Assessments of changes in soil organic carbon (SOC) stocks depend heavily on reliable values of SOC content obtained by automated high-temperature C analysers. However, historical as well as current research often relies on indirect SOC estimates such as loss-on-ignition (LOI). In this study, we revisit the conversion of LOI to SOC using soil from two long-term agricultural field experiments and one arable field with different contents of SOC, clay and particles <20 μm (Fines20). Clay-, silt- and sand-sized fractions were isolated from the arable soil. Samples were analysed for texture, LOI (500°C for 4 hours) and SOC by dry combustion. For a topsoil with 2 g C and 30 g clay 100 g-1 soil, converting LOI to SOC by the conventional factor 0.58 overestimated the SOC stock by 45 Mg C ha-1. The error increased with increasing contents of clay and Fines20. Converting LOI to SOC by a regression model underestimated the SOC stock by 5 Mg C ha-1 at small clay and Fines20 contents and overestimated the SOC stock by 8 Mg C ha-1 at large contents. This was due to losses of structural water from clay minerals. The best model to convert LOI to SOC incorporated clay content. Evaluating this model against an independent dataset gave a root mean square error and mean error of 0.295 and 0.125 g C 100 g-1, respectively. To avoid misleading accounts of SOC stocks in agricultural soils, we recommend re-analysis of archived soil samples for SOC using high-temperature dry combustion methods. Where archived samples are not available, accounting for clay content improves conversion of LOI to SOC considerably. The use of the conventional conversion factor 0.58 is antiquated and provides misleading estimates of SOC stocks.
Highlights: Assessment of SOC contents is often based on less accurate methods such as LOI.Reliable accounts of changes in SOC stocks remain high on the agenda (4‰ initiative).Conversion of LOI to SOC is considerably improved by accounting for clay content.Converting LOI to SOC by the conventional factor 0.58 leads to grossly overestimated SOC stocks.