Methodological uncertainty in estimating carbon turnover times of soil fractions

Improving predictions of soil organic carbon (SOC) dynamics by multi-compartment models requires validation of turnover times of different SOC pools. Techniques such as laboratory incubation and isotope analysis have been adopted to estimate C turnover times, yet no studies have systematically compared these techniques and assessed the uncertainties associated with them. Here, we tested whether C turnover times of soil fractions were biased by methodology, and how this changed across soil particle sizes and ecosystems. We identified 52 studies that quantified C turnover times in different soil particles fractionated either according to aggregate size (e.g., macro- versus micro-aggregates) or according to soil texture (e.g., sand versus silt versus clay). C turnover times of these soil fractions were estimated by one of three methods: laboratory incubation (16 studies), δ¹³C shift due to C₃-C₄ vegetation change (25 studies), and ¹⁴C dating (19 studies). All methods showed that C turnover times of soil fractions generally increase with decreasing soil particle size. However, estimates of C turnover times within soil fractions differed significantly among methods, with incubation estimating the shortest turnover times and ¹⁴C the longest. The short C turnover times estimated by incubation are likely due to optimal environmental conditions for microbial decomposition existing in these studies, which is often a poor representation of field conditions. The ¹³C method can only be used when documenting a successive C₃ versus C₄ vegetation shift. C turnover times estimated by ¹⁴C were systematically higher than those estimated by ¹³C, especially for fine soil fractions (i.e., silt and clay). Overall, our findings highlight methodological uncertainties in estimating C turnover times of soil fractions, and correction factors should be explored to account for methodological bias when C turnover times estimated from different methods are used to parameterize soil C models.