Harmonizing data from different soil organic matter fractionation protocols

• Different physical fractionation methods yield different results. • These differences depend on soil characteristics. • Transfer functions are proposed to harmonize data obtained using different size fractionation methods. • Results obtained by size versus density fractionation cannot be harmonized. • Size and density fractionation results should not be combined in meta -analysis. Physical fractionation is a prevailing tool for studying the biogeochemical stability of soil organic carbon (SOC). Fractionation methods are straightforward to implement in the laboratory, and allow for the separation of SOC into at least two distinct fractions, either by size (OC fine vs. OC coarse ) or density (light vs. heavy fraction), and make it possible to distinguish fractions with contrasting average biogeochemical stability. Methods tend to differ in the protocol for soil dispersion (shaking in water and/or sodium hexametaphosphate (SHMP), shaking with glass beads, or using ultrasonic dispersion), and in the fraction separation (through different sieve/mesh sizes or using chemical density). Several meta -analyses combining results obtained using different physical fractionation methods are emerging. In this context, it is important to determine whether the results obtained using different methods are directly comparable or, at a minimum, whether correction functions can be developed to harmonize the results obtained with different methods. In this study, we compared the results obtained by fractionating 61 wide-ranging soil samples using five methods: dispersion in water with glass beads and sieving at 50 µm (M1) or 63 µm (M2); dispersion in SHMP with glass beads and sieving at 50 µm (M3); ultrasonic dispersion (USD) in water and sieving at 50 µm (M4); and USD in sodium polytungstate and separation by density at 1.6 g cm −3 (M5). Our results show that the five methods provide significantly different results and that these differences depend on common soil characteristics: pH, SOC, and texture. For instance, SHMP is the most effective dispersion method for acidic soils, but less effective in alkaline soils. We successfully developed functions that harmonize the data obtained using the different size fractionation methods (M1 to M4). However, results obtained by density fractionation are significantly different from those obtained by size fractionation and therefore cannot be harmonized effectively at this stage. Nevertheless, when excluding samples with high sand content (>70%), the difference between size and density fractionation measurements remains relatively high, but falls within a more acceptable range. Ultimately, our work shows that combining fractionation results is not a straightforward task and several considerations should be kept in mind. When combining density- and size-based results, sandy soils need to be excluded from the datasets, allowing for a significant reduction in the bias between the two measurements. Furthermore, prior to performing any (meta-)analyses, we therefore strongly recommend the harmonization of results obtained using different size-based fractionation methods, and we urge caution when interpreting results from meta -analyses that combine different fractionation methods.