Soil organic matter content (SOM) has an important role in modulating atmospheric CO2, while soil organic carbon (SOC) is the best marker of SOM. Soil components are interdependent, thus they carry only relative information, belonging to the type of data so-called 'compositional data' (CoDa). VisNIRS modelling with soil reference data needs therefore specific CoDa methods. Soil moisture (SM) changes supposes an important difficulty to VisNIRS measurement of SOC. The CoDa methods integrate SOC prediction along with the other soil parts considered, including SM. Here are presented the final results from different soil compositional approaches carried out within ProbeField Project https://ejpsoil.eu/soil-research/probefield, (EJP-Soil programme), focused on proximal sensing techniques for SOC measurement. Different approaches for 6, 5, 4, 3, and 2 soil parts were assessed. The soil components considered were SOM, SM, soil inorganic carbon (SIOC), the textural fractions clay (C) and silt (S), and the sand content which was considered in all the approaches together the rest of soil mass (Other). The results provided correlation coefficients between SOM predictions and the corresponding reference values oscillating from r = 0.7s with the traditional PLS calibration to predict SOM to r = 0.87 with the 4-Part 'Clay' CoDa approach, which was the best . Therefore, this method to estimate SOM could be satisfactory. More research is ongoing to verify this approach within the frame of the ProbeField Project.