基于小波分析的大豆叶绿素a含量高光谱反演模型

doi:10.3773/j.issn.1005-264x.2008.01.017

摘要/Abstract

摘要：

2003和2004年分别在长春市良种场和中国科学院海伦黑土生态实验站实测了大田耕作与水肥耦合作用下大豆(Glycine max)冠层高光谱反射率与叶绿素a含量数据,对光谱反射率、微分光谱与叶绿素a含量进行了相关分析;采用归一化植被指数(Normalized difference vegetation index, NDVI)、土壤调和植被指数(Soil-adjusted vegetation index, SAVI)、再归一植被指数(Renormalized difference vegetation index, RDVI)、第二修正比值植被指数(Modified second ratio index, MSRI)等建立了大豆叶绿素a反演模型;应用小波分析对采集的光谱反射率数据进行了能量系数提取,并以小波能量系数作为自变量进行了单变量与多变量回归分析,对大豆叶绿素a进行了估算。研究结果表明,大豆叶绿素a与可见光光谱反射率相关性较好,并在红光波段取得最大值(R²>0.70),但在红边处,微分光谱与大豆叶绿素a的相关性较反射率好得多,在其它波段则相反;由NDVI、SAVI、RDVI、MSR等植被指数建立的估算模型可以提高大豆叶绿素a的估算精度(R²>0.75);小波能量系数回归模型可以进一步提高大豆叶绿素a含量的估算水平,以一个特定小波能量系数作为自变量的回归模型,大豆叶绿素a回归决定系数R²高达0.78;多变量回归分析结果表明,大豆叶绿素a实测值与预测值的线性回归决定系数R²均高达0.85。以上结果表明,小波分析可以对高光谱进行特征变量提取,并可在一定程度上提高大豆生理参数反演精度。

关键词: 大豆, 高光谱, 叶绿素a含量, 植被指数, 小波能量系数

Abstract:

Aims A growing number of studies have focused on evaluating spectral indices in terms of their sensitivity to vegetation biophysical parameters. We use a regression model, based on wavelet-transformed reflectance, and vegetation indices (VI) to estimate a wide range of soybean (Glycine max) canopy reflectances to study the sensitivity of wavelet-transformed reflectance and vegetation indices to soybean chlorophyll a concentration. We modify some VI to enhance their sensitivity to variations in chlorophyll a concentration.
Methods We collected soybean canopy hyperspectral reflectance and chlorophyll a concentration data in 2003 and 2004 at two sites in the black soil belt of China. We correlated reflectance, derivative reflectance and soybean chlorophyll a concentration and regressed vegetation indices (NDVI, SAVI, RDVI and MSRI) and soybean chlorophyll a concentration. We transformed soybean canopy reflectance with wavelet analysis and applied extracted wavelet energy coefficient in a regression model for estimation of chlorophyll a concentration.
Important findings Soybean canopy reflectance shows a negative correlation with chlorophyll a concentration in the visible spectral region, while it shows a positive correlation with soybean chlorophyll a concentration in the near-infrared region. Reflectance derivative has a strong relationship with soybean chlorophyll a concentration in the blue, green and red edge spectral region, with maximum correlation coefficient in the red-edge region. Four vegetation indices have strong correlations with soybean chlorophyll a concentration, with R² >0.75. The single variable regression model based upon wavelet-extracted reflectance energy can accurately estimate soybean chlorophyll a concentration, with R ² about 0.75, while R ² was 0.85 with the multivariate regression model. Our study indicated that wavelet analysis can be applied to in-situ collected hyperspectral data for soybean chlorophyll a concentration estimation with accurate prediction and in the future wavelet analysis methods should be applied to hyperspectral data for estimation of other vegetation biophysical and biochemical parameters.

Key words: soybean (Glycine max), hyperspectral, chlorophyll a concentration, vegetation index, wavelet energy coefficient

宋开山, 张柏, 王宗明, 刘殿伟, 刘焕军. 基于小波分析的大豆叶绿素a含量高光谱反演模型. 植物生态学报, 2008, 32(1): 152-160. DOI: 10.3773/j.issn.1005-264x.2008.01.017

SONG Kai-Shan, ZHANG Bai, WANG Zong-Ming, LIU Dian-Wei, LIU Huan-Jun. SOYBEAN CHLOROPHYLL A CONCENTRATION ESTIMATION MODELS BASED ON WAVELET-TRANSFORMED, IN SITU COLLECTED, CANOPY HYPERSPECTRAL DATA. Chinese Journal of Plant Ecology, 2008, 32(1): 152-160. DOI: 10.3773/j.issn.1005-264x.2008.01.017

导出引用管理器 EndNote|Ris|BibTeX

链接本文: https://www.plant-ecology.com/CN/10.3773/j.issn.1005-264x.2008.01.017

https://www.plant-ecology.com/CN/Y2008/V32/I1/152

图/表 9

表1 本研究所采用的植被指数的相关信息

Table 1 Vegetation indices evaluated in the present study

植被指数 Vegetation index	植被指数采用的公式及波段 Formula and bands applied	文献来源 References
归一化植被指数 Normalized difference vegetation index (NDVI)	(R₈₀₁-R₆₇₀)/(R₈₀₁+R₆₇₀)	Rouse et al., 1974
土壤调和植被指数 Soil-adjusted vegetation index (SAVI)	(1+L)×(R₈₀₁-R₆₇₀)/(R₈₀₁+R₆₇₀+L)	Huete, 1988
再归一植被指数 Renormalized difference vegetation index (RDVI)	RDVI=(R₈₀₁-R₆₇₀)/ $(R 801 + R 670)$	Roujean & Breon, 1995
第二修正比值植被指数 Modified second ratio index (MSR)	MSRI=( $R 801 R 670$ -1)/ $(R 801 R 670 + 1)$	Chen & Cihlar,1996

表1 本研究所采用的植被指数的相关信息

Table 1 Vegetation indices evaluated in the present study

植被指数 Vegetation index	植被指数采用的公式及波段 Formula and bands applied	文献来源 References
归一化植被指数 Normalized difference vegetation index (NDVI)	(R₈₀₁-R₆₇₀)/(R₈₀₁+R₆₇₀)	Rouse et al., 1974
土壤调和植被指数 Soil-adjusted vegetation index (SAVI)	(1+L)×(R₈₀₁-R₆₇₀)/(R₈₀₁+R₆₇₀+L)	Huete, 1988
再归一植被指数 Renormalized difference vegetation index (RDVI)	RDVI=(R₈₀₁-R₆₇₀)/ $(R 801 + R 670)$	Roujean & Breon, 1995
第二修正比值植被指数 Modified second ratio index (MSR)	MSRI=( $R 801 R 670$ -1)/ $(R 801 R 670 + 1)$	Chen & Cihlar,1996

图1 光谱及其导数与叶绿素a逐波段回归的决定系数(R2)变化趋势

Fig.1 Coefficient of determination (R2) derived from a linear regression of reflectance and derivative against chlorophyll a concentration

图2 所有NDVI、SAVI波段组合与叶绿素a的回归决定系数的变化趋势 NDVI、SAVI: 见表1

Fig.2 Coefficient of determination for all combinations of wavelength used for linear regression analysis of NDVI and SAVI against chlorophyll a concentration,respectively NDVI、SAVI: See Table 1

图3 所有MSRI、RDVI波段组合与叶绿素a的回归决定系数的变化趋势 MSRI、RDVI: 见表1

Fig.3 Coefficient of determination for all combinations of wavelength used for linear regression analysis of MSRI and RDVI against chlorophyll a concentration,respectively MSRI、RDVI: See Table 1

表2 各种植被指数的最佳波段组合与叶绿素a的线性与非线性回归结果

Table 2 Linear and non-linear regression models based on optimum spectral vegetation indices against chlorophyll a concentration

植被指数 Vegetation indices	线性估算模型 Linear regression model (n=100)					模型验证 Model validation (n=44)
植被指数 Vegetation indices	回归公式 Regression formula	R²		RMSE (mg·g^-1)		R²	RMSE (mg·g^-1)
NDVI (441,223)	y=5.317x-3.223		0.782		0.168	0.754	0.173
SAVI (502,32)	y=-20.621x+1.644		0.762		0.171	0.734	0.181
RDVI (488,28)	y=-17.073 0x+1.757 8		0.759		0.172	0.722	0.189
MSRI (257,128)	y=10.358 0x-9.512 9		0.769		0.170	0.731	0.182
植被指数 Vegetation indices	非线性估算模型 Non-linear regression model (n=100)					模型验证 Model validation (n=44)
植被指数 Vegetation indices	回归公式 Regression formula	R²		RMSE (mg·g^-1)		R²	RMSE (mg·g^-1)
NDVI (441,223)	y=0.019e^4.722^x		0.793		0.164	0.726	0.186
SAVI (502,32)	y=2.077e^-18.03^x		0.798		0.162	0.732	0.182
RDVI (488,28)	y=1.851e^-14.91^x		0.778		0.171	0.717	0.193
MSRI (257,128)	y=0.880x^5.81		0.805		0.158	0.747	0.176

表3 各级小波能量系数与叶绿素a的回归模型及验证

Table 3 Regression models based on different levels of wavelet energy coefficients of soybean spectra and their validation

小波能量系数 Wavelet energy coefficient	估算模型 Regression model (n=100)					模型验证 Model validation (n=44)
小波能量系数 Wavelet energy coefficient	回归公式 Regression formula	R²		RMSE (mg·g^-1)		R²	RMSE (mg·g^-1)
第一系数 First coef.	y=0.378 2x-0.058 0		0.656		0.203	0.648	0.211
第二系数 Second coef.	y=1.191 2x+0.455 0		0.674		0.192	0.653	0.203
第三系数 Third coef.	y=2.871 3x+0.251 9		0.76		0.170	0.755	0.173
第四系数 Fourth coef.	y=4.915 6x+0.203 0		0.708		0.188	0.699	0.191
第五系数 Fifth coef.	y=23.110 0x-0.302 8		0.491		0.248	0.479	0.257
第六系数Sixth coef.	y=80.002 0x+0.107 5		0.784		0.162	0.778	0.167
第七系数 Seventh coef.	y=333.510 0x+0.055 7		0.658		0.203	0.649	0.208
第八系数Eighth coef.	y=544.510 0x+0.621 7		0.230		0.305	0.210	0.312
第九系数 Ninth coef.	y=483.310 0x+0.993 0		0.029		0.343	0.007	0.348

表4 不同小波函数获得的小波能量系数与叶绿素a的回归模型及验证

Table 4 Regression models based on wavelet energy coefficients from various mother functions and their validation

小波母函数 Mother wavelet functions	估算模型 Regression model (n=100)					模型验证 Model validation (n=44)
小波母函数 Mother wavelet functions	回归公式 Regression formula	R²		RMSE (mg·g^-1)		R²	RMSE (mg·g^-1)
db2	y=71.020 0x+0.115 0		0.780		0.162	0.759	0.171
db4	y=6.626 4x+0.471 5		0.732		0.180	0.724	0.221
db6	y=23.437 0x+0.101 5		0.762		0.169	0.757	0.172
db8	y=16.055 0x-0.213 8		0.763		0.168	0.759	0.171
bior33	y=5.864 1x+0.282 1		0.755		0.172	0.750	0.173
bior68	y=19.846 0x+0.651 2		0.749		0.174	0.743	0.176
rbio33	y=52.416 0x+0.079 0		0.807		0.157	0.790	0.160
ciof5	y=8.524 6x+0.224 8		0.751		0.173	0.742	0.178
dmey	y=19.333 0x+0.268 2		0.757		0.172	0.748	0.175
sym8	y=18.266 0x+0.609 8		0.755		0.173	0.747	0.175

表5 基于小波分析的多元回归方程的决定系数(R2)与剩余残差(RMSE)的变化趋势

Table 5 The trend of determination coefficient (R2) and root mean square error (RMSE) from multivariate regression model based on soybean canopy reflectance wavelet transforms energy and chlorophyll a concentration

小波母函数 Mother wavelet functions	估算模型 Regression model (n=100)					模型验证 Model validation (n=44)
小波母函数 Mother wavelet functions	回归公式 Regression formula	R²		RMSE (mg·g^-1)		R²	RMSE (mg·g^-1)
db2	y=0.292+164.4x₆-0.296x₁-184.5x₇+4.46x₅		0.831		0.142	0.825	0.145
db4	y=0.692+34.72x₆-0.77x₁+8.08x₃		0.803		0.152	0.810	0.150
db6	y=0.459+10.525x₅-0.399x₁+7.41x₃-2.22x₂		0.814		0.148	0.807	0.151
db8	y=0.342+27.72x₄-19.1x₅-0.369x₁+127.50x₈		0.817		0.147	0.819	0.146
bior33	y=0.51+10.66x₄-0.231x₁+76.66x₇		0.800		0.153	0.790	0.156
bior68	y=0.432+26.72x₅-89.19x₇+8.89x₃-5.64x₂		0.843		0.132	0.854	0.139
rbior33	y=0.045+103.76x₆-173.59x₇-3.10x₃		0.855		0.130	0.859	0.129
ciof5	y=0.377+54.97x₄+0.188x₁-6.73x₂-11.70x₃-31.21x₆		0.828		0.142	0.822	0.144
dmey	y=0.499+41.14x₄-0.234x₁-77.90x₇-6.61x₂+8.61x₃		0.818		0.146	0.810	0.150
sym8	y=18.266x₅+0.610		0.755		0.173	0.747	0.175

图4 基于小波变换特征量的多元回归模型预测叶绿素a与实测值之间的关系

Fig.4 Relationship between multivariate regression based on wavelet-transformed soybean canopy spectral reflectance predicted chlorophyll a and measured chlorophyll a concentration

参考文献 21

[1]	Adams ML, Norvell WA, Peverly JH, Philpot WD (1993). Fluorescence and reflectance characteristics of Manganese deficient soybean leaves:effects of leaf age and choice of leaflet. Plant and Soil, 155/156, 235-238.
[2]	Bannari A, Morin D, Bonn F, Huete AR (1995). A review of vegetation indices. Remote Sensing Review, 13, 95-120.
[3]	Blackburn GA (1998). Spectral indices for estimating photosynthetic pigment concentrations:a test using senescent tree leaves. International Journal of Remote Sensing, 19, 657-675.
[4]	Bruce LM, Li J (2001). Wavelet for computationally efficient hyperspectral derivative analysis. IEEE Transactions on Geosciences and Remote Sensing, 39, 1540-1546.
[5]	Bruce LM, Koger CH, Li J (2002). Dimensionality reduction of hyperspectral data using discrete transform feature extraction. IEEE Transactions on Geosciences and Remote Sensing, 40, 2331-2338.
[6]	Chen J, Cihlar J (1996). Retrieving leaf area index of boreal conifer forests using Landsat TM images. Remote Sensing of Environment, 55, 153-162.
[7]	Gupta RK, Woolley JT (1971). Spectral properties of soybean leaves. Agronomy Journal, 63, 123-126.
[8]	Huang WJ (黄文江), Wang JH (王纪华), Liu LY (刘良云), Zhao CJ (赵春江), Song XY (宋晓宇), Ma ZH (马智宏) (2004). Correlation between grain quality indicators and spectral reflectance properties of wheat canopies by using hyperspectral data from winter wheat. Transactions of the Chinese Society of Agricultural Engineering (农业工程学报), 20, 203-207. (in Chinese with English abstract)
[9]	Huete AR (1988). A soil vegetation adjusted index (SAVI). Remote Sensing of Environment, 25, 295-309.
[10]	Jacquemoud S, Bacour C, Poilve H, Frangi JP (2000). Comparison of four radiative transfer models to simulate plant canopies reflectance:direct and inverse mode. Remote Sensing of Environment, 74, 417-481.
[11]	Lichtenthaler HK (1998). The stress concept in plants:an introduction. Annals of the New York Academy of Sciences, 851, 187-198.
[12]	Milton NM, Ager CM, Eiswerth BA, Power MS (1989). Arsenic- and Selenium-induced changes in spectral reflectance and morphology of soybean plants. Remote Sensing of Environment, 30, 263-269.
[13]	Myneni RB, Hall FG, Sellers PJ (1995). The interpretation of special vegetation indexes. IEEE Transactions on Geosciences and Remote Sensing, 33, 481-486.
[14]	Pu RL, Gong P (2004). Wavelet transform applied to EO-1 hyperspectral data for forest LAI and crown closure mapping. Remote Sensing of Environment, 91, 212-224.
[15]	Roujean JL, Breon FM (1995). Estimating PAR absorbed by vegetation from bidirectional reflectance measurements. Remote Sensing of Environment, 51, 375-384.
[16]	Rouse JW, Haas RH, Schell JA, Deering DW (1974). Monitoring the Vernal Advancements and Retrogradation of Natural Vegetation. NASA/GSFC, Final Report, Greenbelt, MD, USA, 1-137.
[17]	Song KS (宋开山), Zhang B (张柏), Li F (李方), Duan HT (段洪涛), Wang ZM (王宗明) (2005). Correlative analyses of hyperspectral reflectance, soybean LAI and aboveground biomass. Transactions of the Chinese Society of Agricultural Engineering (农业工程学报), 21, 36-40. (in Chinese with English abstract)
[18]	Wang D, Wilson C, Shannon M (2002). Interpretation of salinity and irrigation effects on soybean canopy reflectance in visible and near-infrared spectrum domain. International Journal of Remote Sensing, 23, 811-824.
[19]	Wang XZ (王秀珍), Huang JF (黄敬峰), Li YM (李云梅), Shen ZQ (沈掌泉), Wang RC (王人潮) (2002). Relationships between rice agricultural parameter and hyperspectral data. Journal of Zhejiang University (Agriculture & Life Sciences Edition) (浙江大学学报 (农业与生命科学版)), 28, 283-288. (in Chinese with English abstract)
[20]	Zhang XY (张兴义), Meng K (孟凯), Sui YY (隋跃宇) (1999). Change of stomatal resistance in soybean canopy at different nutritional level. System Sceinces and Comprehensive Studies in Agriculture (农业系统科学与综合研究), 15, 302-305. (in Chinese with English abstract)
[21]	Zhao CJ (赵春江), Huang WJ (黄文江), Wang JH (王纪华), Yang MH (杨敏华), Xue XZ (薛绪掌) (2002). Studies on the red edge parameters of spectrum in winter wheat under different varieties, fertilizer and water treatments. Scientia Agricultura Sinica (中国农业科学), 35, 980-987. (in Chinese with English abstract)