【作者】 王涛；

【导师】 刘兆刚； 董灵波；

【作者基本信息】 东北林业大学 ， 森林经理学， 2024， 博士

【摘要】 森林空间格局作为森林结构的重要组成部分,在生产实践中扮演着重要角色,一头连接生产实践,一头连接森林功能。但受限于传统地面调查,森林空间格局研究集中在样地尺度,而林场等生产经营的最小单位为小班,这造成了森林空间格局理论研究和生产实践尺度上的不匹配,给森林空间格局理论应用带来一定困难。近年来,随着机载、无人机等小型轻量化-多传感器遥感技术的不断发展,遥感技术为森林空间格局研究进行时空尺度扩展,进而为解决理论研究和生产实践尺度不匹配问题带来了可能。本研究以黑龙江省帽儿山实验林场为研究区,以2种常见森林恢复方式即人工恢复(n=30,2.01 ha)和自然恢复(n=50,3.35 ha)以及天然次生林(n=55,3.685 ha)为例,利用3种遥感数据即机载激光雷达(2015年)、高分辨率正射影像-DAP点云(数字立体摄影测量,Digital Aerial Photography,2015年)和历史航片-DAP点云数据(2003年,1993年和1983年),基于森林空间格局集群算法,在样地尺度上分析了多源遥感数据用于森林空间格局分析的可行性。在对比3种遥感数据产出森林空间格局数据精度的基础上,量化帽儿山林场不同时期(2015年和2003年)随机和不同聚集程度林木(弱度、中度、高度和极度聚集)及林隙的数量和面积特征,不同垂直分布状态林木的数量和面积特征(更新层、亚冠层、林冠下层、林冠中层和林冠上层)(第三章至第五章)。在此基础上,将森林空间格局分析进行时空尺度扩展,时间尺度上由单一年份空间格局(2015年或2003年)扩展至森林空间格局动态(2003-2015年)(第六章),空间尺度上由样地尺度(n=135,共9.045 ha)扩展至小班尺度(n=2911,共22,587.76 ha)以及林场尺度(共26,400 ha)(第七章)。最后,将多源遥感数据提取的森林空间格局信息应用于森林弹性评价(第八章),从树种多样性、水平格局异质性和垂直格局复杂度3个方面,评价天然恢复和人工造林恢复弹性恢复效果,量化林分、环境和种源距离3类变量对弹性恢复结果的影响,阐明弹性恢复的关键影响因子,为森林天然恢复和人工林近自然经营提供理论支持。研究主要结论如下:(1)样地尺度上(2015年),相同森林类型相同分布状态的林木数量特征差异较大,例如人工恢复,既有随机分布林木占较高的林地,也有聚集程度较高的林地,可能与恢复时间和种源距离等相关。不同森林类型空间格局不同,自然恢复样地林木聚集程度最高,林木个数比例为60%-91%,天然次生林林隙个数最多(1-10),人工恢复样地随机分布林木比例显著高于自然恢复和天然次生林。此外,人工恢复分布于更新层(树高1.5-5 m)林木个数比例最高(62%-100%),无林冠上层林木分布(树高>20 m)。(2)时间动态上(2003-2015年),不同森林类型不同水平和垂直分布状态的林木变化趋势不同。自然恢复样地林隙个数无显著变化,天然次生林和人工恢复样地林隙个数下降。自然恢复样地随机分布林木下降,林木聚集程度上升,弱度和中度聚集林木分别由31%-56%和45%-59%上升至44%-75%和72%-87%,出现高度聚集林木。天然次生林随机分布林木上升,弱度聚集林木下降,中度和高度聚集林木上升,极度聚集林木消失。人工恢复样地中随机分布林木上升,弱度聚集林木上升,中度、高度和极度聚集林木消失,可能由于林木枯损,使超大规模集群转换至小规模集群。(3)小班尺度上,3种森林类型即天然次生林-封山育林、天然次生林-经营和人工恢复各类分布状态林木无显著差异,区别于样地尺度,随机分布林木个数比例20%-45%,弱度聚集25%-45%,中度聚集10%-30%,高度聚集10%-50%,极度聚集60%-80%。各垂直分布状态林木无显著差异,更新层林木个数比例50%-75%,亚冠层(树高5-10 m)15%-30%,林冠下层(树高10-15 m)2%-15%,林冠中层(树高15-20m)1%-5%,林冠上层0.1%-1%。林场尺度上(26,400 ha)共分割出378,305株单立木,998,831株‘树木近似物体’分布在138,717个不同大小的集群中。(4)应用上,本文将多源遥感数据提取的森林空间格局信息应用于森林弹性评价,40%的人工造林恢复样地恢复效果较好,52%的自然恢复样地无恢复结果。机器学习评价模型表明,人工恢复样地天然树种恢复、水平格局异质性和垂直格局复杂度恢复更多受到‘大树’断面积(BAL)(DBH>5 cm)的影响,而自然恢复更多受到种源距离的影响。当种源距离较近时,人工恢复样地可以完成天然树种恢复,但人工恢复水平格局异质性和垂直格局复杂度恢复效果较自然恢复差,人工恢复经营时应该将水平格局异质性和垂直格局复杂度提升放在树种多样性提升同等地位。更多还原

【Abstract】 As an important part of forest structure,forest spatial pattern plays an important role in production practice,connecting production practice and forest function.However,limited by traditional ground surveys,forest spatial pattern research focuses on the plot scale,and the smallest unit of production and management such as forest farms is a sub-compartment.This has resulted in a mismatch between the scale of theoretical research on forest spatial pattern and production practice,and has given rise to the problem of forest spatial pattern.The application of theory brings certain difficulties.In recent years,with the continuous development of small,lightweight and multi-sensor remote sensing technologies such as airborne and unmanned aerial vehicles,remote sensing technology has expanded the spatial and temporal scale of forest spatial pattern research,thereby bringing new solutions to the problem of scale mismatch between theoretical research and production practice.This study takes the Maoershan Experimental Forest Farm in Heilongjiang Province as the research area,and uses three common forest types,namely natural recovery forests(n = 50,3.35ha),natural secondary forest(n = 55,3.685 ha)and plantation(n = 30,2.01 ha).For example,three types of remote sensing data were used Namely,airborne lidar(2015),high-resolution orthophoto-DAP point cloud(Digital Aerial Photography,2015)and historical aerial photosDAP point cloud data(2003,1993 and 1983 year),based on the forest spatial pattern clustering algorithm,the feasibility of using multi-source remote sensing data for forest spatial pattern analysis was analyzed at the plot scale.On the basis of comparing the accuracy of forest spatial pattern data produced by three types of remote sensing data,we quantified the random and different aggregation levels of trees(weakly,moderately,highly and agglomerated)in Maoershan Farm in different periods(2015 and 2003).The number and area characteristics of opening,random and different aggregation of trees and different vertical distribution states(renewal layer,sub-canopy layer,lower canopy,middle canopy and upper canopy)(Chapter 3to Chapter 5).On this basis,the forest spatial pattern analysis was expanded to a spatial and temporal scale,from the spatial pattern of a single year(2015 or 2003)to the forest spatial pattern dynamics(2003-2015)(Chapter 6).The spatial scale was expanded from the plot(n = 135,9.045 ha in total)to the sub-compartment(n = 2911,22,587.76 ha in total)and the forest farm(26,400 ha in total)(Chapter 7).Finally,the forest spatial pattern information extracted from multi-source remote sensing data is applied to forest resilience evaluation(Chapter 8)to evaluate the resilience recovery effects of active and passive restoration from three aspects: tree species diversity,spatial pattern and vertical structure complexity,quantify the impact of three types of variables: stand,environment and seed source distance on resilience recovery results,clarify the key influencing factors of resilience recovery,and provide theoretical support for passive restoration and near-natural management plantation.The main conclusions of the study are as follows:(1)At the plot scale(2015),the quantitative characteristics of trees in the same forest type and the same distribution state are quite different.For example,in plantation,there are tree with a high proportion of randomly distributed a high degree of aggregation.This may be related to restoration time and seed source distance.Different forest types have different spatial patterns.Passive restored forests have the highest concentration of trees,with a proportion of 60%-91%.Natural secondary forests have the largest number of openings(1-10).The proportion of randomly distributed trees in plantation is significantly higher than that in passive restoration forests and natural secondary forests.In addition,plantation distributed in the regeneration layer(tree height 1.5-5 m),with the highest proportion of trees(62%-100%),and there are no trees in the upper canopy(tree height > 20 m).(2)In terms of time dynamics(2003-2015),the changes of trees in different horizontal and vertical distribution states of different forest types are different.There was no significant change in the number of opening in passive restoration forests,while the number of opening in natural secondary forests and plantations decreased.The number of randomly distributed trees in passive restoration forests has decreased,and the degree of tree aggregation has increased.The number of weakly and moderately agglomerated trees has increased from 31%-56% and 45%-59% to44%-75% and 72%-87% respectively,indicating a high degree of aggregation.Randomly distributed trees in natural secondary forests increased,weakly clustered trees declined,moderately and highly clustered trees increased,and extremely clustered trees disappeared.Randomly distributed trees in plantation increased,weakly aggregated trees increased,and moderately,highly,and extremely agglomerated trees disappeared.This may be due to the loss of trees,causing the conversion of very large-scale clusters to small-scale clump.(3)On a spatial scale,2911 sub-compartment(22,587.76 ha)divided a total of 143,407 openings,406,676 individuals,138,162 small-scale clumps,12,583 medium-scale clumps,1647large-scale clumps,and 169 ultra-large-scale clumps.There is no significant difference in the distribution status of the three forest types,namely natural secondary forest-no management,natural secondary forest-management and plantation.The proportion of randomly distributed trees is 20%-45%,weak aggregation 25%-45%,moderate aggregation 10%-30%,high aggregation 10%-50%,extreme aggregation 60%-80%.There is no significant difference in the number of trees in each vertical distribution state.The proportion of trees in the regeneration layer is 50%-75%,the sub-canopy(tree height 5-10 m)is 15%-30%,and the lower canopy(tree height 10-15 m)is 2%-15%,1%-5% in the middle canopy(tree height 15-20 m),0.1%-1% in the upper canopy.A total of 378,305 individual were segmented at the farm scale(26,400ha),and 998,831 ‘tree approximate object’ were distributed in 138,717 clump of different sizes.(4)In terms of application,this paper applies forest spatial pattern information extracted from multi-source remote sensing data to forest resilience evaluation.Forty percent of active restoration plots have good recovery results,while 52% of passive restoration plots have no results.The machine learning evaluation model shows that the restoration of natural tree species,horizontal pattern heterogeneity and vertical pattern complexity in plantations are more affected by the basal area(BAL)of ’big trees’(DBH > 5 cm),while active restoration is more affected by the influence of seed source distance.When the seed source distance is close,plantation can complete the restoration of natural tree species,but the restoration effect of horizontal pattern heterogeneity and vertical pattern complexity in plantation is worse than passive restoration.Horizontal pattern heterogeneity and vertical pattern complexity should be combined in forest management.The improvement of complexity is placed on the same level as the improvement of tree species diversity.更多还原