Improving the accuracy of soil carbon models using a Rock-Eval-based initialization method

In most soil organic carbon (SOC) dynamics models, SOC is divided into pools to which different mineralization rates are ascribed. The lack of a reliable, operational and fully validated method to initialize the size of the different SOC kinetic pools is a limitation for the accuracy of predictions of SOC stocks evolution provided by these models. AMG is a simple, well established French model, successfully used to simulate the evolution of C stocks for a large network of long-term monitored sites with agricultural experiments (LTEs). Initial conditions, namely the size of the stable C pool (C_S) at the onset of the simulation, have been shown to be important for the accuracy of the model. Recently, Rock-Eval 6® (RE) thermal analysis has been proposed as a new method for direct determination of SOM stability. Based on this technique, a random forests model (RE model) was developed, calibrated on Long Term Bare Fallow data, which allows the estimation of the size of the centennially persistent SOC fraction (CP_SOC) in a sample. Here, we first aimed at evaluating the performance of the RE model on fully independent soil samples. For this purpose, we compared the CP_SOC values of 73 samples from 7 LTEs calculated with the RE model with the corresponding C_S values optimized from AMG ex-post simulations. Then, we used the CP_SOC values given by the RE model to define the size of the stable C pool of the AMG model (C_S) at the onset of AMG simulations for the 7 sites. We show that the CP_SOC (RE model) and optimized C_S (ex-post AMG simulations) fractions are in good agreement (slope b=1.01, intercept a=0.04 / spearman ρ=0.88). This observation serves as a successful independent validation of the RE model. Finally, we show that the use of the RE based model improves the accuracy of the AMG model compared to default initialization (mean RMSE decreased by 13.5%), especially for sites with complex land-use history and long-term organic matter amendment. Our study therefore provides an operational method suitable to initialize the AMG model that can be expanded to other SOC dynamics models.