State and parameter estimation of the AquaCrop model for winter wheat using sensitivity informed particle filter

• A Bayesian calibration method is introduced for state and parameter estimation. Parameter sensitivity information embedded into particle filter. Multi-spectral UAV images are to derive canopy cover by machine learning. Three filtering methods compared with different levels of prior information. Both MC simulations real-world experiments validate the performance. Crop models play a paramount role in providing quantitative on crop growth field management. However, its prediction performance degrades significantly presence unknown, uncertain parameters noisy measurements. Consequently, simultaneous estimation (SSPE) model required maximize potentials. This work aims develop an integrated dynamic SSPE framework AquaCrop leveraging constrained filter, analysis remote sensing. Monte Carlo simulation one winter wheat experimental case study performed proposed framework. It shown that: (i) state/parameter bound outperforms conventional filter both simulations; (ii) experiment, approach achieves smallest root mean squared error among three algorithms using day forward-chaining validation method.