【作者】 韩爱惠；

【导师】 孙玉军；

【作者基本信息】 北京林业大学 ， 森林经理， 2009， 博士

【摘要】 森林作为陆地生态系统的主体,其生物量和碳储量的变化反映了森林的演替、人类活动、自然干扰、气候变化和人为污染等的影响,在全球陆地生态系统碳循环和气候变化研究中具有重要意义。本研究以东北三省为试验区,利用CBERS—WFI和MODIS数据,在提取森林分布信息的基础上,结合森林资源清查样地数据,以及地理、气象因子数据,采用多元线性回归和BP神经网络的方法,研究构建森林生物量估测模型,并对东北三省范围内提取的森林生物量和碳储量进行估测,同时采用两个模型估计值交叉对比的方法,得到试验区的森林生物量和碳储量动态变化信息。研究结果表明:(1)基于CBERS-WFI数据RGB组合图像提取森林信息的方法,是局部区域快速宏观提取森林信息的一种行之有效的技术方法,森林提取精度能达到80%。基于MODIS与CBERS-WFI数据年植被指数阈值分割法提取森林信息,分别不同区域设定不同的阈值,能有效地提高森林提取精度,东北三省森林提取面积与第六次清查森林面积差值低于16%。(2)采用多元回归和神经网络的方法,结合地理气象因子,研建了以CBERS-WFI数据和MODIS数据为主要信息源的森林生物量估测模型。通过样本检验分析,东北三省总森林生物量估测精度达到了85%以上,但不同区域其估测精度有所差异,这主要与森林植被类型有关。(3)森林生物量与植被指数NDVI、年均降水量、年均蒸散量、年均湿润度呈明显正相关,且相关系数较大;与年均积温和经度也呈正相关,相关系数比其它因子的要小;而与纬度呈明显的负相关,随着纬度增加,每公顷生物量减少。(4)从森林生物量监测遥感数据源分析,MODIS数据优于WFI数据;从估测模型分析,神经网络模型比多元线性回归模型估测结果更接近于样地测算值。采用神经网络模型,利用MODIS数据估测的2004年东北三省总森林生物量为2404.58×10~6t,森林碳储量为1202.29×10~6t;利用WFI数据估测的2008年的总森林生物量为2368.20×10~6t,森林碳储量为1184.10×10~6t。(5)采用MODIS和WFI两种遥感数据及两种森林生物量模型估测值交叉对比分析的方法,提取森林生物量和碳储量变化的区域。东北三省森林生物量和碳储量明显增长的面积为462.03万公顷,占东北三省森林面积的15.5%;生物量和碳储量减少的面积为145.16万公顷,占东北三省森林面积的4.87%。本研究首次利用了CBERS-WFI数据进行森林生物量和碳储量定量估测研究,拓展了森林宏观定量估测的遥感数据源。在森林生物量遥感估测模型研建过程中,首次在东北三省的范围内应用了森林资源清查样地数据,样本数量呈现较大的突破,并且通过应用样地地类面积等级因子,提高了样地与遥感像元的匹配程度,大大提高了森林生物量和碳储量估测精度。对试验区森林生物量和碳储量动态变化分析过程中,首次采用两种数据、两个模型交叉对比分析的方法,提取两期森林生物量和碳储量增长、减少和变化不明显的区域,森林生物量和碳储量变化区域提取更加严谨、科学。更多还原

【Abstract】 Forest,as the major component of the terrestrial ecological system,and the change of forest biomass and carbon storage reflects the influences of forest succession,human activities,natural interference,climate change and man made pollution,and it has significant meaningful to the study of carbon cycle and climate change in global terrestrial ecological system.The paper uses 3 northeast provinces as the trial area,studies on the technological methods of extracting forest information and generating biomass model based on CBERS-WFI and MODIS data.Two methods used in the study are Multiple Linear Regression and BP neural network which are based on the plot data of national forest inventory,geographic and climate factors.And forest biomass and carbon storage in 3 northeast provinces are estimated.Furthermore,the change of forest biomass and carbon storage in the trial area is estimated with the comprehensive approach by comparing the two estimated result with the 2 different models.The study results are shown as below:(1) The technical method of extracting forest is based on RGB composite image of CBERS-WFI data,and is an effective method for macro extracting forest information rapidly.The forest extracting precision is can reach 80%.The forest extracting method of year NDVI slicing based on CBERS-WFI and MODIS,should define different threshold value for different area,and it can improve the precision of forest extracting efficiently,its precision can reach 84%.(2) Forest biomass estimation model based on CBERS-WFI and MODIS data is studied and generated with multiple linear regression and neural network modeling.The test result with test samples shows the forest biomass estimated precision is reached 85%,but different area has different estimated precision because there are distributed different forest types.(3) There is a positive correlation between Forest biomass and NDVI,Annual Average Precipitation,Annual Average Evapotranspiration,Annual Average Wettability,and their correlation coefficient is relative high.There is also positive correlation between Forest biomass and Annual Average accumulated temperature,longitude,and their correlation coefficient is a little low.There is a negative correlation between Forest biomass and latitude,the biomass per hectare will be decreased with the increasing of latitude.(4) In forest biomass estimation,MODIS data is better than WFI data,and BP neural network model gets better estimated result than Multiple Linear Regression.Using neural network model,the total forest biomass and carbon storage in 3 northeast provinces are estimated based on MODIS in 2004 and based on WFI in 2008.In 2004,the total forest biomass is 2404.58×10~6t,the carbon storage is 1202.29×10~6t.In 2008,the total forest biomass is 2368.20×10~6t,the carbon storage is 1184.10×10~6t.(5)The change area of Forest biomass and carbon storage is extracted by comprehensive comparing based on two estimated model and two different data which including CBERS-WFI and MODIS data.The obvious increasing area of forest biomass and carbon storage is 4.62 million hectares, is occupied 15.5%of the forest area in 3 northeast provinces.And the decreasing area of forest biomass and carbon storage is 1.45 million hectares,is occupied 4.87%of the forest area in 3 northeast provinces.The paper is firstly studied on forest biomass and carbon storage estimation model based on CBERS-WFI data,developed remote sensing resources that can be selected for forest macro monitoring. In the process of generating forest biomass estimation model,plot data of national forest inventory is firstly be used in 3 northeast provinces,and the amount of samples is larger than before studies,and the area grade factor of plot is first time be used to improve the matching degree between plots and remote sensing image pixels,all these improve the estimated precision of forest biomass and carbon storage greatly.In the analysis process of forest biomass and carbon storage change,it is the first time to use two types of low resolution RS data and two type of estimation model together to extract the change area of forest biomass and carbon storage.The change area is extracted more exactly and more scientifically.更多还原