Leading image

Digital agriculture: Making the most of machine learning on farm

Dossier: Artificial intelligence in agriculture

AtlasAI integrates satellite information and AI to provide data on agricultural outcomes, as shown with maize yields in Kenya

© AtlasAI

The ability of agricultural equipment to think, predict and advise farmers via a variety of artificial intelligence (AI) applications presents Africa with the potential to achieve food security.

 “AI is the broader concept of machines being able to carry out tasks in a way that is considered smart. The smart processes include machines being able to function automatically, reason and learn by themselves,” explains Claudia Ayin, an independent ICT consultant. Machine learning is the aspect of AI that allows computers to learn by themselves. “Machine learning is therefore a branch of AI that is able to process large data sets and let machines learn for themselves without having been explicitly programmed,” she adds.

According to MarketsandMarkets, an Indian research company, in 2018 the worldwide AI in agriculture market was valued at €545 million and, by 2025, is expected to reach €2.4 billion as more and more smallholder farmers adopt new, data-driven technologies. With the help of data scientists and big tech companies, small-scale farmers in ACP countries are increasingly benefiting from the predictive abilities of AI and machine learning in order to access finance and insurance, predict yields and tackle pests and diseases, to run more profitable and ‘smarter’ sustainable farms.

AtlasAI integrates satellite information and AI to provide data on agricultural outcomes, as shown with maize yields in Kenya

AtlasAI integrates satellite information and AI to provide data on agricultural outcomes, as shown with maize yields in Kenya

© AtlasAI

Data matters

In order to develop effective AI solutions and understand how smallholder farmers use AI and machine learning, agri-tech companies need high-quality data. The future of farming therefore lies in collecting and analysing quality agriculture data in order to maximise efficiency.

Availability of data is crucial. For example, climate uncertainty increases risk for farmers. “It’s not raining less, it’s just more variable,” explains Wesley Black, a farm planner from Bloemfontein, South Africa. “AI will become essential when it comes to helping small-scale farmers tackle climate change.” So for climate resilience, crop and livestock insurance is a key element. However, data is crucial for insurers who need to know the likelihood of crop failure; lenders need to know the likelihood of default, and traders need to know where surplus and deficit regions are. But few, if any, data sources exist that provide this kind of mass information at a broad scale.

African scientists can now have access to free and open source satellite data as a result of a deal signed by the African Union with the European Commission’s Copernicus programme in 2018. But using satellite data to predict weather patterns is no easy task. IBM, for example, processes data from multiple satellites using Watson’s Decision Platform for Agriculture, which aims to combine predictive analytics, artificial intelligence, weather data, and Internet of Things sensors to give farmers insights on ploughing, planting, spraying, and harvesting. Each satellite provides a digital image at different intervals, be it vegetation, soil and water cover, sea and land surface temperature or weather patterns. Using varied AI techniques and algorithms, IBM puts together all the data points to create a more in-depth and usable image of a farm: “Every day we receive around 45 terabytes of data but the data cannot be used as is. Every satellite gives you a portion of a farm, but not one gives you an actual representation of the farm. With AI, we fuse all these images together to get the full picture,” explains Kommy Weldemariam, IBM Research Africa’s chief scientist.

AI potential for protecting against pests

Besides climate change, pests and diseases are a key challenge for small-scale farmers and is one that will be further exacerbated by climate variability. Each year, according to CABI estimates, about 50% of Africa’s crops are lost to pest and diseases.

“Hundreds of millions of African farmers are already suffering from the effects of climate change,” says David Hughes, an entomologist from Pennsylvania State University, and the leader of the project that created Nuru, an Android tool, which has been developed to diagnose crop diseases even without an internet connection. Developed by Penn State’s Plant Village and the International Institute of Tropical Agriculture, Nuru is used in several African countries, including in Kenya in collaboration with SelfHelp Africa to diagnose mite and viral diseases in cassava, as well as to identify fall armyworm infections in maize. Advice from experts – mainly at CGIAR, FAO and governments – is sent offline and in local languages (currently in Swahili, French, Twi, Hindi and English). Although still in beta testing, 28,000 cassava farmers across seven counties in Kenya will benefit from the tool this year. “Digital tools are increasingly becoming integral components… of identification, monitoring, training, and decision-making of globally-important crop pests and diseases,” Hughes states.

A new AI tool that can predict crop growth and help protect vital food supplies from intensifying heat is being added to Nuru. It uses data from a UN satellite that tracks a decade’s worth of information about water availability, along with weather forecasting, to determine crop productivity. “AI offers the potential to get a single set of eyes to look at this problem,” says Hughes. “Nuru is like an extension officer that is always there for farmers, in their fields.”

Hughes believes that in low-income countries that lack human capital in fields like agricultural science, there is an opportunity to use AI to help break the cycle of poverty. Founded in August 2018, Agrix Tech, based in Yaoundé, Ghana, is also using AI to help farmers tackle pests and diseases. Using a mobile phone app, farmers scan the leaf of an infected crop. The app uses an AI library to analyse the issue and provide treatment recommendations via text and voice messages, in customised African local languages, for those who cannot read. According to Adamou Nchange Kouotou, Agrix Tech’s founder and CEO, the app has a 99% accuracy rate and, most importantly, does not need the internet to function.

By utilising AI, Hello Tractor’s platform provides farmers with timely and relevant information to increase their yields

By utilising AI, Hello Tractor’s platform provides farmers with timely and relevant information to increase their yields

© Hello Tractor

Banking on AI

In order to deliver economic and agricultural insights to farmers throughout Africa, AtlasAI – a Silicon Valley tech company that addresses economic data and market intelligence needs in developing countries – uses technology that integrates satellite information and AI with high-quality data from the field. AtlasAI currently generates data for all African countries and is working with organisations that serve governments and farmers in multiple countries: “At AtlasAI, we use cutting edge AI and satellite data to provide granular, accurate, and scalable data on agricultural outcomes across the continent,” explains Marshall Burke, a professor at Stanford University and one of AtlasAI’s three co-founders.

For example, smallholder farmers are underserved by most financial markets; they have difficulty borrowing, they are unable to buy insurance, and they are often at a disadvantage in non-competitive trading environments. With AI and the right data sources, this is an issue AtlasAI looks to solve: “Having accurate, low-cost data on smallholder farmers allows companies to actually design products and services that fit their needs,” adds Burke.

Moving from products to homegrown services that incorporate advanced agricultural analytics and AI is something Hello Tractor, a US start-up based in Kenya and Nigeria, has made work. When Hello Tractor first launched in 2014, their flagship product was an affordable, ultra-low horsepower, two-wheel tractor fitted with monitoring technology. “We sold these to enterprising farmers or cooperatives, who then accessed our tractor-sharing platform to identify and service additional demand from smallholders,” explains Jehiel Oliver, Hello Tractor’s CEO.

In January 2017, Hello Tractor made the strategic decision to focus more on their application than on the tractors themselves. It proved to be an effective model, allowing Hello Tractor to capture 75% of private commercial tractor inflows to Nigeria, expand to five markets across Africa through strategic partnerships, and touch the lives of over 250,000 farmers.

Hello Tractor, in partnership with IBM, is now piloting an advanced agricultural analytics and decision-making tool that cuts across the mechanisation ecosystem. Their data sets are used for fertiliser, seed and financial companies to access real-time, unfiltered information about a farm. “Utilising AI, farmers on the Hello Tractor platform gain access to timely and relevant information to increase their yields, tractor fleet owners receive insights to save time and earn more, and banks are empowered with information for better underwriting and portfolio management,” explains Oliver. “More specifically, we can apply machine learning to not only help predict when farmers should receive their tractor services, but this data can also be mined to develop advice on what inputs should be applied and when,” he says.

Thinking of Africa

From IBM to Deloitte, Amazon Web Services and Google, there are many key corporate players which are working throughout the African continent and partnering with smaller companies and farmers to create locally-relevant AI-focussed solutions. “With AI, we have the potential to unlock increased yields, higher revenues and lower losses,” says Isaac Sesi, the co-founder of Sesi Technologies, an agri-tech company which develops hardware and software solutions for farmers and agribusinesses in Africa. “Yet, it is necessary that these AI tools are developed taking into consideration the context of local agriculture in Africa to ensure that these solutions are relevant and applicable to African agricultural systems, thus necessitating that Africans, who best understand African problems more intimately, be at the forefront of the development of these tools.”

Microsoft, through its 4Afrika Initiative, supports the digital transformation of African agriculture.

“A while back we noticed that agricultural companies were trying to make sense of things like big data, AI, machine learning and analytics. These things rapidly moved their way into agriculture and most companies were not equipped to handle or deal with these new technologies. The need for technology companies like Microsoft to help them navigate this shift is critical,” says Amrote Abdella, Microsoft 4Afrika’s regional director. “We believe technology advancements have the potential to drive significant economic growth and societal impact, specifically because technology is able to bridge gaps in infrastructure that have previously kept people locked out of the formal economy and unable to access essential services on the continent.”

Working with Felix Musau from Kenya, Microsoft helped to develop AGIN, a mobile and Azure cloud-based service that connects farmers to much-needed credit services. Using mobile phones, farmers can capture information like farm size, location, soil composition and crops grown. Using Azure’s built-in AI and machine learning tools, AGIN then helps farmers to establish a credit profile, allowing them to access resources like small loans, with credit lines they can use without ever visiting a bank.

AGIN has served over 140,000 farmers in Kenya and facilitated over US$1.3 million (€1.2 million) per month in transactions, including loans and insurance. After receiving financial and technical support from Microsoft 4Afrika, AGIN now has hopes to expand its reach to 300 million farmers in sub-Saharan Africa by 2020.

Microsoft 4Afrika also works closely with Tulaa, a commerce solution for rural African farmers. Tulaa uses mobile technology and mobile money to enable farms to save and borrow to purchase inputs and agricultural advice, and market their crops at harvest. Though 4Afrika’s AI PopUp Lab in Kenya, Microsoft assisted Tulaa with integrating machine learning into their model to assess credit worthiness. “AI and machine learning offer unprecedented opportunities to reach millions of farmers far more efficiently than we could in the past. At Tulaa, we are only beginning to realise the potential of this technology to transform supply chains and the lives of smallholder farmers,” says Hillary Miller-Wise, Tulaa’s CEO.

“The ability for platforms like Tulaa to scale will be driven in part by our ability to harness the power of AI. The zeitgeist now is around bundled services and platform solutions for smallholder farmers. One of the major reasons that these models are emerging is the availability of AI and machine learning that was not there even 5 years ago,” adds Miller-Wise. CTA is also encouraging young developers to embrace data analytics and AI. During the 2019 edition of Pitch AgriHack – CTA’s competition supporting young digital agriculture start-ups – a special prize on “Data Analytics” was awarded.

The future of agriculture is set to be more automated and data-driven. With innovative, African-driven solutions that make use of AI, smallholder farmers throughout the continent will be able to make smarter, data-driven decisions, becoming more proactive and profitable, as they farm for the future. “We’re just at the start of a revolution in AI,” says Tom Ilube, a futurist and the founder of the African Science Academy in Ghana. “At the heart of AI is the algorithms that drive it. These algorithms define the future. If Africa is going to be a part of the future, we need to be defining the algorithms that tell us what the future looks like.”