近红外光谱(NIRS)技术快速测定湖泊沉积物营养组分研究  

Rapid Determination of Nutrient Components in Lake Sediments Using Near Infrared Spectroscopy

作　　者：昝逢宇[1,2] 霍守亮[3] 席北斗[3] 李青芹[3] 刘鸿亮[3] 

机构地区：[1]北京师范大学环境学院,北京100875 [2]安徽师范大学环境科学学院,安徽芜湖241000 [3]中国环境科学研究院,北京100012

出　　处：《光谱学与光谱分析》2010年第10期2624-2627,共4页Spectroscopy and Spectral Analysis

基　　金：国家科技重大专项项目(2009ZX07106-001);国家重点基础研究发展计划项目(2008CB418206)资助

摘　　要：湖泊沉积物保存了湖泊环境演化的重要信息,但目前还没有建立起快速、准确的湖泊沉积物营养组分组合分析方法。该文利用近红外光谱(NIRS)技术,采用一阶导数、小波滤噪、正交信号校正、小波滤噪+正交信号校正、一阶导数+正交信号校正、正交信号校正+小波压缩6种光谱预处理技术和偏最小二乘(PLS)法相结合在国内建立了NIR光谱测定湖泊沉积物柱芯样品中总碳(TC),总氨(TN),总有机碳(TOC)和总磷(TP)的校正模型,结果表明虽然NIR光谱校正模型对TOC的预测效果不理想,但一阶导数+正交信号校正光谱校正模型对TC和TN,正交信号校正光谱校正模型对TP的预测效果较好,预测相关系数分别为0.759 7(TC),0.865 0(TN)和0.811 2(TP),预测误差(RMSEP)分别为0.13%(TC),0.008 2%(TN)和0.012%(TP)。该研究对于推动我国湖泊沉积物的光谱学特性研究具有重要意义。The lake sediments record important environmental evolution information of lake in recent 100 years.However,a rapid and precise combination analytical method measuring nutrient components in lake sediments can not be established.The near infrared reflectance spectra（NIRS） of sediment coring samples were determined by near-infrared reflectance spectrometry.The respective near NIRS calibration models for predicting total carbon（TC）,total nitrogen（TN）,total organic carbon（TOC） and total phosphorus（TP） were built first in China using partial least squares（PLS） algorithm with six spectral pretreatment tools including first-order derivate,wavelet denoise,orthogonal signal correction（OSC）,wavelet denoise combining orthogonal signal correction（OSC）,first-order derivate combining orthogonal signal correction（OSC）,and orthogonal signal correction（OSC） combining wavelet compression.The results showed that although NIR all calibration models can not well predict TOC,the first-order derivate combining OSC spectral calibration model had a good prediction for TC and TN,and for TP OSC spectral calibration model was good.The correlation coefficients between measuring values and predicted values in validation set for TC,TN and TP were 0.76,0.87 and 0.81,respectively.RMSEP（Root mean square error of the prediction）for TC,TN and TP were 0.13%,0.008 2% and 0.012%,respectively.The study has an important significance of driving the domestic researches on spectroscopy characteristic of lake sediments and establishment of rapid analytical technique determining nutrient components of lake sediments.

关 键 词：近红外光谱(NIRS) 湖泊沉积物 营养物 光谱预处理 偏最小二乘(PLS) 

分 类 号：Q657.3[生物学—生物物理学]