A Unified Quantitative Framework for Assessing Uncertainty in Microbial-Explicit Soil Carbon Models under Diverse Data Availability of Field Observations

Abstract

Abstract Microbial-explicit soil carbon models offer mechanistic representations of soil carbon turnover, but their application to the field remains limited, which is commonly attributed to the scarce microbial observations. However, usually there are significant mismatches between model data requirements and the diverse data availability of field observations, which may greatly affect the uncertainty of soil carbon models and have been largely overlooked. Here, we developed a unified framework that systematically quantified how the field observation type, sampling frequency, and measurement uncertainty jointly constrained model parameters and long-term predictions of three representative microbial-explicit soil carbon models. We quantified how well observations of soil carbon stocks, soil respiration, microbial biomass carbon, and functional gene abundance can reduce the parameter and prediction uncertainties across a comprehensive set of field scenarios. High-frequency, low-error soil respiration observations, when combined with soil carbon measurements, dominated uncertainty reduction across carbon pools, while interestingly, microbial observations only provided conditional improvements for microbial and enzyme-related states. More microbial data did not necessarily yield more effective calibration information, and improvements in prediction accuracy required microbial observations with sufficiently small measurement errors. The unified quantitative framework would also help to design a data strategy when developing new soil carbon models to accurately predict soil carbon cycling.