Modeling and Verification of Soil Salt Content Based on Hyperspectral Characteristic Parameter Optimization

doi:10.6048/j.issn.1001-4330.2021.12.022

Abstract

Abstract:

【Objective】 To explore the feasibility of soil salinity inversion model under different dimensional spectral transformations.【Methods】 The lakeside oasis on the west of Bosten Lake in Xinjiang was taken as the research area, the correlation analysis between 17 one-dimensional mathematical transformation spectra and 3 two-dimensional transformation spectral indexes of ASD hyperspectral data and the measured soil salinity were conducted to obtain the preliminarily optimized spectral characteristic parameters at the significance test level of 0.01. Then, the PLSR model was constructed based on the VIP criteria and selected into the optimal independent variable and the accuracy was verified.【Results】 The average reflectance of dry soil was higher than that of wet season with the increase of salt content, especially at 590, 800, 1,810 nm and 2,150 nm. Among the 17 one-dimensional single-band spectral transformations, the first derivative of logarithmic reciprocal (1/lgR) had the best correlation with soil salinity, the peak sensitive band was 1083nm, and the absolute value of correlation coefficient was up to 0.63. Among the three two-dimensional two-band spectral transforms, the normalized spectral index NDSI(R_{1 780}, R_{1 742}) had the best correlation with soil salt content, and the maximum value of correlation analysis determination coefficient R ² was 0.57. The PLSR estimation model based on characteristic normalized spectral index and VIP criterion for independent variable screening had the best effect. The determination coefficient $R V 2$ of soil salt modeling set and verification set was 0.77, the root mean square error was 0.64 g/kg, and the relative analysis error was 2.11.【Conclusion】 Using the normalized spectral index (NDSI) to establish PLSR hyperspectral model could effectively estimate the soil salinity in the study area.

Key words: soil salinity; spectral transformation; spectral index; characteristic parameters; partial least squares regression

摘要： 目的研究不同维度光谱变换下土壤盐分反演模型及其验证。方法以博斯腾湖西岸湖滨绿洲为研究区,面向ASD高光谱数据,利用17种一维数学变换光谱和3种二维变换光谱指数,分别与实测土壤盐分进行相关分析,得到0.01显著性检验水平下初步优选的光谱特征参数,基于VIP准则选入最佳自变量实现PLSR模型构建,进行精度验证。结果研究区干季土壤平均反射率随含盐量的增加而高于湿季土壤平均反射率,尤其体现在590、800、1 810、2 150 nm处;17种一维单波段光谱变换中,对数倒数的一阶微分(1/lgR)变换与土壤盐分含量相关性最好,峰值敏感波段为1 083 nm,相关系数绝对值|r|最高达0.63;3种二维两波段光谱变换中,归一化光谱指数NDSI(R_{1 780},R_{1 742})与土壤盐分含量相关性最好,相关分析决定系数R ²最大值为0.57;基于特征归一化光谱指数结合VIP准则进行自变量筛选的PLSR估算模型效果最佳,土壤盐分建模集和验证集的决定系数 $R V 2$ 达0.77,均方根误差RMSEV为0.64 g/kg,相对分析误差RPD为2.11。结论利用归一化光谱指数NDSI建立PLSR高光谱模型可有效地对研究区土壤盐分进行定量估算。

关键词: 土壤盐分, 光谱变换, 特征参数, 偏最小二乘回归

CLC Number:

LI Zhi, SU Wuzheng, LI Xinguo, WANG Yinfang, MAO Donglei, Mamattursun Eziz. Modeling and Verification of Soil Salt Content Based on Hyperspectral Characteristic Parameter Optimization[J]. Xinjiang Agricultural Sciences, 2021, 58(12): 2342-2352.

李志, 苏武峥, 李新国, 王银方, 毛东雷, 麦麦提吐尔逊·艾则孜. 基于高光谱特征参数优选的土壤盐分含量建模及其验证[J]. 新疆农业科学, 2021, 58(12): 2342-2352.

Figures/Tables 14

Fig.1 Location of the sampling area and its sample layout

Fig. 2 Statistics of salt content frequency distribution in model set and validation set

Table 1 Definition of parameters

参数Parameter	定义Definition
R_λ	在波段λ处的土壤反射光谱 Soil reflectance spectrum at bandλ
DSI(i,j)	DSI=R_j-R_i
RSI(i,j)	RSI=R_j/R_i
NDSI(i,j)	NDSI=(R_j-R_i)/(R_j+R_i)

Table 2 Parameters of model accuracy

参数 Parameter	简称 Abbreviation	公式 Formula		描述 Description
建模决定系数 Modeling determination factor	$R C 2$	$\sum^{N}_{i=1}(x_{i}-\bar{x})(y_{i}-\bar{y})/ \sqrt{\sum^{N}_{i=1}(x_{i}-\bar{x})^{2}+\sum^{N}_{i=1}(y_{i}-\bar{y})^{2}}$		当值越接近1,模型越稳定
预测决定系数 Predictive determination coefficient	$R V 2$			当值越接近1,模型越稳定
校正均方根误差 Correct the root mean square error	RMSEC	$\sqrt{\sum^{N}_{i=1}(y_{i}-\bar{x})^{2}/N}$		当值越接近0,模型精度越高
验证均方根误差 Verify the root mean square error	RMSEV	$\sqrt{\sum^{N}_{i=1}(y_{i}-\bar{x})^{2}/N}$		当值越接近0,模型精度越高
相对分析误差 Relative prediction deviation	RPD	$S SD / r RMSEV$	0~1.4,差;1.4~2.0,可用;大于2.0,好

Table 2 Parameters of model accuracy

参数 Parameter	简称 Abbreviation	公式 Formula		描述 Description
建模决定系数 Modeling determination factor	$R C 2$	$\sum^{N}_{i=1}(x_{i}-\bar{x})(y_{i}-\bar{y})/ \sqrt{\sum^{N}_{i=1}(x_{i}-\bar{x})^{2}+\sum^{N}_{i=1}(y_{i}-\bar{y})^{2}}$		当值越接近1,模型越稳定
预测决定系数 Predictive determination coefficient	$R V 2$			当值越接近1,模型越稳定
校正均方根误差 Correct the root mean square error	RMSEC	$\sqrt{\sum^{N}_{i=1}(y_{i}-\bar{x})^{2}/N}$		当值越接近0,模型精度越高
验证均方根误差 Verify the root mean square error	RMSEV	$\sqrt{\sum^{N}_{i=1}(y_{i}-\bar{x})^{2}/N}$		当值越接近0,模型精度越高
相对分析误差 Relative prediction deviation	RPD	$S SD / r RMSEV$	0~1.4,差;1.4~2.0,可用;大于2.0,好

Fig.3 Distribution and variation of salinity in soil profile

Fig.4 Characteristics of spectral reflectance curves of soil profiles

Fig.5 One-dimensional correlation analysis of soil salinity content and spectral reflectance of different transformed forms

Table 2 Statistical analysis of characteristic bands of soil salinity under different spectral transformation forms

光谱变换 Spectrum transform	极显著相关系数绝对值超出的数量 The number of extremely significant correlation coefficients exceeding the absolute value (\|r\|≥0.45,P<0.01)	相关系数绝对值的最值 (\|r\|_max)	相关系数绝对值((\|r\|)的最值波段位置 The band position of\|r\|_max
R	0	0.27	1 406 nm
R'	46	0.62	1 230 nm
( $R$ )'	35	0.60	1 230 nm
(1/lgR)'	53	0.63	1 083 nm
lg(1/R)'	18	0.58	2 307 nm

Table 2 Statistical analysis of characteristic bands of soil salinity under different spectral transformation forms

光谱变换 Spectrum transform	极显著相关系数绝对值超出的数量 The number of extremely significant correlation coefficients exceeding the absolute value (\|r\|≥0.45,P<0.01)	相关系数绝对值的最值 (\|r\|_max)	相关系数绝对值((\|r\|)的最值波段位置 The band position of\|r\|_max
R	0	0.27	1 406 nm
R'	46	0.62	1 230 nm
( $R$ )'	35	0.60	1 230 nm
(1/lgR)'	53	0.63	1 083 nm
lg(1/R)'	18	0.58	2 307 nm

Fig. 6 R2 matrix of spectral index DSI, RSI, NDSI and soil salinity

Table 3 Soil salt characteristic band combination based on spectral index

光谱指数 Spectral indices	R²超出的光谱指数数量(R²≥0.4,P<0.01) The number of spectral indices R² exceeds 0.4	R²最大值 $R max 2$	R²最大值的对应波段组合 The band position of $R max 2$
差值指数(DSI) Differential spectral index	32	0.54	R_{1 780}-R_{1 765}
比值指数(RSI) The ratio of spectral index	25	0.48	R_{1 785}/R_{1 756}
归一化指数(NDSI) Normalized spectral index	38	0.57	(R_{1 780}-R_{1 742})/(R_{1 780}+R_{1 742})

Table 3 Soil salt characteristic band combination based on spectral index

光谱指数 Spectral indices	R²超出的光谱指数数量(R²≥0.4,P<0.01) The number of spectral indices R² exceeds 0.4	R²最大值 $R max 2$	R²最大值的对应波段组合 The band position of $R max 2$
差值指数(DSI) Differential spectral index	32	0.54	R_{1 780}-R_{1 765}
比值指数(RSI) The ratio of spectral index	25	0.48	R_{1 785}/R_{1 756}
归一化指数(NDSI) Normalized spectral index	38	0.57	(R_{1 780}-R_{1 742})/(R_{1 780}+R_{1 742})

Fig.7 Screening the best model spectral band based on the VIP algorithm

Fig. 8 Screening the best model spectral index based on VIP algorithm

Table 4 Prediction results of soil salt content by PLSR model

光谱特征参数 Spectral characteristic parameter	自变量个数 Number of independent variables	建模集Modeling set
光谱特征参数 Spectral characteristic parameter	自变量个数 Number of independent variables	$R C 2$	RMSEC	$R V 2$	RMSEV	RPD
R'	15	0.78	0.74	0.73	0.79	1.80
( $R$ )'	8	0.74	0.79	0.66	0.82	1.68
(1/lgR)'	16	0.82	0.61	0.77	0.69	2.01
[lg(1/R)]'	7	0.62	0.83	0.57	0.92	1.51
DSI	12	0.81	0.51	0.72	0.73	1.88
RSI	8	0.74	0.75	0.63	0.95	1.62
NDSI	12	0.86	0.46	0.77	0.64	2.11

Table 4 Prediction results of soil salt content by PLSR model

光谱特征参数 Spectral characteristic parameter	自变量个数 Number of independent variables	建模集Modeling set
光谱特征参数 Spectral characteristic parameter	自变量个数 Number of independent variables	$R C 2$	RMSEC	$R V 2$	RMSEV	RPD
R'	15	0.78	0.74	0.73	0.79	1.80
( $R$ )'	8	0.74	0.79	0.66	0.82	1.68
(1/lgR)'	16	0.82	0.61	0.77	0.69	2.01
[lg(1/R)]'	7	0.62	0.83	0.57	0.92	1.51
DSI	12	0.81	0.51	0.72	0.73	1.88
RSI	8	0.74	0.75	0.63	0.95	1.62
NDSI	12	0.86	0.46	0.77	0.64	2.11

Fig. 9 Verification results based on one-dimensional and two-dimensional optimal response parameterspectral estimation models

References 25

[1]	祁亚琴, 张显峰, 张立福, 等. 基于高光谱数据的农田土壤养分含量估测模型研究[J]. 新疆农业科学, 2018,55(3):490-495.
	QI Yaqin, ZHANG Xianfeng, ZHANG Lifu, et al. Research on soil nutrient content estimated model by hyperspectral remote sensing data[J]. Xinjiang Agricultural Sciences, 2018,55(3):490-495.
[2]	郭全恩, 王益权, 马忠明, 等. 植被类型对土壤剖面盐分离子迁移与累积的影像[J]. 中国农业科学, 2011,44(13):2711-2720.
	GUO Quanen, WANG Yiquan, MA Zhongming, et al. Effect of vegetation types on soil salt ions transfer and accumulation in soil profile[J]. Scientia Agricultura Sinica, 2011,44(13):2711-2720.
[3]	徐明星. 江苏沿海滩涂地区典型剖面土壤性质演化及其高光谱响应研究[D]. 南京:南京大学, 2011.
	XU Mingxing. Evolution of soil basic properties and its hyperspectral response in typical profiles of coastal tidal region of Jiangsu province, China[D]. Nanjing: Nanjing University, 2011.
[4]	童庆禧, 张兵, 郑兰芬. 高光谱遥感—原理、技术、应用[M]. 北京: 高等教育出版社, 2006.
	TONG Qingxi, ZHANG Bing, ZHENG Lanfen. Hyperspectral Remote Sensing-Principles, Techniques, Applications [M]. Beijing: Higher Education Press, 2006.
[5]	翁永玲, 宫鹏. 土壤盐渍化遥感应用研究进展[J]. 地理科学, 2006,26(3):369-375.
	WENG Yongling, GONG Peng. A review on remote sensing technique for salt-affected soils[J]. Scientia Geographica Sinica, 2006,26(3):369-375.
[6]	洪永胜, 朱亚星, 苏学平, 等. 高光谱技术联合归一化光谱指数估算土壤有机质含量[J]. 光谱学与光谱分析, 2017,37(11):3537-3542.
	HONG Yongsheng, ZHUYaxing, SU Xueping, et al. Estimation of soil organic matter content using hyperspectral techniques combined with normalized difference spectral index[J]. Spectroscopy and Spectral Analysis, 2017,37(11):3537-3542.
[7]	李焱, 王让会, 管延龙, 等. 基于高光谱反射特性的土壤全氮含量预测分析[J]. 遥感技术与应用, 2017,32(1):173-179.
	LI Yan, WANG Ranghui, GUAN Yanlong, et al. Prediction analysis of soil total nitrogen content based on hyperspectral[J]. Remote Sensing Technology and Application, 2017,32(1):173-179.
[8]	Chernousenko G I, Kalinina N V, Khitrov N B, et al. Quantification of the areas of saline and solonetzic soils in the Ural Federal Region of the Russian federation[J]. Eurasian Soil Science, 2011,44(4):367-379.
[9]	Mashimbye Z E, Cho M A, Nell J P, et al. Model-based integrated methods for quantitative estimation of soil salinity from hyperspectral remote sensing data: A case studyof selected South African soils[J]. Pedosphere, 2012,22(5):640-649.
[10]	Kumar S, Gautam G, Saha S K. Hyperspectral remote sensing data derived spectral indicesin characterizing salt-affected soils: a case study of Indo-Gangetic plains of India[J]. Environmental Earth Sciences, 2015,73(7):3299-3308.
[11]	彭杰, 迟春明, 向红英, 等. 基于连续统去除法的土壤盐分含量反演研究[J]. 土壤学报, 2014,51(3):459-469.
	PENG Jie, CHI Chunming, XIANG Hongying, et al. Inversion of soil salt content based on continuum-removal method[J]. Acta Pedologica Sinica, 2014,51(3):459-469.
[12]	张贤龙, 张飞, 张海威, 等. 基于光谱变换的高光谱指数土壤盐分反演模型优选[J]. 农业工程学报, 2018,34(1):110-117.
	ZHANG Xianlong, ZHANG Fei, ZHANG Haiwei, et al. Optimization of soil salt inversion model based on spectral transformation from hyperspectral index[J]. Transactions of the Chinese Society of Agricultural Engineering, 2018,34(1):110-117.
[13]	王宁, 熊黑钢, 马利芳, 等. 新疆有无人为干扰下土壤盐分估算的比较[J]. 干旱区研究, 2019,36(2):323-330.
	WANG Ning, XIONG Heigang, MA Lifang, et al. Estimation and comparison of soil salinityunder different intensities of human disturbance in Xinjiang[J]. Arid Zone Research, 2019,36(2):323-330.
[14]	翁永玲, 戚浩平, 方洪宾, 等. 基于PLSR方法的青海茶卡-共和盆地土壤盐分高光谱遥感反演[J]. 土壤学报, 2010,47(6):1255-1263.
	WENG Yongling, QI Haoping, FANG Hongbin, et al. PLSR-based hyperspectral remote sensing retrieval of soil salinity of Chaka-Gonghe Basin in Qinghai province[J]. Acta Pedologica Sinica, 2010,47(6):1255-1263.
[15]	蒲智, 于瑞德, 尹昌应, 等. 干旱区典型盐碱土壤含盐量估算的最佳高光谱指数研究[J]. 水土保持通报, 2012,32(6):129-133.
	PU Zhi, YU Ruide, YIN Changying, et al. Optimal hyperspectral indices for soil salt content estimation on typical saline soil in arid areas[J]. Bulletin of Soil and Water Conservation, 2012,32(6):129-133.
[16]	朱赟, 申广荣, 项巧巧, 等. 基于不同光谱变换的土壤盐含量光谱特征分析[J]. 土壤通报, 2017,48(3):560-568.
	ZHU Yun, SHEN Guangrong, XIANG Qiaoqiao, et al. Spectral characteristics of soil salinity based on different pre-processing methods[J]. Chinese Journal of Soil Science, 2017,48(3):560-568.
[17]	李新国, 李和平, 任云霞, 等. 开都河流域下游绿洲土壤盐渍化特征及其光谱分析[J]. 土壤通报, 2012,43(1):166-170.
	LI Xinguo, LI Heping, REN Yunxia, et al. Analysis on the characteristics of the oasis soil salinization and soil spectrum in the lower reaches of Kaidu River Basin, Xinjiang[J]. Chinese Journal of Soil Science, 2012,43(1):166-170.
[18]	李志, 李新国, 毛东雷, 等. 博斯腾湖西岸湖滨带不同植被类型土壤剖面盐分特征分析[J]. 西北农业学报, 2018,27(2):260-268.
	LI Zhi, LI Xinguo, MAO Donglei, et al. Analysis of salinity characteristics of different vegetation types in soil profile in westernside of Bosten Lake of Xinjiang[J]. Acta Agriculturae Boreali-Occidentalis Sinica, 2018,27(2):260-268.
[19]	李和平, 樊自立, 程心俊, 等. 采用土地资源利用限制因素指标进行土壤基层分类(以新疆干旱土纲基层分类为例)[J]. 干旱区研究, 2000,17(2):28-33.
	LI Heping, FAN Zili, CHENG Xinjun, et al. Study on basic soil classification form limited factor index of land resources development—an example from basic categories of arid sols order[J]. Arid Zone Research, 2000,17(2):28-33.
[20]	林鹏达, 佟志军, 张继权, 等. 基于CWT的黑土有机质含量野外高光谱反演模型[J]. 水土保持研究, 2018,25(2):46-52,57.
	LIN Pengda, TONG Zhijun, ZHANG Jiquan, et al. Inversion of black soil organic matter content with field hyperspectral reflectance based on continuous wavelet transformation[J]. Soil and Water Conservation Research, 2018,25(2):46-52,57.
[21]	鲍士旦. 土壤农化分析[M]. 北京: 中国农业出版社, 2000.
	BAO Shidan. Analysis of Soil Agrochemicals [M]. Beijing: China Agriculture Press, 2000.
[22]	陈弈云, 赵瑞瑛, 齐天赐, 等. 结合光谱变换和Kennard-Stone算法的水稻土全氮光谱估算模型校正集构建策略研究[J]. 光谱学与光谱分析, 2017,37(7):2133-2139.
	CHEN Yiyun, ZHAO Ruiying, QI Tianci, et al. Constructingrepresentative calibration dataset based onspectral transformation and Kennard-Stone algorithmfor VNIR modeling of soil total nitrogen in paddy soil[J]. Spectroscopy and Spectral Analysis, 2017,37(7):2133-2139.
[23]	章文龙, 曾从盛, 高灯州, 等. 闽江河口湿地土壤全磷高光谱遥感估算[J]. 生态学报, 2015,35(24):8085-8093.
	ZHANG Wenlong, ZENG Congsheng, GAO Dengzhou, et al. Estimating the soil total phosphorus content based on hyperspectral remote sensing data in the Min River estuarine wetland[J]. Acta Ecologica Sinica, 2015,35(24):8085-8093.
[24]	徐庆, 马驿, 蒋琦, 等. 水稻叶片含水量的高光谱遥感估算[J]. 遥感信息, 2018,33(5):1-8.
	XU Qing, MA Yi, JIANG Qi, et al. Estimation of rice leafwater content based on hyperspectral remote sensing[J]. Remote Sensing Information, 2018,33(5):1-8.
[25]	尼加提·卡斯木, 茹克亚·萨吾提, 师庆东, 等. 基于优化光谱指数的土壤有机质含量估算[J]. 农业机械学报, 2018,49(11):155-163.
	Nijiati Kasimu, Rukeya Sawuti, SHI Qingdong, et al. Estimation of soil organic matter content based on optimized spectral index[J]. Transactions of the Chinese Society of Agricultural Machinery, 2018,49(11):155-163.