Novel approaches to irrigation scheduling are becoming critically important as water resources in the Basin come under increased scrutiny. It has been well known for a long time that through increased measurement of plant water requirements, one can dramatically improve yield and quality with less water. However, the challenge has been in implementing the science in a way that accurate and easy to use. This challenge clearly remains since the current uptake of smart irrigation scheduling system is less than 10%. One important reason for this is the lack of irrigation scheduling techniques that can be adapted or scaled across a range of cropping systems with consistent outcomes.

Between 2004 and 2008, the Victorian Government, through the STI grants, funded a joint research initiative between National ICT Australia, The University of Melbourne and Uniwater called Smart Irrigation. The purpose of this project was to develop low‐cost wireless sensor networks and irrigation scheduling algorithms for dairy, viticulture and horticulture. This project demonstrated up to 30% water savings in dairy irrigation and up to 75% increases in yield in horticulture. The project laid the foundations for future research in the application of intelligent sensor networks and control algorithm to irrigation science.

Current approaches to irrigation scheduling rely on point‐scale measurements of soil moisture. However, due to crop, soil and micro‐climate variability, decisions based on point‐scale data may not be optimal for the entire field. Overcoming this limitation requires a sensing system with a wider spatial coverage and a rethink of the algorithms that use this data, as well as existing point‐scale measurements.

In this project, we propose to develop a new scalable approach to irrigation scheduling and innovative algorithm for developing field‐scale water demand prediction models. Our algorithm will use more than one point‐scale measurements of soil moisture combined with low‐cost, low‐resolution thermal images and local micro‐climate data, to predict short‐term water demand. By combing point‐scale soil moisture data with spatial images, the algorithm will enables users to calculate water application rates that are optimum at the field‐scale, rather than focused on a single plant. This approach avoids the bias towards a single measurement point that is common in most irrigation scheduling techniques in use today.

Due to the fact of long experimental turnarounds and wide spatial distribution of sensing devices, the approach to separate device management will raise the cost and complexity of experiment. Alternatively, our approach employs low‐cost to establish a sensing system for data collection and remote storage. It provides user with a lower cost and more efficient way toreal‐time data access and remote monitoring in contrast with conventional ones.