Prediction of various soil properties for a national spatial dataset of Scottish soils based on four different chemometric approaches: A comparison of near infrared and mid-infrared spectroscopy

• Published in: Geoderma Vol. 396; p. 115071
• Main Authors: Haghi, R.K., Pérez-Fernández, E., Robertson, A.H.J.
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 15.08.2021
• Subjects: ATR-FTIR; bulk density; carbon; Chemometric; chemometrics; clay; CNN; Cubist; data collection; exchangeable magnesium; exchangeable potassium; Fourier transform infrared spectroscopy; inventories; least squares; neural networks; NIR; PLS; prediction; reflectance spectroscopy; sand; Scotland; SHAP; silt; spatial data; spectral analysis; Spectroscopy; SVR; total nitrogen; Scotland; Spectroscopy; CNN; SHAP; ATR-FTIR; NIR; PLS; Chemometric; Cubist; SVR

•ATR-FTIR in MIR outperformed DR spectroscopy in the NIR for prediction of soil properties.•ATR-FTIR data significantly improve the accuracy for the prediction of carbon.•CNN-1D and cubist techniques generated higher predictive ability than other models.•Feature impacts on predicted values were expl...