免费文献传递   相关文献

A near-infrared spectral index for estimating soil organic matter content.

一种估测土壤有机质含量的近红外光谱参数


通过系统分析我国中、东部地区5种不同类型土壤风干样本的有机质含量与近红外(1000~2500 nm)光谱反射率和一阶导数两波段构成的比值、差值及归一化指数之间的关系,构建了适合土壤有机质含量估测的光谱参数及定量反演模型.结果表明:用多元散射校正及Savitzky-Golay平滑方法对原始光谱反射率进行预处理后,其两波段组成的光谱参数与土壤有机质含量的相关性明显优于原始光谱反射率组成的光谱参数,而由上述预处理后的反射率一阶导数的两波段构成的光谱参数介于二者之间;不同类型光谱参数构成形式中,以差值指数的预测性最好,其次为比值和归一化指数;与土壤有机质含量相关程度最高的光谱参数,是以近红外合频区1883和2065 nm 2个波段的反射率经多元散射校正和Savitzky-Golay平滑后构建而成的差值指数DI(CR1883, CR2065),两者呈极显著的直线相关.经不同类型土壤的观测资料检验,模型的决定系数为0.837,均方根误差为4.06;与偏最小二乘法的全谱建模结果相比,尽管DI(CR1883, CR2065)的预测精度略低于后者,但该指数只使用了2个波段的反射率,且所建模型比较简单,能为便携式监测仪的研制提供更有效的信息,可作为一种良好的土壤有机质估测光谱参数.

Taking the air-dried samples of five soil types from middle and easte
rn China as test materials, the correlations of their organic matter content wit
h the spectral reflectance of near-infrared (1000〖KG-*2〗-〖KG-*7〗2500 nm), a
nd with the ratio index (RI), difference index (DI), and normalized diff
erence index (ND) of the first derivative values of the reflectance between
two bands were studied. Based on this, the key spectral indices and the quantita
tive models for estimating soil organic matter (SOM) content were developed. Aft
er corrected with Multiplicative Scatter Correction (MSC) and Savitzky-Golay (S
G) smoothing methods, the spectral reflectance of near-infrared had an obviousl
y high correlation with SOM, compared with the original spectral reflectance, wh
ile the corrected spectral indices of the first derivative values of the reflect
ance between two bands took the intermediate position. The correlation of the sp
ectral indices with SOM was in the order of was DI>RI>ND, regardless the c
omposition of the original spectral reflectance or the first derivative spectra.
 The DI of the reflectance of near-infrared between 1883 and 2065 nm correc
ted with MSC and SG smoothing methods [DI(CR1883, CR2065)]
 had the best linear correlations with SOM. The test of the moni
toring model based on DI(CR1883, CR2065) with the independen
t datasets of SOM showed that the R2 and RMSE validation values were 083
7 and 406, respectively. Comparing with the results from the Partial Least Squ
are (PLS) method, the monitoring model based on DI (CR1883, CR
2065) was somewhat inferior. However, the DI (CR1883, CR2065
) only needed two reflectance bands, and the monitoring model was
 simpler, being able to provide more available information for developing portab
le instruments, and a good spectral index for estimating SOM content.