【作者】 徐萍；

【导师】 徐天蜀；

【作者基本信息】 西南林学院 ， 森林经理学， 2008， 硕士

【摘要】 以“3S”技术为基础的森林资源动态监测方法,具有宏观性、综合性、周期短和成本低的特点,已成为当今森林资源动态监测技术发展的主要方向。探索适合林业生产实际需要的森林资源动态监测技术方法,研究解决3S技术在森林资源动态监测中的应用,特别是对森林资源调查监测的两大重要指标,面积与蓄积量,探讨其高效与高精度的估测方法,对于加快林业信息化建设的步伐,提高林业发展和经营决策水平,促进林业可持续发展都具有重要的意义。本研究以高黎贡山自然保护区为研究对象,利用印度遥感卫星和Landsat TM遥感数据,通过数据预处理、遥感图像解译分类、数据空间叠加、蓄积量模型建立与优化等技术环节,对研究区的森林资源土地利用/土地覆盖变化和森林蓄积量遥感估测模型进行研究,主要研究内容和成果如下:（1）建立研究区域数字高程模型DEM,并以DEM为基础提取用于森林蓄积量遥感定量估测的海拔、坡度、坡向等因子。（2）对印度遥感卫星IRS-P6卫星数据的5.8米分辨率的全色光谱影像和23.5米分辨率的多光谱影像经校正后,采用IHS方法进行融合,结果表明融合后图像既有较高空间分辨率,也包含较丰富的光谱信息。（3）参照《土地利用现状分类》,结合研究区域和遥感分类的特点,制定了适合研究区的以森林植被类型为主的土地利用/土地覆盖分类系统。采用目视解译的方法对研究区遥感图像进行解译分类,总体分类精度达88.27%,获得了研究区域1987年与2006年两期土地利用/土地覆盖分类图,从分类图上可以看出研究区内面积最大的地类是阔叶林,其次是农地,其中阔叶林主要分布在高黎贡山自然保护区内,农地则分布于保护区外。（4）对两期分类图统计分析得出研究区面积变化较大的土地类型是阔叶林和灌草丛,其中阔叶林面积增加了1078.49公顷,而灌草丛面积减少了1559.74公顷;面积百分比变化较大的类型是裸岩和灌草丛,其中裸岩增加面积占原有面积的72.34%,灌草丛减少面积占原有面积的23.56%。（5）对两期分类图的矢量图层进行空间叠加,计算得出研究区1987年至2006年土地利用/土地覆盖动态变化转移矩阵,进而分析森林资源土地利用/土地覆盖变化,包括地类间面积的变化和类型的变化,分析结果表明:1987年至2006年转移面积最大的地类是灌草丛,有3105公顷发生变化,主要流向为阔叶林和农地;2006年由1987年补给而来的面积最大地类是阔叶林,补给面积达2567.27公顷,补给来源主要为灌草丛和农地;居民地转移概率和补给概率则是所有地类中最大的,转移概率达74.68%,补给概率达78.89%。（6）以研究区阔叶林为例,以2006年一期遥感影像为基础建立森林蓄积量遥感估测模型,通过散点图法和相关系数分析法对数据进行预分析,得出公顷蓄积量与各项变量因子间存在线性关系,可以通过建立多元线性回归方程对阔叶林的公顷蓄积量进行估测。（7）对初步建立的线性回归模型通过共线性诊断和异常样地探测进行优化,并对线性回归模型进行回归关系检验,残差正态性检验,等方差性检验和模型拟合优度检验,结果证明森林公顷蓄积量估测模型中,公顷蓄积量与自变量因子包括遥感因子(B2、B5、NDVI、B_3/4 )和GIS因子（海拔、坡度、坡向）之间存在较强的线性回归关系,同时也满足多元线性回归模型的假设,回归模型的决定系数、修正决定系数和复相关系数分别为0.63、0.527、0.794,模型达到较好的拟合效果。（8）通过实测样地对逐步回归和多次优化所得到的线性回归模型进行检验,各项检验指标值说明了模型的适用性,其中,预估精度P在95%的置信水平下为76.89%,总体看来所建阔叶林的公顷蓄积量估测模型可以在研究区域高黎贡山自然保护区内应用。（9）本研究在高黎贡山自然保护区共提取阔叶林小班996块,总面积为15507.502 hm²,结合小班面积,通过所建公顷蓄积量估测模型估测出的阔叶林小班总蓄积量是3750741.778 m³。林分蓄积量大,说明了整个保护区内资源丰厚,在生态平衡、科学研究、社会经济效益方面有着特殊的意义和作用,保护区应注重对森林资源的保护,在研究区内以3S技术为基础进行森林蓄积量遥感定量估测的方法是切实可行的。更多还原

【Abstract】 “3S”technology-based methods for forest resources dynamic monitoring has the features of macroscopic, comprehensive, cycle short and low-cost and become the main direction of development of update forest resources dynamic monitoring technology. Probe for appropriate methods of forest resources dynamic monitoring technology, research the application of“3S”technology in forest resources dynamic monitoring technology, especially the most important two indicators for forest resources surveying, area and volume, discuss the rapidly and accurately prediction methods, it has important significance for speeding up the pace of forestry informationization construction, improve forestry management decision-making levels and promote the sustainable development of forestry.This study take Gaoligong Mountain Nature Reserve as the research area, using Indian remote sensing satellite and Landsat TM remote sensing data, research the area change of forest resources land use/land cover and the model of forest volume remote sensing estimation by data pretreatment, imagery interpretation and classification, overlying space data, build and optimize volume model, the main content and conclusion of this research is as follows:(1) Establish the digital elevation model DEM of the research area, and extract elevation, slope, aspect which used for build the model of forest volume based on DEM.(2) Fusion images about panchromatic image and multi-spectral image of IRS-P6 by IHS, the result indicate that the fusion image have higher spatial resolution, also includes rich spectral information.(3) Set up land use/land cover classification system relay mainly on forest type according to the《land use classification》,the characteristics of the research area and the image. Classificy by visual interpretation, the overall accuracy is 88.27 percent, acquired land use/land cover classification map in 1987 and in 2006, from the classification map can be seen the broadleaf’s area is the largest, followed by farmland, the broadleaf mainly distributed in the Gaoligong Mountain Nature Reserve and the farmland distributed outside.(4) Analyse the changing statistics of land use/land cover area of the research area from 1987 to 2006, the results indicate: the broadleaf’s and shrubbery’s area have larger change, the broadleaf’s area increased 1078.49hectares, the shrubbery’s area decreased 1559.74 hectares; the uncovered rock and shrubbery have a larger change of the area percentage, the uncovered rock increased 72.34 percent, and the shrubbery decreased 23.56 percent.(5) Overlaying the two classification map, a convertion matrix of land use/land cover from 1987to 2006 is obtain, analysis the area and the type change of land use/land cover, the results indicate: the shrubbery’s area have the largest divert from 1987 to 2006, divert area is 3105 hectares and mostly divert to broadleaf and farmland; the broadleaf’s area have the largest supply from 1987 to 2006, supply area is 2567.27 hectares and mostly supply by shrubbery and farmland; the resident have the largest divert probability and supply probability, the divert probability is 74.68%, supply probability is 78.89%.(6) Example by broadleaf, build remote sensing estimation model of forest volume by one image of 2006. The scatter graph and the correlativity analyse indicate the volume per ha have linearity relation with the independent variable factor and estimation volume by multivariate linear regression model.(7) Optimize the initial model by multicolinearities diagnostics and the influence point detection, and make linear regression test, residual normal school test, equal-variance test, the test results prove the volume per ha have a stronger linearity relation with the independent variable factor, and the hypothesis of regress model is right. The coefficient of determine, the modificatory coefficient of determine and the correlation coefficients are 0.63, 0.527, 0.794, model have a good fitting effect.(8) check up the multivariate linear regression models by sample, the indexs of the check up show applicability of the model, the estimation accuracy P = 76.89% when confidence is 95%, satisfy the accuracy requirements of the Category II of forest inventory, the model is applicable for Gaoligong Mountain Nature Reserve.(9) This research pick up 996 broadleaf sub-compartments, all of the volume is 3750741.778 m~3. The volume is very large, so the entire forest resources in Nature Reserve is rich, which play important role in ecological balance, scientific research and social and economic benefits, the Nature Reserve should pay attention to protect forest resources and the method that estimate the forest volume using 3S technology is feasible.更多还原