Premium
Combining Vis–NIR spectroscopy and advanced statistical analysis for estimation of soil chemical properties relevant for forest road construction
Author(s) -
Mousavi Fatemeh,
Abdi Ehsan,
Knadel Maria,
Tuller Markus,
Ghalandarzadeh Abbas,
Bahrami Hossein Ali,
Majnounian Baris
Publication year - 2021
Publication title -
soil science society of america journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.836
H-Index - 168
eISSN - 1435-0661
pISSN - 0361-5995
DOI - 10.1002/saj2.20253
Subject(s) - partial least squares regression , cation exchange capacity , diffuse reflectance infrared fourier transform , soil test , near infrared spectroscopy , environmental science , soil organic matter , mean squared error , spectroscopy , artificial neural network , adaptive neuro fuzzy inference system , soil science , remote sensing , soil water , mathematics , computer science , chemistry , statistics , machine learning , artificial intelligence , fuzzy logic , geology , biochemistry , physics , fuzzy control system , photocatalysis , quantum mechanics , catalysis
A thorough quantification of soil chemical properties is essential for assessing the engineering properties of forest soils for road design, construction, and maintenance. Here, we investigate the applicability of visible–near‐infrared (Vis–NIR) spectroscopy in conjunction with advanced statistical analysis for estimation of soil chemical properties. Sixty forest soil samples were collected and analyzed for pH, electrical conductivity (EC), CaCO 3 , organic matter (OM), and cation exchange capacity (CEC) with established laboratory methods. The spectral measurements were performed with a Vis–NIR spectrometer within a range of 350–2,500 nm. To estimate abovementioned soil properties from reflectance spectra, advanced statistical techniques including partial least squares regression (PLSR), hybrid partial least squares and artificial neural networks (PLS–DI–ANN) models, hybrid partial least squares and adaptive neural fuzzy inference system (PLS–DI–ANFIS) models, as well as narrow band spectral indices were applied. The obtained results indicate that the PLS–DI–ANFIS models show great potential for the estimation of pH, EC, OM, and CEC from reflectance spectra and their first derivatives, exhibiting higher R 2 values and lower RMSE than the other investigated models. The estimation accuracy for CaCO 3 , however, was low for all applied methods. The results confirm that Vis–NIR spectroscopy may be applied as a rapid and cost‐efficient alternative to standard chemical soil analysis techniques, aiding forest road design, construction, and maintenance.