A spatial co-simulation based forest carbon estimation for Kaihua County

YUAN Zhenhua; ZHANG Maozhen; GUO Hanru; QIN Lihou

doi:10.11833/j.issn.2095-0756.2016.03.003

Volume 33 Issue 3

Oct. 2016

Turn off MathJax

Article Contents

Article Navigation > Journal of Zhejiang A&F University > 2016 > 33(3): 384-393

YUAN Zhenhua, ZHANG Maozhen, GUO Hanru, QIN Lihou. A spatial co-simulation based forest carbon estimation for Kaihua County[J]. Journal of Zhejiang A&F University, 2016, 33(3): 384-393. doi: 10.11833/j.issn.2095-0756.2016.03.003

Citation:

YUAN Zhenhua, ZHANG Maozhen, GUO Hanru, QIN Lihou. A spatial co-simulation based forest carbon estimation for Kaihua County[J]. Journal of Zhejiang A&F University, 2016, 33(3): 384-393. doi: 10.11833/j.issn.2095-0756.2016.03.003

A spatial co-simulation based forest carbon estimation for Kaihua County

doi: 10.11833/j.issn.2095-0756.2016.03.003

YUAN Zhenhua^{1,2
,},
ZHANG Maozhen^{1,2
,
,},
GUO Hanru,
QIN Lihou^1,2

1.
Zhejiang Provincial Key Laboratory of Carbon Cycling in Forest Ecosystems and Carbon Sequestration, Zhejiang A & F University, Lin’an 311300, Zhejiang, China;
2.
School of Environmental and Resources Science, Zhejiang A & F University, Lin’an 311300, Zhejiang, China

Received Date: 2015-05-13
Rev Recd Date: 2015-07-08
Publish Date: 2016-06-01

Abstract

To estimate the spatial distribution of aboveground forest carbon storage, an important index of forest ecosystem productivity and the foundation of regional forest carbon sink measurements, and carbon density in Kaihua County, Zhejiang Province, ZY-3 image data from 2013, National Forest Inventory data from 2014, and the Sequential Gaussian Co-simulation Method were used. Estimated results of above ground forest carbon were analyzed by four indicators: mean error (ME), residual square sum (RSS), mean relative error (MRE), and root mean square error (RMSE). A correlation analysis between sample plots of remote sensing and forest carbon density was also conducted. Results showed that the above ground carbon was 7.221 573 Tg, estimated values of spatial co-simulation ranged from 0 to 109.178 Mg·hm^-2, mean carbon density was 32.376 4 Mg·hm^-2, mean relative error based on the 15% test sample was 4.565%, and forest carbon estimated by simulation was in the confidence interval range measured by sample plots. Also, the correlation analysis varied with the resolution of remote sensing image changes. Thus, a key point of further research would be to improve the estimation accuracy of forest carbon reserves.[Ch, 5 fig. 3 tab. 29 ref.]
- forest ecology,
- forest carbon storage,
- Sequential Gaussian Co-simulation,
- ZY-3 image,
- forest resource inventory

Relative Article

[1]	JIAN Yongqi, WU Jiasen, SHENG Weixing, NIE Guohui, ZHENG Cheng, JIANG Peikun. Effects of thinning and stand types on litter stock and soil water-holding capacity . Journal of Zhejiang A&F University, 2021, 38(2): 320-328. doi: 10.11833/j.issn.2095-0756.20200355
[2]	LI Zuoyu, DONG Hongxian, LIU Leilei, LEI Zupei. Evaluation of forest ecosystem service value in Wuyanling National Nature Reserve of Zhejiang Province . Journal of Zhejiang A&F University, 2020, 37(5): 891-897. doi: 10.11833/j.issn.2095-0756.20190573
[3]	FU Jie, ZHANG Hua, HE Hong, LÜ Rui, WANG Ying, ZHU Yeping. Species composition characteristics of forest community in the paleo-stone stream of the eastern Liaoning mountainous region . Journal of Zhejiang A&F University, 2019, 36(3): 533-540. doi: 10.11833/j.issn.2095-0756.2019.03.014
[4]	WANG Yanfang, LIU Ling, YUE Feixue, LI Dong, SHANGGUAN Zhouping. Forest aboveground carbon storage with the Grain for Green Program in Henan Province . Journal of Zhejiang A&F University, 2019, 36(3): 507-514. doi: 10.11833/j.issn.2095-0756.2019.03.011
[5]	ZHAO Qing, QIAN Wanhui, TANG Honghui, YANG Qing, YAN Jun. Differences of health care functions of six forest stands in Yunyong Forest Park, Guangdong . Journal of Zhejiang A&F University, 2018, 35(4): 750-756. doi: 10.11833/j.issn.2095-0756.2018.04.021
[6]	ZHANG Yong, HU Haibo, WANG Zeng, HUANG Yujie, LÜ Aihua, ZHANG Jinchi, LIU Shenglong. Varieties of active soil organic carbon of four forest types with varying incubation temperatures in Fengyang Mountain . Journal of Zhejiang A&F University, 2018, 35(2): 243-251. doi: 10.11833/j.issn.2095-0756.2018.02.007
[7]	ZHU Yuyi, XIE Weijia, HUANG Huaguo. Modeling sensible flux and latent flux in Heihe and boreal forests based on a 3D ENVI-met model . Journal of Zhejiang A&F University, 2018, 35(3): 440-452. doi: 10.11833/j.issn.2095-0756.2018.03.007
[8]	HE Tao, SUN Yujun. Dynamic monitoring of forest carbon stocks based on the InVEST model . Journal of Zhejiang A&F University, 2016, 33(3): 377-383. doi: 10.11833/j.issn.2095-0756.2016.03.002
[9]	NIU Xiaodong, JIANG Hong, WANG Fan. Stable isotope composition for atmospheric water vapor in the forest ecosystem of Mount Tianmu . Journal of Zhejiang A&F University, 2015, 32(3): 327-334. doi: 10.11833/j.issn.2095-0756.2015.03.001
[10]	GONG Hede, YAN Teng, PENG Yihang, ZHENG Li, MA Yuewei. Seedling growth of Symplocos ramosissima with manually controlled experimental conditions in the Ailaoshan Mountains, Yunnan Province . Journal of Zhejiang A&F University, 2015, 32(2): 237-243. doi: 10.11833/j.issn.2095-0756.2015.02.010
[11]	GUO Hanru, ZHANG Maozhen, XU Lihua, YUAN Zhenhua, CHEN Tiange. Geographically weighted regression based on estimation of regional forest carbon storage . Journal of Zhejiang A&F University, 2015, 32(4): 497-508. doi: 10.11833/j.issn.2095-0756.2015.04.002
[12]	ZHANG Yang, SHEN Jiahong, ZHANG Fanggang. Structure and dynamics of forest community in the West Lake Scenic of Hangzhou, China . Journal of Zhejiang A&F University, 2014, 31(4): 589-596. doi: 10.11833/j.issn.2095-0756.2014.04.015
[13]	Lü Kunlong, RAO Liangyi, LI Feifei, LI Huijie, ZHU Mengxun, ZHU Zhenya, ZHOU Jian. Storage and determinants of coarse woody debris in China’s forests . Journal of Zhejiang A&F University, 2013, 30(1): 114-122. doi: 10.11833/j.issn.2095-0756.2013.01.017
[14]	WEI Shujing, SUN Long, WEI Shuwei, HU Haiqing. Coarse woody debris in forest ecosystems: a review . Journal of Zhejiang A&F University, 2013, 30(4): 585-598. doi: 10.11833/j.issn.2095-0756.2013.04.019
[15]	LIU Jiamin, ZHANG Hui, HUANG Xiufeng, XU Huachao. Insect diversity to monitor and evaluate forest environmental quality . Journal of Zhejiang A&F University, 2013, 30(5): 719-723. doi: 10.11833/j.issn.2095-0756.2013.05.013
[16]	SUN Chong, LIU Qijing. Carbon storage changes for major forest types in Beijing . Journal of Zhejiang A&F University, 2013, 30(1): 69-75. doi: 10.11833/j.issn.2095-0756.2013.01.010
[17]	CAI Xia, WANG Zu-hua, CHEN Li-juan. Zoning with spatial division of forest ecosystem services in Chun’an County，Zhejiang Province . Journal of Zhejiang A&F University, 2011, 28(5): 727-734. doi: 10.11833/j.issn.2095-0756.2011.05.007
[18]	CHEN Ya-feng, YU Shu-quan, YAN Xiao-su, YI Li-ta, BAO Chun-quan. Community structure of three different forests in Tonglu，Zhejiang Province . Journal of Zhejiang A&F University, 2011, 28(3): 408-415. doi: 10.11833/j.issn.2095-0756.2011.03.010
[19]	杨廉雁, 张树斌, 郑征. Research progress on the hollow－bearing trees in forest ecosystems . Journal of Zhejiang A&F University, 2010, 27(6): 928-934. doi: 10.11833/j.issn.2095-0756.2010.06.021
[20]	WANG Bin, YANG Xiao-sheng, ZHANG Biao, LIU Mou-cheng. Dynamics of ecosystem services in China during 1973 - 2003 . Journal of Zhejiang A&F University, 2009, 26(5): 714-721.

References

[1]	FANG Jingyun, CHEN Anping. Dynamic forest biomass carbon pools in China and their significance [J]. Acta Bot Sin, 2001, 43(9): 967-973.
[2]	SCHIMEL D S, HOUSE J I, HIBBARD K A, et al. Recent patterns and mechanisms of carbon exchange by terrestrial ecosystems [J]. Nature, 2001, 414(6860): 169-172.
[3]	TANG Xuguang, LIU Dianwei, WANG Zongming, et al. Estimation of forest aboveground biomass based on remote sensing data: a review [J]. Chin J Ecol, 2012, 31(5): 1311-1318.
[4]	XU Xiaojun, DU Huaqiang, ZHOU Guomo, et al. Review on correlation analysis of independent variables in estimation models of vegetation biomass based on remote sensing [J]. Remote Sens Technol Appl, 2008, 23(2): 123-131.
[5]	FAZAKAS Z, NILSSON M, OLSSON H. Regional forest biomass and wood volume estimation using satellite data and ancillary data [J]. Agric For Meteorol, 1999, 98(1): 417-425.
[6]	MAKELA H, PEKKARINEN A. Estimation of forest stands volumes by Landsat TM imagery and stand level field inventory data [J]. For Ecol Manage, 2004, 196(2/3): 245-255.
[7]	CAO Qingxian, XU Daping, JU Hongbo. Biomass estimation of five kinds of mangrove community with the KNN method based on the spectral information and textural features of TM images [J]. For Res, 2011, 24(2): 144-150.
[8]	WANG Shujun, GUAN Dongsheng. Remote sensing method of forest biomass estimation by artificial neural network models [J]. Ecol Environ, 2007, 16(1): 108-111.
[9]	LIU Zhihua, CHANG Yu, CHEN Hongwei. Remote sensing method of forest biomass estimation by artificial neural network models [J]. Chin J Appl Ecol, 2008, 19(9): 1891-1896.
[10]	WANG Shaohua, ZHANG, Maozhen, ZHAO Pingan, et al. Modelling the spatial distribution of forest carbon stocks with artificial neural network based on TM images and forest inventory data [J]. Acta Ecol Sin, 2011, 31(4): 998-1008.
[11]	LEE N J, NAKANE K. Forest vegetation classification and biomass estimation based on Landsat TM data in a mountainous region of west Japan[M]//GHOLZ, H L, NAKANE K, HIMODA H. The Use of Remote Sensing in the Modeling of Forest Productivity [J]. [s. l.]: Springer Netherlands, 1997: 159-171.
[12]	YANG Cunjian, LIU Jiyuan, ZHANG Zengxiang. Study on the estimation of the tropical forest vegetation biomass by using remote sensing data ,meteorological data and topographical data [J]. Geogr Geo-Inf Sci, 2004, 20(6): 22-25.
[13]	WANG Guanghua, LIU Qijing. Estimating carbon sequestration of Beijing’s forests based on TM images and forest inventory data [J]. J Basic Sci Eng, 2013, 21(2): 224-235.
[14]	NIU Zhichun, NI Shaoxiang. Study on models for monitoring of grassland biomass around Qinghai Lake assisted by remote sensing [J]. Acta Geogr Sin, 2003, 58(5): 695-702.
[15]	LI Jian, SHU Xiaobo, CHEN Shuisen. Establishment of wetland vegetation biomass model by in-situ and remote sensing observation in Poyang Lake area [J]. J Guangzhou Univ Nat Sci Ed, 2005, 4(6): 494-498.
[16]	MYEONG S, NOWAK D J, DUGGIN M J. A temporal analysis of urban forest carbon storage using remote sensing[J]. Remote Sens Environ, 2006, 101(2): 277-282.
[17]	FENG Yiming, TANG Shouzheng, LI Zengyuan. Simulation of spatial distribution pattern of forest types by using sequential indicator simulation [J]. Acta Ecol Sin, 2004, 24(5): 946-952.
[18]	WANG Guangxing, OYANA T, ZHANG Maozhen, et al. Mapping and spatial uncertainty analysis of forest vegetation carbon by combining national forest inventory data and satellite images [J]. For Ecol Manage, 2009, 258(7): 1275-1283.
[19]	ZHANG Maozhen, WANG Guangxing, GE Hongli, et al. Estimation of forest carbon distribution for Xianju County based on spatial simulation [J]. Sci Silv Sin, 2014, 50(11): 13-22.
[20]	REESE H, NILSSON M, SANDSTROM P, et al. Applications using estimates of forest parameters derived from satellite and forest inventory data [J]. Comp Elect Agric， 2002, 37(2): 37-55.
[21]	XU Xinliang, CAO Mingkui. An analysis of the applications of remote sensing method to the forest biomass estimation [J]. Geo-Inf Sci, 2006, 8(4): 122-128.
[22]	SHEN Xi, ZHANG Maozhen, QI Xiangbin. Comparison of regional forest carbon estimation methods based on regression and stochastic simulation [J]. Sci Silv Sin, 2011, 47(6): 1-8.
[23]	ZHANG Maozhen, WANG Guangxing, ZHOU Guomo, et al. Mapping of forest carbon by combining forest inventory data and satellite images with co-simulation based up-scaling method [J]. Acta Ecol Sin, 2009, 29(6): 2919-2928.
[24]	ZENG Weisheng, XIAO Qianhui, HU Jue, et al. Establishment of single-tree biomass equations for pinus massoniana in southern China [J]. J Cent South Univ For Technol, 2010, 30(5): 50-55.
[25]	ZHU Tangjun, SHEN Chuchu, JI Biyong, et al. Research on biomass expansion factor of chinese fir forest in Zhejiang Province based on LULUCF greenhouse gas Inventory [J]. Acta Ecol Sin, 2013, 33(13): 3925-3932.
[26]	SHEN Chuchu. The Research on Biomass Expansion Factors of The Dominant Tree Species in Zhejiang Province [D]. Lin’an: Zhejiang A & F University, 2013.
[27]	CHEN Hui, HONG Wei, LAN Bin, et al. Study on biomass and productivity of Phyllostachys heterocycla cv. pubescens forest in the north of Fujian [J]. Sci Silv Sin, 1998, 34(1): 60-64.
[28]	XIONG Qiufen, HUANG Mei, XIONG Minquan, et al. Cross-validation error analysis of daily gridded precipitation based on China meteorological observation [J]. Plateau Meteorol, 2011, 30(6): 1615-1625.
[29]	SONG Yongkang, SHU Xiao, WANG Bingjie. Selection and assessment of geological prediction models based on cross validation [J]. Petrochem Ind Appl, 2013, 32(12): 65-88.

Proportional views

通讯作者: 陈斌, bchen63@163.com

1.
沈阳化工大学材料科学与工程学院沈阳 110142

Figures(5) / Tables(3)

Get Citation

PDF

XML

Article views(3250) PDF downloads(450) Cited by()

Proportional views

HTML

森林是陆地生态系统的主体，也是生物碳的主要载体，每年固定的碳占整个陆地生态系统碳总量的2/3，对减缓温室效应起着决定性作用^[1-2]。精确计量和监测区域森林碳储量及其分布，不仅是区域碳源汇功能评价的基础，也对环境发展和相关决策有重要参考意义。目前，对森林生物量/碳储量，传统估算方法主要有收获法、测树学法、森林资源清查法。传统估计方法以实测数据为基础，需要进行大量的实地调查，工作量大、周期性长，在推测大面积林分生物量时，待测林分的每木检尺数据往往难以获得^[3]。遥感技术作为一门综合性探测技术，可以在短时间内对同一地区进行重复探测，获取大面积同步观测数据，并且不受地形阻隔等限制，弥补了传统样地调查方法的不足，具有良好的综合性和现势性^[4]。在基于遥感的森林碳储量估测过程中，根据光谱信息与森林植被关系构造的不同遥感信息模型进行分类，可概括为非参数模型和参数模型2种方法。非参数模型法主要包括：KNN(K-nearest neighbor)法，作为一种较早被国内外学者使用的方法，主要用应用于区域森林蓄积量和生物量的估计^[5-7]；人工神经网络(ANN)模型估测法，在近几十年中被应用和发展，通过与遥感信息和野外调查数据结合对研究区的森林蓄积量和生物量估计，并在有效性和精确性上得到验证^[8-10]。参数模型法有：线性模型估测法，该方法原理简单且应用广泛，以遥感影像信息和实测样地生物量间的线性关系为依据，对森林生物量进行估算^[11-13]；非线性模型估测法则通过建立两者的非线性模型实现对森林生物量的估计及其动态变化^[14-16]，也是森林生物量估算的常用方法之一；序列高斯仿真模型估测法，在过去的研究中主要运用在土壤、采矿、油田等方面，近年来在林业应用中有所发展，通过将模型与林业实际生产过程需求结合^[17]，建立遥感信息及样地清查数据之间的空间关系^[18-19]，对区域森林碳储量及其分布进行探究。非参数估计中人工神经网络模型外延性较差，并需要依据大量的野外调查数据以及多次测算作为依据^[5]，KNN法则在原理上过于依赖极限定理，致使在像元尺度上生物量估计的误差较大^[20-21]。参数估计方法中主要以回归模拟为主导，采用最小二乘法以降低局部真实性使模拟结果达到整体最优，具有较高的平滑效应^[22]。序列高斯协同仿真模拟不仅克服了回归模型的平滑缺陷，真实再现区域森林碳分布的局部特征，还提供了估计结果的不确定性，通过对不确定性传播模型的模拟分析实现对森林碳估计结果的评价^[23]。目前，空间仿真方法在区域森林碳储量估计方面的应用研究较少。本研究以浙江省开化县为研究区，基于2014年森林资源清查样地数据与2013年资源3号遥感影像数据(2 m × 2 m分辨率融合影像)，采用序列高斯协同仿真算法，对该区域地上部分森林碳储量及其分布进行空间估计，并对模拟误差进行分析。

2. 研究方法

2.1. 实测样地森林植被碳储量估计

地面实测数据包含样地数据和样木数据(含单株毛竹数据)2个部分，每个样地的森林生物量、碳储量由每株样木(或毛竹)的生物量和碳储量累加而来。以97个地面样地每木检尺记录为基础，按照研究区主要树种的立木胸径—生物量模型(表 2)计算单株样木地上部分生物量，因而选择不含树高的一元模型计算单株生物量。最后根据广泛接受的生物量—碳储量转换系数0.5，将累加样地内所有样木生物量得到的样地森林地上部分生物量转换为样地地上部分碳储量。其中，扣除检尺类型为采伐木、多测木以及枯倒木的部分，由于样木调查不包含树高。

树种类型	生物量回归方程	文献
马尾松	$ w = 0.112243{D^{2.36501}}$	^[24]
杉木	$ w = 0.086904{\left\{ {{D^2}\left[ {1.3 + {{\left( {0.232467 + \frac{{2.362912}}{D}} \right)}^{ - 25}}} \right]} \right\}^{0.819180}}$	^[25]
硬阔类	$ w = 0.108921{\left\{ {{D^2}\left[ {1.3 + {{\left( {0.2791 + \frac{{1.7661}}{D}} \right)}^{ - 25}}} \right]} \right\}^{0.901726}}$	^[26]
软阔类	$ w = 0.097822{\left\{ {{D^2}\left[ {1.3 + {{\left( {0.3063 + \frac{{1.9783}}{D}} \right)}^{ - 25}}} \right]} \right\}^{0.893738}}$	^[26]
毛竹	$w = 386.4951{D^{1.6579}} $	^[27]
说明：w表示碳储量，单位为Mg·hm^-2；D表示胸径，单位为cm。

Table 2. Carbon density and biomass calculated in the aboveground of the study area

2.2. 序列高斯协同仿真模型模拟估计

序列高斯协同仿真模拟(sequential Gaussian co-simulation)，以地统计学中变异函数为基础，将已知点x处的观测值Z(x)作为随机变量构造一个概率空间，并依照区域内距离观测点为h(有方向、大小的向量)的所有值Z(x＋h)，计算此概率空间内随机变量的期望和方差，可得到与该随机变量Z(x)相符的高斯分布概率密度函数以及条件累积分布函数。进而根据Monto-Carlo原理随机抽取累积分布函数中一个值作为服从该分布的一次模拟结果。该模拟方法不但能保持变量的空间相关性不变，还能使观测数据条件化，参与模拟点越多，模拟结果越接近客观实际。

针对森林的分布和自身地理特征，序列高斯模拟可以很好地对区域森林碳储量结构性和随机性进行描述和分析。其中，协方差函数和半变异函数是以区域化变量为理论基础的地统计学基本函数，用这2个模型对森林碳储量进行描述能够兼顾其结构性与随机性。同时，2个相交的半变异函数可以度量其对应的2个随机变量Z(xα)与Z(xα+h)的空间相关性。随机函数的半变异函数γ_Z(h)和协方差C_Z(h)表示为：

在式(1)和式(2)中：h是一个有方向性的向量，决定了2个随机变量之间的距离和位置； N(h)为给定向量h上样本数据对的数量；Z(x_α)与 Z(x_α+h)分别为在 x_α与 x_α+h 的2个位置的森林碳储量统计值；m_-h与m_+h表示结尾和初始数据的均值。

在区域森林碳估计中，仅采用实测样地数据信息相对单一。王广兴等^[18]则在此基础上引入信息丰富的遥感影像数据。假设遥感影像的光谱变量为Y，则Y(x)为空间x处的一个随机函数，通过样地与影像相应位置的交叉半变异函数γ_ZY(h)和交叉协方差函数C_XY(h)更加精确地描述出森林碳储量的分布和空间变异情况。2个函数分别表示为：

式(3)和(4)中：Y(x_α),Y(x_α+h)分别表示在 x_α和 x_α+h这2个位置的光谱值。

序列高斯协同仿真模拟的程序运行步骤可概括为：①采用随机抽样法抽取一个像元作为待估点并设置影像像元的估算顺序；②对已知样点拟合得到半变异函数，以确定模型模拟的变程、基台值、块金值参数；③根据点协同克里格法得到变程内已知样本的均值和方差，以此得到待估像元对应的条件累积分布函数；④在累积分布函数中随机抽取一个值作为该像元的碳密度模拟值；⑤重复步骤③~④遍历整个影像直至得出研究区内每个像元的估计值，即可获得一幅研究区森林碳分布图；⑥重复③~⑤步骤N次，得到N幅森林碳密度分布图，通过计算碳分布图对应像元N次模拟的均值和方差，进而得到用于不确定性分析的森林碳分布方差图和概率图。在本研究中，N的经验值为200。

2.3. 精度验证

根据以下几种误差衡量指标来进行统计描述^[28-29]。

2.3.1. 平均误差E_ME(mean error，ME)

所有测量值估计误差的平均值，用来衡量估值结果的准确性。

2.3.2. 残差平方和E_RSS(residual square sum，RSS)

真实值与估测值之差的平方和，表示随机误差的效应。

2.3.3. 平均相对误差E_MRE(mean relative error，MRE)

是一种比较常用的指标，但是需要与其他指标相结合共同完成模型精度的评估。

2.3.4. 均方根误差E_RMSE(root mean square error，RMSE)

衡量估计值与真实值之间的偏差，能够很好的反映出估测的精度。

2.3.5. 相对均方根误差E_RRMSE(relative root mean square error，RRMSE)

通过相对均值的离散程度来反映模型的模拟精度。

式(5)~式(9)中：z(x_i)为样地真实值，$\hat z\left( {{x_i}} \right) $为样地估计值，$z\left( {\bar x} \right) $为碳密度均值，n为样地个数。

4. 结论与讨论

本研究以森林资源清查数据与资源三号遥感影像数据为基础，对开化县2014年森林碳储量及其分布进行序列高斯协同仿真模拟，得到总碳储量估计结果7.221 573 Tg，碳密度模拟值为0~109.178 0 Mg·hm^-2，碳密度平均值为32.376 4 Mg·hm^-2。以实测样地数据估计结果作为真实参考，总碳储量为5.953 610 Tg，样地碳密度值为0~142.112 2 Mg·hm^-2，碳密度均值26.650 0 Mg·hm^-2，总体估计精度达78.7%。

由于开化县拥有森林资源丰富的国家森林公园和自然保护区，因此估计得到较高的碳储量。受人类活动范围的影响，其碳分布主要集中在研究区北部、西北部以及南部、西南部地区，与碳分布的估计结果基本吻合。根据精度分析得出空间仿真模拟对区域碳储量估计具有满意结果，仿真估计的碳总量落在实测样地估算总值的置信区间内。

由于序列高斯协同仿真算法在森林碳储量估计过程中要求样地数据和遥感数据尺度的一致性，因此需要将2 m分辨率的遥感影像重采样至30 m与样地匹配，然而将高分辨率影像降至低分辨率会使信息量减少，并产生一定误差。本研究中，2 m和30 m分辨率遥感影像信息与实测样地碳储量相关性分别为0.352和0.402，表明重采样可以增加影像与样地的匹配度，从而可以提高两者相关性；对重采样前后2种影像进行一元二次回归验证得出，2 m分辨率影像模拟研究区森林碳总量为631.102 8 Mg，平均碳密度28.249 9 Mg·hm^-2，30 m分辨率影像模拟得到的森林碳总量为627.296 0 Mg，平均碳密度28.079 5 Mg·hm^-2，表明30 m分辨率影像模拟得到的森林碳密度均值和碳总量更接近于样地实测值。重采样后回归模型的拟合优度高于重采样前40.7%，进一步说明对遥感影像重采样的合理性。本研究还发现，将融合影像以5 m为一个间隔重采样至5，10，15，20，25，30 m分辨率影像，地面样地森林碳储量与影像的相关性随空间分辨率的变化而变化，表明样地与影像分辨率不统一对森林碳储量的精确估计有一定阻碍，因此，如何使不同尺度的2种数据有效结合，是下一步需要完善和改进的重点。此外，空间协同仿真模拟依赖于参与模拟的样地数量，样地的多少决定了模型参数量，对模拟结果会造成一定影响，对样地多少的控制也值得进一步探索和研究。

Reference (29)

[1]	方精云, 陈安平. 中国森林植被碳库的动态变化及其意义[J]. 植物学报, 2001, 43(9): 967-973.	FANG Jingyun, CHEN Anping. Dynamic forest biomass carbon pools in China and their significance[J]. Acta Bot Sin, 2001, 43(9): 967-973.
[2]	SCHIMEL D S, HOUSE J I, HIBBARD K A. Recent patterns and mechanisms of carbon exchange by terrestrial ecosystems[J]. Nature, 2001, 414(6860): 169-172.
[3]	汤旭光, 刘殿伟, 王宗明. 森林地上生物量遥感估算研究进展[J]. 生态学杂志, 2012, 31(5): 1311-1318.	TANG Xuguang, LIU Dianwei, WANG Zongming. Estimation of forest aboveground biomass based on remote sensing data: a review[J]. Chin J Ecol, 2012, 31(5): 1311-1318.
[4]	徐小军, 杜华强, 周国模. 基于遥感植被生物量估算模型自变量相关性分析综述[J]. 遥感技术与应用, 2008, 23(2): 123-131.	XU Xiaojun, DU Huaqiang, ZHOU Guomo. Review on correlation analysis of independent variables in estimation models of vegetation biomass based on remote sensing[J]. Remote Sens Technol Appl, 2008, 23(2): 123-131.
[5]	FAZAKAS Z, NILSSON M, OLSSON H. Regional forest biomass and wood volume estimation using satellite data and ancillary data[J]. Agric For Meteorol, 1999, 98(1): 417-425.
[6]	MAKELA H, PEKKARINEN A. Estimation of forest stands volumes by Landsat TM imagery and stand level field inventory data[J]. For Ecol Manage, 2004, 196(2/3): 245-255.
[7]	曹庆先, 徐大平, 鞠洪波. 基于TM影像纹理与光谱特征和KNN方法估算5种红树林群落生物量[J]. 林业科学研究, 2011, 24(2): 144-150.	CAO Qingxian, XU Daping, JU Hongbo. Biomass estimation of five kinds of mangrove community with the KNN method based on the spectral information and textural features of TM images[J]. For Res, 2011, 24(2): 144-150.
[8]	王淑君, 管东生. 神经网络模型森林生物量遥感估测方法的研究[J]. 生态环境, 2007, 16(1): 108-111.	WANG Shujun, GUAN Dongsheng. Remote sensing method of forest biomass estimation by artificial neural network models[J]. Ecol Environ, 2007, 16(1): 108-111.
[9]	刘志华, 常禹, 陈宏伟. 基于遥感、地理信息系统和人工神经网络的呼中林区森林蓄积量估测[J]. 应用生态学报, 2008, 19(9): 1891-1896.	LIU Zhihua, CHANG Yu, CHEN Hongwei. Remote sensing method of forest biomass estimation by artificial neural network models[J]. Chin J Appl Ecol, 2008, 19(9): 1891-1896.
[10]	汪少华, 张茂震, 赵平安. 基于TM影像森林资源清查数据和人工神经网络的森林碳空间分布模拟[J]. 生态学报, 2011, 31(4): 998-1008.	WANG Shaohua, ZHANG , Maozhen , ZHAO Pingan. Modelling the spatial distribution of forest carbon stocks with artificial neural network based on TM images and forest inventory data[J]. Acta Ecol Sin, 2011, 31(4): 998-1008.
[11]	LEE N J, NAKANE K. Forest vegetation classification and biomass estimation based on Landsat TM data in a mountainous region of west Japan[M]//GHOLZ, H L, NAKANE K, HIMODA H. The Use of Remote Sensing in the Modeling of Forest Productivity [J]. [s. l.]: Springer Netherlands, 1997: 159-171.
[12]	杨存建, 刘纪远, 张增祥. 热带森林植被生物量遥感估算探讨[J]. 地理与地理信息科学, 2004, 20(6): 22-25.	YANG Cunjian, LIU Jiyuan, ZHANG Zengxiang. Study on the estimation of the tropical forest vegetation biomass by using remote sensing data ,meteorological data and topographical data[J]. Geogr Geo-Inf Sci, 2004, 20(6): 22-25.
[13]	王光华, 刘琪璟. 基于TM影像和森林资源二类调查数据的北京森林碳汇估算[J]. 应用基础与工程科学学报, 2013, 21(2): 224-235.	WANG Guanghua, LIU Qijing. Estimating carbon sequestration of Beijing’s forests based on TM images and forest inventory data[J]. J Basic Sci Eng, 2013, 21(2): 224-235.
[14]	牛志春, 倪绍祥. 青海湖环湖地区菜地植被生物量遥感监测模型[J]. 地理学报, 2003, 58(5): 695-702.	NIU Zhichun, NI Shaoxiang. Study on models for monitoring of grassland biomass around Qinghai Lake assisted by remote sensing[J]. Acta Geogr Sin, 2003, 58(5): 695-702.
[15]	李健, 舒晓波, 陈水森. 基于Landsat-TM数据鄱阳湖湿地植被生物量遥感监测模型的建立[J]. 广州大学学报:自然科学版, 2005, 4(6): 494-498.	LI Jian, SHU Xiaobo, CHEN Shuisen. Establishment of wetland vegetation biomass model by in-situ and remote sensing observation in Poyang Lake area[J]. J Guangzhou Univ Nat Sci Ed, 2005, 4(6): 494-498.
[16]	MYEONG S, NOWAK D J, DUGGIN M J. A temporal analysis of urban forest carbon storage using remote sensing[J]. Remote Sens Environ, 2006, 101(2): 277-282.
[17]	冯益明, 唐守正, 李增元. 应用序列指示条件模拟算法模拟森林类型空间分布[J]. 生态学报, 2004, 24(5): 946-952.	FENG Yiming, TANG Shouzheng, LI Zengyuan. Simulation of spatial distribution pattern of forest types by using sequential indicator simulation[J]. Acta Ecol Sin, 2004, 24(5): 946-952.
[18]	WANG Guangxing, OYANA T, ZHANG Maozhen. Mapping and spatial uncertainty analysis of forest vegetation carbon by combining national forest inventory data and satellite images[J]. For Ecol Manage, 2009, 258(7): 1275-1283.
[19]	张茂震, 王广兴, 葛宏立. 基于空间仿真的仙居县森林碳分布估算[J]. 林业科学, 2014, 50(11): 13-22.	ZHANG Maozhen, WANG Guangxing, GE Hongli. Estimation of forest carbon distribution for Xianju County based on spatial simulation[J]. Sci Silv Sin, 2014, 50(11): 13-22.
[20]	REESE H, NILSSON M, SANDSTROM P. Applications using estimates of forest parameters derived from satellite and forest inventory data[J]. Comp Elect Agric, 2002, 37(2): 37-55.
[21]	徐新良, 曹明奎. 森林生物量遥感估算与应用分析[J]. 地球信息科学, 2006, 8(4): 122-128.	XU Xinliang, CAO Mingkui. An analysis of the applications of remote sensing method to the forest biomass estimation[J]. Geo-Inf Sci, 2006, 8(4): 122-128.
[22]	沈希, 张茂震, 祁祥斌. 基于回归于随机模拟的区域森林碳分布估计方法比较[J]. 林业科学, 2011, 47(6): 1-8.	SHEN Xi, ZHANG Maozhen, QI Xiangbin. Comparison of regional forest carbon estimation methods based on regression and stochastic simulation[J]. Sci Silv Sin, 2011, 47(6): 1-8.
[23]	张茂震, 王广兴, 周国模. 基于森林资源清查、卫星影像数据与随机协同模拟尺度转换方法的森林碳制图[J]. 生态学报, 2009, 29(6): 2919-2928.	ZHANG Maozhen, WANG Guangxing, ZHOU Guomo. Mapping of forest carbon by combining forest inventory data and satellite images with co-simulation based up-scaling method[J]. Acta Ecol Sin, 2009, 29(6): 2919-2928.
[24]	曾伟生, 肖前辉, 胡觉. 中国南方马尾松立木生物量模型研建[J]. 中南林业科技大学学报, 2010, 30(5): 50-55.	ZENG Weisheng, XIAO Qianhui, HU Jue. Establishment of single-tree biomass equations for pinus massoniana in southern China[J]. J Cent South Univ For Technol, 2010, 30(5): 50-55.
[25]	朱汤军, 沈楚楚, 季碧勇. 基于LULUCF温室气体清单编制的浙江省杉木林生物量换算因子[J]. 生态学报, 2013, 33(13): 3925-3932.	ZHU Tangjun, SHEN Chuchu, JI Biyong. Research on biomass expansion factor of chinese fir forest in Zhejiang Province based on LULUCF greenhouse gas Inventory[J]. Acta Ecol Sin, 2013, 33(13): 3925-3932.
[26]	沈楚楚. 浙江省主要树种组生物量转换因子的研究[D]. 临安：浙江农林大学, 2013.	SHEN Chuchu. The Research on Biomass Expansion Factors of The Dominant Tree Species in Zhejiang Province [D]. Lin’an: Zhejiang A & F University, 2013.
[27]	陈辉, 洪伟, 兰斌. 闽北毛竹生物量与生产力的研究[J]. 林业科学, 1998, 34(1): 60-64.	CHEN Hui, HONG Wei, LAN Bin. Study on biomass and productivity of Phyllostachys heterocycla cv. pubescens forest in the north of Fujian[J]. Sci Silv Sin, 1998, 34(1): 60-64.
[28]	熊秋芬, 黄玫, 熊敏诠. 基于国家气象观测站逐日降水格点数据的交叉检验误差分析[J]. 高原气象, 2011, 30(6): 1615-1625.	XIONG Qiufen, HUANG Mei, XIONG Minquan. Cross-validation error analysis of daily gridded precipitation based on China meteorological observation[J]. Plateau Meteorol, 2011, 30(6): 1615-1625.
[29]	宋永康, 舒晓, 王兵杰. 基于交叉检验法的地质预测模型优选[J]. 石油化工应用, 2013, 32(12): 65-88.	SONG Yongkang, SHU Xiao, WANG Bingjie. Selection and assessment of geological prediction models based on cross validation[J]. Petrochem Ind Appl, 2013, 32(12): 65-88.

数据	样本数量	碳密度/(Mg·hm^-2)			标准差	偏值	峰值	变异系数
数据	样本数量	最小值	最大值	平均值	标准差	偏值	峰值	变异系数
训练样本(85%)	80	0	142.11	26.700	30.333	1.509 8	5.540 2	1.136 1
检验样本（15%)	15	0	103.31	26.402	30.201	1.163 0	3.799 1	1.143 9

检验样本/个	平均误差	残差平方和	平均相对误差/％	均方根误差
15(15%)	9.702 2	7 667.262 0	4.565 8	22.608 6
95(100%)	5.272 3	33 505.518 8	0.558 0	18.780 0

A spatial co-simulation based forest carbon estimation for Kaihua County

doi: 10.11833/j.issn.2095-0756.2016.03.003