What role can artificial intelligence play in agronomy?

Grains Research and Development Corporation manager of transformational technologies, Liam Ryan, said AI was likely to add value to various agronomic decisions differently, but would fall into three main categories of informing the decisions, guiding decisions and prescribing decisions.

Speaking at the GRDC Updates in Wagga Wagga earlier in the year, Mr Ryan said GRDC had been investing in AI and would work with companies to bring their capability to market in user friendly products and services.

This was in line with the organisation's strategy to accelerate the development of autonomous machinery across the next decade.

He said some forms of AI were adept at retrieving, synthesising and summarising vast amounts of information, such as large language models (LLMs), which include current platforms like Chat GPT by Open AI and Bard by Google.

"The capabilities of LLMs are evolving and their future influence on agronomic decision making could be profound," he said.

Mr Ryan said these platforms were worth experimenting with to retrieve information such as optimum planting times for specific varieties or disease ratings, however, they had biases and errors that needed to be accounted for.

In terms of guiding decisions, Mr Ryan said this was where they had been doing most of their investment.

He said it this mainly boiled down to producing good quality data and user friendly formats that could help support agronomic decision making.

One of the main projects had focused on crop phenology and predicting variety specific flowering times, using a variety of information and methods including AI.

The AI tool was used to predict phenology parameters required for APSIM Next Gen, a yet-to-be released predictive program written from scratch to be compatible with multiple platforms (i.e. Windows, Linux, etc), to simulate flowering time using genetic data.

"This is an example of a recurring theme where we use AI as a tool in the toolbox under the guise of relevant domain expertise to do really impactful things," he said.

Other projects include looking at frost and heat stress, weed mapping, disease management and plant water availability, using AI to add value.

These projects are using AI for a variety of things, including collating data to develop analytics methods for predictions and create models.

When it came to prescribing decisions Mr Ryan said AI could have a significant and strategic impact, but would require large volumes of on-farm data.

He said one example was the Future Farms project which worked with growers to implement large-scale strip trials and resulted in a method to predict the economically optimum nitrogen rate.

Mr Ryan said the quality and quantity of on-farm production data like this was the driver of AI in agronomy.

However, he said it did not need to be complicated.

"Just simple strip trials, working with quality assured yield data and good soil based information alongside that," he said.

"They can start to generate a really powerful positive feedback loop."

Mr Ryan said where AI was heading was difficult to predict.

"I think it's going to be heavily dependent on the extent to which we see greater collaboration between people who have relevant technical credentials and scientific expertise working with those companies and agronomists and others who can translate that into user friendly products," he said.