长期以来由于受到人类活动、外来物种入侵和全球气候变化等多重因素的影响，自然生态系统面临着巨大的生境退化和丧失风险。通过遥感手段对自然生态系统的生境结构进行监测和评估已经成为生物多样性保护的共识，然而传统的光学遥感手段无法获取精细完整的森林结构，对于进一步提高森林生境结构准确度存在较大的局限性。尤其在温带针阔混交林，由于复杂和多层级结构，基于机载和卫星平台的遥感很难获取完整精确的林下结构信息，从而限制了对林下生境结构的监测和研究。地基激光雷达技术（Terresrial Laser Scanning，TLS）凭借其能够快速、直接地获取高精度的三维结构信息能力，并提供垂直结构信息以及数据的智能化处理方面的优势，逐渐成为森林生境监测的重要工具。 
温带针阔混交林支撑和维持着独特和多样的野生动植物区系，位于该区域的东北虎豹国家公园是中国境内最大的以保护旗舰物种东北虎和东北豹为主的优先区域，同时也面临着过量放牧等较为严重的人类活动干扰，导致许多地区存在不同程度的生境退化和破碎化现象。因此，有必要对野生动物主要栖息的林下生境结构进行精准监测，了解人类活动及其他干扰导致的生境退化状况。本研究通过TLS技术开展了精细尺度下林下重要生境监测指标的提取和估算研究，并将提取的指标用于不同林分生境结构特征的量化和对比，为国家公园科学的森林管理和物种保护提供科学支撑。主要研究结果如下：
（1）利用TLS技术提取林下植被结构，结合生物量回归模型，对森林生态系统小于2 m的林下层生物量进行预测，比较了基于TLS的变量和实地测量变量的预测精度，并根据最优预测模型，绘制了不同食草动物密度影响下的林下生物量分布。研究结果表明，基于TLS的数据比实地测量数据能更准确预测林下层植被生物量，TLS提取的冠层盖度具有最佳的草本层生物量预测精度（R2 = 0.72，RMSE = 12.73 g/m2），TLS提取的冠层体积获得最高灌木层生物量预测的精度（R2 = 0.69，RMSE = 43.64 g/m2）。不同鹿密度样地林下草本层生物量差异显著（p < 0.001），灌木层生物量差异不显著（p = 0.22）。
（2）通过TLS技术对公园内一些典型林分（家畜放牧林、梅花鹿啃食林、次生林和混交林）的三维结构进行了量化和对比研究，提取的参数包括基本的林分结构参数（DBH，冠层高度，乔木数和材积），水平和垂直结构（植被面积指数[VAI]和郁闭度），以及林下空隙度（understory gap）和可见度（visibility）。结果显示，家畜放牧林的冠层高度，乔木数量及材积都明显低于其他林分。不同林分之间以及不同高度之间在垂直和水平结构上都表现出明显的不同。干扰较多的家畜放牧林和梅花鹿啃食林在低冠层部分（0-5 m）VAI和郁闭度要明显低于次生林和混交林。而次生林的VAI在中间冠层（5-10 m）要明显低于其他林分类型。相比之下，混交林的VAI和郁闭度在的不同高度层都维持在一个较高的水平，说明层级的结构多样性高。由于食草动物啃食，家畜放牧林和梅花鹿啃食林的林下空隙和可见度一般都远高于次生林和混交林。
（3）利用TLS数据获取局部点云的几何和辐射特征，通过随机森林模型来对温带阔叶森林的林下层枝干和叶片进行分类。结果显示，家畜放牧林的平均总体分类精度为87.50%，梅花鹿啃食林平均分类精度为77.50%，次生林平均分类精度为75.00%。其中平均叶片分类精度为86.58%，枝干分类精度为76.42%。辐射特征与几何特征相结合的分类方法总体分类精度为80.00%，平均Kappa系数为0.71，单独使用几何特征和辐射特征的总体分类精度分别为73.16%和50.58%。样地分类结果表明，林下植被密度与分类精度显著相关，随着林下植被密度的增加，总体的分类精度也在下降。
（4）根据枝叶分离结果，量化了公园内不同生境条件下，不同食草动物对林下枝叶结构的影响。通过对30个不同林地环境的样地进行TLS枝叶分离，结果显示，牛（Bos taurus domestica）和梅花鹿活动均显著降低了林下植被密度，但二者在影响的程度和模式上有明显区别。牛和梅花鹿啃食均显著减少了林下叶片密度，分别为87.15%和70.32%，但对木质密度的影响并未有明显的变化。同时，牛与梅花鹿活动造成的林下枝叶密度差异因环境条件的不同而有差异。阳坡、山脊地区，牛对叶片和木质密度减少程度与梅花鹿活动相比并无显著差异，但在阴坡、沟谷地区，牛对林下枝叶密度减少程度明显大于梅花鹿活动造成的林下枝叶减少。
综上所述，本文通过TLS技术对温带针阔混交林林冠以下一些关键的生境指标进行了提取和估算研究，包括林下层生物量分布，VAI，林下空隙，可见度以及林下枝叶分离等。这些精细和新颖的指标能够更好的量化和描述林下生境结构的细微差别，相较于传统的光学遥感和野外调查，TLS能够准确的评价食草动物或其他干扰对林下生境三维结构的影响。这些指标的量化为植被的恢复，物种生境质量评价提供科学依据，同时也促进激光雷达遥感在物种保护和生物多样性领域更加广泛和深入的应用。

Natural ecosystems are at risk of habitat degradation and loss as a result of the multiple impacts of human activities，alien species invasion and global climate change. The use of remote sensing to monitor and quantify the habitat structure of natural ecosystems has become the accepted practice in biodiversity conservation. However，traditional optical remote sensing methods are incapable of obtaining fine and intact forest structure，there are great limitations for further accuracy improvements in forest habitat structure monitoring. It is difficult to obtain complete and accurate understory information from remote sensing based on airborne or satellite platforms，especially in temperate mixed forest with complex and multi-level structure，which limits our monitoring and research of understory habitat structure. Terresrial Laser Scanning （TLS） has become an important tool in monitoring the structure beneath canopy because of its ability to quickly and directly acquire high-precision 3-D structural information，particularly vertical structural information and subsequent intelligent data processing. 
Temperate mixed forest support and maintain unique and diverse wildlife fauna，as China's largest priority area for the conservation of flagship spcies--the Amur tiger and Amur leopard，the Northeast Tiger and Leopard National Park suffered serious human disturbances，such as overgrazing，resulted in varying degree of habitat degradation and fragmentation in different forest stands. Thus，it is necessary to carry out accurate monitoring of forest structure beneath the canopy，where wildlife primarily inhabits，and to understand the changes in habitat quality caused by human activities and their effects on the survival and persistence of wildlife. In this study，TLS technology was used to extract and estimate important forest habitat monitoring indicators on a fine scale，and the extracted indicators were used to quantify and compare the habitat structure characteristics of different forest stands，so as to provide scientific support for forest management and conservation strategies in national parks. The main results were as follows:     
（1） We used TLS data and biomass regression model to predict the forest understory biomass （understory ≤ 2 m），and then compared the prediction accuracy of variables based on TLS and field measurement. Finnaly，the understory biomass distribution of different herbivore densities according to the optimal prediction model. Our results demonstrated that TLS-derived data were more accurate than field measurements in predicting understory biomass，that TLS-derived canopy cover yielded the highest herb layer biomass estimation accuracy（R2 = 0.72，RMSE = 12.73 g/m2），and that TLS-derived vegetation volume obtained the highest accuracy assessment for the shrub layer biomass prediction （R2 = 0.69，RMSE = 43.64 g/m2）. There was a significant difference in the understory herb layer biomass in different deer-density plots （p < 0.001），but no significant difference in shrub layer biomass （p = 0.22）. 
（2） The 3-D structures of different stand types（grazed forest，deer-browsed forest，secondary forest and mixed forest）in the park were quantified and compared by TLS. The extracted parameters include basic stand structure parameters（DBH，canopy height，tree number and volume），horizontal and vertical structure（vegetation area index [VAI] and canopy cover），understory gap and visibility. The results showed that the canopy height，trees number and the stem volume of grazed forest were significantly lower than those in other stand types. The horizontal and vertical structure varied greatly between different stand types and height intervals. The VAI and canopy cover of lower canopy height（0 - 5 m）was significantly lower in the grazed forest and the deer-browsed forestcompared to secondary forest and mixed forest with less disturbed. The VAI of secondary forest in the middle canopy（5 - 10 m）was significantly lower than that of other stand types. In contrast，the VAI and canopy cover of mixed forest at different heights maintained a higher level，which indicated high structural diversity of canopy. Due to intensive foraging by large herbivores，the understory gap and visibility of grazed forest and deer-browsed forest were generally much higher than that of secondary forest and mixed forest. 
（3） We used TLS data to extract geometric and radiation feature of local point clouds，and combined these data with random forest model to conduct wood and foliage separation in forest understory. The results showed that the average classification accuracy of grazed forest，deer-browsedforestand secondary forest was 87.50%，77.50% and 75.00% respectively. The average classification accuracy of foliage and wood was 86.58% and 76.42% respectively. The overall classification accuracy of the method combining radiation and geometric features is 80.00%，and the average Kappa coefficient is 0.71，while overall classification accuracy by using geometric features and radiation features alone is 73.16% and 50.58%，respectively. The classification results showed that understory vegetation density was significantly correlated with the classification accuracy. With the increase of understory vegetation density，the overall classification accuracy also decreased.        
 （4） Based on the classification results above，we quantified the influence of herbivory on understory foliage and wood under different habitat conditions in the park. 30 forest plots in different environments were scanned by TLS，the results showed that both cattle grazing and sika deer browsing significantly reduced understory density，but the degree and patterns of their impact on understory were significant differences. Cattle grazing and sika deer browsing significantly reduced understory foliage density，which were 87.15% and 70.32%，respectively，but there was no significant change in wood density. The variation of understory foliage and wood density caused by cattle and sika deer activities was different at different environmental conditions. In sunny slope and ridge area，the reduction rates of foliage and wood density resulted by cattle activities was no significant different with the rates caused by sika deer activies. But in the shady slope and ravine area，the decrease of foliage and wood density caused by cattle activities was significantly greater than that caused by sika deer activities.
In summary，TLS techniques was used to extract and predict some key habitat indicators below forest canopy，including understory biomass，VAI，understory gap，visibility，and understory foliage and wood density. These fine and novel indicators are capble to descible and quantify the subtle variation of forest habitat structure. Compared with the traditional optical remote sensing or field survey，TLS can accurately assess the effects of different herbivores or disturbance on 3-D structures of forest understory. The quantification of these indicators can contribute to provide a scientific basis for vegetation restoration and habitat quality evaluation in this park，and also promote the application of LiDAR remote sensing in the field of species conservation and biodiversity.