D3.2_20240117_ProbeField_Preprosessed_Spectra_V1

Deliverable 3.2 consists of a dataset of pre-processed field spectra. The dataset includes 583 spectra collected by nine EJP Soil Probefield’s partners in eight countries in 214 sampling points. Nine different VNIR spectroradiometers equipped with contact probes were used to collect the spectra (350 – 2500 nm). Each spectrum corresponds to the average reflectance of five repetitions .

Soil samples were collected from each sampling point and brought to the laboratory for determining the organic content (TOC). The raw spectra were reduced in the range between 400 and 2450 to remove the spectrum tails usually containing noisy signal and the following spectral pre-treatments and transformation were applied:

1.      Spline interpolation of the values of the edges of the middle sensor to correct steps

2.      Savitzky-Golay smoothing filter using a second polynomial order and a window of 11 nm

3.      Detrending spectral data by applying a SNV transformation

4.      Spectral alignment applying a different correction factor for each spectroradiometer using the Lucky Bay sands as internal soil standard

5.   The spectra were further transformed using the External Parameter Orthogonalization (EPO) matrix obtained from DT_EPO.csv to remove the effects of soil moisture

Therefore, three different datasets were extracted:

a      D3.2_20240117_ProbeField_Preprosessed_Spectra_DT_raw.csv: data extracted at the end of point 3

b     D3.2_20240117_ProbeField_Preprosessed_Spectra_DT_ISS.csv: data extracted at the end of point 4

c      D3.2_20240117_ProbeField_Preprosessed_Spectra_DT_EPO.csv: data extracted at the end of point 5