Optimising nitrogen management for climate-smart agriculture: A hybrid modelling approach in wheat-maize rotations

Optimising nitrogen management has the potential to enhance crop productivity while mitigating greenhouse gas emissions. Nevertheless, it has low adoption rates, due to the complex interactions of crop types, environments (climate and soil) and management combinations, posing significant challenges to advancing climate-smart agriculture. In this study, a hybrid modelling approach was developed to target a minimum of 90% of the potential yield, while simultaneously increasing nitrogen use efficiency and optimising N inputs, reducing net GHG emissions and GHG intensity. A 30-year field trial was conducted on a wheat-maize rotation system in the North China Plain. The observations (annual yields, SOC and N₂O emissions) were then used to validate the process-based DNDC model, and the NSGA-Ⅲ machine learning algorithm was applied for multi-objective optimisation. This hybrid modelling approach simulated and optimised three levels of nitrogen management under future climate scenarios (level 1: fertilizer rates; level 2: fertiliser rates, timing, frequency, and crop schedules; level 3: level 2 plus irrigation and residue retention). From 1990 to 2100, the optimised practice combinations were identified: delaying and reducing basal fertilization (+5 d, -52.8 kg N ha^-1) while advancing top-dressing in wheat (-5 d) and both events in maize (-9 d, -3 d); postponing wheat sowing (+5 d) and advancing maize sowing (-9 d); aligning irrigation event with fertilization, and adding one irrigation event during the maize bell stage; and lowering residue retention (-0.2). Integrating additional practices with fertiliser rates (levels 2 and 3) proves effective in meeting these climate-smart objectives. Under SSP245 and SSP585, the optimal level 3 practices, compared to maintaining current practices unchanged (conventional practices), increase annual crop yields by 5.6% and 1.7%, respectively, while concurrently reducing net GHG emissions by 9.4% and 8.4%, respectively. Optimal level 3 practices, in comparison to level 2, increased yields by only 0.7%, but significantly reduced net GHG emissions by 8.7%. Furthermore, the implementation of optimal level 3 practices, compared to conventional practices, led to a reduction in N inputs, irrigation water use and residue inputs by 17.2%, 6.7% and 20.0%, respectively. The findings of this study demonstrate that the optimal practices continually adapted in order to respond to the changing climate conditions. It is imperative for decision-makers to consider the trade-off between achieving greater GHG reductions and the potentially higher implementation costs associated with adjusting practices, given the minimal yield differences but significant GHG emission disparities across levels.