Spectral reflectance based indices for soil organic carbon quantification
We investigated 40 samples from nine different soil types, originating from several climatic zones and a large variety in SOC content (0.06-45.1%). Spectral measurements for all soil samples were performed in a controlled laboratory environment. We tested the performance of several spectral indices which have been developed to detect biochemical constituents (e.g., cellulose, lignin) for their ability to retrieve SOC, and compared it to PLS. Good relations were found for indices based on the visible part of the spectrum(R-2=0.80) and for the absorption features related to cellulose (around 2100 nm) (R-2=0.81). The best index based relations were compared to the results for PLS (R-2=0.87). Cross validation was used to evaluate the predictive capacity of the spectral indices. The results demonstrate that it is feasible to use spectral indices derived from laboratory measurements to predict SOC in various soil types. However, a large variance in SOC is required for the calibration of the prediction model, since extrapolation beyond the SOC range in the training dataset results in large errors. PLS proves to be much less sensitive towards extrapolation of the model beyond the mineralogy and SOC levels used during the calibration. (c) 2008 Elsevier B.V. All rights reserved.
| Main Authors: | , , , , , |
|---|---|
| Format: | Article/Letter to editor biblioteca |
| Language: | English |
| Subjects: | agricultural soils, calibration, least-squares regression, matter, moisture, near-infrared spectroscopy, nitrogen, nm, world, |
| Online Access: | https://research.wur.nl/en/publications/spectral-reflectance-based-indices-for-soil-organic-carbon-quanti |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | We investigated 40 samples from nine different soil types, originating from several climatic zones and a large variety in SOC content (0.06-45.1%). Spectral measurements for all soil samples were performed in a controlled laboratory environment. We tested the performance of several spectral indices which have been developed to detect biochemical constituents (e.g., cellulose, lignin) for their ability to retrieve SOC, and compared it to PLS. Good relations were found for indices based on the visible part of the spectrum(R-2=0.80) and for the absorption features related to cellulose (around 2100 nm) (R-2=0.81). The best index based relations were compared to the results for PLS (R-2=0.87). Cross validation was used to evaluate the predictive capacity of the spectral indices. The results demonstrate that it is feasible to use spectral indices derived from laboratory measurements to predict SOC in various soil types. However, a large variance in SOC is required for the calibration of the prediction model, since extrapolation beyond the SOC range in the training dataset results in large errors. PLS proves to be much less sensitive towards extrapolation of the model beyond the mineralogy and SOC levels used during the calibration. (c) 2008 Elsevier B.V. All rights reserved. |
|---|