Data-driven approaches to soil microbial modeling

Microorganisms catalyze almost all transformation processes of organic carbon in soil and are largely responsible for changes in soil carbon cycle feedback to climate change. To account for the microbial role in regulation of carbon-climate feedback, several dozens of microbial models have been developed in the past decades, mostly based on an idea that microbial biomass or microbial extracellular enzymes control decomposition of soil organic carbon (SOC). However, these idea-based models may or may not be well supported by empirical evidence. This presentation will show how data have been used to develop and test microbial models with three case studies. The first case study is to infer microbial mechanisms from observed patterns of lignin decomposition. Our study indicates that time-dependent growth and mortality of the microbial community, instead Michaelis-Menten kinetics, control microbial decomposition of lignin. The second case is to incorporate observed mechanisms into a carbon cycle model. Our meta-analysis indicates that changes in SOC under experimental warming and nitrogen addition are closely related to changes in microbial oxidative enzyme activities but not in hydrolytic enzyme activities. We directly incorporated this observed mechanism into a terrestrial ecosystem model to predict SOC changes. The third case study is to confront microbial models with nearly 58,000 vertical profiles of SOC over the globe to identify mechanisms underlying global SOC storage. Overall, scientists have developed different microbial models to explore all kind of possibilities while data offer reality. The data-model integration helps identify the most probable mechanisms under a Bayesian inference framework.