It’s an unusually frigid December morning in Gurugram and I’m sitting with Ranveer Chandra, chief scientist, Microsoft Azure Global, at Microsoft India’s tenth floor office. Down below, a monorail train cuts through the haze, which is now almost permanently suspended over the city. Later in the week, the region will witness the coldest day in more than a century.
“Climate change will affect small-holder farmers the most. Even a few degrees of change in temperature or untimely rainfall will create havoc in their lives,” he says.
Chandra is the founder of FarmBeats, which started as a Microsoft Research project in 2014 to enable data-driven farming and is now available as a platform for digital agriculture in the form of Azure FarmBeats.
According to the United Nations, the world population is projected to reach 9.8 billion by the year 2050, which means there will be more mouths to feed with same or lower arable land and unpredictable weather patterns due to climate change.
“We need to grow more food. But not just food, we need to grow more nutritious food without harming the planet. We need to do all this with an amount of land that isn’t increasing and water levels that are receding,” Chandra explains.
Taking the guesswork out of farming
Be it a farmer from the Great Plains region in the United States or Punjab in India, many of the farming decisions are taken by guesswork. From the amount of fertilizer to be used to when one should sow, water, and harvest the crops, some of the most critical decisions in agriculture are based on experience and instinct.
With Azure FarmBeats, Chandra has two goals. First is to remove the guesswork in agriculture and provide farmers with data-driven insights. The second phase is to make data-driven agriculture more affordable.
“Last year’s yield is not going to tell you how much nitrogen is needed in different parts of the field. Even though farmers know the benefits of data-driven agriculture, the technology didn’t exist,” Chandra tells me.
The team has already delivered on its first goal with Azure FarmBeats which is now out in preview. The system creates a digital map of a farm with drone or satellite imagery as well as a grid of sensors spread across the farm that monitor multiple parameters in the soil ranging from temperature and moisture to carbon and nitrogen levels.
Armed with all the data, farmers can get actionable insights with the help of artificial intelligence (AI) and machine learning (ML) based models, which are built by partners on Azure FarmBeats. The solution is already live in some farms in the United States, where farmers are starting to see the benefits.
Building non-existent tech
Chandra’s interest in agriculture wasn’t a coincidence. While growing up in Jamshedpur in eastern India, he’d spend his summer and winter vacations at his grandparents’ village in Bihar, one of India’s most impoverished states.
“Back then, I hated agriculture like anything. The farms in the village would have no electricity, not even toilets. You had to go out in the fields even in cold weather like this,” he recalls. “But as it happens, I spent a lot of time in the village in the first 18 years of my life. I have seen extreme poverty and those images have stuck to my mind.”
“What can we, as technologists, do to improve the lives of these farmers?” he asks.
In its current avatar, Azure FarmBeats works for large farms that are spread over tens or hundreds of acres. The AI models are built to ingest satellite and drone imagery and churn out insights about the farm’s health. These are combined with data from a network of sensors buried in the soil that provide real-time information from ground zero.
Unfortunately, current technologies for getting data from the farm are too expensive for small-holder farmers like those in India to be able to afford.
“One of the things we’ve noticed is these farmers grow multiple crops in the same acre or two of land that they have. Satellite images cover thousands of acres planted with the same crop, where every acre corresponds to just tens of pixels, which are not enough to see the leaves or pests or weeds to get meaningful insights,” Chandra explains.
To solve this challenge, the FarmBeats team came up with a solution called “Tethered Eye”, which replaces drones with helium filled balloons and a smartphone with its camera facing down, flying at a height of 200 feet. The assembly is tethered on to a stick and one can have a person walk the length and breadth of the field to create an aerial map of the field using computer vision algorithms.