【作者】 黄兴召；

【导师】 张守攻；

【作者基本信息】 中国林业科学研究院 ， 森林培育， 2014， 博士

【摘要】 森林生物量是陆地生物量的主要组成部分，如何准确地估算森林的生物量和碳储量是森林生态系统研究的主要内容之一。本研究以全国6个落叶松人工林主产区的209块标准地的310株单木生物量实测数据为基础，对落叶松人工林生物量预估模型的技术方法和精确估算碳储量进行系统的研究。详细阐述贝叶斯原理在生物量建模中的运用，比较了贝叶斯方法和贝叶斯分层方法，为生物量建模提供了一种新的可行的技术方法；基于贝叶斯分层方法建立单木及各组分生物量模型，分析了区域间单木生物量的变化规律；并结合二类资源调查数据，利用生物量转换因子连续函数估算林分生物量；以实测单木各组分的碳系数估算各区域单木和林分的碳储量，分析它们的差异变化，为全国性落叶松生物量和碳储量建模与监测提供可靠的理论和技术支持。通过本项研究得出以下主要结论：（1）收集了80篇文献，共计304个关于地上部分生物量的异速生物量模型。虽然这些模型具有广泛的分布范围，而且是对不同树种或者不同区域相同树种的模拟，但模型参数a、b值变化范围较小，符合二元正态分布。将80篇文献按照不同的属对其模型参数a、b进行方差分析，参数a在属间差异不显著，参数b存在显著性差异。（2）以异速生物量方程为基础，比较贝叶斯方法和最小二乘法的建模效果，在此基础上分析贝叶斯方法和贝叶斯分层方法的建模效果。结果表明，在拟合样本≥50时，最小二乘法和贝叶斯方法的拟合效果无显著差异；当拟合样本＜50时，贝叶斯方法的拟合效果优于最小二乘法。进一步分析贝叶斯方法和分层方法的建模效果发现，贝叶斯分层方法建模的偏差信息准则DIC明显低于贝叶斯方法，且其决定系数R2均有所提高。通过F值来检验两种方法建模的差异性，表明贝叶斯方法和贝叶斯分层方法在单木生物量、干材生物量、树根生物量、树枝生物量和树叶生物量模型中均存在显著差异。因此，贝叶斯分层方法优于贝叶斯方法。（3）基于贝叶斯方法建立的单木及各组分生物量模型，结合胸径和树高生长方程，分析树龄对生物量分配格局的影响。干材生物量占单木生物量的比例呈现逐年上升趋势，树根、树叶、树皮和树枝生物量占单木生物量的比例均成逐年下降趋势。比较6个区域间单木生物量和各组分生物量的变化规律，结果表明，单木生物量和干材生物量为湖北的日本落叶松﹥甘肃的日本落叶松﹥辽宁的日本落叶松﹥黑龙江的长白落叶松﹥河北的华北落叶松﹥内蒙的兴安落叶松，其它组分在区域间也无明显规律。（4）以贝叶斯分层方法为基础，建立生物量转换因子方程，并结合湖北、甘肃和辽宁3个区域的二类资源调查数据，估算落叶松林分生物量，区域内林分每公顷生物量的平均增长量均呈现为中龄林﹥幼龄林﹥近熟林﹥成熟林的规律，区域间的生物量呈现为湖北日本落叶松﹥甘肃日本落叶松﹥辽宁日本落叶松的规律。3个区域的落叶松人工林生物量的平均值分别为61.34t/hm2、54.91t/hm2和50.51t/hm2；林分总生物量分别为723941.83t、984962.26t和8263708.41t。（5）通过实测不同区域落叶松各组分的碳系数，通过方差分析，表明碳系数在不同组分和不同区域间存在显著差异。因此，以区域为单位确定了区域间各组分的碳系数，并以各组分生物量占单木生物量的比值为权重，得出不同区域落叶松单木的碳系数。（6）基于单木生物量模型、生物量转换因子方程和实测碳系数，计算不同区域落叶松人工林的碳储量。区域间单木和干材的碳储量的变化规律为湖北的日本落叶松﹥甘肃的日本落叶松﹥辽宁的日本落叶松﹥黑龙江的长白落叶松﹥河北的华北落叶松﹥内蒙的兴安落叶松，其它各组分的碳储量在区域间的变化规律不明显。湖北省、甘肃省和辽宁省3个区域林分每公顷碳储量在不同龄组间的变化规律为湖北日本落叶松﹥甘肃日本落叶松﹥辽宁日本落叶松；3个区域林分每公顷的碳储量和总的碳储量分别为33.19t/hm2、26.30t/hm2、23.79t/hm2和391652.52t、471796.92t和3909044.37t。（7）根据单木、林分的生物量和碳储量的年增长变化规律，确定了单木、林分生物量和碳储量增长量的峰值。落叶松单木生物量的年增长量在23年之前逐渐增大，23-26年之间年增长量达到最大，26年之后增长量逐年减少。林分生物量和碳储量的年增长量的峰值在中龄林，其变化规律为中龄林﹥幼龄林﹥近熟林﹥成熟林。更多还原

【Abstract】 Forest ecosystem is a mainly body of the earth’s terrestrial biosphere，and accuratelyestimate the forest biomass and carbon sequestration plays an important role in global carboncycle. In this study, data of tree biomass were obtained from destructive sampling of310treesin six different regions of larch plantation in China. Technical and precies for estimating andpredicting the biomass and carbon storage for larch plantation conducted a study of a system inthe paper. Bayesian approach was used to establish the biomass model for comparing theBayesian no-hierarchical and Bayesian hierarchical approach. Bayesian hierarchical approachprovided a new method to establish the biomass modeles. Baesd Bayesian hierarchicalapproach, the continuous biomass expansion factor (BEF) method was used to estimate thestand biomass which combined National Forest Resource Inventory Data. The carbon storageof trees and stands were estimated under measured carbon content from the each componentsamples. The variation in different regions of the biomass and carbon storage were alsoanalyzed. These provided technical and theorical support for accounting and monitoring theChinese forest biomass and carbon stocking.The mainly conclusions were as foollows:(1)304functions of80papers was collectted which represented the allometryaboveground biomass model. Although these models have a broad distribution which are thesame species or different species in different regions, but the values of parameters a and b werea small range of variation. It found the distribution of the parameters a and b were wellapproximated by a bivariate normal. ANOVA was tested to parameters in different Genus. Theresult showed that there was only significant difference in parameter b.(2) Allometric biomass equations of the total and each component were fitted used byminimum-least-square regression and Bayesian approach. Meanwhile the Bayesianno-hierarchicaland and Bayesian hierarchical approaches was also used to built and compared the biomass model. The result showed that with sample size was more than50, both bayesianmethod and minimum-least-square regression was no significant difference in the meanabsolute error. And it was less than50, bayesian method was better than minimum-least-squareregression. Compared to Bayesian no-hierarchical and Bayesian hierarchical approaches,Deviance information criterion (DIC) reduced and the coefficient of determination (R2)elevated when used Bayesian hierarchical approach. F test was used to analyze the differencesof effect models in the two methods. There were a significantly differences except theallometric equations of skin. Therefore, Bayesian hierarchical approach was a better approachthan Bayesian no-hierarchical when they used to estimate the biomass in large scales.(3) Based Bayesian hierarchical approaches to build the bioamss model of the total andeach component, and the diameter and height growth equations was also built for predicted thetree’s total and each component biomass. The proportions of each component biomass werecalculatedand analyzed the changing order with ages in six regions. The result showed that theproportion of stem biomass was a rising trend, and the proportions of root, skin, branch andleaf biomass exhibited a declining trend. The each component and total biomass were alsoanalyzed the changing with the ages. The result showed that root, skin, branch and leaf biomasshad not a significant trend. But stem and total biomass had a consistent variation whichfollowed L. kaempferi of Hubei province﹥L. kaempferi of Gansu province﹥L. kaempferi ofLiaoning province﹥L. olgensis of Heilongjiang province﹥L. principis—rupprechtii of Hebeiprovince﹥L. gmelinii of Inner Mongolia.(4) Continuous biomass expansion factor of stand biomass were built that used byBayesian hierarchical approach. Combined National Forest Resource Inventory Data toestimated the standing biomass of different regions. The biomass of per hectare in each regionwas followed middle stands﹥young stands﹥pre-mature stands﹥mature stands, and thebiomass of per hectare in different regions was followed L. kaempferi of Hubei province﹥L.kaempferi of Gansu province﹥L. kaempferi of Liaoning province.The biomass of per hectare were respectively61.34t/hm2,54.91t/hm2and50.51t/hm2, and the total standing biomasswere respectively723941.83t,984962.26t and8263708.41t.(5) The carbon content of each component was measured in different regions whichplayed an important role in accurately estimate the carbon stocking. ANOVA was tested tocarbon content of each component. It found that there was significant variation in differentregions of each component. Therefore, The carbon content of each component was determinedin every region，and the carbon content of tree was calculated by the weighted whose based onthe proportions of each component biomass.(6) The carbon storage of larch plantation was calculated which based on trees models,standing models and carbon content. The tree’s carbon stocking were also analyzed thechanging with the age in six regions. These showed that stem, skin, branch and leaf carbonstorage had not a significant trend. But total and stem carbon storage had a consistent variation,the followed order was L. kaempferi of Hubei province﹥L. kaempferi of Gansu province﹥L.kaempferi of Liaoning province﹥L. olgensis of Heilongjiang province﹥L.principis—rupprechtii of Hebei province﹥L. gmelinii of Inner Mongolia. The carbon storageof per hectare was followed order: L. kaempferi of Hubei province﹥L. kaempferi of Gansuprovince﹥L. kaempferi of Liaoning province. The carbon storage of per hectare wererespectively33.19t/hm2,26.30t/hm2,23.79t/hm2, and thetotal standing carbon storage wererespectively391652.52t、471796.92t and3909044.37t。(7) The growth inflections of trees and stands were determined from analyzed the biomassand carbon storage growth in each year. Annual biomass and carbon storage growth of treeswere increased before23years, and the increasement reached maximum between23-26years,and Annual increasing growth decreased after26years. The growth inflection of stands wasmidel stand. Annual biomass and carbon storage growth of per hectare were followed middlestands﹥young stands﹥pre-mature stands﹥mature stands更多还原