Data requirements for automated model-based control of irrigation and fertiliser application

Model-based adaptive control strategies can be used to determine site-specific irrigation and fertiliser volumes with the aim of maximising crop water use efficiencies and/or yield. These strategies use a crop model to predict the crop's response to climate and management throughout the crop season, and identify which irrigation and fertiliser application volume and timing produces the desired crop response or condition (e.g. maximum yield). The model can be calibrated with infield weather, soil and crop measurements to ensure the model predictions accurately reflect infield measurements. However, data collection of soil and plant parameters spatially over a field and throughout the crop season will potentially lead to a large sensed data requirement which may be impractical in a field implementation. In addition, not all the data may be required to calibrate the crop model with sufficient accuracy. A smaller dataset consisting of only the most influential sensor variables may be sufficient for adaptive control purposes. This paper reports on a simulation study which evaluated the relative significance of weather, soil and plant variables (either individually or in combination) for calibration of the APSIM french bean crop production model. This involved comparing the outputs of a crop with variations in evaporative demand, soil moisture, leaf area, canopy size and/or thermal time for growth stages. The most significant parameters for the simulated pod size, height, soil moisture and total nitrogen were the thermal times for each growth stage