【作者】 张丽琼；

【导师】 秦武明；

【作者基本信息】 广西大学 ， 森林生态学， 2012， 硕士

【摘要】 以高峰林场桉树林为研究对象,利用Landsat TM影像、数字高程模型,结合2009年森林资源二类调查数据,探讨桉树人工林生物量遥感估测方法。从研究区的桉树小班中随机抽取346个小班作为样地小班,提取各遥感数据和地形数据与桉树生物量进行相关性分析。采用随机抽样的方法,从样地小班中选取242个样地用于回归模型构建和BP人工神经网络模型的训练样本；71个样地用于BP人工神经网络模型的模拟样本；剩余的33个样地进行模型的精度检验和误差分析。研究结果显示：①从遥感图像和数字高程模型中派生的19个变量因子中,与桉树林生物量显著相关的有10个,相关系数大小排序为：PVI>GVI>TM4>DVI>NDVI> TM3> MSAVI>TM7>SAVI> RVI,其中在0.01水平上显著相关的有8个因子；在0.05水平上显著相关的有SAVI、RVI,最高相关系数仅为0.577。②利用桉树样地生物量与通过相关分析筛选得到的10个变量建立一元线性回归、一元非线性回归和多元线性回归模型。通过比较,获得桉树林生物量最优回归模型为Y=246.808+16.899TM3-13.729TM4+1.767TM7+29.735RVI+488.234NDVI+13.617GVI+184.261SAVI-514.677MSAVI,R为0.571,在0.05水平上显著。③利用BP人工神经网络建立桉树林生物量非线性预测模型与样地生物量的相对误差为10.1%,估算精度达到89.9%；而桉树生物量最优回归模型的估算精度仅为84.7%,说明BP人工神经网络的非线性理论能较真实的反映研究区桉树林生物量的实际情况。④利用BP人工神经网络构建桉树生物量遥感估测模型对研究区桉树生物量进行估测,研究区内桉树林总生物量为159066.511t。更多还原

【Abstract】 Using Landsat TM images, digital elevation models and the2009Forest Resource Inventory datas to study the object, Eucalyptus forest of Gaofeng forest farm, which aimed at Inves tigating the remote sensing estimation methods of Eucalyptus Plantation biomass. Selecting346small patches randomly from the Eucalyptus study area as the sampling classes to do the correlation analysis with the Eucalyptus measured biomass. Then selecting242samples to build regression model and as training samples of BP artificial neural network model as the principles of random sampling. Last,71samples could be used for the simulation of samples of artificial neural network model and the remaining33samples from the sample in small classes could be analyzed the accuracy of the model testing and errors.The results shown:①elating to the Eucalyptus forest biomass significantly about10within the19variable factors which derived form the remote sensing images and digital elevation model.They can be sortse in order like: PVI>GVI>TM4>DVI>NDVI>TM3> MSAVI>TM7>SAVI> RVI. There are8factors isplaying significant relation at0.01level such and The remaining dowell at0.05level sush as SAV、RVI within the10variable actors> the bighest correlation coefficient is0.577.②Respectively building linear regression equation, nonlinear regression and multiple linear egression models by using the the Eucalyptus biomass and10variables which filtered through the correlation analysis. By comparison, the best Eucalyptus forest biomass regression model is like this: Y=246.808+16.899TM3-13.729TM4+1.767TM7+29.735RVI+488.234 NDVI+13.617GVI+184.261SAVI-514.677MSAVI, which R is0.571, significantly at0.05level.③the relative errors between non-linear model of Eucalyptus biomass established by using BP artificial neural networks and measured biomass is10.1%, estimation about89.9%. However, accuracy of the optimal regression model of Eucalyptus biomass estimation was only84.7%, indicating that the nonlinear theory of artificial neural network could reflect Eucalyptus forest biomass about time turthly.④Using BP artificial neural networks to estimate Eucalyptus forest biomass in the eara, there are totally159066.511t.更多还原