Improved intact soil-core carbon determination applying regression shrinkage and variable selection techniques to...

Area of Science:

• Soil Science
• Analytical Chemistry
• Spectroscopy

Background:

• Laser-induced breakdown spectroscopy (LIBS) is a promising technique for rapid, in situ soil carbon analysis.
• Previous studies suggested ultraviolet (UV) spectrum LIBS may lack sufficient elemental data for accurate soil organic carbon (SOC) and inorganic carbon (IC) discrimination.
• Intact soil core analysis presents challenges due to sample heterogeneity.

Purpose of the Study:

• To compare the efficacy of complete spectrum (245-925 nm) LIBS versus UV spectrum (200-300 nm) LIBS for determining soil total carbon (TC), IC, and SOC in intact cores.
• To evaluate the performance of partial least squares 2 (PLS2), least absolute shrinkage and selection operator (LASSO), and sparse multivariate regression with covariance estimation (MRCE) for soil carbon prediction.
• To identify key wavelengths for accurate soil carbon quantification using LIBS.

Main Methods:

• Analysis of 60 intact soil cores from six wheat fields using a custom complete spectrum core-scanning LIBS instrument.
• Development and comparison of predictive multi-response partial least squares (PLS2) models using full and reduced spectrum LIBS data.
• Application of LASSO and MRCE for variable selection and improved calibration accuracy in soil carbon prediction.

Main Results:

• Complete spectrum LIBS reduced the standard error of prediction (SEP) by 15% for TC and 19% for IC compared to UV spectrum LIBS when using PLS2.
• LASSO and MRCE approaches significantly improved calibration accuracy, reducing SEP by 32-55% over UV spectrum PLS2 models.
• Complete spectrum LIBS demonstrated superior performance for intact soil core analysis without pretreatment.

Conclusions:

• Complete spectrum LIBS is more effective than UV spectrum LIBS for direct soil carbon measurement in intact cores.
• LASSO and MRCE offer enhanced prediction accuracy over PLS2, warranting further validation with diverse soil types and analytes.
• Measurement errors related to intact core properties like density and surface roughness require additional investigation.