反射光谱估算滨海土壤黏粒含量

探明反射光谱估算土壤黏粒含量的成因是实现黏粒含量测定、提高估算精度的基础。该文以江苏省滨海平原的150个土壤样品为研究对象,将测得的原始光谱数据进行平滑、一阶导数、连续统去除和倒数等数据变换,采用逐步多元线性回归(stepwise multiple linear regression,SMLR)和偏最小二乘回归(partial least squares regression,PLSR)方法估算黏粒含量,并在此基础上分析建模的影响波段,探讨反射光谱估算土壤黏粒含量的成因。结果表明,连续统去除光谱数据的SMLR分析估算精度最高,建模集和验证集决定系数分别为0.941和0.750。360~900、1 800~2 490 nm是黏粒含量的重要建模影响波段,该建模影响波段主要包括铁离子(410 nm附近)、土壤有机质(500~800 nm)、层状硅酸盐中的结晶水(1 900 nm附近)、绿泥石和蛭石等黏土矿物(2 325 nm)的吸收特征波段;PLSR分析表明,1 400 nm附近回归系数出现的双峰特征源于高岭石的双峰吸收。黏粒中的黏土矿物、黏粒对铁离子的吸附特性以及黏粒与有机质的高度相关性是实现反射光谱估算滨海土壤黏粒含量的原因。

Coastal soil clay content estimation using reflectance spectroscopy

Abstract: Clay content is an important soil property that affects the structure, nutrient supply and other characteristics of soils. Variations in clay content can indicate the degree of soil development or soil age. In traditional chemical analyses of soil properties, the extractant interacts in the solution and at the solution-particle interface, thus altering the equilibrium between the soil solid and solution phases. Soil reflectance spectroscopy has been developed as an effective alternative method of measuring soil properties primarily because it requires minimal sample preparation and it is fast, cost-effective, non-destructive and non-hazardous to the soil. In recent decades, research on the use of reflectance spectroscopy in soil science has achieved rapid advances. Reflectance spectroscopy can be successfully applied to estimate the soil clay content. However, the mechanisms of soil clay content estimation using reflectance spectroscopy are not very clear. The goals of this study were to identify the bands within the range of 360-2490 nm that can be used to estimate the clay content and explore the mechanisms of the clay content estimation using reflectance spectroscopy. A total of 150 coastal soil samples were collected. The soil reflectance spectra were measured in a dark room using a FieldSpec 3 portable spectrometer. Raw spectral data were pre-processed by smoothing (R) and then by first derivative (FD), continuum removal (CR) or reciprocal transformation (DS). Calibration (75 soil samples) and validation datasets (75 soil samples) were obtained from 1,000 random selections of the data. Stepwise multiple linear regression (SMLR) and partial least squares regression (PLSR) were performed to estimate the soil clay content and to further identify the bands useful for modeling this parameter. The results indicated that the SMLR analysis of CR and R spectra and the PLSR analysis of R and FD spectra were characterized by good calibration and validation accuracies regarding the soil clay content. The frequency of a SMLR-selected band and the regression coefficient in the PLSR regression equation indicated the impact of the clay content on the reflectance spectroscopy. The bands of 360-900 nm and 1800-2490 nm were important for the clay content estimation. The absorption bands near 1900 nm were caused by crystal water in phyllosilicates, whereas the absorption at 2325 nm was attributed to the combined effects of absorption by chlorite and vermiculite. The 500-800 nm absorption bands were caused by the soil organic matter (SOM). And the relationship between the clay content and reflectance spectra was a secondary relationship attributed to the high correlation between SOM and clay content. The absorption near 410 nm was produced by the iron ions. The mutual adsorption of iron ions and clay was consistent with the relatively high model contribution. The regression coefficients in the PLSR yielded dual peaks of absorption near 1400 nm, which was attributed to the dual absorption peaks of kaolinite. In summary, the clay mineral composition, adsorption of clay and iron ions and correlations between clay and organic matter were the causes of the clay content estimation using reflectance spectroscopy.