AI-powered personal weather forecasts promise to boost crop yields

The AgTech Trends 2023 survey recently unveiled​ by the Yield revealed the concerning the impact of weather unpredictability. Over 90% of agribusinesses acknowledged that climate change is affecting their crop yields, with 36.7% observing it makes yields highly unpredictable.

The need for more accurate weather forecasting is therefore evident, with nearly 80% of specialty crop respondents believing improved forecasts would enhance crop quality. The research added that even with data from established sources, trust in weather predictions remains low. Enhanced and precise weather forecasting stands out as a potential solution, with many seeing it as a way to boost crop quality and reduce weather-related risks.

Solutions are coming as quick as a change in the weather. One is from San Francisco, California-based start-up Benchmark Labs. It is leveraging AI and machine learning to create environmental forecasts that can better inform agricultural operations. 

“Farms have multiple microclimates”, CEO Carlos Gaitan told AgTechNavigator. Farms have changes in elevation; might be close to lakes; even the type of grass or crop creates different conditions. Current weather forecasts at regional or grid level, meanwhile, are outdated and ineffective in the face of these multiple microclimates and changing weather, which can lead to significant financial and operational losses.

The company is training more than 1,500 machine learning models a per week tailored to each specific location so farmers get more actionable alerts. The technology has helped California growers to save 10% of water and increase agricultural yields 50-100%, it claims. 

The company currently serves customers in Europe, South America and North America, typically specialty crop growers. Customers include wineries in Catalonia and the US West coast and citrus and avocado growers in South America. Temperature changes are one of the biggest weather challenges they are facing, he told us. 

In 2021, for example, winemakers in France were hit by the worst frost in decades. Burgundy lost around 80% of its crops. “Frost can devastate wineries,” said Gaitan, who earned his PhD in atmospheric science and is a former member of the American Meteorological Society's AI Committee.

Warmer temperatures – 2023 was the hottest year on record – can be equally devasting to crops and can impact labour costs too if, for example, it is too hot for workers to be outdoors.

How can AI help? “What we do differently is create weather forecast tailored to each specific location. Traditional weather forecasting all over the world divides the world in boxes an grids where everybody in that box receives exactly the same forecast. That's a non-local approach and that works for many applications but if you are an owner of a fixed asset like a farm, wind, solar, agriculture farm, what you care about is the forecast at your specific location.”

How it works​

Benchmark collects data from on-site sensors. It then analyses this data using its proprietary AI software to provide more accurate forecasts tailored to a specific location.

“Farmers might have sensors located 200 metres apart that have different environmental conditions” explained Gaitan. “Our AI tech allows them to get two different forecasts. We have cases where users have one sensor on top of the other to get the temperature above and below the canopy. So it’s at the same location but different height. That's the advantage of our approach. We can provide you forward-looking data so you are better prepared for what's going to happen in the hours and days ahead.” 

Benchmark Labs estimates there are already 22 million sensors in agricultural land globally. The plan is therefore to get the owners of these existing sensors using its AI-powered forecasts instead of the ones provided by the nearest weather station. “The margins in agriculture are thin and sometimes it’s hard to go and ask a grower to replace a sensor that they have already invested in just to get analytics,” explained the CEO. The company, however, is still looking to partner with hardware companies with a view to selling both sensors and its weather data to farms too.

Labour, water, pesticides and fertiliser benefits ​

Accuracy depends on location, Gaitan said. Users have seen improvements of up to 84% in comparison with forecasts they had been getting from the national weather service in the US, he claimed.

One group to benefit were land managers in Minnesota and Nebraska carrying out controlled burns to prevent forest fires. “We improved their relative humidity and wind speed forecast. That’s important if you are doing controlled burns.”

Gaitan claims that better weather forecasts can improve a host of metrics including labour, water for irrigation, pesticides and fertiliser management. Farmers can lose thousands of dollars if they apply pesticides before a rainfall, for example. More accurate forecasts can therefore further assist runoff and leaching management.  

The aforementioned AgTech Trends 2023 survey also highlighted the problem of agribusiness workers often struggling to derive actionable insights from their data. Gaitan is aware of this challenge. Users can access forecasts every hour via a web interface (or app) or via an API on a computer. “Our whole goal is to make it actionable and very easily to use,” he said.