Computing the mean residence time of soil carbon fractions using stable isotopes: impacts of the model framework

Soils contain the largest carbon (C) reservoir on Earth, but the mean residence time (MRT) of soil C is often poorly estimated, despite its importance for assessing the efficiency with which soils may serve as a sink for atmospheric C. The objective of this study was to evaluate how the structure of simple models of soil C dynamics affects the MRT determined from isotope‐mixing experiments, using data from field studies with either artificial labelling (FACE) or C3/C4 vegetation change. We first highlighted theoretically how non‐steady‐state conditions and the model structure (one single, two successive, or two parallel C pools) can have an impact on the MRT assessment. We then tested these different model structures against published data on the dynamics of different soil organic matter separates and their constituents. Our findings indicated that many of the reviewed studies assumed wrongly that the system was at steady state or could be described by a single‐pool approach. To select the correct model, exact knowledge of C input rates and several data points are needed from the beginning of the experiment. For steady‐state conditions an apparent temporal change of MRT computed from a single‐pool model is thus a clear indicator that a two‐pool approach must be chosen. The errors made by the wrong choice of model varied with the length of the experiment and usually resulted in an over‐estimate of MRT by a factor of 1.15 for some data published on physical size separates, but by a factor of up to 11 for individual microbial biomarkers such as muramic acid.