Can SOC modelling be improved by accounting for pedogenesis?

Recent findings suggest that soil organic carbon mineralization and stabilization depend to a substantial degree on the soil geochemistry and the degree of weathering. We hypothesized that this dependence can be translated to decay rate modifiers in a model context, and used data from the Merced chronosequence (CA, U.S.A., 100 yr-3 Myr), representing a weathering sequence, to test, on a 1000-year time scale for model spin-up, a simple soil organic carbon (SOC) model based on the RothC26.3 model concepts. Model performance was tested for four levels of information: (1) known decay rates for each model SOC pool at individual chronosequence locations, obtained by calibrating the model to measured SOC-fractions and measured site-specific C-inputs; (2) average decay rates for each SOC-pool, corrected per location with rate modifiers based on geochemical proxies and measured site-specific C-inputs; (3) uncorrected average decay rates per SOC-pool and measured site-specific C-inputs; (4) uncorrected average decay rates per SOC-pool and averaged C-inputs. A lumped root mean square error (RMSE) statistic was calculated for each information level. We found that using local measurements of fresh C-input led to a decrease in RMSE of near 15% relative to information level (4). Applying geochemical rate modifiers led to a further reduction of 20%. Thus, we conclude that there is a benefit of including geochemical rate modifiers in this SOC-model. We repeated this analysis for a five-pool and a four-pool SOC model that either included or excluded an inert organic matter pool. In terms of the lumped RMSE both models performed similarly, but by comparing measured and simulated percentage Modern Carbon (pMC) for bulk SOC we concluded that measured pMC was best approximated using a four-pool SOC model (without an Inert Organic Matter pool). Furthermore, it is likely that a five-pool model including a very slowly decaying pool would further improve model performance.