基于PROSAIL模型的山地草原叶面积指数高光谱反演

doi:10.6048/j.issn.1001-4330.2022.02.023

摘要/Abstract

摘要：

【目的】 研究基于PROSAIL模型监测天然草地的动态变化,掌握草地的质量与数量。【方法】 研究使用地物光谱仪连续3年在天山北坡中段的2个山地草原样区采集光谱数据和配套数据,基于PROSAIL模型进行冠层LAI的高光谱反演,重点研究应用不同代价函数、植被种类变化对反演精度的影响。【结果】 多数代价函数反演LAI的决定系数(R²)在0.54~0.55,均方根误差(RMSE)在0.23~0.25,归一化均方根误差(NRMSE)在17~19。在9个来自不同统计类型的代价函数中,常用的RMSE代价函数的反演精度相对不高。将获取的427个样方数据依据种类数分成组,然后用PROSAIL进行LAI反演。种类数越多,RMSE在增大,R²在减少,反演精度越差。但精度的下降幅度不是均匀的,种类数≤2的组和种类数≤3的组之间精度差异最大。【结论】 在利用物理模型反演天然草地的叶面积指数时,不同代价函数获得的反演精度差别比较大;随着植被种类数量的增多,反演的精度是下降的。

关键词: 叶面积指数; 高光谱; 草原; PROSAIL

Abstract:

【Objective】 Therefore, the monitoring of dynamic changes in natural grasslands is important for both ecology and agriculture. 【Method】 This study focused on the impact of changes in plant species on the inversion accuracy.Spectral data for two mountain steppe plots were collected using a spectrometer over three years and concurrently, supporting data were obtained. 【Result】 The results showed that when the physical model PROSAIL was used to invert the mountain steppe LAI, the inversion error was very large for a single solution.Adding random noise and averaging multiple solutions when solving significantly improved the LAI inversion accuracy.The coefficient of determination (R²) for the LAI inversion for most cost functions was between 0.54 and 0.55, root mean square error (RMSE) was between 0.23 and 0.25, and normalized root mean square error (NRMSE) was between 17 and 19.In nine cost functions from different statistical types, the inversion accuracy of the commonly used RMSE cost function was relatively low.The 427 samples obtained were divided into four groups according to the number of species.The inversion results indicated that a greater number of species corresponded to an increasing RMSE, decreasing R², and poorer LAI inversion accuracy, although the precision decrease was not uniform.Groups with up to two species and groups with up to three species had the greatest difference in inversion accuracy. 【Conclusion】 When using physical model to invert LAI of natural grassland, the precision of inversion obtained by different cost functions is different.As the number of vegetation species increases, the precision of inversion decreases.

Key words: leaf area index; hyperspectral; grassland; PROSAIL

中图分类号:

S812

贠静, 郑逢令, 安沙舟, 阿斯娅·曼力克, 李超, 艾尼玩·艾麦尔, 田聪. 基于PROSAIL模型的山地草原叶面积指数高光谱反演[J]. 新疆农业科学, 2022, 59(2): 451-457.

YUN Jing, ZHENG Fengling, AN Shazhou, Asiya Manlike, LI Chao, , TIAN Cong. Hyperspectral Inversion of Leaf Area Index in Mountain Steppe Ecosystems Based on the PROSAIL Model[J]. Xinjiang Agricultural Sciences, 2022, 59(2): 451-457.

图/表 6

参考文献 17

[1]	Darvishzadeh R, Atzberger C, Skidmore A K, et al. Leaf area index derivation from hyperspectral vegetation indices and the red edge position[J]. International Journal of Remote Sensing, 2009, 30(23):6199-6218. DOI URL
[2]	童庆禧, 张兵, 张立福. 中国高光谱遥感的前沿进展[J]. 遥感学报, 2016, 20(5):689-707.
	TONG Qingxi, ZHANG Bing, ZHANG Lifu. Current progress of hyperspectral remote sensing in China[J]. Journal of Remote Sensing, 2016, 20(5):689-707.
[3]	Verrelst J, Rivera G P, Leonenko G, et al. Optimizing LUT-based radiative transfer model inversion for retrieval of biophysical parameters using hyperspectral data[C]// Geoscience and Remote Sensing Symposium.IEEE, 2012: 7325-7328.
[4]	Masemola C, Cho M A, Ramoelo A. Comparison of Landsat 8 OLI and Landsat 7 ETM+ for estimating grassland LAI using model inversion and spectral indices:case study of Mpumalanga, South Africa[J]. International Journal of Remote Sensing, 2016, 37(18):4401-4419. DOI URL
[5]	SI Yali, Schlerf M, Zurita-Milla R, et al. Mapping spatio-temporal variation of grassland quantity and quality using MERIS data and the PROSAIL model[J]. Remote Sensing of Environment, 2012, 121:415-425. DOI URL
[6]	Atzberger C, Darvishzadeh R, Immitzer M, et al. Comparative analysis of different retrieval methods for mapping grassland leaf area index using airborne imaging spectroscopy[J]. International Journal of Applied Earth Observation and Geoinformation, 2015, 43:19-31. DOI URL
[7]	杨灿灿, 吴见, 王春, 等. 基于HJ-1B影像的内蒙古草地叶面积指数反演[J]. 测绘工程, 2015, 24(5):29-32.
	YANG Cancan, WU Jian, WANG Chun, et al. Research on leaf area index quantitative inversion of grassland in Inner Mongolia based on HJ-1B data[J]. Engineering of Surveying and Mapping, 2015, 24(5):29-32.
[8]	昌梦雨, 魏晓楠, 王秋悦, 等. 植物叶绿素含量不同提取方法的比较研究[J]. 中国农学通报, 2016, 32(27):177-180.
	CHANG Mengyu, WEI Xiaonan, WANG Qiuyue, et al. A comparative study on different extraction methods for plant chlorophyll[J]. Chinese Agricultural Science Bulletin, 2016, 32(27):177-180.
[9]	朱高龙. 基于LAI-2200的草地LAI测量与聚集度系数分析[J]. 农业机械学报, 2016, 47(5):307-314.
	ZHU Gaolong. Validation of grassland leaf area index and clumping index retrievals from LAI-2200[J]. Transactions of the Chinese Society of Agricultural Machinery, 2016, 47(5):307-314.
[10]	Jacquemoud S, Verhoef W, Baret F, et al. PROSPECT+SAIL models: A review of use for vegetation characterization[J]. Remote Sensing of Environment, 2009, 113(S1):S56-S66. DOI URL
[11]	Verrelst J, Romijn E, Kooistra L. Mapping vegetation density in a heterogeneous river floodplain ecosystem using pointable CHRIS/PROBA data[J]. Remote Sensing, 2012, 4(9):2866-2889. DOI URL
[12]	Leonenko G, Los S, North P. Statistical distances and their applications to biophysical parameter estimation: Information measures, M-estimates, and Minimum contrast methods[J]. Remote Sensing, 2013, 5(3):1355-1388. DOI URL
[13]	GU Chengyan, DU Huaqiang, MAO Fangjie, et al. Global sensitivity analysis of PROSAIL model parameters when simulating Moso bamboo forest canopy reflectance[J]. International Journal of Remote Sensing, 2016, 37(22):5270-5286.
[14]	Darvishzadeh R, Atzberger C, Skidmore A, et al. Mapping grassland leaf area index with airborne hyperspectral imagery: A comparison study of statistical approaches and inversion of radiative transfer models[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2011, 66(6):894-906. DOI URL
[15]	Verrelst J, Rivera J P, Leonenko G, et al. Optimizing LUT-based RTM inversion for semiautomatic mapping of crop biophysical parameters from Sentinel-2 and -3 data: Role of cost functions[J]. IEEE Transactions on Geoscience and Remote Sensing, 2014, 52(1):257-269. DOI URL
[16]	Rivera J P, Verrelst J, Leonenko G, et al. Multiple cost functions and regularization options for improved retrieval of leaf chlorophyll content and LAI through inversion of the PROSAIL model[J]. Remote Sensing, 2013, 5(7):3280-3304. DOI URL
[17]	Darvishzadeh R, Skidmore A, Schlerf M, et al. Inversion of a radiative transfer model for estimating vegetation LAI and chlorophyll in a heterogeneous grassland[J]. Remote Sensing of Environment, 2008, 112(5):2592-2604. DOI URL

模型参数 Model parameter		单位 Unit	范围 Range	步长 Step length	分布类型 Distribution type
叶片模型:Prospect-4
N	叶片结构指数	/	1.3~2	0.1	Uniform
LCC	叶片叶绿素含量	μg/cm²	20~50	/	Gaussian(x:33,SD:5)
C_m	叶片干物质含量	g/cm²	0.004 5~0.005 5	0.000 5	Uniform
C_w	等效水厚度	cm	0.007~0.014	0.001	Uniform
冠层模型:4SAIL
LAI	叶面积指数	m²/m²	0.1~1.80	/	Gaussian(x:0.84,SD:0.3)
α_soil	土壤比例因子	/	0.1~1	0.1	Uniform
ALA	平均叶倾角	^°	40~70	5	Uniform
HotS	热点参数	m/m	0.05~0.5	0.05	Uniform
skyl	散射光占比	%	5	固定值	/
θs	太阳天顶角	^°	20~40	5	/
θv	观测天顶角	^°	0	固定值	/
φ	太阳-仪器方位角	^°	0	固定值	/

模型参数 Model parameter		单位 Unit	范围 Range	步长 Step length	分布类型 Distribution type
叶片模型:Prospect-4
N	叶片结构指数	/	1.3~2	0.1	Uniform
LCC	叶片叶绿素含量	μg/cm²	20~50	/	Gaussian(x:33,SD:5)
C_m	叶片干物质含量	g/cm²	0.004 5~0.005 5	0.000 5	Uniform
C_w	等效水厚度	cm	0.007~0.014	0.001	Uniform
冠层模型:4SAIL
LAI	叶面积指数	m²/m²	0.1~1.80	/	Gaussian(x:0.84,SD:0.3)
α_soil	土壤比例因子	/	0.1~1	0.1	Uniform
ALA	平均叶倾角	^°	40~70	5	Uniform
HotS	热点参数	m/m	0.05~0.5	0.05	Uniform
skyl	散射光占比	%	5	固定值	/
θs	太阳天顶角	^°	20~40	5	/
θv	观测天顶角	^°	0	固定值	/
φ	太阳-仪器方位角	^°	0	固定值	/

查找表尺寸 Look up table size	代价函数 Cost function	最优解占比 Optimal solution ratio(%)	RMSE	NRMSE	R²
1 000	RMSE	3	0.26	18.87	0.50
10 000	RMSE	3	0.26	18.58	0.50
30 000	RMSE	3	0.25	18.45	0.50
50 000	RMSE	3	0.25	18.49	0.50

查找表尺寸 Look up table size	代价函数 Cost function	最优解占比 Optimal solution ratio(%)	RMSE	NRMSE	R²
1 000	RMSE	3	0.26	18.87	0.50
10 000	RMSE	3	0.26	18.58	0.50
30 000	RMSE	3	0.25	18.45	0.50
50 000	RMSE	3	0.25	18.49	0.50

代价函数 Cost function	噪声占比 Noise ratio(%)	样本占比 Sample to sample ratio(%)	RMSE	NRMSE	R²
K(x)={log(x)^2}	3	3	0.23	17.08	0.55
K(x)=x{log(x)}-x	2	2	0.24	17.35	0.55
Kullback-leibler	1	3	0.25	18.10	0.55
Pearson chi-square	1	1	0.25	18.13	0.55
Hellinger distance	4	4	0.25	18.17	0.54
Least absolute error	0	10	0.25	18.36	0.45
Geman and McClure	0	10	0.25	18.52	0.47
RMSE	1	3	0.26	18.99	0.54
K(x)=-log(x)+x	0	1	0.47	34.06	0.52