Improvement of the RothC model for Andosol paddy fields based on long-term Japanese monitoring data

Assessing the carbon sequestration potential of agricultural soils with accuracy is essential for global food security and climate change mitigation. We aimed to improve the Rothamsted Carbon (RothC) model for application in Andosol paddy fields by integrating two RothC-based models previously developed in Japan: one tailored for paddy fields and one for Andosol upland fields. The core refinement involved adjusting the decomposition rate of the humified organic matter (HUM) pool within the paddy model using the phosphate adsorption coefficient (PAC). Validation was performed using long-term continuous field data (11–48 years) from five Andosol paddy sites throughout Japan, encompassing control plots (chemical fertilizer only) and plots with varying organic matter amendments. Comparing simulated and observed soil organic carbon (SOC) stocks (to 30 cm), the performance of the improved Andosol-Paddy model showed substantial gains relative to the paddy-specific RothC model. The improved model yielded an average root mean square error (RMSE) of 9.99 and an average absolute mean error (ME) of 7.49, with values < 50% of the original model error (RMSE: 20.32; absolute ME: 16.69). This evidence clearly demonstrates that incorporating PAC-based HUM decomposition into the paddy-specific RothC model significantly elevates the predictive accuracy of SOC dynamics in Andosol paddy fields.