In this image, you’re looking at about 5,000 agricultural fields and a few hundred farms or so. Specifically, this is in Iowa, so probably corn & soybeans.
Each little white square you see here is a farm - a family. 20% don’t have internet access, they have an average land size of about 350 acres, and only 8% of the farmers are under 35 years old. But this isn’t just Iowa; this applies similarly to farms across the world.
Whether we like it or not, millions of data points are being generated and collected on every single agricultural field, not just in this image, but on this planet, 50 million square kilometers - every single day. From satellites photographing the earth, soil probes in the ground, weather stations tracking every parameter, one thing we have now that we didn’t have ten years ago is data. And we’ve got an awful lot of it.
Agriculture is one of the very few industries today where a minimal increase in production or efficiency of say 1 or 2 percent can equate to feeding millions more with the same inputs.
But the big question is - how do we achieve that 1%?
At CropSafe, we believe the answer to that question is with data but making data accessible, not embedded within a dozen Excel spreadsheets. Making it easy to understand, not transcribed behind 50-year-old terms and expressions. Most importantly, making it easy to access, not limited, and behind a paywall that only large corporations and farms can afford.
But why now?
Over the past 20 years, the cost to launch a satellite to orbit has plummeted, prices dropping more than 90%, and we expect this to reduce tenfold in the following decade once more. Following Moore’s Law, our computing ability is only growing in parallel too.
We’re coming to a critical intersection in agriculture where the needs of farms are changing. At this intersection, we’re seeing more regular satellite mapping, more accurate soil probes, and affordable personal weather stations for the first time. At a fraction of the cost, this influx of new technologies allows the average, everyday farmer to take advantage of them is a crucial turning point.
In this photo, we’re looking at one of SpaceX’s Falcon 9 rockets; on top of it, about 44 miniature earth observation satellites, about the size of a shoebox, which will join a constellation of about 200 more shoeboxes in orbit. Launched by a planet company, these satellites are just another data source of the thousands we have access to.
All this data is excellent, but even still, the mast majority of it is siloed and has yet to see the light of day. Most - never touched. And to be honest, some data probably won’t be helpful for years to come. But it’s still important to collect.
Training machine learning & visual recognition models are tricky; they require a lot of input and an incredible amount of fine-tuning. We’re dumping vast amounts of information such as crop pest probabilities, locations, times, and dates and asking the computer to help us predict where it thinks the next outbreak may be. But this is all historical data, collected maybe five to ten years ago. Although that data may not have been much help back then, it is now. Because now, we’re able to save crop yields of farms across Iowa by sending them something as simple as a text message warning.
Where are we headed? What’s next?
In the past one hundred years, we’ve gone from horse-drawn carts, often a three-person operation, to tractors that pretty much drive themselves - no human intervention at all. From field scouting, a process that can eat up days, to a process where a tiny digital camera 250 miles above our heads, traveling 4.7 miles a second, can snap a photo and tell us in a few seconds which square meter of our 350-acre farm needs attention. The difference in incredible... going from waking up at 5 am, putting on your wellington boots, walking your fields, line by line, checking each plant for any contaminations, diseases, pests, to having all this data at your fingertips within seconds.
It might seem like we are optimized to the max and that there’s very little to innovate in this space, but we’re still only at the bottom of this curve. Today, we’re seeing only a fraction of what will be possible in the next decade. But this new wave won’t be driven by hardware - but by data - historical data. We have connected tractors on the farm, controlled irrigation systems, weather stations. We collect masses of data, but now we need to make the best use of it through software. When provided to a farm, their data alone is valuable - but becomes multiple times more reliable when crossed referenced with other sources.
An on-farm weather station may read the precipitation today and what is forecasted this week. Your on-farm weather station data crossed with satellite data can tell you that, as well as where on your farm is most likely to get waterlogged or flooded. Using A.I. and traveling data sources can help us achieve much more productive yields than ever before.
But there are a few things we need to focus on first
Our vast silos of data, through newer advancements in machine learning and artificial intelligence, are slowly becoming immensely powerful in how we optimize yields. However, we must keep our eye on who and where we distribute this technology.
The first version of CropSafe - was great; it allowed farmers to view the most recent satellite imagery on their farm, check weather patterns, and log field history, all within one app. We shipped our app to a few dozen farms on our waiting list, and thought that is was all we needed to do. But when checking user analytics, we dug deeper to see what our users were looking at and found something interesting. They weren’t using our software to look at satellite maps, weather patterns, field history. Every time they logged in, they had a particular question at the top of their head.
“Should I spray today?”
“Is the Shore Rd field water-logged?”
“When is the best time to harvest?”
Farmers don’t have time to look at graphs, excel sheets, tables or maps. Our focus lies in translating all this data we have into simple, actionable insights and suggestions we can provide farms. Making this data accessible is our current challenge. We’ve got the self-driving tractors, satellites and soil probes already. Now we have to translate the information they’re producing into helpful output.
Keeping crops safe through A.I. isn’t about how much data we can produce and collect, generate insights from - but rather, how we optimize to push real-life action on the ground. We shouldn’t be building for data scientists, but for the 95% of family farms, the 92% of farmers over 35 years old and the 20% that don’t have internet access.
It’s not large corporations that produce the majority of our food; it’s the smaller, 350-acre family homes, just like the one I grew up on - these farms power the world.
