Application of mid-infrared photoacoustic spectroscopy in monitoring carbonate content in soils

• Published in: Sensors and actuators. B, Chemical Vol. 188; pp. 1167 - 1175
• Main Authors: Du, Changwen, Ma, Zhaoyang, Zhou, Jianmin, Goyne, Keith W.
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.11.2013
• Subjects: absorption; Chemometrics; China; Erosion; least squares; Loess soil; monitoring; neural networks; Photoacoustic spectroscopy; prediction; principal component analysis; reflectance spectroscopy; Soil carbonate; soil properties; soil sampling; China; FTIR-PAS; RMSEP; RPD; GRNN; Photoacoustic spectroscopy; Soil carbonate; PLSR; Loess soil; Chemometrics; PCA; Erosion
• ISSN: 0925-4005; 1873-3077
• DOI: 10.1016/j.snb.2013.08.023

•Loess soils were firstly characterized using infrared photoacoust spectroscopy.•We verified the featured absorption band of carbonate around 1410cm−1.•Photoacoustic spectroscopy shows potential in the application of soil sensing.•Carbonate content can be well predicted combining chemometrics methods. Infrared photoacoustic spectroscopy provides an alternative to conventional infrared reflectance spectroscopy for rapidly estimating a wide array of soil properties. The objective of this study was to investigate the application of Fourier transform mid infrared (500–4000cm−1)–photoacoustic spectroscopy (FTIR-PAS) to estimate soil carbonate content in samples collected from the Loess Plateau of China. Principal component analysis (PCA), partial least squares regression (PLSR) and generalized regression neural network (GRNN) models were used to calibrate and validate soil carbonate analysis using FTIR-PAS. Absorption bands for carbonate were observed in the FTIR-PAS spectra. Even though most bands associated with carbonate were subject to interference from other soil components, significant relationships were observed between carbonate content and FTIR-PAS spectral components, particularly in the range of 1000–2000cm−1. Among the chemometric approaches applied, the GRNN model demonstrated the best performance [root mean square error (RMSEP)=1.21% with a ratio of standard deviation to prediction error (RPD)=3.83] for predicting soil carbonate content. This work demonstrates that FTIR-PAS is suitable for analyzing solid soil samples exhibiting great IR absorption, and the technique permits accurate and rapid determination of soil carbonate content.