|WEINDORF, DAVID - Texas Tech University|
|CHAKRABORTY, SOMSUBHRA - Indian Institute Of Technology Kharagpur|
|MOORE-KUCERA, JENNIFER - Texas Tech University|
|BIN, LI - Louisiana State University|
|FULTZ, LISA - Louisiana State University Agcenter|
|CHENHUI, LI - Texas Tech University|
Submitted to: International Journal of Bioresource Science
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 3/11/2018
Publication Date: 6/13/2018
Citation: Weindorf, D.C., Chakraborty, S., Moore-Kucera, J., Bin, L., Fultz, L., Acosta Martinez, V., Chenhui, L. 2018. Advanced modeling of soil biological properties using visible near infrared diffuse reflectance spectroscopy. International Journal of Bioresource Science. 5/1-20. https://doi.org/10.30954/2347-9655.01.2018.1.
Interpretive Summary: Studying soil microbes in the laboratory often require chemicals that prevent reusing the samples for other types of analysis. However, spectroscopy techniques are becoming of interest in certain cases because they are non-destructive and provide fast results via a direct reading from the soil sample without any pre-treatment. Scientists from Texas Tech University and USDA-ARS in Lubbock TX, Louisiana State University, and Indian Institute of Technology in Kharagpur, India combined their knowledge on different instrumentation and statistical models. By analyzing several (123) soil samples with a spectroscopy technique, they were able to relate 12 biological parameters using a statistical model (the random forest predictive model). The approach will facilitate predicting distribution of different microbial groups in soil and have potential to be used in the field, without additional equipment or treatment.
Technical Abstract: Although visible near infrared diffuse reflectance spectroscopy (VisNIR DRS) is an emerging, rapid, non-destructive, and cost effective technology to predict a host of soil biological parameters, the traditional chemometric partial least squares regression (PLS) model often poses challenges during sensor development. In an effort to identify alternatives to the PLS model, three multivariate machine learning algorithms [PLS, penalized spline regression (PSR), and random forest (RF) regression]in conjunction with two spectral preprocessing methods [Savitzky–Golay first derivative and absorbance (ABS)] were compared with respect to 12 soil biological parameters of 123 soil samples. The RF model with ABS spectra successfully predicted all biological parameters with residual prediction deviation (RPD) ranging from 2.60 to 3.60 and outperformed PSR and PLS models. The best PSR model was obtained for total bacteria with an RPD of 2.70 and an r2 of 0.86 and among other variables, only Gram positive bacteria (RPD=2.63, r2=0.85), Gram negative bacteria (RPD=2.58, r2=0.85), and SOM (RPD=2.67, r2=0.86) were satisfactorily predicted, exhibiting r2>0.80 and RPD>2.5. Conversely, all variables except SOM (RPD=2.07) were poorly predicted by PLS models which had an RPD<2. Furthermore, linear discriminant analysis qualitatively clustered soils with different levels of microbial parameters. Summarily, the RF model with ABS spectra showed great promise in characterizing soil microbial communities with potential for such analysis in-situ.