Towards a farm-scale digital twin for carbon farming: regenerative management scenarios and decision support for a German arable farm

Carbon farming aims to sequester soil organic carbon (SOC) in agroecosystems by increasing soil organic matter content to improve soil health, while contributing to reducing greenhouse gas emissions. In this context, a promising approach is the use of agroecosystem models that can inform and optimize farmer’s practices but also offer a holistic perspective to enhance both agronomic and environmental outcomes while accounting for climate change. In this study, we developed a farm-scale digital twin of 14 fields with a combined area of 80.4 ha for an arable farm in western Germany. The twin uses a spatialized version of the process-based agroecosystem model AgroC, driven by detailed soil, climate, and management data since 2009.

In its current version, the digital twin provides a reconstruction of past SOC dynamics and can be used to explore future management scenarios focused on regenerative agricultural practices, such as cover cropping, harvest residue management, and the application of organic amendments. It can thus provide decision support for optimizing carbon sequestration at the field-to-farm scale. The analysis of simulated SOC development for the past 16 years showed that simulated SOC stocks increased by 30.9 Mg C y^-1 in the period 2009-2025. This is equivalent to a farm sequestration rate of 0.4 Mg C ha^-1 y^-1, corresponding to a relative gain of 6.5‰ y^-1. To attribute the SOC sequestration achieved by regenerative management to individual practices, we compared SOC trajectories simulated for a system without regenerative management, the farm’s actual management, and each practice considered separately. Across all regenerative practices, cover cropping accounted for 4.2% of the additionally sequestered SOC, harvest residue retention for 69.7%, and organic fertilizer applications for 15.1%, with the remaining 11% attributable to interaction effects among these practices. For a moderate future climate scenario (RCP 4.5 ensemble), an additional carbon sequestration potential of 7.5±0.7 Mg C ha^-1 by 2050 is predicted for a representative field under present management. Sequestration rates were found to slow after around 2028–2032 and to stagnate thereafter.

Beyond these results for current management, the digital twin provides a decision-support environment in which farmers can explore future SOC management options, such as effects of new cover crops, altered crop rotation sequences, and improved cultivars. The available options can be compared for their long-term SOC sequestration potential, and the farmer can select the strategies that fit the farm-specific objectives and constraints best. When new observations on SOC, biomass, and yield become available, the digital twin can be updated and used to reevaluate the scenarios.

In conclusion, the digital twin for carbon farming introduced here is a promising tool to identify locally adapted, farm-specific management strategies that further increase and sustain SOC and to evaluate their robustness under current and future climate conditions. These management changes may also have implications for other ecosystem services, including nutrient leaching and agricultural productivity. Our process-based approach already represents these processes, and in future work we aim to extend the digital twin to explore trade-offs and synergies between a range of ecosystem services beyond carbon sequestration.