INSTRUCTION FOR THE CHALLENGE 2024 -

For the 2024 challenge, only 1 parameter has to be predicted. The calibration data set is a public data set of 19 036 spectra of soil samples collected in whole Europe. The data set is usually called the LUCAS dataset. 

The dried and ground samples were scanned on a near infrared spectrometer from 400 to 2498 nm.

The Y parameter to be calibrated is the content of Total Organic Carbon (TOC) expressed in % of dried soil.

The 2 test sets are soil samples scanned on two other instruments.  These 2 instruments are NOT represented in the CAL set.

The goal is to predict accurately the Y values (only COT) of the 2 validation sets.

The criterion to classify the results will be the RMSEP of the 2 merged sets. 

The participants are invited to send their results (1 solution and 2 vectors of 226 and 229 values in TXT or Excel format) to Pierre Dardenne at dardennepaj@gmail.com  by the Wednesday 21th of February, midnight. The participants will join a short PPT or DOC file with their methodology. The participants with the best results will be asked to present their results during the meeting in 5-10 minutes depending of the number of answers. Good luck. P. Dardenne Tel : +32 478 92 15 95.

We thank, from CRAW-Gembloux, Vincent Baeten for providing the validation data and Louis Paternostre for the data setting up.  

Files to be downloaded in the zip file: HERE

Instructions for the challenge 2024.docx

LUCAS_SOIL_corr_desc.xlsx  19036 sample description

LUCAS_SOIL_corr_spc.CSV   19036 spectra

LUCAS_SOIL_corr_spc.xlsx           ‘’

lucas_SOIL_SPC.mat                      ‘’

LUCAS_SOIL_Y_OC.xlsx         19036 Y values , parameter to be calibrated

VAL_CRA-SET_1.xlsx               validation set 1 (226 spectra)

VAL_CRA-SET_2.xlsx               validation set 2 (229 spectra)