The project will develop an open-source information, prediction and advisory platform for viticulture (VitiVisor) to support on-farm decision making, increase information transferability and access, which will improve farm outcomes (i.e. gross margins/profitability and sustainability measures). The key focus will be on water usability and efficiency.
Digital agriculture has the potential to boost the value of production by 25 per cent compared to 2014–15 levels (Leonard et al, 2017, ‘Accelerating precision agriculture to decision agriculture: Enabling digital agriculture in Australia’), which translates to an estimated $706 million for the grape and wine sector.
Evidence collected during the Pilot project reveals that uptake of existing technologies remains relatively low in practice, with incomplete (and often paper-based) documentation of historical management actions and end-of-season outcomes, and difficulty in converting existing AgTech solutions to management actions. Extensive consultation with growers from the Riverland wine region during the Pilot, together with an audit of existing AgTech solutions, revealed some of potential barriers to wider adoption:
- a narrow focus on bio-physical data streams, rather than operational decisions and/or financial data that are critical to support management actions
- insufficient attention to the inter-operability between proprietary solutions, creating ‘lock-in’ and data ownership concerns
- a limited emphasis of existing solutions on provision of guidance and advice, with most solutions focusing exclusively on information display, and
- a focus on single or small numbers of use-cases, leading many growers to require a multitude of incompatible and/or overlapping tools to manage vineyard operations.
This project seeks to overcome these barriers by proposing an open-source information, prediction and advisory platform that will be co-created with Riverland growers.
The project will establish a data management system with the capacity to integrate established and newly developed data streams using sensors installed on trial vineyards throughout the Riverland. A monitoring sensor network will be built for the detection of irrigation system failures including identifying leaks and blockages.
Vision data collected at key stages of a growing season. Image segmentation based on machine learning of key vine attributes will be validated in order to develop models for the prediction of canopy growth/ architecture, seasonal irrigation requirements and yield.
Preliminary predictions from the machine learning and process-based modelling, and proof-of-principle model will be used to identify management strategies to achieve targeted vineyard outcomes.
The focus on reducing costs of production will lead to a direct emphasis on optimising resource consumption (including water, chemical and energy consumption). In particular, the project will better align water application with quality/quantity/disease management targets, optimising the use of this valuable resource and the energy required to access it.
VitiVisor will be an open source platform, expandable over time, allowing anyone to add new applications.