Can AI help Nigeria reduce agricultural losses during flooding, nip disease outbreaks?

Technology adoption, particularly in Agriculture has often been slow in Nigeria, yet the potentials to make tremendous impact exists. The use of Artificial Intelligence in some parts of the world, may offer solutions to some of the problems confronting agricultural productivity in Nigeria, such as recurrent flooding and disease outbreaks.

The flooding this year across 12 states has been reported to have caused over 100 deaths, destroyed thousands of houses, as well as several thousands of hectares of farmland.

The ACAPS Flood Briefing Report in September, stated new estimates indicate that 122,653 hectares of agricultural land have been flooded across central and southern Nigeria. Crops were destroyed before the harvesting season begins in October. This makes the affected population more vulnerable to food insecurity, and negatively impacts the livelihoods of farmers. Flooding is also likely to affect other livelihoods, such as fishers and petty traders.

READ ALSO: The fury of floods and food insecurity

The flooding of farmlands as recorded this year is becoming a recurring, annual event, and this is putting food security at risk. It is also leaving already poor (smallholder) farmers further impoverished. However, if the use of Artificial Intelligence in India in a partnership led by Google can be replicated in Nigeria, then managing the flood outbreaks may become easier to achieve.

Bridget Gosselink, head of Product Impact at Google.org, said in an interview during the Google Making AI event which took place in Amsterdam last week, that the way the current models are built, is best at forecasting river base flooding. “If flooding was caused by rivers, then great, because we are actually good at that,” said Gosselink.

The model works by combining data from stream gauges; both current and historic data as provided by the water commissions. It then uses traditional models from existing work in modelling of floods, as well as information that can be extracted from terrain maps that provide information on elevation changes. This, combined with other data sets are put together, and using some predictive elements, makes it possible to get a much more granular forecast of flooding and how it will happen.

As Gosselink explained, the current way (flood forecasting) is done tends to be very broad, but this new method provides more accurate probabilities so that individuals can take action to save lives, properties, and salvage as much investments as possible; in this case agricultural produce.

Yossi Matias, vice president of Engineering at Google, who leads efforts in Search, Research and Crisis Response explained that the Flood Focusing model has been deployed in India, and with lessons from its usage, can be scaled globally with time.

Matias who said he was in Lagos last year during the severe floods in some parts of the Lagos Island, is already familiar with the situation in Nigeria, even though the disaster caused in agricultural communities far exceed what he witnessed in Lagos.

He explained that the Flood Focusing initiative was set up a little over a year and half ago, to experiment how Machine learning and other technologies can be used for better forecasting of flood.

Activated for the first time in India this past September, Matias explained the project is a collaboration with the Indian government, which is providing the measurements for water levels in the river. Google on its part then uses machine learning technologies and other simulations in the cloud of other hydraulic models in collaboration with some research groups in the academia, to help achieve floods forecasting with over 90 percent accuracy. The use of satellite imagery to analyse water levels instead of relying on measurements is also being explored.

All of this knowledge BusinessDay learnt, is transferrable, but the right partnerships are required to trigger the decision to setup in Nigeria.

Apart from flood prediction, Artificial Intelligence can also serve in early detection and treatment of crop diseases. This is at least going by the present use case in Tanzania where farmers can take photographs of disease-infected areas of Cassava crops, which is then analysed to indicate what particular disease it is, and how it can be treated.

Jeff Dean, head of Google’s AI division, in a video chat, explained that a research team at Pennsylvania State university in the United States and the International Institute of Tropical Agriculture (IITA) in Tanzania, worked together in developing this model. According to him, they worked together to create a machine learning model that used computer vision of pictures taken of a cassava plant and can detect what kind of disease that plant had, and how to treat it. It runs entirely on mobile phones without any network challenge, even on low-end android phones.

“So, in the middle of the field, a farmer can take a picture of a plant and get helpful advice on how best to treat the disease they are seeing their crops having. This is an example where AI is helping ensure better food security,” Dean said.

There is also a collaboration on famine between Google, the World Bank, United Nations and some other organisations, to try and build models to identify famine situations. This will make it possible for governments, agribusinesses and even individuals to take action before crisis occurs. In all these use cases, Nigeria can benefit from the applications of AI to safeguard agricultural productivity, but predicated on being positioned to attract the right partnerships into the country to deliver these technologies.

CALEB OJEWALE