Differential responses of soil organic carbon fractions to warming: Results from an analysis with data assimilation

This study was aimed to assess the decomposition temperature sensitivity (Q10) of C fractions cycling from yearly through decadesʹ and up to centennial timescales using a data assimilation approach. A three-pool C-cycling model was optimally fitted with previously-published data from a 588-day long soil incubation experiment conducted at two temperatures (25 and 35 °C) for 12 soils collected from six sites arrayed across a mean annual temperature gradient from 2.0 to 25.6 °C. Three sets of key parameters of the model, which are initial C pool fractions, decomposition rates and Q10 of individual pools, were estimated with a Markov chain, Monte Carlo technique. Initial C pool fractions were well constrained with pool 1 (the most labile pool), pool 2 (more recalcitrant pool) and pool 3 (the most recalcitrant pool) accounting for 4.7% ± 2.6% (mean ± SD), 22.4% ± 16.1% and 72.9% ± 17.6%, respectively, of the total initial C pools. Mean residence time (MRT) was 0.19 ± 0.17, 2.71 ± 2.34 and 80.15 ± 61.14 years for pool 1, pool 2 and pool 3, respectively. Q10 values increased from pool 1 to pool 3 for individual soils or across all the soils. When Q10 values were plotted against MRT after the data were log-transformed, Q10 for the three pools formed three clusters and increased with MRT. Higher Q10 for decades-old C fractions implies that a major portion of soil C may become a source of atmospheric CO2 under global warming in the 21st century.