 |
| Titel |
Combining mid infrared and total X-ray fluorescence spectroscopy for prediction of soil properties |
| VerfasserIn |
Erick Towett, Keith Shepherd, Andrew Sila, Ermias Aynekulu, Georg Cadisch |
| Konferenz |
EGU General Assembly 2015
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 17 (2015) |
| Datensatznummer |
250101651
|
| Publikation (Nr.) |
 EGU/EGU2015-836.pdf |
|
|
|
|
|
| Zusammenfassung |
| Mid-infrared diffuse reflectance spectroscopy (MIR) can predict many soil properties but
extractable nutrients are often predicted poorly. We evaluated the potential of MIR and total
elemental analysis using total X-ray fluorescence spectroscopy (TXRF), both individually
and combined, to predict results of conventional soil tests. Total multi-elemental analysis
provides a fingerprint of soil mineralogy and could predict some soil properties and help
improve MIR predictions. A set of 700 georeferenced soil samples associated with the Africa
Soil Information Service (AfSIS) (www.africasoils.net) from 44 stratified randomly-located
100-km2 sentinel sites distributed across sub-Saharan Africa were analysed for
physico-chemical composition using conventional reference methods, and compared to MIR
and TXRF spectra using the Random Forests regression algorithm and an internal out-of-bag
validation. MIR spectra resulted in good prediction models (R2 >0.80) for organic C and
total N, Mehlich-3 Ca and Al, and pH. To test the combined spectroscopic approach, TXRF
element concentration data was included as a property predictor along with the
first derivative of MIR spectral data using the RF algorithm. Including TXRF did
not improve prediction of these properties. TXRF was poorer (R2 0.86) as these elements are
not directly determined with TXRF, however the variance explained is still quite
high and may be attributable to TXRF signatures relating to mineralogy correlated
with protection of soil organic matter. TXRF model for Mehlich-3 Al had excellent
prediction capability explaining 81% of the observed variation in extractable Al
content and was comparable to that of MIR (R2 = 0.86). However, models for pH and
Mehlich-3 exchangeable Ca exhibited R2 values of 0.74 and 0.79 respectively and thus
had moderate predictive accuracy, compared to MIR alone with R2 values of 0.82
and 0.84 respectively. Both MIR and TXRF methods predicted soil properties that
relate to nutrient buffering capacity, including some exchangeable bases, pH, and
organic matter content, and fingerprint basic soil mineralogy. Further work should
investigate whether MIR and TXRF fingerprinting could better predict soil nutrient
supply capacity, as determined by crop nutrient uptake, than conventional soil tests. |
|
|
|
|
|
|