The Global Turnover Time Distribution of Soil Carbon Derived from a Meta-analysis of Radiocarbon Profiles

Monday, 14 December 2015: 15:25
2004 (Moscone West)
Yujie He1, James Tremper Randerson1, Steven D Allison2, Margaret S Torn3, Jennifer W Harden4, Lydia J Smith5, Tessa van der Voort6 and Susan Trumbore7, (1)University of California Irvine, Department of Earth System Science, Irvine, CA, United States, (2)University of California Irvine, Irvine, CA, United States, (3)Lawrence Berkeley National Laboratory, Berkeley, CA, United States, (4)USGS Geological Survey, Menlo Park, CA, United States, (5)University of California Berkeley, Berkeley, CA, United States, (6)ETH Zurich, Zurich, Switzerland, (7)Max Planck Institute for Biogeochemistry, Jena, Germany
Soil is the largest terrestrial carbon reservoir and may influence the sign and magnitude of carbon cycle feedbacks under climate change. Soil carbon turnover times provide information about the sensitivity of carbon pools to changes in inputs and warming. The spatial and vertical distribution of soil carbon turnover times emerges from the interplay between climate, vegetation, and soil properties. Radiocarbon levels of soil organic matter can be used to estimate soil carbon turnover using models that take into account radioactive decay over centuries to millennia and inputs of 14C from atmospheric weapons testing (“bomb carbon”) during the second half of the 20th century. By synthesizing more than 200 soil radiocarbon profiles from all major biomes and soil orders, we 1) explored the major controlling factors for soil carbon turnover times of surface and deeper soil layers; 2) developed predictive models (tree-based regression, support vector regression and linear regression models) of ∆14C that depends on depth, climate, vegetation, and soil types; and 3) extrapolated the predictive model to produce the first global distribution of soil carbon turnover times to the depth of 1m.

Preliminary results indicated that climate and depth were primary controls of the vertical distribution of ∆14C, contributing to about 70% of the variability in our model. Vegetation and soil order exerted similar level of controls (about 15% each). The predictive model performed reasonably well with an R2 of 0.81 and RMSE (root-mean-squared error) of about 50‰ for topsoil and 100‰ for subsoil, as estimated using cross-validation. Extrapolation of the predictive model to the globe in combination with existing soil carbon information (e.g., Harmonized World Soil Database) indicated that more than half of the global total soil carbon in the top 1m had a turnover time of less than 500 years. Subsoils (30-100cm) had millennium-scale turnover times, with the majority (70%) turning over between 1000 and 5000 years. This study provides a data-constrained estimate of the global distribution of soil carbon turnover times and may help to constrain the performance of Earth system models used to evaluate future scenarios of change.