Furrow irrigation is one of the oldest techniques of surface irrigation and is the most popular method for the irrigation of row crops. In Australia the method is
widely used for the irrigation of cotton and in some years it has accounted for about 95% of the total production. The furrow system is however often associated with high labour requirement and low water used e�ciency. In furrow irrigation
the soil is used both as a medium for infiltration and also for conveyance of water from one end of field to the other. However, the spatial and temporal soil infiltration variability causes non-uniformity in water absorption rates and furrow stream advance rates. This significantly reduces water use e�ciency because current design and management practices do not take this variability into account.
Most operations in the furrow system are undertaken manually, and are hence labour-intensive.
Real time optimisation and control of furrow irrigation has been proposed for the management of infiltration variability. The system estimates the soil infiltration
characteristics in real time and uses the data to control the same irrigation, potentially leading to improvement of water use e�ciency. The major goal of this research was therefore to develop, prove and demonstrate an automated system for real time optimisation of furrow irrigation. The hypotheses of the research were that: (i) use of real time optimisation and control in furrow irrigation can lead to signi�cant improvement in irrigation performance, and (ii) automation of furrow irrigation is feasible.
The system developed in this research is an integration of a simulation model and associated automation hardware and consists of �five main components: (i) a water delivery system, (ii) an inflow measurement system, (ii) a water sensor to monitor advance of water along the furrow, (iv) computer running the simulation model (AutoFurrow), and (v) a radio telemetry system to facilitate communication among the system components. AutoFurrow uses a scaling technique to adjust the soil infiltration characteristic and determine the soil conditions prevailing for the particular irrigation. Hence it optimises the current irrigation to
satisfy the soil moisture deficit and other user-defined objectives (for example target effi�ciency, uniformity and run-o�) and determines the time to end the irrigation in suffi�cient time for effective control of the irrigation.
Trials to test and prove the new system were undertaken on two separate commercial cotton properties over two consecutive irrigation seasons. The system implemented for the field trials was not fully automated, and operations such as starting and cutting of flow was achieved manually. Apart from evaluations of the optimisation system, full advance data and other measurements were taken for all trials to enable a post-irrigation complete (actual) irrigation evaluation to be undertaken. Performances expected as per the grower's irrigation management practices were also evaluated. The SISCO simulation model was used for analysis of data.
The results suggested that the optimisation system was successful in delivering irrigation performance significantly better than achieved by the grower. However,
in the 2010/11 irrigation season this performance (predicted by the optimisation system) was found to be slightly higher than the actual performance and much
less than that suggested by a post irrigation optimisation undertaken using the full measured data for each irrigation. This suggested that the system had not
reached its full potential and further improvements were necessary. Factors investigated for their possible contribution to performance of the real time optimisation
system were: flow rate, objective function, selection of the model curve, and the infiltration scaling process. The investigations involved an exhaustive series of simulations using the SISCO model, varying each of these factors in turn. The key changes in the evaluation methodology effected as a result of these investigations and used for the 2011/12 irrigation season trials were: the adoption of a simpler objective function consisting only of RE, and (ii) taking the average shape of the previous in�filtrations curves and using it as the model curve. The benefit of these changes was clearly evident in the results obtained from the 2011/12 trials - the performance of the optimisation system improved and the difference between the actual performance predicted by the optimisation system was reduced to � 4%.
This research has therefore achieved its overall goal of designing and testing a real time optimisation system for furrow irrigation. It has also successfully demonstrated the potential benefits of real time optimisation and shown that the automation of the furrow system is feasible. Further research has been recommended including a comprehensive economic analysis and the trialling of the
system in bay irrigation.