Large differences in estimates of soil organic carbon turnover in density fractions by using single and repeated radiocarbon inventories

Radiocarbon measurements are frequently used to model the turnover of soil organic carbon (OC) fractions. The assumption of homogeneous turnover in these fractions is typically stated, but consequences of its violation have not been tested. We used samples of a repeated soil inventory of a German beech forest from the litter layer to 50 cm depth in 2004 and 2009 to determine the suitability of short-term repeated radiocarbon inventories for estimating the turnover times of OC in soil fractions. Samples from 0–5 and 10–20 cm soil depth were density separated into a free light fraction (fLF), an occluded light fraction (oLF), and a heavy, mineral-associated fraction (HF). Samples were analysed for radiocarbon (∆14C), OC, and total nitrogen (TN) contents. Similar portions of OC were stored in the HF, but contributions of fLF and oLF varied between study years, probably due to interannual variations or methodological constraints. Following declining atmospheric CO2–14C, also ∆14C values at 0–5 cm depth declined significantly between 2004 and 2009. Exchange of old for new OC was largest in the fLF and smallest in the HF, which confirms slow turnover of OC associated with minerals. Model results revealed that turnover time estimates based on single-pool models were not in agreement with observed changes in any of the fractions, suggesting all of them to be mixtures, to varying degrees, of fast and slow cycling pools. While single-pool models suggest average turnover times of 115 years for HF-OC at 0–5 cm depth, thus being a stable fraction, fitting a two-pool model to the two-point measurements of radiocarbon suggested the presence of a fast cycling pool of 15–25 years of turnover time. It was however only possible to constrain the portion of this fast pool as being between 50 and 85% of total HF-OC. Increasing ∆14C in bulk soil and density fractions between study years at 10–20 cm depth suggest that OC enters deeper soil layers with a longer time lag than topsoil layers, e.g., by slow transport, and 5 years was not enough to induce significant changes. Even 40 years after the bomb peak, radiocarbon time series are still suitable to detect OC fractions of decadal turnover and hint at time lags and translocation processes. Nevertheless, they do not allow for fixing the portions of fast and slow cycling OC in two-pool models and their turnover times