Farmers Don’t Have Enough Water. Can AI Help?

For the fourth time in 10 years, farmers I know in California are facing a harsh reality — they won’t see a drop of water from federal government reserves to supplement the little bit they’ll get from Mother Nature.

Water allocations have become a hot-button issue throughout the state, as citizens, environmentalists and farmers fight for their fair share in a drought that’s made it impossible to please everyone.

With no help coming from reserves, farms have been left to draw water from the ground where they can. Working with a fraction of their usual supply, many farmers have no choice but to leave fields fallow, a devastating hit to their bottom line. For smaller farms, that can be the beginning of the end.

But I’ve also seen a very different approach.

Precision agriculture — the use of like networked sensors and artificial intelligence — is helping farmers get by without the water they once had. The efficiencies are real, and the impact is tangible. I’ve seen up close how precision agriculture is making a difference for farms facing extreme drought.

But embracing this technology isn’t always easy. In fact, it requires fundamentally revisiting our relationship with water in agriculture.

When it comes to irrigation, precision agriculture gives farmers a leg up in two areas — understanding how their water is being utilized, and maximizing delivery to stay alive.

On the understanding front, data is supplementing — and in some cases rewriting — irrigation practices that were developed over generations. Irrigation was, and still is, viewed as an art form. Farmers have long relied on rules of thumb based on visual signs of water stress in crops, or insights gained from working the land for decades. This ultimately led to irrigation being based on a general feeling, and not much more.

But that artform is changing into a science as we gain access to concrete data, both at the level of individual plants and in the aggregate. This data is stronger than the circumstantial evidence they used to base decisions on, giving them the ability to see what used to be hidden. I’ll share an example that’s close to home.

My organization now has more than a billion trees under observation in orchards around the world, with sensors reporting back data every 10 minutes on variables like soil moisture, water absorption, and trunk diameter. With that much information, key correlations emerge that were previously unnoticed. Most of these sensors and the data they collect isn’t new — but they have always been presented separately and required a trained eye and purposeful time to draw out the meaning. That’s precious time growers don’t have.

Now, with immense information available in one place, these systems can anticipate how soil moisture will be affected by factors such as temperature, humidity and wind, and translate that into predictive algorithms. Herein lies the real potential of this technology: a prescription of exactly where to water and when. The result is an ability to maximize “crop-per-drop” at a level unthinkable even a few years ago.

But knowing that a particular row of trees needs water means very little unless you have the technology to get the water there.

That leads me to the second point — maximizing delivery.