Improve maize simulation capacity with reduced calibration by assimilation of soil moisture and canopy cover data

Abstract:

Crop modelling is a powerful tool in crop growth simulation, yield prediction and field management. Calibration is normally required prior to model simulation to fit different observation data, such as crop phenology and yield. Despite the importance of model calibration, it is data-demanding and labour-intensive, and is also dependent on practitioner’s expertise as well as data quality. In this study, a data assimilation framework was proposed to improve AquaCrop maize simulation performance with reduced calibration workload. The hybrid-/cultivar-specific phenological parameters were uniformly scaled from a previous calibration study for a different maize hybrid in another location, as different hybrids/cultivars of the same crop species often demonstrate similarities in the growing stages. Other key model parameter values were taken from model defaults. To reduce the simulation uncertainties, soil moisture and/or canopy cover observations were assimilated into the model using the ensemble Kalman filter. The methodology was tested for multiple years in a field in Nebraska. Results suggested that data assimilation achieved an accuracy similar to that with traditional parameter calibration, and that the joint soil moisture and canopy cover assimilation outperformed single-variable assimilation. On this basis, more sophisticated state-parameter update assimilation strategies were tested. The final goal is to extend this methodology to large-scale applications using remote sensing data.