Enhanced predictions of measurable soil carbon pools in the Australian rangelands with the Millennial v2 model

The Millennial model is a next-generation soil carbon (C) model that incorporates recent advances in understanding soil C dynamics, including microbial decomposition, mineral association and aggregation. Despite these advances, the model’s simulations remain uncertain. This is due to limited data on soil organic C fractions, structural complexity, and the many parameters. Recent progress has been made, but there is a notable lack of reporting on the model’s validation against empirical data. The growing availability of soil C data along with information on climate, vegetation, and soil properties, presents an opportunity to constrain the model and enhance simulations and predictions. Our objective here is to use the Millennial v2 model to simulate soil C dynamics in the Australian rangelands. We use site-specific soil physical and chemical measurements and forcing inputs such as net primary productivity (NPP), soil moisture, and soil temperature. We first employed a multiple-objective global sensitivity analysis to identify influential parameters. Then, we calibrated the model using the shuffled complex evolution (SCE-UA) algorithm to reduce parameter uncertainty. The model’s equation to estimate maximum sorption capacity was updated using a frontier line analysis and soil-type-specific capacities. Simulations were conducted across rangeland ecosystems using three calibration schemes: global (uniform parameters across rangelands), bioregional (parameters grouped by bioclimatic zone), and site-specific (parameters optimised by site). Compared to the bioregional calibration (root mean square error, RMSE = 6.15 MgC ha − 1 ; concordance correlation coefficient, ρ c = 0.61), the site-specific calibration scheme (RMSE = 2.79 MgC ha − 1 ; ρ c = 0.94) produced a 55% improvement in the predictions. The methods presented here offer a practical approach to reducing uncertainty in applications of the Millennial v2 model. This enables monitoring of potential short- and long-term changes in soil C composition across different ecosystems, with implications for C sequestration, nutrient availability, and ecosystem sensitivities to global change. • Multi-objective global sensitivity analysis for influential parameter identification. • Frontier line analysis updated maximum sorption capacity. • Three parameter optimisation schemes to capture spatial variability. • Enables precise monitoring of short- and long-term soil C composition changes.