Aarhus University Seal / Aarhus Universitets segl

Mogens Humlekrog Greve

Prediction of Clay/Organic Carbon Ratio Using On-The-Go Proximal Spectroscopy

Publikation: KonferencebidragKonferenceabstrakt til konferenceForskning

Standard

Prediction of Clay/Organic Carbon Ratio Using On-The-Go Proximal Spectroscopy. / Tabatabai, Salman; Knadel, Maria; Greve, Mogens Humlekrog; Marakkala Manage, Lashya P.

2017. Abstract fra 18th International Conference on Near Infrared Spectroscopy, Copenhagen, Danmark.

Publikation: KonferencebidragKonferenceabstrakt til konferenceForskning

Harvard

Tabatabai, S, Knadel, M, Greve, MH & Marakkala Manage, LP 2017, 'Prediction of Clay/Organic Carbon Ratio Using On-The-Go Proximal Spectroscopy' 18th International Conference on Near Infrared Spectroscopy, Copenhagen, Danmark, 11/06/2017 - 15/06/2017, .

APA

Tabatabai, S., Knadel, M., Greve, M. H., & Marakkala Manage, L. P. (2017). Prediction of Clay/Organic Carbon Ratio Using On-The-Go Proximal Spectroscopy. Abstract fra 18th International Conference on Near Infrared Spectroscopy, Copenhagen, Danmark.

CBE

Tabatabai S, Knadel M, Greve MH, Marakkala Manage LP. 2017. Prediction of Clay/Organic Carbon Ratio Using On-The-Go Proximal Spectroscopy. Abstract fra 18th International Conference on Near Infrared Spectroscopy, Copenhagen, Danmark.

MLA

Vancouver

Tabatabai S, Knadel M, Greve MH, Marakkala Manage LP. Prediction of Clay/Organic Carbon Ratio Using On-The-Go Proximal Spectroscopy. 2017. Abstract fra 18th International Conference on Near Infrared Spectroscopy, Copenhagen, Danmark.

Author

Tabatabai, Salman ; Knadel, Maria ; Greve, Mogens Humlekrog ; Marakkala Manage, Lashya P. / Prediction of Clay/Organic Carbon Ratio Using On-The-Go Proximal Spectroscopy. Abstract fra 18th International Conference on Near Infrared Spectroscopy, Copenhagen, Danmark.1 s.

Bibtex

@conference{8410226dfa1547daa843b564498d0878,
title = "Prediction of Clay/Organic Carbon Ratio Using On-The-Go Proximal Spectroscopy",
abstract = "Previous studies have outlined the critical value of clay to organic carbon (OC) ratio (n-index) as a functional soil attribute in soil science. Traditional lab methodologies for obtaining clay and OC and calculating the n-index are time consuming and costly. Successful determination of n-index using laboratory visible near-infrared spectroscopy (Vis-NIRS) for a wide range of soils was recently published. In this study, we tested a direct prediction of n-index in the field using a proximal spectrometer. Spectral reflectance of eight agricultural fields was measured in the range of 350-2200 nm using a commercial mobile sensor platform (Veris Technologies, USA). Principal components analysis was performed on spectra followed by fuzzy c-means clustering to select 15 representative sampling locations on each field. Clay and OC were determined for all samples using pipette and ignition methods, respectively and n-index was calculated (range: 1.16-8.45). Partial least squares (PLS) models were calibrated using pretreated vis-NIR spectra as predictors and n-index as the response. Ventian Blinds (VB) cross validation (CV) using 15 segments, one-field-out (OFO) CV and regression prediction using a Kennard-Stone 70-30{\%} data split were used to validate the models. VB CV had validation R2 0.96, RMSEV 0.40 and ratio of interquartile range to prediction (RIQP) 3.13. OFO CV showed a significantly lower performance with R2 0.84, RMSEV 0.88 and RIQP 1.42. In the regression model, 70{\%} of the data was used to make a VB cross-validated model, which was used to predict the remaining 30{\%} independently. The prediction resulted in R2 0.88, RMSEP 0.47 and RIQP 1.37. The similar low performance of OFO CV and Regression Prediction indicate that the current models are not capable of yielding reliable results for n-index prediction in a hypothetical blind field and calls for further investigations.",
author = "Salman Tabatabai and Maria Knadel and Greve, {Mogens Humlekrog} and {Marakkala Manage}, {Lashya P}",
year = "2017",
language = "English",
note = "null ; Conference date: 11-06-2017 Through 15-06-2017",
url = "http://icnirs2017.com/",

}

RIS

TY - ABST

T1 - Prediction of Clay/Organic Carbon Ratio Using On-The-Go Proximal Spectroscopy

AU - Tabatabai, Salman

AU - Knadel, Maria

AU - Greve, Mogens Humlekrog

AU - Marakkala Manage, Lashya P

PY - 2017

Y1 - 2017

N2 - Previous studies have outlined the critical value of clay to organic carbon (OC) ratio (n-index) as a functional soil attribute in soil science. Traditional lab methodologies for obtaining clay and OC and calculating the n-index are time consuming and costly. Successful determination of n-index using laboratory visible near-infrared spectroscopy (Vis-NIRS) for a wide range of soils was recently published. In this study, we tested a direct prediction of n-index in the field using a proximal spectrometer. Spectral reflectance of eight agricultural fields was measured in the range of 350-2200 nm using a commercial mobile sensor platform (Veris Technologies, USA). Principal components analysis was performed on spectra followed by fuzzy c-means clustering to select 15 representative sampling locations on each field. Clay and OC were determined for all samples using pipette and ignition methods, respectively and n-index was calculated (range: 1.16-8.45). Partial least squares (PLS) models were calibrated using pretreated vis-NIR spectra as predictors and n-index as the response. Ventian Blinds (VB) cross validation (CV) using 15 segments, one-field-out (OFO) CV and regression prediction using a Kennard-Stone 70-30% data split were used to validate the models. VB CV had validation R2 0.96, RMSEV 0.40 and ratio of interquartile range to prediction (RIQP) 3.13. OFO CV showed a significantly lower performance with R2 0.84, RMSEV 0.88 and RIQP 1.42. In the regression model, 70% of the data was used to make a VB cross-validated model, which was used to predict the remaining 30% independently. The prediction resulted in R2 0.88, RMSEP 0.47 and RIQP 1.37. The similar low performance of OFO CV and Regression Prediction indicate that the current models are not capable of yielding reliable results for n-index prediction in a hypothetical blind field and calls for further investigations.

AB - Previous studies have outlined the critical value of clay to organic carbon (OC) ratio (n-index) as a functional soil attribute in soil science. Traditional lab methodologies for obtaining clay and OC and calculating the n-index are time consuming and costly. Successful determination of n-index using laboratory visible near-infrared spectroscopy (Vis-NIRS) for a wide range of soils was recently published. In this study, we tested a direct prediction of n-index in the field using a proximal spectrometer. Spectral reflectance of eight agricultural fields was measured in the range of 350-2200 nm using a commercial mobile sensor platform (Veris Technologies, USA). Principal components analysis was performed on spectra followed by fuzzy c-means clustering to select 15 representative sampling locations on each field. Clay and OC were determined for all samples using pipette and ignition methods, respectively and n-index was calculated (range: 1.16-8.45). Partial least squares (PLS) models were calibrated using pretreated vis-NIR spectra as predictors and n-index as the response. Ventian Blinds (VB) cross validation (CV) using 15 segments, one-field-out (OFO) CV and regression prediction using a Kennard-Stone 70-30% data split were used to validate the models. VB CV had validation R2 0.96, RMSEV 0.40 and ratio of interquartile range to prediction (RIQP) 3.13. OFO CV showed a significantly lower performance with R2 0.84, RMSEV 0.88 and RIQP 1.42. In the regression model, 70% of the data was used to make a VB cross-validated model, which was used to predict the remaining 30% independently. The prediction resulted in R2 0.88, RMSEP 0.47 and RIQP 1.37. The similar low performance of OFO CV and Regression Prediction indicate that the current models are not capable of yielding reliable results for n-index prediction in a hypothetical blind field and calls for further investigations.

M3 - Conference abstract for conference

ER -