First order decay models fail to represent soil organic carbon dynamics in both single-pool and in multiple-pool models based on size fractionation

PIÑEIRO, GERVASIO; JOBBAGY, E.G.; RIZZOTTO, MARCOS; JACKSON, ROBERT B.

San Francisco, California, USA

Congreso; Fall Meeting-American Geophysical Union; 2008

American Geophysical Union

Soil organic carbon (SOC) models often rely on analytical solutions derived from litter bag experiments in which C losses follow first–order kinetics of the type (Ct= Co . e-kt). In this model, k is the decomposition or decay factor and is considered constant over time. Several authors pointed out that one-pool exponential SOC models fail to reproduce long-term SOC dynamics, and that multiple pools or time-dependent k values are needed to capture SOC behavior given its heterogeneous composition and turnover. As a consequence, models that separate organic matter in multiple pools with their own specific k values, expected to remain constant through time, have been proposed. Whether these multiple pool models with constant k values through time are able to represent SOC trajectories is unknown. We explored the stability of k values through time for total SOC and its more labile fraction known as particulate organic carbon (POC) across 15 grasslands sites in which shifts in the stable isotopic composition of C inputs after grazing exclosure (change from C4 to C3 dominated vegetation) allowed us to trace SOC dynamics. Time elapsed after grazing exclusion ranged from 4 to 30 years. We sampled soils and roots in grazed-ungrazed paired plots and measured SOC, POC and 13C composition.  Using 13C and SOC changes we estimated k and carbon mean residence times (MRT) after grazing exclusion. As expected k (mass of annual C losses per unit of mass of C storage) values for total SOC were variable between sites but always lower (0.03 to 0.11) than k values of the POC fraction (0.09 to 0.19), and the opposite was true for MRT. For the total SOC, k values varied with time elapsed since grazing removal (r2= 0.71), suggesting that a variable k through time is needed to reproduce SOC dynamics with a one-pool model. Surprisingly, k values for the POC fraction were also significantly correlated with time (r2= 0.75), suggesting that even the POC fraction can not be accurately modeled with a first order kinetic model with a constant k  that is successfully applied to litter bag decomposition experiments. Our results suggest that the widely used POC fraction does not separate a sufficiently homogenous pool with constant k values, or either that the exponential decay model does not apply for SOC dynamics.