The challenge

With nitrogen fertiliser  making up 30 to 40 per cent of input costs for cereal growers, decisions made around nitrogen application are a major determinant of profitability and productivity.

Additionally, GRDC’s More Profit from Crop Nutrition program found that a grower’s risk is closely related to the magnitude and effectiveness of nitrogen fertiliser expenditure.

Optimising the efficient use of nitrogen is, therefore, an important goal, which leaves the sector routinely investing in opportunities to innovate nitrogen application decision-making tools.

Among the strategies developed is the use of new precision agriculture (PA) analytics to deliver on two key components of the ‘Four Rs’, which involves putting the right amount of the right product in the right place at the right time. The focus for PA analytics is on the right amount of nitrogen fertiliser in the right place.

However, optimising nitrogen application strategies requires the use of data specific to a location (geospatial data). In turn, this typically requires a substantial investment in time spent processing and analysing data. Tools developed to aid in this process are often not easy to use.

There can also be difficulties incorporating all the available data layers – such as soil tests, yield maps and remotely sensed imagery – when trying to make a nitrogen application prescription that matches nitrogen supply with demand.

A need was, therefore, identified to improve the way in which multiple layers of soil and crop sensor data can be used to inform nitrogen management decisions and a range of other input decisions.

A requirement was also identified to automate the entire process, from data acquisition through to analysis and generation of prescription maps to make it easier for agronomists and growers to make use of all the available data layers for input decisions.

The response

GRDC invested in the Future Farm Initiative to explore new ways of using crop and soil data layers to automate complex input decisions and maximise the profitability and sustainability derived from those decisions. The focus was on nitrogen management decisions as the key use case.

Phase 1 of the Future Farm Initiative was a desktop study that explored different innovations in sensing and analytics globally to identify promising opportunities in automated agronomy and autonomous machines. That work led to a field-based experimentation program called Future Farm Phase 2.

In Phase 2, data collected from large-scale on-farm experiments across Australia were used to develop 11 different methods for predicting optimal mid-season nitrogen fertiliser rates for wheat and barley crops. These were then compared at different spatial resolutions and against an industry benchmark method and the collaborating grower’s current methods for making mid-season nitrogen decisions.

These methods were assessed in terms of the net partial profit (NPP) derived from nitrogen fertiliser applications with each recommendation method. The economic optimum nitrogen rate (EONR) – the rate at which nitrogen supply was perfectly matched to demand across the paddock – was used as a hypothetical benchmark for comparison.

The nitrogen recommendation methods were also assessed using the root mean square error (RMSE), which determines how close each method is to perfectly matching nitrogen supply with demand relative to the EONR, were the RMSE was zero kilograms of nitrogen per hectare.

This research and development work was led by CSIRO in collaboration with the University of Sydney, the University of Southern Queensland, the Queensland University of Technology, and Agriculture Victoria. The Cotton Research and Development Corporation led a similar project tailored to cotton, with researchers across both projects meeting to share results, ideas and perspectives.

The impact

The comparison of nitrogen application tools found that a machine learning method that makes use of data derived from multiple years of nitrogen-rich and nitrogen-nil strips performed exceptionally well. The potential gross benefit possible from this site-specific application method relative to the farmer practice was approximately $57 per hectare. As a result, the Future Farm Initiative is now in its third, commercialisation phase.

The overall potential economic benefit of project outcomes was further assessed using a base case scenario approach using conservative estimates.

The potential net present value of this investment could amount to approximately $155 million if the technology makes its way into growers’ hands. Work is underway with a market-leading PA analytics company to make that happen. That company has already been evaluating the method with leading growers across Australia as part of the commercialisation effort.

The project is likely to result in a benefit cost ratio of 13:1. Sensitivity analysis showed that these overall results were found to be robust to changes in some of the key parameters. This highlights that this investment into innovations in crop sensing, soil sensing and automated analytics could provide substantial returns to grain growers across Australia.

The investment could also return environmental benefits by reducing the potential for nitrogen loss from the system. The method also produces a robust on-farm dataset that could be used to report on improvements in greenhouse gas emissions intensity.

Additionally, this method has the potential to help counter some of the risk aversion among grain growers in making fertiliser decisions and help agronomists and PA consultants realise even-greater value from variable-rate application technologies.

The economic analysis was performed by Bob Farquharson & Associates.

More information: view more Delivering impact case studies.