Adaptability of wheat model algorithms integration platform in North China Plain

Wheat model algorithms integration platform (WMAIP) has been integrated various algorithms from the different modules, including CERES-Wheat, APSIM-Wheat, WOFOST, SWAT, AquaCrop, and WheatSM, in order to simulate the soil water stress and winter wheat growth and development indices. The WMAIP can be potential to the prospects of application in the wheat model algorithm comparison and improvement, integrated simulation, and climate change impact assessment. The WMAIP verification and algorithm comparison have been carried out to focus mainly on a single site so far. The platform can be expected to guide the water management of local winter wheat. However, the current WMAIP cannot evaluate in the regional multi-point production of winter wheat. Therefore, it is a high demand to evaluate the adaptability of WMAIP in North China Plain (NCP). In this study, two algorithms were selected from the different modules in WMAIP. The phenology algorithm included the clock model of WheatSM module and the thermal time of APSIM-Wheat module, whereas, the biomass algorithm contained the carbon assimilation of WOFOST and canopy photosynthesis of WheatSM model, and the water stress algorithm consisted of the actual to potential transpiration ratio of CERES-Wheat model and the water supply to demand ratio of APSIM-Wheat model. 16 combined models were formed to combine the different algorithms from different modules using WMAIP. These models were verified using experimental data from four typical NCP experimental stations for several years. The optimal model was selected using the normalized root mean squared error (NRMSE). Finally, the adaptability of WMAIP integrated model in the NCP was evaluated to simulate the soil water dynamics, as well as the growth and development indicators of winter wheat. The results showed that all 16 models using the WMAIP platform were effectively simulated these parameters, with a simulation error of the development period below 4.2% and the soil water storage in the 2 m soil layer below 7.0%. The simulation errors of biomass and yield were ranged from 17.3% to 23.7% and 10.8% to 20.8%, respectively. However, the simulation performance of a single model was unstable, and there were differences between the calibration and validation for the optimal model. Nevertheless, the model integration was significantly reduced the simulation error of winter wheat yield in the NCP. Meanwhile, the simulation errors were similar to the field experiments using the integration of six models. Overall, the WMAIP ensemble model was able to simulate the response of winter wheat yield to the sowing date and irrigation treatment in the NCP. Among them, the NRMSE of biomass and yield values were 18.7% and 11.8%, respectively. The multi-year experimental data was effectively evaluated the WMAIP simulation on the winter wheat growth and development in different years, sowing dates, and irrigation treatments in the North China Plain. The excellent adaptability was achieved in the NCP for the wheat production management and climate change impact assessment.