【作者】 周洋；

【导师】 唐代生；

【作者基本信息】 中南林业科技大学 ， 森林经理学， 2011， 硕士

【摘要】 立地质量是指在某一立地上既定森林或者其他植被类型的生产潜力,所以立地质量与树种关联,并有高低之分。立地质量的评价分为直接评价和间接评价,直接评价法包括根据林分蓄积或收获量进行评定和利用林分树高进行评价,间接评价法包括多元地位指数法和植被指数法等。因树高易于测定,而且受林分密度影响较小,因此,利用林分高编制地位指数表的是林业上最常采用的评定方法。但是定量化的方法进行森林生态系统立地分类还存在许多困难,例如野外样点的采集和调查非常耗时和耗力,许多森林类型缺乏长期定位观测的样地资料,大部分结果主要还是来源于以点为基础的小范围调查的工作。所以寻找一个简便、直观和高精度的方法来估计地位指数是非常有意义的。遥感技术的发展和数学方法的渗透,为地位指数的研究提供了很好的平台,卫星遥感为大面积研究森林生态系统生产力和及其空间分布提供了一条重要的途径,从遥感角度分析地表植被覆盖情况最有研究价值的是植被指数。20世纪80年代以来,卫星资料在国内外植被研究中得到广泛应用,例如：利用NDVI研究叶片面积指数的空间分布特点及其变化,研究森林生态系统净初级生产力的空间分布等。利用典型的植被指数和地位指数的关系估计和分析森林地位指数及其动态变化对于制定合理的造林方案和经营管理措施具有十分重要的意义。但由于理论、技术与人力和物力等条件的限制,该领域的研究还处于起步阶段,相关报道很少。“地学信息图谱”概念提出给了作者启发,它是按照一定指标递变规律或分类规律排列的一组能够反映地球科学时空信息规律的数字形式的地图、图表、曲线或图像。于是提出了地位指数信息图谱这个概念,根据树种的不同,按照一定的时间规律排列的能够反映研究区域地位指数时空规律的数字形式地图、图表、曲线或者图像。本文选择建立一个分析模型对其进行初步的探索研究,为提高模型拟合的精度,采用多尺度分析的方法,它也是地学信息图谱研究中一个重要方法之一。考虑到数据的收集以及地域的限制等问题,综合考虑将实验区域定为湖南攸县黄丰桥林场。利用TMM遥感图像光谱信息良好的综合性以及归一化指数(NDVI)是植物生长状态以及植被空间分布密度的最佳指示因子这一特性,借助遥感软件(ENVI)和地理信息系统软件(GIS),结合MATLAB小波分析和SPSS逐步回归分析方法,对湖南省株洲攸县黄丰桥林场杉木地位指数遥感估测的尺度进行选择,并对拟合的模型进行分析和精度评估。研究发现,植被指数在进行5重分解重构后基本达到分解末端,在进行逐步回归分析后,只有植被指数原始数据以及一级分解重构的数据保留,其余4个尺度的数据都被剔除。拟合的模型中包括两个自变量,一是植被指数原始数据,个其一级分解重构后的数据,即一尺度数据。最后对两个拟合模型的精度评估,检验模型的使用价值。因在林业上的地位指数是一个正整数,而拟合的模型不可能精确的算出这个数字,只能算出一个近似的数字,结合小班的具体情况来得出最终的正整数,所以本文采用建立1：1的比例分析图对模型拟合结果进行精度分析,在SPSS中建立1：1的精度分析图,结果表明两个模型拟合的相关度都在0.9以上,这说明利用遥感指数预估地位指数可以达到一定的精度,具有较高的使用价值。但由于国内少有此类研究,理论尚不成熟,有待深化。更多还原

【Abstract】 Site quality is established in a stand on the floor of forests or other vegetation types of production potential, so the associated site quality and tree line, and high and low points. Evaluation of site quality assessment is divided into direct and indirect assessment, direct assessment method including the Stand or harvest were evaluated and assessed by stand height, indirect assessment of the status of law, including multiple index and vegetation index method. Easy to determine because of tree height, and less affected by stand density, therefore, prepared using high-index table Stand is the most commonly used assessment of forestry methods.But the quantitative methods of site classification of forest ecosystems, there are still many difficulties, such as sample collection and field surveys are time consuming and labor-intensive, and many forest types in the lack of long-term observation plots targeting information, most of the results mainly from A point-based small-scale investigations. So looking for a simple, intuitive and accurate method to estimate the position index is very meaningful.Remote sensing technology and the penetration of mathematical methods for the study site index provides a good platform for the large area of satellite remote sensing of forest ecosystem productivity and its spatial distribution provides an important way, from the perspective of remote sensing of vegetation coverage of most of the vegetation index value.20th century,80 years, satellite data of vegetation at home and abroad has been widely used, for example:the use of leaf area index of NDVI spatial distribution and changes of forest net primary productivity of ecosystems such as the spatial distribution. Use of vegetation index and status of the typical relationship between the estimation and analysis of Indices of forest status index and its dynamics program for the development of the Planting and management measures is of great significance. However, theory, technology and human and material resources and other conditions, researchers in the field still in its infancy, very few relevant reports."Geo-information map of put forward the concept of inspiration to the author, which is in accordance with the classification of certain indicators of Homologous or a group of regular arrangement to reflect the laws of earth science spatial information in digital form of maps, charts, curves or images. So I think the concept of site index map information, according to the different species, in accordance with the laws of a certain time arranged to reflect the status of the study area in digital form the law of exponential time and space maps, charts, curves or images.However, theory, technology and human and material resources and other conditions, use of remote sensing vegetation indices reflect the site quality and its spatial and temporal dynamics research in the field still in its infancy, few relevant reports, so this model was chosen to establish a preliminary analysis The exploration and research, to improve the accuracy of model fitting, this multi-scale analysis method chosen, it is also the study of geo-information spectrum method is an important one. Taking into account data collection, and geographical constraints and other issues, this selection of You County, Hunan Toyohashi yellow cedar forest data as experimental data.Spectral information of TM remote sensing images using well-integrated and normalized difference index (NDVI) is a plant growth status, and spatial distribution of vegetation density factor of this feature the best indication, with remote sensing software (ENVI) and geographical information systems software (GIS), Combined with MATLAB wavelet analysis and SPSS stepwise regression analysis method, You County, Zhuzhou, Hunan Toyohashi yellow fir forest site index to choose the scale of remote sensing estimation, and model fitting analysis and accuracy assessment.Found that decomposition of vegetation index during the 5 re-decomposition of the reconstructed basically reached the end, during the stepwise regression analysis, only the raw data of vegetation index and the level decomposition and reconstruction of data retention, and the remaining four scales of data have been excluded. Fitting model includes two independent variables, namely the raw data of vegetation index, a level of decomposition and reconstruction of their data after that data of a scale.Finally, the accuracy assessment of the two simulation models to test the model using the value. Because of the position index in forestry is a positive integer, and can not fit the model calculates the exact figure can only calculate an approximate figure, combined with the specific circumstances of small classes to come to the end of the positive integers, so this is used to establish Figure 1:1 ratio of precision of the results of model fitting analysis in SPSS to establish the accuracy of analysis of Figure 1:1, the results show that the correlation of two model fitting are above 0.9, indicating that the use of remote sensing index forecast Position index can reach a certain precision, with high use value. But because few such studies, theory is not yet mature, to be deepened.更多还原