A Comparison of Linear Conventional and Nonlinear Microbial Models for Simulating Pulse Dynamics of Soil Heterotrophic Respiration in a Semi-Arid Grassland

As a critical process in regulating terrestrial feedback to climate change, soil heterotrophic respiration is commonly simulated with the first-order kinetics in current Earth system models. Compared with the first-order kinetic models, explicit microbial models are expected to better simulate nonlinear carbon (C)-cycle phenomena, such as the pulse dynamics of soil heterotrophic respiration driven by dry-rewetting cycles in grasslands. However, these two types of models (i.e., linear conventional and nonlinear microbial models) have never been compared based on in situ observations of soil heterotrophic respiration in the semi-arid grassland, which is significantly affected by the dry-rewetting events. Here, based on the field data of soil heterotrophic respiration in a semi-arid grassland in northern China, we first showed that the shift from a conventional linear model to a nonlinear microbial model did not substantially improve the simulation of soil heterotrophic respiration. Then, we quantified the contributions of different moisture-response functions to the uncertainty in simulating the soil C dynamics. The results showed that the selection of moisture-response functions combined with parameterization in the soil C models dominated the modeled uncertainties in soil heterotrophic respiration. These findings suggest that both the conventional linear model and nonlinear microbial model can simulate well the pulse dynamic of soil heterotrophic respiration in grasslands with an improved parameterization of water regulation on soil carbon decomposition. This study also calls for more observations of nonlinear C phenomena for reducing the simulation uncertainty on soil C cycling in Earth system models.