云南松林分生长模型和生物量的遥感估测
郎荣
学位类型硕士
导师许建初
2010-11
学位授予单位中国科学院研究生院
学位授予地点北京
学位专业植物学
摘要云南松广泛分布于我国西南地区,是云南省主要用材树种之一。以预测林分生长动态为目的建立的林分生长模型,不但可以为森林的经营和管理提供决策支持,对于森林碳储量及碳汇潜力的研究也有重要意义。较为系统的云南松生长模型研究还较少,多数有关生长模型的研究只关注林分某一方面的生长过程,有必要对云南松的生长模型做深入的研究。遥感技术在森林生物量估算中的应用可有效实现样地生物量至区域尺度的上推。本文以保山市隆阳区杨柳彝族白族乡实地调查的91块云南松林样地数据研究了林分生长模型。使用非线性拟合的分析方法拟合和优选常用经验及理论生长方程,建立了包括地位指数、密度与林龄、平均直径、平均高和蓄积量的生长模型。依据一定公式和参数转换实测样地蓄积量至地上生物量、地下生物量和碳储量,以SPOT 4影像计算的各种指数为自变量,分析了地上生物量与各遥感变量间的相关关系,并建立了基于遥感数据的地上生物量模型。同时比较了3种不同的影像处理水平对建立生物量模型的影响,包括未经辐射校正的融合影像、辐射校正后的表观反射率影像和大气校正后的地表反射率影像。分析中使用的遥感变量包括各波段反射率、归一化植被指数、比值植被指数、主成分等共18个变量。研究的主要结论如下:(1)云南松林分生长模型各备选方程中Schumacher方程对林分生长的描述能力更好,特别是参数变异系数较其余非线性模型低,模型具有更好的稳定性。地位指数、林龄单因变量的平均直径、平均高和蓄积量生长模型均以Schumacher方程的拟合效果最好。密度与林龄的拟合以S函数为最优模型。(2)与林龄单因变量的平均直径生长模型相比,再次参数化引入地位指数和密度指数后模型的拟合度略有提高,逐步回归建立的包括平均高、株数密度和林龄为自变量的模型拟合度最高。引入地位指数和密度指数可大幅提高林龄单因变量平均高生长模型的拟合度。(3)林龄单因变量的蓄积量生长模型决定系数低,引入地位指数和密度指数再次参数化的生长模型可显著提高模型拟合度。逐步回归时林龄不满足变量的入选标准,以林龄代替与之相关性最高的平均高逐步回归时,包括林分变量的模型决定系数高于单因变量模型,但共线性检验中容限值偏低,方差扩大因子稍大。(4)遥感影像的中红外波段与地上生物量的关系非常密切,大气校正有助于提高地上生物量模型的拟合度。未经辐射校正的融合影像计算的各指数中,中红外波段与红波段计算的比值植被指数与地上生物量的相关性最高;表观反射率和地表反射率影像则以中红外波段和红波段计算的归一化植被指数的相关性最高;经比较地上生物量的多变量线性模型、逐步回归模型和单变量非线性模型,以单变量S函数模型的拟合效果最好。基于地表反射率影像的S函数地上生物量模型的误差评价显示模型的均方根误差较大,相对偏差很小,适于较大区域的生物量分布特征模拟。最后以大气校正影像的S函数模型为云南松地上生物量模型估算了研究区域云南松林的地上生物量和碳储量。
资助项目Pinus yunnanensis is one of the main timber species in Yunnan province, and widely distributed in southwest of China. Forest stand growth models are built for predicting stand growth, which could provide useful quantitative information for forest management. Understanding stand growth is also important for estimating forest carbon stock and carbon sequestration. Several studies of growth model for Pinus yunnanensis have been reported, however, most of which focus on a specific part of growth model. Therefore, it is necessary and meaningful to further the study of growth model for important stand factors. As a scaling-up tool for the observational data, remote sensing technique makes scaling biomass estimation from ground plot to a large scale more feasible and effective. Stand growth model was studied based on 91 sample plots data collected in the Yangliu Township, Longyang District, Baoshan Prefecture. Nonlinear fitting was used in fitting and selecting alternative growth functions, models of site index, density and stand age, average DBH, average tree height and stock volume were built in this study. Stock volume calculated from plot data was converted to aboveground biomass, belowground biomass and carbon stock according to selected equations and parameters. A total of 18 remote sensing variables were derived from SPOT 4 imagery, including reflectance of each band, normalized difference vegetation index, ration vegetation index, etc. The correlationship between aboveground biomass and remote sensing variables was analyzed, based on their significance of correlation, multiple regression, stepwise regression, nonlinear fitting were used to establish the aboveground biomass model. The effect of image processing levels on biomass estimation was compared, 3 types of images were analyzed including fusioned image without radiometric correction, radiometrically corrected to top of atmosphere (TOA) reflectance, and atmospherically corrected to surface reflectance. The main conclusions of this study are as follow: (1) Schumacher function fitted the stand growth better in site index, single variable average DBH, average tree height and stock volume growth model. Models fitted from Schumacher were more stable as the coefficient of variation was much lower than other alternative functions. S function was the best model in the fitting between stand density and age. (2) Comparing with the single variable average DBH growth model, fitting was improved a little after reparameterizing the model by introducing site index and stand density index variables. Stepwise regression improved the fitting significantly, the model with average tree height, density and age variables had best fitting without colinearity between variables. Repameterizing the average tree height growth model improved the fitting results a lot by introducing site index and density index variables. (3) Single age variable stock volume growth model had low coefficient of determination, but it was improved a lot by introducing site index and stand density index in reparameterization. Stand age was excluded in the stepwise regression modles, if raplace average tree height with stand age, the stepwise regression model that included age variable had higher coefficient of dertmination, but might have collinearity between variables when using a higher colinearity caritia. (4) Image analysis indicated that middle infrared was an important band for biomass estimation, and atmpspheric correction could improve fitting results of aboveground bimass model. Modified ratio vegetation index calculated from middle infrared and red band had highest correlation coefficient with aboveground biomass in pansharpened image, while modified NDVI derived from middle infrared and red band had highest coefficient in both radometrically corrected image. Comparing the fitting results of multiregression, stepwise regression and single variable nonlinear fitting, S biomass model was the most suitable model in all images. The final aboveground biomass S model was built from the atmospherically corrected image with modified NDVI as variable. Biomass estimation error of S model with surface reflectance image was evaluated, results shown estimation had high root of mean square error (RMSE), and quite small relative bias, which indicating that this method was suitable for mapping biomass spatial distribution on large scale. Finally, above ground biomass and carbon stock of Pinus yunnanensis were estimated with S model and surface reflectance image.
文献类型学位论文
条目标识符http://ir.kib.ac.cn/handle/151853/16078
专题昆明植物所硕博研究生毕业学位论文
推荐引用方式
GB/T 7714
郎荣. 云南松林分生长模型和生物量的遥感估测[D]. 北京. 中国科学院研究生院,2010.
条目包含的文件
文件名称/大小 文献类型 版本类型 开放类型 使用许可
郞荣.pdf(1482KB) 限制开放--请求全文
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[郎荣]的文章
百度学术
百度学术中相似的文章
[郎荣]的文章
必应学术
必应学术中相似的文章
[郎荣]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。