The contribution of old soil C (SOM) to total soil respiration (R_S) in forest has been a crucial topic in global change research, but remains uncertain. Isotopic methods, such as natural variations in carbon isotope composition (δ¹³C) of soil respiration, are more frequently being applied, and show promise in separating heterotrophic and autotrophic contributions to R_S. However, natural and artificial modification of δ¹³C_Rs can cause isotopic disequilibria in the soil-atmosphere system generating a mismatch between what is usually measured and what process-based models will predict. Here we report the partitioning of the soil surface CO₂flux in a warm Mediterranean forest into components derived from root, litter/humus, and SOM sources using a new, three end-member mixing model, and compare this with the conventional partitioning into autotrophic and heterotrophic components. The three end-member mixing model takes into account both the discrimination during CO₂ respiration/decomposition of the three components, as well as the fractions of their CO₂ fluxes integrated over the total soil profile mass. In addition, we used a novel dual-chamber technique to ensure that δ¹³C_Rs was subjected to minimal artefacts during measurement.

We observed that by using measured soil surface CO₂ concentrations as a baseline level for the dual-chamber operation, it was possible to achieve and monitor the necessary conservation of the soil CO₂steady-state diffusion conditions during the measurements, without using permanent collars inserted deeply into the soil. When R_S (8.64 g CO₂ m² d⁻¹) was partitioned into two components, the mean autotrophic and heterotrophic respiration was 56 and 44%, respectively. When R_S was partitioned using the three-way model, however, roots, litter/humus, and SOM contributed 30, 33, and 37% of the total flux. Our results confirm that to improve the estimates of the partitioning method, it is important to distinguish the fractional contribution of the long-term SOM-derived flux from younger and more labile sources.