【作者】 温小荣；

【导师】 佘光辉；

【作者基本信息】 南京林业大学 ， 森林经理学， 2017， 博士

【摘要】 论文主要针对南方集体林区的森林资源调查与监测的关键技术与方法问题开展研究,选取了试验区1(浙江省建德市)、试验区2(江西省吉水白沙林场)、试验区3(江苏省东台林场)作为试验区。这些区域都处于南方集体林区的亚热带常绿阔叶林和针阔混交林经营区,南方集体林区具有较好的森林生长自然条件,森林资源集约化监测与管理对提升森林质量具有重要的意义。森林资源二类调查成果是指导和规范科学经营森林的重要依据,是森林质量精准提升的基础。森林资源监测的高新技术不断涌现,无人机遥感技术、地基激光雷达、机载激光雷达技术开始在森林资源二类调查中应用。现代化森林调查监测技术广泛应用了航天遥感、航空遥感、全球定位系统、数据库技术及计算机网络技术等高新技术,由此引发新形势下对高时效、高精度、多层次的森林资源二类调查的新要求。论文主要研究内容如下:1、构建了一个提取有林地小班地类变化的综合相似度指数FSi计算公式综合相似度指数FSi描述发生变化小班与入样有林地小班对象因子特征值之间的相似程度,FSi值越大表示该小班地类有极大的可能性发生变化。综合相似度指数计算公式如下:FSi=(?),Fzi=bzi-m1/σi式中,bzi为特征波段i中第z个小班波段值,m1、δ i分别为特征波段i中入样有林地小班对象波段均值和标准差,N为特征波段个数,FZi为构建的相似度指数统计量,FSi为综合相似度指数。试验区(建德市)2013-2014年landsat8 OLI遥感影像分析结果表明:2014年的各入样有林地小班对象的Band2影像特征值、Band3影像特征值、2013-2014的NDVI差值和其主成分分析第一主成分PC1差值的FZi,其趋势均呈现近似正态分布规律。利用该特征构建综合相似度指数FSi,实现试验区2013-2014年变化小班的提取,在不区分小班类型时正确率、漏检率、错检率分别为86.79%、13.21%、84.91%,区分小班不同的坡度和坡向类型时,其正确率都达到90%以上。该方法应用于同一地区2014-2015年小班变化信息的提取得到较好的效果,其正确率都达到80%以上。该方法为小班地类变化信息提取提供了一种改进的方法,为森林资源年度变更调查、森林资源二类调查的复查、小班空间数据获取提供支撑,具有较好的应用价值。2、研究了基于无人机遥感数据的森林蓄积量双重回归估计方法采用基于无人机遥感样地的模型预估蓄积量值作为双重回归估计中辅助因子,地面实测样地的蓄积量值作为双重回归估计中主因子(目标变量),论文提出了双重回归估计中辅助因子的几种估测方案。结果表明:五种辅助因子获取方案其估计精度都在90%以上,方案一、方案三、方案五其R2都在0.68以上,有利于提高估计精度,5种方案其估计区间也较为一致,说明基于无人机遥感数据获取辅助因子并进行双重回归估计是可行的,方法的研究为无人机遥感技术在区域森林资源二类调查和监测开辟了新的途径。由方案一双重回归估计得到试验区(东台林场)杨树人工林公顷平均蓄积量为142.6m3,公顷平均蓄积量其估计区间为133.8～151.4m3,其蓄积量总量估计区间为94265.1m3～106628.1 m3。该方案的估计精度为93.85%。由方案五双重回归估计得到东台林场杨树人工林公顷平均蓄积量为143.0m3,该方案的估计精度为93.26%,试验区杨树人工林蓄积量总量估计区间为94031.8m3～107371.0 m3。3、探讨了基于无人机遥感数据的森林生物量双重回归估计方法根据论文研建的冠幅和树高模型W = 0.0039Cw1.1153h2.8713,对样地生物量进行测算,作为其辅助因子。该方案基于无人机遥感影像获取的样地平均冠幅和林分平均高,本次试验用模型预估值代替林分平均高。应用样地平均冠幅和林分平均高推算无人机遥感样地单株平均生物量,根据无人机遥感样地获取的株数乘以单株平均生物量,得到样地的生物量。根据双重回归估计得到试验区平均单位面积的公顷生物量为73098.5247 kg。试验区杨树人工林地上部分生物量其总量估计为5.1486×107kg,估计区间为(4.7985×107～5.4987×107kg),估计精度为93.2%。4、提出了小班ΠPS抽样和分层ΠPS抽样的估计方法本文提出了小班ΠPS抽样总体总量的估计,并给出了小班ΠPS抽样的近似方差的估计量计算公式,并对试验区杨树人工林的蓄积量进行估计。论文研究结果可知,小班ΠPS抽样不分层的情况下,试验区杨树人工林总体总量的估计为98114.40 m3,估计区间为86348.08 m3～109880.72 m3,精度达到88.00%。对小班组合类型的分层ΠPS抽样估计得到较好的效果。小班分层ΠPS抽样对杨树人工林总体蓄积量的估计为99327.15 m3,其估计精度达到92.24%。在相同样本量的情况下,小班分层ΠPS抽样比不分层的小班Π PS抽样的精度要高。5、研究了小班Π PS抽样的森林生物量抽样估计方法对于试验区杨树人工林总体而言,小班分层Π PS抽样估计森林生物量其总量的估计为 51945846.68 Kg,估计区间为 47916655.21～55975038.16Kg。精度达到 92.24%。在相同样本量的情况下,小班分层Π PS抽样比不分层小班Π PS抽样估计森林生物量的精度要高。森林资源二类调查中各小班单元大小不等,应用不等概抽样效率高的优点进行森林资源二类调查中小班不等概抽样达到对调查总体提供可靠的估计,使得森林资源二类调查自成体系并有一定精度保证,是森林资源调查小班抽样需要解决的技术难题之一。本文研究的无放回小班不等概抽样(小班ΠPS抽样)在试验区杨树人工林蓄积量、生物量、林木总株数的估计中都取得了较好的效果,能达到森林资源二类调查规程规定的精度。论文研究有利于补充和完善小班不等概抽样理论与方法,形成小班ΠPS抽样的森林资源监测体系。总之,深入研究森林资源二类调查的关键技术与方法,将有助于推动地方森林资源监测技术进展。无论从森林资源监测的实际需要和该理论方法的解决等方面,该项研究都是有积极意义的。更多还原

【Abstract】 With the development of information technology,it is very important to establish an efficient forest resources survey and monitoring system,which can obtain fast and reliable information on forest resources and can promote the development of ecological forestry and the people’s livelihood forestry.To study the critical techniques and methods of forest management inventory of collective forest in southern China,three study sites were selected as Jiande city in Zhejiang province(site 1),Baisha forest farm in Jishui county,Jiangxi province(site 2),Dongtai forest farm in Jiangsu province(site 3).These sites are all located in the subtropical region with broad leaved evergreen forest and mixed forest.There are good natural conditions for forest growth.Intensive monitoring and management of forest resources is of important significance for improving forest quality.The data acquired by forest management inventory are important basis for guiding and standardizing the scientific management of forests,which is the basis for improving forest quality.Advanced forest resources monitoring technologies are constantly emerging.New technologies such as Unmanned aerial vehicle(UAV)remote sensing,airborne laser scanning,and terrestrial laser scanning have been used in forest second type inventory.The wide application of remote sensing,GNSS,database management,and internet technology provides opportunities and challenges to forest management inventory to achieve high time-efficiency,high-accuracy and multi-scale.The research studied the critical techniques and methods of forest management inventory and the main contents of the thesis are as follows:1.The method to extract forest sub-compartment change based on the comprehensive similarity index(FSi)equationThe comprehensive similarity index describes the similarity between the eigenvalues for changed sub-compartment and the forest sub-compartment which is sampled.The bigger the FSi value is,the higher the possibility that the land cover of the sub-compartment may change.The FSi equation is:FSi=(?),Where,bzi is the value of sub-compartment z in characteristic wave band i.Respectively,m1 and5 i are mean and standard deviation of the characteristic wave band iof the sampled forest sub-compartments.N is the number of characteristic wave bands.FZi is the similarity index statistics and FSj is the comprehensive similarity index.For study site 1(Jiande city),the results of Landsat 8 OLI data(2013-2014)analysis showed a normal distribution for the value of band 2 and band 3 of sampled sub-compartmentforest,the difference of NDVI and the main component of PC1,FZi.Based on this characteristic,comprehensive similarity index(FSi)was calculated to extract land cover change for forest sub-compartments in 2013-2014.When forest sub-compartment types were not differentiated,detection rate,omission ratio,and the false detection rate were 86.79%,13.21%,and 84.91%,respectively.When forest sub-compartments were classified based on slope and aspect,detection rate was over 90%.The comprehensive similarity index calculated by combination of characteristic variables of remote sensing data has different effects on improving the detection accuracy of forest sub-compartment change.The method was applied to the forest sub-compartments in the same area(2014-2015)and detection accuracy was above 80%.The method can be used to detect land cover change of forest sub-compartment,thus to help annual forest resources monitoring and forest management inventory.2.The method to estimate forest volume with double regression sampling technique based on UAV high resolution image dataThe forest volume estimation was based on a double regression estimation method consisting of many plots from which both ground and UAV high resolution image data were obtained.UAV images were used to estimate forest volume using double regression procedure.The procedure requires that the auxiliary variable x(UVA images in this case)is observed in a larger phase 1 sample of size n while the primary variable y(forest volume in sample plots)is observed in a phase 2 sub-sample of size n.The paper proposed several estimation schemas for auxiliary variable in double-regression sampling technique.The result shows that the estimation accuracy for the five variable schemas is all above 90%and R2 for schema 1,3 and 5 is all above 0.68.The forest volume estimations of the five schemas were relatively consistent which confirms that the combination of obtaining auxiliary variable based on UAV Image with double-regression sampling technique is feasible.The research shows a new way to survey and monitor regional forest resources by UAV technology.Based on schema 1,the average forest volume of study site 3(Dongtai state forest farm)was estimated at 142.6m3 per hectare for poplar plantation,the estimation interval was 133.8 to 151.4 i3 per hectare and the total forest volume for study site was between 942,65.1 m3 and 106628.21m3 with an accuracy of 93.85%.Based on schema 5,corresponding values were 143.0 m3 per hectare,133.5 m3 to 152.4 m3 per hectare,94,031.8 m3 to 107371.0m3,and an accuracy of 93.26%.Thus it is practical to use UAV remote sensing data to estimate forest volume with double regression sampling method.3.The method to estimate forest biomass with double regression sampling technique based on UAV high resolution image dataForest biomass was estimated based on a double regression sampling method consisting of many plots from which both ground and UAV high resolution image data were obtained.The model was given as W = 0.0039Cw1.1153 h2 8713 and average crown diameter and mean height were used in the equation.UAV data were used to derive mean crown diameter and mean height of sample plot.The average individual-tree biomass was calculated using average crown diameter and mean height of sample plot.Then the biomass of sample plot was obtained by multiplying the number of trees with average individual-tree biomass.The aboveground biomass was estimated at 73098.5247 kg per hectare according to double-regression sampling technique.The total aboveground forest biomass of poplar plantation of the study site was estimated to be 5.1468 × 107 kg,the estimation interval was between 4.7985 × 107 and 5.4987 × 107 kg with an accuracy of 93.2%.The improvement in precision of regression estimates depends on the strength of correlations between forest biomass and UAV images.4.The method to estimate forest volume with Π PS sampling and stratified Π PS sampling for forest sub-compartmentThe Estimator for unequal probability sampling without replacement of subcompartment was studied in the paper.The equation to estimate forest volume based on Π PS sampling of sub-compartment was presented and the total forest volume was estimated for the poplar plantation in Dongtai forest farm(study site 3).The results showed that the estimated forest volume of poplar plantation was 98,114.40 m3,estimation interval was 86,348.08 m3～109,880.72 m3 with an accuracy of 88.00%when the Π PS sampling of sub-compartment is not stratified.The accuracy of forest volume estimation by stratified Π PS sampling was better.With stratified Π PS sampling of forest sub-compartment,the estimated forest volume of poplar plantation was 9899,327.15 m3 with an accuracy of 92.24%.With the same sample size,the accuracy of forest volume estimation by the stratified Π PS sampling of forest sub-compartment has higher accuracy than that by the un-stratified Π PS sampling of sub-compartment.This provides an improved method for sampling estimation with unequal probabilities for forest resources monitoring of different types of sub-compartment.The results showed that for forest volume estimation,the Π PS sampling of sub-compartment is feasible and efficient for forest farm.The stratified Π PS sampling of sub-compartment significantly improves sampling efficiency and accuracy and serves as a new method for sub-compartment sampling estimation with unequal probabilities.5.The method to estimate forest biomass with Π PS sampling and stratified Π PS sampling for forest sub-compartmentFor the poplar plantation in the study area,the forest biomass was estimated to be 51,945,846.68 Kg with stratified Π PS sampling of sub-compartment,and estimation interval was between 47,916,655.21 and 55,975,038.16 Kg,with an accuracy of 92.24%.With the same sample size,the accuracy of forest biomass estimation by the stratified Π PS sampling of sub-compartment is higher than that of un-stratified Π PS sampling of sub-compartment.In summary,based on the non-replacement sub-compartment sampling method with unequal probabilities(Π PS sampling of sub-compartment),the accuracy of forest volume estimation,biomass estimation and total tree number estimation of poplar plantation is good and meets the accuracy requirement of forest management inventory.The research complemented the theory and method of sub-compartment sampling with unequal probabilities to form forest resources monitoring system of Π PS sampling of forest sub-compartment.更多还原