Understanding the response of soil organic carbon (SOC) to environmental and management factors is necessary for estimating the potential of soils to sequester atmospheric carbon. Changes over time in the amount and distribution of SOC fractions with different turnover rates can be estimated by means of soil SOC models such as RothC, which typically consider two to five SOC pools. Ideally, these pools should correspond to measurable SOC fractions. The aim of this study was to test the relationship between SOC pools used in RothC and fractions separated through a fractionation procedure. A total of 123 topsoil samples from agricultural sites (arable land, grassland and alpine pasture) across Switzerland were used. A combination of physical and chemical methods resulted in two sensitive (particulate organic matter and dissolved organic carbon), two slow (carbon associated to clay and silt or stabilized in aggregates) and one passive (oxidation-resistant carbon) SOM fractions. These fractions were compared with the estimated equilibrium model pools when the corresponding soils were modelled with RothC. Analysis revealed strong correlations between SOC in measured fractions and modelled pools. Spearman’s rank correlation coefficients varied between 0.82 for decomposable plant materials (DPM), 0.76 for resistant plant materials (RPM), 0.99 for humified organic matter (HUM) and biomass (BIO), and 0.73 for inert organic matter (IOM). The results show that the proposed fractionation procedure can be used with minor adaptations to identify measurable SOC fractions, which can be used to initialize and evaluate RothC for a wide range of site conditions.