Top Three Ways Data Science Will Change AgTech in 2020

As part of the data science team here at Granular, I recently had the opportunity to convene a session on advances in agriculture at the largest earth sciences conference in the world, hosted by the American Geophysical Union (AGU). This conference brought more than 20,000 scientists from a multitude of disciplines together to discuss the latest in their respective areas of research. Here are some of my key take-aways:

1 — Remote Sensing is Improving In-Season Crop Health

Many in agriculture are becoming accustomed to near real-time imagery showing the crop health of their fields. Advances on the horizon include much higher temporal resolutions than many have been exposed to. New cubesat swarms, miniature satellites that can be deployed in large numbers, will soon enable daily ~3m data to be collected across the planet. This higher resolution imagery will allow the detection of crop health issues in near real-time at an unprecedented level of detail.

Images from the National Agricultural Imagery Program (NAIP) show the range of difference between 1m imagery and 250m

Our latest Granular product, Granular Insights, leverages the latest high frequency, high resolution satellite imagery available today and couples it with directed scouting, which provides growers vegetation indexing, auto-prioritized field rankings, and weekly email notifications. We’re excited to leverage the best remote sensing options available to power Granular Insights, now and going forward.

2 — Access to Ag Data is Exploding

In the next 5 years, YEARLY new public data alone is expected to increase by ~48 petabytes a year and the total stored volume will increase to nearly half an exabyte. For reference, more than 15,000 iphones would be required to store 1 petabyte of information. This huge increase in data volume will require new techniques and approaches to handle this vastly increased data sets. In addition to higher resolution imagery, more remote sensing data will unlock new opportunities for monitoring and analysis in many areas of agriculture.

A key example of how farm management software (FMS) is leveraging these increasing datasets is with yield modeling. By importing new data on a regular basis, some FMS providers are able to provide farmers updates on expected yield throughout the growing season. We’re looking forward to providing our customers even more insights into their crop’s potential via our yield model in several Granular products later this year.

3 — Machine Algorithms on Overdrive

When it comes to leveraging large datasets, the best way to make them actionable is often to apply advanced and complex algorithms. For example, our yield modeling pipelines use sophisticated machine learning algorithms to detect small variances in vast datasets to distill actionable information in near real-time. We’re relying on these algorithms more than ever to enable insights into incredibly large and complex datasets that standard statistical analyses are not well suited for.

The main takeaway from my time at AGU is that there is an ever-evolving body of work from the academic arena to improve agriculture in a variety of ways, ranging from nitrogen use and application to seeding rates. However, while there’s immense work being conducted, the complexity of implementing many of the ideas being tested in practice remains. We in data science must remain mindful of the practicality of implementing research.

This is a key area where Granular, and more specifically our Granular Data Science team, works. Granular DS has the research and academic training to understand, model, and distill these bodies of work. And by partnering with others throughout Granular and Corteva, we’re looking forward to helping farmers be more successful by extracting useful information from these data streams in a fast and meaningful way.

About Matt

Matt grew up in rural Michigan before joining the Marine Corps. Following his service, Matt studied Wildlife Ecology in his undergraduate days and rangeland ecology and hydrology for his masters work. Matt specializes in applying statistical and machine learning techniques to large scale spatial data sets. In this work he has driven insights into various areas from land cover changes in the sierra nevadas in California, golden eagle migration corridors, novel approaches to analyzing wildlife tracking data, modeling crop yield over large spatial areas and integrating new techniques and algorithms to drive insights in a variety of systems. Matt is currently a Data Scientist at Granular working to derive insights into farmers’ most pressing questions.

Stay in the know!

Subscribe to get new blog posts sent straight to your email


In Search of a Leader!

Jolene Brown, CSP, CPAE, December 27, 2019