AI and the future agriculture

Almost half the world's population lives in households that rely on agrifood systems as their main source of employment, according to the United Nations. And with the world's population projected to go from 7.3 billion to 9.7 billion by 2050, water scarcity and crop failures due to climate change could lead to catastrophic global food shortages.

To keep up with the growing population, the world will need to boost agricultural production by 60% over the next 25 years-and even then we may still have 300 million going hungry. At the same time, row crop yields are expected to drop by 11% because of more severe weather and more pests. Right now, it takes about $136 million in the US and around 12 years to develop a new crop trait.

The world's heating up, there's less fresh water, and the successful crops of yesteryear won't be dependable for much longer. Can AI help mitigate the impending agricultural crisis we'll be facing over the next few decades?

Dr. Abhisesh Silwal, a systems scientist at Carnegie Mellon University whose research focuses on AI and robotics in agriculture, thinks so. "AI could lead to more accurate and timely predictions, especially for spotting diseases early," he explains, "and it could help cut down on carbon footprints and environmental impact by improving how we use energy and resources."

Here are a few agritech businesses and nonprofits working to beat the clock on climate change.

Better, faster phenotyping

In Tanzania, David Guerena, an agricultural scientist at the International Center for Tropical Agriculture, is using AI to kick plant evolution into overdrive.

Guerena's project, called Artemis, uses AI and computer vision to speed up the phenotyping process. "People are not good at providing reliable quantitative estimates of things we see," he explains. "It is not possible for a human to accurately count flowers on hundreds of plants across thousands of plots. We get tired, lose our focus, or just physically can't see all that we need to. A computer doesn't have these problems. Well-trained computer vision models produce consistent quantitative data instantly."

When Guerena's team first started working with smartphone images, they used convolutional neural networks (CNNs). They did the job, but they were hungry for data-needing thousands of labeled examples to train them, which was expensive and time-consuming. They aimed to develop models for various traits (like counting flowers) across different crops, but CNNs were too demanding for that.

A few years back, vision transformers and open-source models like YOLO and Segment Anything came out and changed the game, Guerena explains. These models drastically cut down the need for labeled data. Now, instead of thousands of examples of crop images, they only need a few hundred. "The vision transformer and foundation models have really enabled us to fast-track model development."

Guerena's team is now working on integrating speech-to-text and natural language processing alongside computer vision in the systems they're building. Interacting with people through natural speech can help overcome language and literacy barriers, he explains. Plus, combining different types of data can create a fuller picture of how plants are doing in the field. "Farmers themselves can use the speech function to describe what some of the challenges they faced over the season were," he says. "If we collect this data across thousands of farms all over the world, we can better piece together how local environmental conditions affect variety performance."

Forecasting weather months ahead

Growing up, Max Evans and Himanshu Gupta saw up close how irregular weather patterns were impacting their communities. "Max grew up on a pineapple farm in Ecuador, witnessing firsthand the impacts of climate change on crops," Gupta tells us. "I was raised in a village in northern India, where we often had to walk a mile for clean water during droughts and deficit monsoon seasons."

Those experiences left lasting impressions on Evans and Gupta and inspired them to found ClimateAi. "While most technology companies aim to improve the traditional 14-day weather forecast, these time frames are not actionable for farmers and food companies," Gupta says. "We innovated in weather forecasting beyond two weeks, deploying patented biophysics-based AI approaches to forecast the risk of extreme weather cheaper, faster, and more reliably than supercomputer models used by meteorological agencies, including NOAA." This allows ClimateAi to provide farmers with ultra-localized weather predictions for the day or for months in advance. They can also recommend optimal times to plant and harvest specific crops and estimate their yields.

ClimateAi ran simulations for farmers in Maharashtra, India, and found that extreme heat and drought would lead to an approximately 30% decrease in tomato output in the region over the next two decades. These insights were then used by a tomato seed company to accelerate trials for launching drought-tolerant seeds for smallholder farmers in the area. They also teamed up with a major food and beverage company in India to roll out adaptation playbooks in 300 villages, helping around 100,000 smallholder farmers. The playbooks offer tips on the best seeds to use, how to manage water, and the best times to plant and harvest, resulting in productivity that has gone up by as much as 40%.

The goal is to make food and water systems more resilient to climate change and to improve lives around the world, Gupta explains, and farmers are a key part of these systems: "[They] are the most critical nodes in food and water value chains.

Explore IBM Environmental Intelligence Suite

Growing broccoli in one-third of the time

North Carolina-based agritech company Avalo is using explainable AI (xAI) to precisely identify the genes linked to complex crop traits. Using detailed genetic profiles to guide breeding enables Avalo to develop crops five times faster and 50 times cheaper than their competitors, according to CEO Brendan Collins. One project Avalo has tackled is vertical farming, which is a method of growing crops indoors all year round. While it's great for continuous crop cycles, it struggles with high energy and fertilizer costs, leading to slim profits.

If these farms could grow more valuable crops faster, Collins notes, they'd see better profits. Right now, the only crop that works well in their system is cheap lettuce. Avalo teamed up with a vertical farm to grow broccoli, which is one of the highest-priced vegetables in grocery stores but takes more than 120 days to grow outside. They looked at more than 500 broccoli varieties to create a version that can be harvested in just 37 days. This quick harvest also meant that pesticides weren't needed, since the time frame was too short for pests to become a problem.

"Plant genomes are crazy," Collins says. "They operate under different rules than animal genomes." While animals have small, efficient genomes to save energy, plants have big, flexible genomes because they can't move to avoid threats, he explains. And unlike the simple diploid genomes of animals, many plants have multiple genome copies-like four in cotton, eight in strawberries or even ten in sugarcane-making them tricky to study. "We're excited about using xAI to make sense of these complex plant genomes."

The IBM Sustainability Accelerator: Empowering farmers through data

The IBM Sustainability Accelerator is partnering with farmers around the world as well, sharing IBM's cloud and AI expertise to help them adapt to a changing planet.

"We aim to help equip farmers with the data and AI insights required to remain profitable, productive and sustainable in the face of climate challenges and environmental threats," says Michael Jacobs, IBM Sustainability and Social Innovation Leader, Corporate Social Responsibility. "As a result, four out of our five projects on agriculture have now concluded with approximately 65,300 direct beneficiaries-farmers and their families using technology to help increase yields and make their operations more resilient. These solutions can be scaled across different geographies and industries."

Here's a quick look at one of their successes.

Smarter irrigation for Texas farmers

Droughts, floods, extreme heat and cold snaps are just some of the ways climate change is affecting Texas, causing ongoing problems for its farmers and ranchers. It's impacting all Texans: last year the state's department of agriculture released a report linking climate change to higher food prices and increased food insecurity.

David Chapin, a farmer and IBMer in Lampasas, Texas, who has worked with IBM to test water management solutions, has spent more than a decade watching climate change wreak havoc in his orchard. He started out with 3,600 olive trees, all of which were killed by Winter Storm Uri in 2021. The next year he received a USDA grant to replace the trees with pears, plums, apricots and peaches, which were more winter-hardy but also required more water-and his irrigation system could not keep up.

"Since it's crucial to monitor moisture at the root level, not just the surface, I'm looking for a precise method to manage this," Chapin explains. "AI could help by analyzing water needs in different areas, but it would require numerous subsurface sensors. Ideally, I'd like real-time soil condition maps and future irrigation predictions to better manage water storage and avoid shortages."

IBM and Texas A&M AgriLife are collaborating to help small-scale farmers like Chapin. The result is Liquid Prep, a tool that aims to give farmers the information they need on where and when to water as efficiently as possible, with plans to deploy and scale the tech across arid US regions.

Liquid Prep uses an IoT sensor combined with a mobile application that runs on IBM Cloud. Farmers can set up a moisture sensor in the ground and link it to their mobile app to monitor soil moisture, then upload it to be analyzed. The project team is currently expanding the app to include weather data, soil types and when-to-water decision support, so farmers can manage irrigation more effectively.

"Small farmers are usually not very technologically advanced, and that's one of the critical differences between small farmers and large farmers," Chapin says. "But people understand how to use a cell phone and use some apps, and Liquid Prep is a pretty simple app to operate…that's where it will change the playing field."

Download the ebook: Your guide to GHG emissions accounting

Public link

Please use the above public link if you want to share this noodl on another website.