Public-Private Examples of Sensor Networks in Australian Agriculture.

Farmers and agri-businesses are keen to discover ways to use on-line data services for production, environmental and marketing opportunities. Easy access to relevant data from public and private sensors is possible, but few managers are aware that it can be tailored to their needs, and the possibilities that it provides for better decision-making. CSIRO and Sense-T conducted several projects in Australia in recent years to develop a platform to capture, manage, analyse and deliver real-time data, and importantly, to learn and demonstrate how such readily-available real-time data can improve decision-making in agriculture from farms to catchment scales. Early experience emphasised the value in federating disparate data from public sensor networks off-farm (e.g. weather, stream-flows, and water quality), private sensors on-farm (e.g. soil water, weather), and other information (e.g. spatial and markets), and then combining these with models to predict farm attributes. Examples from pasture-based livestock farms include developing a cloud service that integrates in situ and regional weather data with a biophysical plant growth model to deliver predictions of pasture growth, and the capture and delivery of real-time cattle live-weight data in the field with external market information to enable graziers to better meet cattle market specifications. Another example shows how access to real-time data on observed and forecast rainfall and stream-flow has enabled the maintenance of environmental flows and water availability for irrigation. Using real-time data, confidence and capacity has increased in this irrigation community to the point where they are now willing to experiment with catchment water management options in collaboration amongst themselves and with the regulator, with the view to increased co-governance of this valuable resource. These examples relied on new links and common understandings developing between software engineers, data analyists, and agriculturalists. Some farmers now see value in sharing private sensor data. Prediction services used machine-learning and deterministic approaches. Statistical inaccuracy and imprecision did not detract from the usefulness of action-learning approaches, but rather exemplified the need for probabilistic forecasts to aid risk-informed decision-making. Data availability and privacy protocols have proven satisfactory, and some data services are now progressing from the research phase to commercial or public good operational models. These sensor-based data services offer new opportunities for the development of agri-environmental businesses that link data and translate it into useful insights.