A systematic study on the application of scatter-corrective and spectral-derivative preprocessing for multivariate prediction of soil organic carbon by Vis-NIR spectra

• Published in: Geoderma 2018-03, Vol.314, p.262-274
• Main Authors: Dotto, Andre Carnieletto, Dalmolin, Ricardo Simao Diniz, ten Caten, Alexandre, Grunwald, Sabine
• Format: Article
• Language: English
• Subjects: Modeling; Pedometrics; Prediction; Soil property; Visible-near infrared spectroscopy
• ISSN: 0016-7061; 1872-6259
• DOI: 10.1016/j.geoderma.2017.11.006

Soil organic carbon (SOC) plays a crucial role as an ecosystems indicator. Its quantification requires an affordable, and less time-consuming method. Visible and near infrared (Vis-NIR) reflectance spectroscopy has demonstrated its applicability to predict SOC over the years. The aims of this study were to i) to compare the influence of preprocessing techniques on prediction performance, ii) assess the modeling performance of a wide range of multivariate methods, and iii) evaluate the potential of Vis-NIR spectroscopy to predict SOC. Soil sampling was conducted over an area of approximately 1800km2 in the Southern region of Brazil, where a total of 595 soil samples were collected. Oxisols are predominant in the area following by Entisols and Inceptisols. The seven preprocessing techniques that were employed can be divided into two categories based on their SOC prediction performance: scatter-correction and spectral-derivatives. A total of nine different methods were evaluated to predict SOC from Vis-NIR spectra. The models that use scatter-corrective preprocessing exhibited superior prediction compared to the spectral-derivatives group. In the scatter-correction group, continuum removal was the most suitable preprocessing method for SOC prediction. Except for random forest (RF), all the multivariate methods presented robust predictions. The best fit and highest model accuracy for SOC models in validation mode were achieved when applying the weighted average partial least-squares (WAPLS) method and normalization by range (NBR) preprocessing (R2=0.82, root mean square error=0.48%, and ratio of the performance to the interquartile range=3.18). Findings from this systematic methodology study identified the reliability of SOC determinations by examining how preprocessing techniques and multivariate methods affect spectral analyses. It also guides future studies to select the most appropriate methods on similar soils. •We tested a combo of 63 spectral preprocessing and multivariate prediction models.•Organic carbon was predicted using 595 soil samples from surface to 2m depth.•Spectral preprocessing by scatter-correction techniques improved prediction.•Vis-NIR spectroscopy reached a plateau with persisting 20% of unexplained variability.•Coupling of spectroscopic data and environmental covariates is needed.