How predictive analytics will make farming more sustainable

Last week I made the 19 hour plane flight(s) from Brisbane, Australia to New Orleans, Louisiana to attend the 2017 Team Summit for the Tulane Nitrogen Reduction Challenge. Tulane University is taking the issue of nitrogen reduction very seriously.

Is a $1,000,000 first prize serious enough?

That’s the reward for the winning AgTech company that successfully reduces nitrogen utilisation in a real-world farming environment without negatively impacting yield or grower profitability.

The aim is to find sustainable and renewable solutions that reduce the use of nitrogen on farmland. The ultimate goal of the challenge is to find a solution to combat hypoxia, a deadly deficiency of oxygen that creates “dead zones” as found in the Gulf of Mexico.

One of the five finalists is AgDNA, an innovative AgTech company who intends to solve the excess nitrogen issue using Predictive Analytics.

The Problem

Nitrogen is one of the greatest contributors to increasing crop yield. Combined with its relative low cost and high potential return, farmers are inclined to over apply nitrogen as an insurance policy to ensure they maximise yield.

As a result, excess nitrogen is lost into waterways with potentially damaging environmental consequences.

One Potential Solution

AgDNA looks to solve the excess nitrogen problem by optimally matching nitrogen requirements to site-specific conditions spatially throughout the field. Therefore intra-field soil variability can be managed as well as seasonal fluctuations in weather using computer modelling.

The company’s underlying technology combines next-generation cloud computing infrastructure, big-data analysis techniques, high-resolution climate data, soil characteristics and as applied agronomic information from in-field machinery. The various inputs are then combined with the latest crop and Precision Nitrogen Management (PNM) models to dynamically predict soil nitrogen requirements.

The resulting insights are delivered as nitrogen application recommendations and variable rate prescriptions (VRx) that can be delivered wirelessly to in-field machinery for execution. This removes friction from the system by providing a seamless two-way communication of data to and from the machinery.

Predictive Analytics Approach

Predictive analytics lends itself to the excess nitrogen issue as the models can be developed to handle a wide range of inputs and spatial variation. However, there are several other factors that must also be considered when developing a predictive analytics solution for agriculture.

1) Understand the entire crop production process

Farming is a complex business with a lot of variables, particularly weather. That’s why there is no silver bullet for farming, instead you need a lot of lead bullets. So the starting point for any analytics solution is to understand the entire process and workflow from the customer’s perspective.

2) Determine the most valuable insights

By understanding the crop production workflow you get an appreciation for the decisions that need to be made throughout the season. This helps to get a feel for timing, cost and potential impact of each decision.

3) Build a broad dataset

Too many solutions in agriculture are very narrow in their application. Often this is due to the limited amount of data any one company is able to access. To exploit Predictive Analytics solution providers need a broad array of data about soil, climate, crop, agronomic practices and real-time data from other in field sensors where available.

4) Generate insights that are actionable

Insights are of greatest value when they can be actioned. In the case of nitrogen reduction, this can be in the form of a prescription recommendation including nitrogen rate, location, method and timing.

5) Deliver insights at the point of highest impact

Farming is highly seasonal with distinct phases such as ground preparation, seeding, application, irrigation and harvest. Therefore any insights that are generated must be in the context of the season and crop stage. Nitrogen recommendations must account for fluctuations in weather and ensure optimal nitrogen is available to the plant when uptake is required most.

6) Keep refining the models

No two seasons are the same in agriculture — dry or wet, hot or cold, early or late. The weather continually changes as do the seed varieties and available data sources. This is where big-data and machine-learning methods come into play and allow continuous improvement of the models as more information becomes available.

Real-World Adoption

Predictive analytics and sophisticated computer models alone are not enough to solve the excess nitrogen problem. There is a critical stakeholder in the system called “the farmer”. For the technology to be widely adopted the grower must be able to clearly identify the following outcomes:

• Insightful
• Actionable
• Easy-to-use
• Value adding

By seamlessly integrating with the grower’s equipment and workflow, predictive analytic tools can become part of the day-to-day farming operation. By improving sustainability and ultimately farm profits, such tools can become one of those “can’t live without” items.

Conclusion

The top five finalists will begin testing their innovative nitrogen management systems in March on a parcel of farmland in Tensas Parish, located in the North-eastern corner of Louisiana. This is where the “thinking” of academia and the “doing” of private industry will converge to tackle one of the world’s significant environmental issues of excess nitrogen.

Concluding in December the winner of the Tulane Nitrogen Reduction Challenge will be judged by its innovative use of technology and its effectiveness in reducing nitrogen in the field. For AgDNA it will be another step forward in the delivery of predictive analytic insights to solve real world challenges for growers around the world.