This is a spectral platform for predictions of soil properties for
soil samples from the Democratic Republic of Congo and includes more
than 1300 individual soil samples.
The samples have been collected from agricultural fields, coffee
plantations, from an arboretum,
from mixed, monodominant
bamboo, lowland, montane primary and secondary
forests and from termite mounts in the provinces Tshuapa, Tshopo,
South Kivu, Bas-Uele, Haut-Katanga and Kongo Central
. The wet chemistry analsyis
has been done at ETH Zurich, Ghent University, Université Catolique
de Louvain, IITA Kalambo, ICRAF and Université de Lubumbashi.
The spectral analysis and the modeling is done in R (R core Team, 2018), and is mainly based on the package
(Baumann, unpublished). It embeds several packages for data
preprocessing, splitting and modeling, such as the
package interface (Kuhn et al., 2016) for model development and predictions and
package (Mevik et al., 2016) for Partial Least Squares Regressions (PLSR).
For the final models, the spectra were resampled to a new wavenumber
range from 500 cm
to 3996 cm
with a spectral resolution of 2 cm
and preprocessed using a
Savitzky-Golay filter (Savitzky and Golay, 1964) with a first derivative
based on third order polynomial approximation and a window size of 21 cm
. For model tuning, a 5 times repeated 10-fold
cross-validation was performed and the maximal number of PLS components
was set to 10. The model evaluation was done by predicting the hold-out data
and computing the residuals between the predicted and observed values,
the Root Mean Square Error (RMSE), R
the Ratio of Performance to Deviation (RPD).
PLSR models for total carbon, total nitrogen, pH and soil texture
Created by Laura Summerauer, ETH Zurich (2019)
Contributors: Philipp Baumann (ETH Zurich), Dr. Matti Barthel (ETH Zurich), Dr. Johan Six (ETH Zurich), Dr. Marijn Bauters (Ghent University), Dr. Kristof Van Oost (Université Catholique de Louvain), Dr. Pascal Boeckx (Ghent University), Dr. Elizabeth Kearsley (Ghent University), Dr. Eric Van Ranst (Ghent University), Dieudonné Chiragaga (IITA Kalambo), Dr. Bernard Vanlauwe (IITA), Dr. Basile Mujinya Bazirake (Université de Lubumbashi), Andrew Sila (ICRAF), Dr. Keith Shepherd (ICRAF), Nathanaël Dériaz (ETH Zurich), Simon Baumgartner (ETH Zurich), Nora Gallarotti (ETH Zurich), Oscar Vercleyen (Ghent University), Kasper Minten (Ghent University), Marco Bravin (Université Catolique de Louvain), Emilien Aldana Jague (Université Catolique de Louvain), Dr. Sebastian Dötterl (ETH Zurich) www.congo-biogeochem.com
Soil map: Jones, A., Breuning-Madsen, H., Brossard, M., Dampha, A., Deckers, J., Dewitte, O., Gallali, T., Hallett, S., Jones, R., Kilasara, M., Le Roux, P., Micheli, E., Montanarella, L., Spaargaren, O., Thiombiano, L., Van Ranst, E., Yemefack, M. , Zougmoré R., (eds.), 2013, Soil Atlas of Africa. European Commission, Publications Office of the European Union, Luxembourg. 176 pp.
1. Recommended soil processing after sampling
- Oven drying: 48 hours, 60°C
- Milling to fine powder (e.g. ball mill: frequency 30 Hz, 50 seconds)
2. Sample labelling
Unique and quickly identifiable sample_id's are crucial but often
underestimated. Do not use any spacing, capital letters or special characters
within the labels (e.g. +/()='* ). R cannot read those characters
and will therefore end in error messages. Use only '_' to separate
words or letters.
3. FT-IR spectroscopy using ALPHA Bruker spectrometer
Standard operating procedures for FT-IR spectrometer, Sustainable Agroecosystems, ETH Zurich
4. R modelingR package
simplerspecby Philipp Baumann
5. Literature, tutorials and other
simplerspecteaching material by Philipp Baumann