In old-growth forests, natural disturbances form a complex mosaic of structures, providing a wide diversity of habitats and functions of great importance. Old-growth forests are still often seen as a homogeneous whole and few remote-sensing approaches have been tested to identify their structural diversity, especially in boreal forests. The aim of this study is to use a combination of airborne LiDAR and satellite imagery to identify and discriminate old-growth forest structures resulting from different disturbance histories. The study area, which was located in the mixed boreal forest of Quebec (Canada), is Monts Valin National Park and adjacent managed territories. Balsam fir (Abies balsamea (L.) Mill) is the dominant species in the study area, but hardwood species such as white birch (Betula papyrifera Marsh.) and trembling aspen (Populus tremuloides Michx.) can also be abundant in the early succession stages. Four forest classes were studied: second-growth (logged between 1970 and 1980); transition old-growth (burned in 1920); undisturbed old-growth (unburned for at least 125 years); and disturbed old-growth forest (unburned for at least 125 years, but severely disturbed by an insect outbreak around 1980). A multivariate Random Forest model was used to discriminate the classes on 6466 1 ha tiles, based on 11 complementary LiDAR and satellite-derived indices describing stand vertical and horizontal structure, together with “greenness” and disturbance history over the last 30 years. This model had high predictive efficiency (AUC = 94.2), with 81.8% of the tiles accurately classified. Interestingly, undisturbed old-growth forests exhibited intermediate characteristics compared to transition and disturbed old-growth forests. This emphasizes that some structural attributes recognized as important for the classification of temperate and tropical old-growth forests, such as high vertical complexity, are of lesser relevance for boreal old-growth forests. In comparison to undisturbed old-growth forests, transition old-growth forests had a taller canopy of high “greenness” due to a greater hardwood abundance; disturbed old-growth forests had a higher gap fraction and heterogeneity in tree size; second-growth forests exhibited a lower and more even canopy. Misclassified tiles were explained by spatial variation in disturbance severity or different levels of forest resistance and resilience to disturbance. These misclassifications are also of ecological interest, as they highlight the nuances in structural diversity that are rarely identified by disturbance mapping. A reasonable combination of LiDAR and satellite indices was effective not only in discriminating old-growth forests from second-growth forests, but also identifying their different structures, which result from specific disturbance histories. This method could contribute to effective monitoring of changes in the areas and characteristics old-growth forest that are caused by anthropogenic and natural disturbances.

在原始森林中，自然干扰形成了复杂的镶嵌结构，提供了非常重要的栖息地和功能的广泛多样性。古老的森林仍然经常被视为一个同质的整体，很少有遥感方法经过测试来确定其结构多样性，尤其是在北方森林中。本研究的目的是结合机载 LiDAR 和卫星图像来识别和区分由不同干扰历史引起的古老森林结构。研究区位于魁北克（加拿大）的北方混交林中，是 Monts Valin 国家公园和邻近的管理领土。香脂冷杉（Abies balsamea (L.) Mill）是研究区的优势树种，但硬木树种如白桦（Betula papyrifera Marsh.）和颤杨（Populus tremuloidesMichx.) 在早期演替阶段也可能很丰富。研究了四种森林类别：第二次生长（1970 年至 1980 年间采伐）；过渡老式生长（1920 年烧毁）；未受干扰的古树（至少 125 年未燃烧）；和受干扰的原始森林（至少 125 年未被烧毁，但在 1980 年左右受到昆虫爆发的严重干扰）。基于 11 个互补的 LiDAR 和卫星衍生的描述林分垂直和水平结构的指数，以及过去 30 年的“绿度”和干扰历史，使用多元随机森林模型来区分 6466 1 公顷瓦片上的类别。该模型具有较高的预测效率（AUC = 94.2），81.8% 的瓷砖准确分类。有趣的是，与过渡和受干扰的老生林相比，未受干扰的老生林表现出中间特征。这强调了一些被认为对温带和热带原始森林分类很重要的结构属性，例如高度的垂直复杂性，与北方原始森林的相关性较小。与未受干扰的原始森林相比，过渡原始森林的树冠更高，“绿度”更高，因为硬木丰度更大；受干扰的原始森林在树木大小方面具有更高的间隙分数和异质性；次生林的树冠更低，更均匀。错误分类的瓦片可以通过干扰严重程度的空间变化或不同水平的森林抵抗力和干扰恢复力来解释。这些错误分类也具有生态学意义，因为它们突出了结构多样性中的细微差别，而这些细微差别很少被干扰映射识别。激光雷达和卫星指数的合理组合不仅可以有效区分老林和次生林，还可以识别它们因特定干扰历史而产生的不同结构。该方法有助于有效监测人为和自然干扰引起的原始森林面积和特征的变化。还要确定它们不同的结构，这些结构是由特定的干扰历史引起的。该方法有助于有效监测人为和自然干扰引起的原始森林面积和特征的变化。还要确定它们不同的结构，这些结构是由特定的干扰历史引起的。该方法有助于有效监测人为和自然干扰引起的原始森林面积和特征的变化。