Remote Sensing Powers New Granular Profitability Insights

July 3, 2019

Remote sensing, or gathering information about a field or subfield area without actually visiting it, is gaining traction quickly within the ag industry as its accuracy and use cases continue to build. Learn more about remote sensing at Granular, along with how it’s being used in Profit Maps, a joint collaboration between John Deere and Granular, in this Q&A with Michael Asher, Data Science Lead for Granular.

Q: Can you please tell us a bit about your role and background?

A: I’ve always wanted to contribute to ag, but I’m better at numbers and computers than using my hands. I studied math and computer science in my home of Australia and then made my way into agriculture. I’ve been on the remote sensing team here at Granular for around 18 months. Our team uses satellite and other sources of continental scale data to model crop growth.

Q: Why is remote sensing such a hot topic in agriculture right now?

A: I think it’s gaining traction quickly for a couple reasons. One, being able to forecast crop yields has potential implications for economics, food security, and human welfare. Having insight into yield months in advance of harvest allows for better logistical and financial planning for growers. Second, the data we can use for remote sensing purposes has improved dramatically in the past decade. With improved data comes improved results, so remote sensing is proving to be an increasingly valuable asset to the ag industry.

Q: What powers remote sensing at Granular?

A: We take all kinds of data points into consideration, including satellite imagery from 6 different satellites, weather data, elevation data, soil data, and historical yields from around a region. By using all those datasets, we model at a high resolution what yield is going to be. It’s something we’ve been working on for years at Corteva, and I’ve been focused on it since I came to Granular.

Q: How can growers experience what you’ve been working on in 2019?

A: Our first grower-focused tool that utilizes the yield model developed by our remote sensing team is in Profit Maps, our financial insights collaboration with John Deere. We’re introducing two Profit Maps new layers this month: an estimated revenue layer and our in-season yield forecast beta group.

This new estimated revenue layer, available in the John Deere Operations Center and all John Deere mobile apps, will be in addition to the cost, revenue and profit layers currently available. This layer is available for 2016, 2017, and 2018 seasons for corn and soybean acres across the majority of the midwest. The estimated revenue layer will display whenever there is a corn or soy seeding event, but no harvest event — making it a valuable asset for back-filling missing harvest data.

For 2019, Profit Maps users will also receive free, exclusive access to the Granular in-season Yield Forecast beta group. Our in-season model utilizes the same proprietary data science to generate an estimate of your 2019 corn and soybean yield potential where you have a seeding event. This model is an estimate and is still in development, but by signing up to be part of our in-season beta group, you can expect bi-weekly yield estimates beginning in mid-July. Given the challenges of this year, we wanted to provide growers another datapoint for consideration as they make in-season decisions.

Q: How can growers use yield modeling?

A: In-season, there’s a few different use cases we’re developing. The obvious one is scouting. You can use this layer the same way you would use crop health imagery to identify issues in field during the seasons. Instead of looking at something abstract like crop health index or NDVI, you’d be looking at a bushels per acre value. It’s not just straight NDVI, it’s more accurate than that. For example, trials should show up early, along with issues related to machinery, irrigation or a particular hybrid in-season. We’re hoping to develop the product down the track further into in-season decisions. For example, you could use it to decide on whether to a fungicide application. There’s obvious implications for grain marketing, although I don’t recommend using the model for that at this time.

Q: Where is the model most accurate?

For right now, the model is most accurate in the corn belt. We’ve back-tested this model all the way back to 2007, and history shows that Iowa is the place it’s most accurate. We just have the most training data and consistent conditions there. For 2019, the model is also only available for corn and soybeans, but we’ll be moving into broader geographies and wheat, cotton and rice in the future.

Q: How will you and the remote sensing team at Granular continue to improve this yield model going forward?

A: Fortunately for our team, the remote sensing space is growing very rapidly, so there’s always more high resolution data imagery coming online. So, we’re continuously incorporating more data as it becomes available.

Right now, the model is a black-box machine learning type approach, which means we’ve taken all the data we can possible find and we’ve thrown it at a computer, and just let the computer figure out all the details. That can be a really powerful approach, but we believe that we can improve the accuracy substantially by including more explicit agronomic principles. With that, we hope to offer insights into things like predicted plating dates, ponding, disease stress, etc.

Q: Where can growers go to learn more about Profit Maps and these new yield model opportunities?

A: For this year, growers who use John Deere equipment can find Profit Maps in the John Deere Operations Center by finding the Granular logo in the tools section. Until July 31, John Deere customers can also receive 20% off Granular Insights, a directed scouting tool featuring high-resolution, high-frequency satellite imagery. I would encourage growers to check it out. We’ve all seen how innovation in agriculture can protect yield and profitability, so this year especially is a good time to use all the information you can to make the best decisions possible.

Michael Asher, Remote Sensing Data Science Lead, can be reached at (734) 277-0537 or Learn more about Profit Maps here. For more information about the exclusive Granular Insights discount for John Deere customers here.

Enter your email to sign up for The Bottom Line, Granular’s monthly Newsletter