The 2024 harvest is complete, and this year, we used a wide range of ag technology to assist with on farm decisions. The most potent piece of information we used were the yield maps from the previous year. We also used satellite imagery to monitor crop growth through the season, and a drone was used to spray crops late in the season. Yield maps were processed in an industry online platform, and satellite imagery was accessed through Google Earth Engine.

Once the yield maps were evaluated, we made the decision to completely ameliorate an entire field. The lowest yielding field was deep ripped. Lime and Gypsum were applied and incorporated. Yield maps confirmed this field routinely yielded 0.5 t/ha less than its neighbours. After the amelioration, yields from this field were comparable to its neighbours.

The decision-making process highlights a challenge for ag-tech. I did not create a sophisticated algorithm to determine whether this field should be ameliorated. I did not conduct sophisticated analytics to confirm the response to the treatment, even though I could. In short, we simply looked at a few yield maps, made a decision, and acted. We evaluated that decision by simply looking at a few yield maps, and concluded, in the most Australian way that “Yeah, that worked”.

I did build an algorithm in Google Earth Engine to quickly identify other parts of the farm that were like that field. No other field needs to be completely ameliorated. This is more complex to organise, and the marginal value of further aggressive amelioration is no where near as substantial as the initial investment, so it can wait. This is a business decision, more than an agronomic one. It can wait until we have another season with high prices and or very high yields.

This creates another challenge for Ag – tech, that at face value, has not yet been supported. Farm management decisions need to be made in the context of the business. They should not be made without considering the business fundamentals. Analytical systems that integrate business models with agronomic models, and information from sensors and yield maps are now feasible. However, farms would need to recognise they are moving into a data ecosystem, that must be organised and managed to deliver these insights. Few farms run sophisticated AWS, Google or Azure accounts that capture all this information, and provide end users with the necessary analytical tools, such as machine learning, optimisation or AI to harness the power of on-farm data. Furthermore, these powerful analytical tools should probably not be deployed for basic input / output agronomic decisions. Rather, they could assist with strategic decision making, such as whether to expand, and purchase another farm, how to co-ordinate enterprise activities around available labour, whether to update machinery, or take on a new lease. These strategic decisions can make or break a farming operation, and ag-tech is still learning how to assist in this space.

In future, ag-tech will need to assist farmers move to an ag data ecosystem to allow sophisticated analytics that manage all aspects of on-farm risk. It will need to communicate the value of this system back to the end user, and this is not easy to do.