Simulation of Site-specific Irrigation Control Strategies with Sparse Input Data

Crop and irrigation water use efficiencies may be improved by managing irrigation application timing and volumes using physical and agronomic principles. However, the crop water requirement may be spatially variable due to different soil properties and genetic variations in the crop across the field. Adaptive control strategies can be used to locally control water applications in response to in-field temporal and spatial variability with the aim of maximising both crop development and water use efficiency. A simulation framework ‘VARIwise’ has been created to aid the development, evaluation and management of spatially and temporally varied adaptive irrigation control strategies (McCarthy et al., 2010). VARIwise enables alternative control strategies to be simulated with different crop and environmental conditions and at a range of spatial resolutions. An iterative learning controller and model predictive controller have been implemented in VARIwise to improve the irrigation of cotton. The iterative learning control strategy involves using the soil moisture response to the previous irrigation volume to adjust the applied irrigation volume applied at the next irrigation event. For field implementation this controller has low data requirements as only soil moisture data is required after each irrigation event. In contrast, a model predictive controller has high data requirements as measured soil and plant data are required at a high spatial resolution in a field implementation. Model predictive control involves using a calibrated model to determine the irrigation application and/or timing which results in the highest predicted yield or water use efficiency. The implementation of these strategies is described and a case study is presented to demonstrate the operation of the strategies with various levels of data availability. It is concluded that in situations of sparse data, the iterative learning controller performs significantly better than a model predictive controller.