三种森林生物量估测模型的比较分析

doi:10.3724/SP.J.1258.2011.00402

摘要/Abstract

摘要：

森林生物量的定量估算为全球碳储量、碳循环研究提供了重要的参考依据。该研究采用黑龙江长白山地区的TM影像和133块森林资源一类清查样地的数据, 选取地学参数、遥感反演参数等71个自变量分别构建多元逐步回归模型、传统BP (back propagation)神经网络模型和基于高斯误差函数的BP神经网络改进模型(Gaussian error function, Erf-BP), 进而估算该地区的森林生物量, 并进行比较分析。结果表明, 多元逐步回归模型估测的森林生物量预测精度为75%, 均方根误差为26.87 t·m^-2; 传统BP神经网络模型估测森林生物量的预测精度为80.92%, 均方根误差为21.44 t·m^-2; Erf-BP估测森林生物量的预测精度为82.22%, 均方根误差为20.83 t·m^-2。可见, 改进后的Erf-BP能更好地模拟生物量与各个因子之间的关系, 估算精度更高。

关键词: 生物量, BP神经网络模型, 基于高斯误差函数的BP神经网络改进模型, 多元逐步回归

Abstract:

Aims Quantitative estimation of forest biomass is significant to studies of global carbon storage and carbon cycle. Our objective is to develop models to estimate forest biomass accurately.
Methods Multi-stepwise regression model, traditional back propagation (BP) neutral network model and BP neutral network model based on Gaussian error function (Erf-BP) were developed to estimate forest biomass in Changbai Mountain of Heilongjiang, China according to TM imagery and 133 plots of forest inventory data. There were 71 dependent variables of geoscience and remote sensing.
Important findings The precisions and root mean square errors of multi-stepwise regression model, traditional BP neutral network model and Erf-BP were 75%, 26.87 t·m^-2; 80.92%, 21.44 t·m^-2 and 82.22%, 20.83 t·m^-2, respectively. Therefore, the relations between forest biomass and various factors can be better modeled and described by the improved Erf-BP.

Key words: biomass, back propagation (BP) neural network model, BP neutral network model based on Gaussian error function (Erf-BP), multi-stepwise regression

范文义, 张海玉, 于颖, 毛学刚, 杨金明. 三种森林生物量估测模型的比较分析. 植物生态学报, 2011, 35(4): 402-410. DOI: 10.3724/SP.J.1258.2011.00402

FAN Wen-Yi, ZHANG Hai-Yu, YU Ying, MAO Xue-Gang, YANG Jin-Ming. Comparison of three models of forest biomass estimation. Chinese Journal of Plant Ecology, 2011, 35(4): 402-410. DOI: 10.3724/SP.J.1258.2011.00402

导出引用管理器 EndNote|Ris|BibTeX

链接本文: https://www.plant-ecology.com/CN/10.3724/SP.J.1258.2011.00402

https://www.plant-ecology.com/CN/Y2011/V35/I4/402

图/表 13

参考文献 21

[1]	Brown S, Sathaye J, Canell M, Kauppi PE (1996). Mitigation of carbon emission to the atmosphere by forest management. Commonwealth Forestry Review, 75, 80-91.
[2]	Chen CG (陈传国), Zhu JF (朱俊凤) (1989). Northeast Main Forest Biomass Handbook (东北主要林木生物量手册). China Forestry Publishing House, Beijing. (in Chinese)
[3]	Ding BY (丁宝永), Sun JH (孙继华) (1989). Studies on biological productivity and nutrient cycling of artificial forest ecosystem of Korean pine. Journal of Northeast Forestry University (东北林业大学学报), 17(Suppl. 2), 18-19. (in Chinese with English abstract)
[4]	Efron B (1979). Bootstrap methods: another look at the jackknife. Annals of Statistics, 7, 1-26.
[5]	Guo QX (国庆喜), Zhang F (张锋) (2003). Estimation of forest biomass based on remote sensing. Journal of Northeast Forestry University (东北林业大学学报), 31(2), 13-16. (in Chinese with English abstract)
[6]	Guo ZH (郭志华), Peng SL (彭少麟), Wang BS (王伯荪) (2002). Estimating forest biomass in western Guangdong using landsat TM data. Acta Ecologica Sinica (生态学报), 22, 1832-1839. (in Chinese with English abstract)
[7]	Hall RJ, Skakun RS, Arsenault EJ, Case BS (2006). Modeling forest stand structure attributes using Landsat ETM+ data: application to mapping of aboveground biomass and stand volume. Forest Ecology and Management, 225, 378-390.
[8]	Haralick RM (1979). Statistical and structural approaches to texture. Proceeding of the IEEE, 67, 786-804.
[9]	Holben B, Tucker CJ, Fan CJ (1980). Spectral assessment of soybean leaf area and leaf biomass. Photogrammetric Engineering and Remote Sensing, 46, 651-656.
[10]	Ingram JC, Dawson TP, Whittaker RJ (2005). Mapping tropical forest structure in southeastern Madagascar using remote sensing and artificial neural networks. Remote Sensing of Environment, 94, 491-507.
[11]	Jiang H (江洪) (1992). Study on Donglingshan Mountain Deciduous Broadleaved Forest Biomass and NPP Research (东灵山落叶阔叶林森林生物量和NPP的研究). Postdoctoral research report, Institute of Botany, Chinese Academy of Sciences, Beijing. (in Chinese with English abstract)
[12]	Lu D, Batistella M (2005). Exploring TM image texture and its relationship with biomass estimation in Rondonia, Brazilian Amazon. Acta Amazonica, 35, 249-257.
[13]	Luo TX (罗天祥) (1996). Patterns of Net Primary Productivity for Chinese Major Forest Types and Their Mathematical Models (中国主要森林类型生物生产力格局及其数学模型). PhD dissertation, The Commission for Integrated Survey of Natural Resources, Chinese Academy of Sciences, Beijing. 181-184. (in Chinese with English abstract)
[14]	Pang Y (庞勇), Li ZY (李增元), Chen EX (陈尔学), Sun GQ (孙国清) (2005). Lidar remote sensing technology and its application in forestry. Scientia Silvae Sinicae (林业科学), 41(3), 129-136. (in Chinese with English abstract)
[15]	Singh KP, Ojha P, Malik A, Jain G (2009). Partial least squares and artificial neural networks modeling for predicting chlorophenol removal from aqueous solution. Chemometrics and Intelligent Laboratory Systems, 99, 150-160.
[16]	Tang SZ (唐守正), Zhang HR (张会儒), Xu H (胥辉) (2000). Study on establish and estimate method of compatible biomass model. Scientia Silvae Sinicae (林业科学), 36(1), 19-27. (in Chinese with English abstract)
[17]	Wang HW (王惠文), Wu ZB (吴载斌), Meng J (孟洁) (2006). Partial Least-Squares Regression-Linear and Nonlinear Methods (偏最小二乘回归的线性与非线性方法). National Defence Industrial Press, Beijing. 57. (in Chinese)
[18]	Xue L (薛立), Yang P (杨鹏) (2004). Summary of research on forest biomass. Journal of Fujian College of Forestry (福建林学院学报), 24, 283-288. (in Chinese with English abstract)
[19]	Yang CJ (杨存建), Liu JY (刘纪远), Zhang ZX (张增祥) (2004). Study on the estimation of the tropical forest vegetation biomass by using remote sensing data, meteorological data and topographical data. Geography and Geo-Information Science (地理与地理信息科学), 20(6), 22-25. (in Chinese with English abstract)
[20]	Zhang HY (张海燕), Feng TJ (冯天瑾) (2002). A study on BP networks with combined activitation functions. Journal of Ocean University of Qingdao (青岛海洋大学学报), 32, 621-626. (in Chinese with English abstract)
[21]	Zeng WH (曾文华) (2006). Texture Information Extraction in Remote Sensing Imageries with Gray Level Co-occurrence Matrix and Wavelet Transform (基于灰度共生法和小波变换的遥感影像纹理信息提取). PhD dissertation, Northeast Normal University, Changchun. 11-14. (in Chinese with English abstract)

树种¹⁾ Species¹⁾	茎生物量 Stem biomass (W_S)	枝生物量 Branch biomass (W_B)	叶生物量 Foliage biomass (W_L)	出处 Derivation	树高 Tree height (H)
1	W_S= 0.0145(D²H)^1.006	W_B = 0.000063(D²H)^1.536	W_L = 0.00104(D²H)^1.148	Ding & Sun, 1989	H = 6.6831exp(0.0515D)
2	W_S = 0.0134(D²H)^1.020	W_B = 0.0105(D²H) ^0.7386	W_L = 0.181D^1.8415	Luo, 1996	H = 5.7584exp(0.0499D)
3	W_S = 0.025(D²H)^0.96	W_B = 0.0021(D²H)^0.8181	W_L = 0.00126(D²H)^0.968	Chen & Zhu, 1989	H = 6.9441exp(0.0468D)
4	W_S = 0.057D^2.4753	W_B = 0.0116D^2.4054	W_L = 0.0083D^2.3733	Chen & Zhu, 1989	H = 6.8879exp(0.0343D)
5	W_S = 0.0238(D²H)^0.936	W_B = 0.005(D²H)^0.9105	W_L = 0.0036(D²H)^0.897	Chen & Zhu, 1989	H = 6.8879exp(0.0343D)
6	W_S = 0.032417D^2.3565	W_B = 0.20893D^1.7082	W_L = 0.29156D^1.2807	Self-established	H = 6.6831exp(0.0515D)
7	W_S = 0.06013(D²H)^0.891	W_B = 0.00652(D²H)^1.169	W_L = 0.0044(D²H)^0.9919	Jiang, 1992	H = 8.4834exp(0.0277D)
8	W_S = 0.02511(D²H)^0.927	W_B = 0.00957(D²H)^0.974	W_L = 0.8725(D²H)^0.2034	Jiang, 1992	H = 10.6exp(0.0169D)
9	W_S = 0.2286(D²H)^0.6938	W_B = 0.0247(D²H)^0.7378	W_L = 0.0108(D²H)^0.8181	Self-established	H = 9.285exp(0.0196D)
10	W_S = 0.3274(D²H)^0.7217	W_B = 0.01349(D²H)^0.7197	W_L = 0.02347(D²H)^0.893	Self-established	H = 7.0338exp(0.0332D)
11	W_S = 0.03146(D²H)^1.032	W_B = 0.007429D^2.6745	W_L = 0.002754D^2.4965	Chen & Zhu, 1989	H = 8.1427exp(0.0249D)
12	W_S = 0.07936(D²H)^0.901	W_B = 0.014167(D²H)^0.764	W_L = 0.01086(D²H)^0.847	Self-established	H = 13.799exp(0.0138D)
13	W_S = 0.14114(D²H)^0.723	W_B = 0.00724(D²H)^1.0225	W_L = 0.0079(D²H)^0.8085	Jiang, 1992	H = 6.2635exp(0.0334D)
14	W_S = 0.03141(D²H)^0.733	W_B = 0.002127D^2.9504	W_L = 0.00321D^{2.473 5}	Chen & Zhu, 1989	H = 8.1877exp(0.0219D)
15	W_S = 0.01275(D²H)^1.009	W_B = 0.00824(D²H)^0.975	W_L = 0.00024(D²H)^0.991	Chen & Zhu, 1989	H = 7.0889exp(0.0349D)
16	W_S = 0.1193(D²H)^0.8372	W_B = 0.002(D²H)^1.12	W_L = 0.000015(D²H)^1.47	Luo, 1996	H = 9.8065exp(0.0246D)
17	W_S = 0.2286(D²H)^0.6938	W_B = 0.0247(D²H)^0.7378	W_L = 0.0108(D²H)^0.8181	Chen & Zhu, 1989	H = 12.136exp(0.0133D)

树种¹⁾ Species¹⁾	茎生物量 Stem biomass (W_S)	枝生物量 Branch biomass (W_B)	叶生物量 Foliage biomass (W_L)	出处 Derivation	树高 Tree height (H)
1	W_S= 0.0145(D²H)^1.006	W_B = 0.000063(D²H)^1.536	W_L = 0.00104(D²H)^1.148	Ding & Sun, 1989	H = 6.6831exp(0.0515D)
2	W_S = 0.0134(D²H)^1.020	W_B = 0.0105(D²H) ^0.7386	W_L = 0.181D^1.8415	Luo, 1996	H = 5.7584exp(0.0499D)
3	W_S = 0.025(D²H)^0.96	W_B = 0.0021(D²H)^0.8181	W_L = 0.00126(D²H)^0.968	Chen & Zhu, 1989	H = 6.9441exp(0.0468D)
4	W_S = 0.057D^2.4753	W_B = 0.0116D^2.4054	W_L = 0.0083D^2.3733	Chen & Zhu, 1989	H = 6.8879exp(0.0343D)
5	W_S = 0.0238(D²H)^0.936	W_B = 0.005(D²H)^0.9105	W_L = 0.0036(D²H)^0.897	Chen & Zhu, 1989	H = 6.8879exp(0.0343D)
6	W_S = 0.032417D^2.3565	W_B = 0.20893D^1.7082	W_L = 0.29156D^1.2807	Self-established	H = 6.6831exp(0.0515D)
7	W_S = 0.06013(D²H)^0.891	W_B = 0.00652(D²H)^1.169	W_L = 0.0044(D²H)^0.9919	Jiang, 1992	H = 8.4834exp(0.0277D)
8	W_S = 0.02511(D²H)^0.927	W_B = 0.00957(D²H)^0.974	W_L = 0.8725(D²H)^0.2034	Jiang, 1992	H = 10.6exp(0.0169D)
9	W_S = 0.2286(D²H)^0.6938	W_B = 0.0247(D²H)^0.7378	W_L = 0.0108(D²H)^0.8181	Self-established	H = 9.285exp(0.0196D)
10	W_S = 0.3274(D²H)^0.7217	W_B = 0.01349(D²H)^0.7197	W_L = 0.02347(D²H)^0.893	Self-established	H = 7.0338exp(0.0332D)
11	W_S = 0.03146(D²H)^1.032	W_B = 0.007429D^2.6745	W_L = 0.002754D^2.4965	Chen & Zhu, 1989	H = 8.1427exp(0.0249D)
12	W_S = 0.07936(D²H)^0.901	W_B = 0.014167(D²H)^0.764	W_L = 0.01086(D²H)^0.847	Self-established	H = 13.799exp(0.0138D)
13	W_S = 0.14114(D²H)^0.723	W_B = 0.00724(D²H)^1.0225	W_L = 0.0079(D²H)^0.8085	Jiang, 1992	H = 6.2635exp(0.0334D)
14	W_S = 0.03141(D²H)^0.733	W_B = 0.002127D^2.9504	W_L = 0.00321D^{2.473 5}	Chen & Zhu, 1989	H = 8.1877exp(0.0219D)
15	W_S = 0.01275(D²H)^1.009	W_B = 0.00824(D²H)^0.975	W_L = 0.00024(D²H)^0.991	Chen & Zhu, 1989	H = 7.0889exp(0.0349D)
16	W_S = 0.1193(D²H)^0.8372	W_B = 0.002(D²H)^1.12	W_L = 0.000015(D²H)^1.47	Luo, 1996	H = 9.8065exp(0.0246D)
17	W_S = 0.2286(D²H)^0.6938	W_B = 0.0247(D²H)^0.7378	W_L = 0.0108(D²H)^0.8181	Chen & Zhu, 1989	H = 12.136exp(0.0133D)

变量 Variable		相关系数 Correlation coefficient	变量 Variable		相关系数 Correlation coefficient	变量 Variable		相关系数 Correlation coefficient
原始波段 Original band	Band1	-0.029	纹理信息 Texture		对角线方向 diagonal direction	相异性 Dissimilarity	Dis1	-0.206
	Band2	-0.508^**	平均值 Mean	Mean1	-0.354^**		Dis2	-0.342^**
	Band3	-0.450^**		Mean2	-0.225^*		Dis3	-0.061
	Band4	0.035		Mean3	-0.122		Dis4	-0.093
	Band5	-0.357^**		Mean4	0.027		Dis5	-0.301^**
	Band7	-0.459^**		Mean5	-0.383^**		Dis7	0.038
波段组合 Band combination	TM73	0.100		Mean7	-0.323^**	熵 Entropy	Ent1	-0.233^*
	TM437	-0.095	方差 Variance	Var1	-0.250^*		Ent2	-0.342^**
	TM452	0.511^**		Var2	-0.308^**		Ent3	-0.137
	TM42	0.591^**		Var3	-0.122		Ent4	-0.136
	差值植被指数 Difference vegetation index (DVI)	0.121		Var4	-0.178		Ent5	-0.336^**
	简单比值指数 Simple ratio (SR)	0.482^**		Var5	-0.305^**		Ent7	-0.017
	归一化植被指数 Normalized difference vegetation index (NDVI)	0.419^**		Var7	-0.062	角二阶矩 Angular second moment	Sec1	0.230^*
	转换型植被指数 Transformed vegetation index (TVI)	0.415^**	均一性 Homogeneity	Hom1	0.206		Sec2	0.340^**
	垂直植被指数 Perpendicular vegetation index (PVI)	0.071		Hom2	0.342^**		Sec3	0.137
	红外指数 Infrared index (II)	0.466^**		Hom3	0.061		Sec4	0.108
	土壤调整植被指数 Soil-adjusted vegetation index (SAVI)	0.418^**		Hom4	0.088		Sec5	0.329^**
	优化的简单比值指数 Modified simple ratio (MSR)	0.469^**		Hom5	0.301^**		Sec7	0.019
	土壤调整植被指数2 Soil-adjusted vegetation index 2 (SAVI2)	0.409^**		Hom7	-0.038	相关性 Correlation	Cor1	0.309^**
	非线性植被指数 Non-linear index (NLI)	0.330^**	对比度 Contrast	Con1	-0.206		Cor2	0.366^**
地学信息 Geographic information	纵坐标Y Y-coordinate	-0.259^*		Con2	-0.342^**		Cor3	0.131
	横坐标X X-coordinate	-0.231^*		Con3	-0.061		Cor4	0.143
	高程 Elevation	0.598^**		Con4	-0.113		Cor5	0.172
	坡向 Slope aspect	-0.280^**		Con5	-0.292^**		Cor7	0.095
	坡度 Slope	-0.125		Con7	0.038	郁闭度 Closure		0.283^**

变量 Variable		相关系数 Correlation coefficient	变量 Variable		相关系数 Correlation coefficient	变量 Variable		相关系数 Correlation coefficient
原始波段 Original band	Band1	-0.029	纹理信息 Texture		对角线方向 diagonal direction	相异性 Dissimilarity	Dis1	-0.206
	Band2	-0.508^**	平均值 Mean	Mean1	-0.354^**		Dis2	-0.342^**
	Band3	-0.450^**		Mean2	-0.225^*		Dis3	-0.061
	Band4	0.035		Mean3	-0.122		Dis4	-0.093
	Band5	-0.357^**		Mean4	0.027		Dis5	-0.301^**
	Band7	-0.459^**		Mean5	-0.383^**		Dis7	0.038
波段组合 Band combination	TM73	0.100		Mean7	-0.323^**	熵 Entropy	Ent1	-0.233^*
	TM437	-0.095	方差 Variance	Var1	-0.250^*		Ent2	-0.342^**
	TM452	0.511^**		Var2	-0.308^**		Ent3	-0.137
	TM42	0.591^**		Var3	-0.122		Ent4	-0.136
	差值植被指数 Difference vegetation index (DVI)	0.121		Var4	-0.178		Ent5	-0.336^**
	简单比值指数 Simple ratio (SR)	0.482^**		Var5	-0.305^**		Ent7	-0.017
	归一化植被指数 Normalized difference vegetation index (NDVI)	0.419^**		Var7	-0.062	角二阶矩 Angular second moment	Sec1	0.230^*
	转换型植被指数 Transformed vegetation index (TVI)	0.415^**	均一性 Homogeneity	Hom1	0.206		Sec2	0.340^**
	垂直植被指数 Perpendicular vegetation index (PVI)	0.071		Hom2	0.342^**		Sec3	0.137
	红外指数 Infrared index (II)	0.466^**		Hom3	0.061		Sec4	0.108
	土壤调整植被指数 Soil-adjusted vegetation index (SAVI)	0.418^**		Hom4	0.088		Sec5	0.329^**
	优化的简单比值指数 Modified simple ratio (MSR)	0.469^**		Hom5	0.301^**		Sec7	0.019
	土壤调整植被指数2 Soil-adjusted vegetation index 2 (SAVI2)	0.409^**		Hom7	-0.038	相关性 Correlation	Cor1	0.309^**
	非线性植被指数 Non-linear index (NLI)	0.330^**	对比度 Contrast	Con1	-0.206		Cor2	0.366^**
地学信息 Geographic information	纵坐标Y Y-coordinate	-0.259^*		Con2	-0.342^**		Cor3	0.131
	横坐标X X-coordinate	-0.231^*		Con3	-0.061		Cor4	0.143
	高程 Elevation	0.598^**		Con4	-0.113		Cor5	0.172
	坡向 Slope aspect	-0.280^**		Con5	-0.292^**		Cor7	0.095
	坡度 Slope	-0.125		Con7	0.038	郁闭度 Closure		0.283^**

变量 Variable	Coef_uns	Coef_s	t_0.05	Sig.	VIF
常数 Constant	125.696 8		6.211	0.000
郁闭度 Closure	18.692 6	0.123 7	2.402	0.018	1.099
高程 Elevation	0.024	0.134 3	2.265	0.026	1.456
X	-0.000 14	-0.266 3	-5.273	0.000	1.057
Band3	-0.057 5	-0.273 1	-4.258	0.000	1.704
Cor1	0.295 4	0.539 5	7.903	0.000	1.931