Armed with AI, Nigerian farmers act before floods

Bukola Adebayo, Reuters

Artificial intelligence and traditional aid are helping farmers like Salihu Ali in northern Nigeria act before their farms are submerged by increasingly regular floods fuelled by climate change. Ali was warned that floods were coming in 2022 and was provided with cash to prepare his farm in Dasin Hausa, a remote village of cowpea, maize, and rice farmers in Adamawa state. In response, he harvested his maize early and moved the barn where he kept seedlings and fertiliser uphill. “We (farmers) contributed money to buy canoes to transport our crops, seedlings, fertilisers and other farm inputs when the water became too high,” Ali said. He was among 1,450 families who were given $450 (195,000 naira) in the pilot project, which relied on AI to develop flood forecasting models based on large amounts of data, ranging from satellite information to historical events to river levels. The project, run by the International Rescue Committee (IRC), a nongovernmental humanitarian organisation, in partnership with Google.org, the tech giant’s philanthropic arm, has now been renewed and expanded with a cash injection of $4.6 million from Google to IRC and the non-profit GiveDirectly. The idea is simple: harness AI to predict floods before they happen and give communities cash before the waters rise so they can prepare their farms and limit the damage to their homes. Forecasts from Google’s AI tool Flood Hub and data from national authorities will be used to strengthen early warning systems and trigger payouts to 7,500 people in flood-prone areas in the northern Adamawa and Kogi states. Google’s Flood Hub uses open-source data, including weather forecasts and satellite imagery, to predict floods up to seven days in advance.

Warming temperatures due to climate change are increasing the intensity and frequency of Africa’s rains, according to United Nations climate experts. Extreme weather events are undermining food security in Nigeria, Africa’s most populous nation. But early warning systems can help reduce fatalities and mitigate the risk of widespread hunger caused by regular floods.

The floods in 2022 destroyed 770,000 hectares of farmland and submerged thousands of homes across several Nigerian states. More than 600 people were killed, and at least 2 million people had to flee their homes. But the damage could have been worse. A few weeks before the rains came, researchers from IRC’s Airbel Impact Lab prepared flood forecasts using hydrological and meteorological data gathered from its flood monitoring system and Nigeria’s meteorological agency. Clare Clingain, research coordinator at the Airbel Impact Lab, said they shared the information with local authorities and IRC workers in Nigeria, who organised meetings and distributed flyers to warn residents. “When designing anticipatory action, everything is based on a forecast, and you want to make sure that forecast is reliable and accurate,” Clingain said.

In Adamawa, more than half of the farmers who received cash said they used the money to harvest crops early, elevate storage barns from flood paths and stockpile food, according to IRC’s evaluation report. The forecasts drew on data from the European Union’s Global Flood Awareness System (GloFAS), an AI model that uses satellite flood detection algorithms, in addition to information from Nigeria’s meteorological and emergency agencies. The IRC says it will now use forecasts from Google’s Flood Hub and a combination of other tools to get accurate predictions.

Clingain said accuracy was key to persuade farmers to act. In the past, forecasts have been unreliable or warnings sometimes came too late for farmers, or aid agencies, to act. Alex Diaz, head of AI for social good at Google.org, said timing was also critical. “If you get an early warning a day before, that might save your life, but if you get it five days to a week in advance, that can save your livelihood when you have the cash to help you do that,” he said. GiveDirectly, which makes cash transfers to the mobile phones of people living in poverty, has also tested the approach in Mozambique. AI tools can help bridge gaps in gathering weather-related data, while early forecasts help aid agencies redefine their disaster responses, said Vera Lummis, a senior manager of digital innovation at GiveDirectly. “The traditional way of delivering disaster aid is that it’s done after a disaster, but unfortunately, in a lot of cases, cash assistance or aid many times doesn’t reach families affected (until) months after a disaster,” she said. “(This) makes it much harder for them to recover and rebuild their lives, especially after floods.”