Page Banner

United States Department of Agriculture

Agricultural Research Service

Title: Evaluation of PLS, LS-SVM, and LWR for quantitative spectroscopic analysis of soils)

Author
item Igne, Benoit
item Reeves iii, James
item Mccarty, Gregory
item Hively, Wells - Dean
item Lund, Eric
item Hurburgh, jr., Charles

Submitted to: Geoderma
Publication Type: Peer reviewed journal
Publication Acceptance Date: 6/18/2010
Publication Date: 7/15/2010
Publication URL: http://handle.nal.usda.gov/10113/59928
Citation: Igne, B., Reeves, J.B., McCarty, G.W., Hively, W.D., Lund, E., Hurburgh, C.R. 2010. Evaluation of PLS, LS-SVM, and LWR for quantitative spectroscopic analysis of soils. Geoderma. 18(3):167-176.

Interpretive Summary: Recently there has been a lot of interest on the potential for soil to remove and sequester carbon from the atmosphere. This would both improve soil quality and potentially help with reduce global warming. If credit or payments are to be made for these efforts, methods are needed to determine how much carbon is being sequestered in the soil over a specific period of time. Traditional methods for determining carbon in soil are very time consuming and expensive and newer methods such as near- and mid-infrared spectroscopy have received a lot of attention as possible methods for measuring soil carbon. Both methods, which utilize the interaction of light, beyond the range of human vision, with the soil are rapid, non-destructive, and generate no chemical wastes, but require complex mathematical methods to relate the resulting spectral data to the composition (carbon content) of the soil. The objective of this research was to investigate several different statistical procedures to determine if any performed better at extracting the information of interest. A comparison of the use of spectral pretreatment as well as the implementation of linear and non-linear regression methods was performed. This study presents an overview of the use of infrared spectroscopy for the prediction of five physical (sand, silt, and clay) and chemical (total carbon and total nitrogen) soil parameters with near and mid infrared units in bench top and field setups. Even though no significant differences existed among pretreatment methods, models using second derivatives performed better. The implementation of partial least squares (PLS), least squares support vector machines (LS-SVM), and locally weighted regression (LWR) for the development of the calibration models showed that the LS-SVM did not outperform linear methods for most components while LWR that creates simpler models performed well. The present results tend to show that soil models are quite sensitive to the complexity of the model. The ability of LWR to select only the appropriate samples did help in the development of robust models. Results also proved that field units performed as well as bench top instruments. This was true for both near-infrared and mid-infrared technology.

Technical Abstract: Soil testing requires the analysis of large numbers of samples in laboratory that are often time consuming and expensive. Mid-infrared spectroscopy (mid-IR) and near-infrared spectroscopy (NIRS) are fast, non-destructive, and inexpensive analytical methods that have been used for soil analysis, in laboratory and on the go, to reduce the need for measurements using complex chemical/physical analyses. A comparison of the use of spectral pretreatment as well as the implementation of linear and non-linear regression methods was performed. This study presents an overview of the use of infrared spectroscopy for the prediction of five physical (sand, silt, and clay) and chemical (total carbon and total nitrogen) soil parameters with near and mid infrared units in bench top and field setups. Even though no significant differences existed among pretreatment methods, models using second derivatives performed better. The implementation of partial least squares (PLS), least squares support vector machines (LS-SVM), and locally weighted regression (LWR) for the development of the calibration models showed that the LS-SVM did not outperform linear methods for most components while LWR that creates simpler models performed well. The present results tend to show that soil models are quite sensitive to the complexity of the model. The ability of LWR to select only the appropriate samples did help in the development of robust models. Results also proved that field units performed as well as bench top instruments. This was true for both near-infrared and mid-infrared technology.

Last Modified: 8/24/2016
Footer Content Back to Top of Page