摘要
以托克托县采集的非盐渍土为研究对象,向其添加盐渍土中常见的易溶性钠盐,采用便携式地物光谱仪SVC HR-1024,测定不同盐分含量的土壤的光谱反射率,构建了土壤盐分与高光谱信息的偏最小二乘回归(PLSR)和多元线性逐步回归(MLSR)的土壤盐分反演模型。结果表明:(1)不同种类及含量的盐分对土壤的光谱曲线影响不同,经Na_(2)CO_(3)、Na_(2)SO_(4)和NaHCO_(3)盐溶液处理的土壤光谱反射率随着土壤含盐量的增加而增加,NaCl溶液处理后的土壤反射率随着土壤含盐量的增加而递减,且各土样反射率均高于未经处理的土样。(2)4种土样原始光谱数据与SSC相关性从高到低依次为:NaCl、Na_(2)SO_(4)、NaHCO_(3)、Na_(2)CO_(3):经过数学变换后,4种土样最大相关系数分别为:R′-NaCl(-0.895)、(√R)′-Na_(2)CO_(3)(-0.781)、R″-Na_(2)SO_(4)(0.767)、(lg R)′-Na_(2)CO_(3)(-0.874):各土样根据相关系数所选敏感波段不同。(3)建模效果最好的是NaCl土样的一阶微分处理的PLSR模型,决定系数(R^(2))和均方根误差分别为0.871和0.764;MLSR模型中效果最好的为经对数一阶微分处理后的NaHCO_(3)模型,决定系数(R^(2))和均方根误差分别为0.824和0.846。经验证,PLSR模型更适宜进行土壤盐分反演。
With the non-saline soil collected in Togtoh County,by adding the common soluble sodium salt in saline soil,we used portable ground object spectrometer SVC HR-1024,to measure the spectral reflectance of soil with different salt content,and established the soil salinity inversion models of partial least squares regression(PLSR)and multiple linear stepwise regression(MLSR).The results showed that:(1)Different types and contents of salt had different effects on the spectral curve of soil.The spectral reflectance of soil treated with Na_(2)CO_(3),Na_(2)SO_(4) and NaHCO_(3) increased with the increase of soil salt content,and the soil reflectance after NaCl solution treatment decreased with the increase of soil salt content,and the reflectance of each soil sample was higher than that of untreated soil.(2)The correlation between SSC and the original spectral data of four soil samples was in the order of NaCl,Na_(2)SO_(4),NaHCO_(3),and Na_(2)CO_(3).After mathematical transformation,the highest correlation coefficients of the four soil samples are R′-NaCl(-0.895),-Na_(2)CO_(3)(-0.781),R″-Na_(2)SO_(4)(0.767),and(lg R)′-Na_(2)CO_(3)(-0.874).According to the correlation coefficient,the sensitive bands of soil samples are different.(3)The PLSR model with first-order differential treatment of NaCl soil sample has the best modeling effect,coefficient of determination(R^(2))and root mean square error is 0.871 and 0.764 respectively;among MLSR models,NaHCO_(3) model with logarithmic first-order differential treatment is the best,coefficient of determination(R^(2))and root mean square error are 0.824 and 0.846,respectively.After verification,PLSR model is more suitable for soil salinity inversion.
作者
穆其尔
杨光
陈昊宇
张天琪
Mu Qier;Yang Guang;Chen Haoyu;Zhang Tianqi(Inner Mongolia Agricultural University,Hohhot 010018,P.R.China)
出处
《东北林业大学学报》
CAS
CSCD
北大核心
2021年第11期68-75,共8页
Journal of Northeast Forestry University
基金
内蒙古自治区科技重大专项(2019ZD003)。
关键词
土壤
盐渍化
高光谱反演
可控实验
回归模型
Soil
Salinization
Hyperspectral inversion
Controllable experiment
Regression model
