 30 米的大型活树（> 50 厘米、> 75 厘米和 > 100 厘米直径）和障碍（> 25 厘米和 > 50 厘米直径）的年度地图. GNN 整合了 Landsat 卫星图像、地理空间气候和地形数据以及美国农业部林务局森林清查和分析数据，以预测研究区域内所有林地的森林属性。在大多数 WHT 中，GNN 分类准确度对于大型活树（Cohen's kappa = 0.2 – 0.6）从差到好，对于障碍（Cohen's kappa = 0.1 – 0.3）从差到公平，尽管在干燥的 WHT 中性能要低得多，那里的大型活树和障碍物很少见。我们的结果强调了从历史到现代，支持大型活树和障碍物的森林的长期减少，尤其是在更潮湿、生产力更高的 WHT 中。相比之下，我们观察到支持大型活树和障碍物的区域在短期（1993-2017 年）有所增加。与非联邦土地相比，联邦森林与参考条件更相似，并且最近在支持大型活树的地区方面表现出更大的增长。因此，俄勒冈州和华盛顿州已经失去了大量包含大型活树和障碍物的森林，最近在区域范围内补充这些树木是一个缓慢的过程，主要发生在联邦土地上。然而，