The 2030 Agenda for Sustainable Development, adopted by all UN member states in 2015, sets a 15-year plan of action to improve everyone’s lives and prospects, and to protect the planet. This plan includes 17 goals and 169 targets, from ending poverty and hunger, to reducing inequalities, and to making cities more sustainable. Some of these goals are overly ambitious and considered by many critics to be unachievable in just a 15-year span. They were nevertheless in part devised as a roadmap to guide the international development agenda, for which they set a direction to work towards. However, halfway through the intended timeline, some targets are now further away than they were in 2015.

This is certainly the case for Sustainable Development Goal (SDG) 2, whose first target is to “end hunger and ensure access by all people, in particular the poor and people in vulnerable situations, including infants, to safe, nutritious and sufficient food all year round”. The right to food is recognized in the 1948 Universal Declaration of Human Rights as part of the right to an adequate standard of living, and it was reiterated in the 1966 International Covenant on Economic, Social and Cultural Rights. Yet, the Global Report on Food Crises 2023 estimates that over a quarter of a billion people were acutely food insecure in 20221. In 2016, they numbered 105 million, a 140% increase in six years, in large part due to the socioeconomic impact of the COVID-19 pandemic, the repercussions of the war in Ukraine, and the calamitous effects of climate change1.

Along with food security, access to safe drinking water, after being recognized as an essential human right by the UN General Assembly in 2010, was also included in the 2030 Agenda as SDG 6, whose first target is to “achieve universal and equitable access to safe and affordable drinking water for all”. According to the State of the World’s Drinking Water, published in October 2022, between 2000 and 2020 the percentage of the global population with access to safely managed drinking water services at home has increased from 62% to 74%2. The report however also remarks that, despite this increase of over 2 billion people, none of the world regions are on track to achieve universal access to safely managed drinking water services by 2030, and that globally, a quadrupling of current rates of progress is needed to meet the aforementioned target.

Furthermore, in the years to come, efforts to achieve food and water security will have to face the rapidly increasing effects of climate change. The Intergovernmental Panel on Climate Change (IPCC)’s sixth assessment report3 warns that climate impacts on food systems are projected to increase undernutrition and diet-related mortality and risks globally, especially in sub-Saharan Africa, South Asia and Central America, where higher temperatures and humidity increase toxigenic fungi on food crops and harmful water-borne diseases. Drought and flood risks are projected to increase with every degree of global warming, and the combined effects of water and temperature changes to cause increased risks to agricultural yields.

Leveraging computational tools

To address the needs of a growing global population facing climate change, conflicts, and economic and health shocks, and achieve access to quality food and water for everyone everywhere, innovative solutions are needed. In these, technology and computational models play a key role, but, as already highlighted by many, technology is not a panacea, and such solutions should always be designed, developed and implemented in close collaboration with community stakeholders and never disregard the local context.

Technological innovation in agriculture was accelerated in the last decade by the increasing use of information and communication technology, which is nowadays widely spread also in the Global South. Examples include real-time data collection through internet of things (IoT) technology to provide farmers with personalized prompt advice, and the development and use of unmanned aerial vehicles for farming purposes (such as application of fertilizers or harvesting of fruits), and, when combined with lightweight hyperspectral cameras, to calculate, for instance, biomass development and fertilization status of crops. Walter et al. suggested that, thanks to these improvements, agriculture is undergoing a fourth revolution that will generate disruptive changes in farming practices4.

Blockchain technology in agriculture has also been explored as a means to track the provenance of food and thus to help create trustworthy food supply chains, and, as a trusted way of storing data, to facilitate the use of data-driven technologies to make farming smarter. However, challenges related to costs and integration with existing systems seem to currently outweigh the potential benefits5.

Food insecurity can also be driven by shortages caused by local shocks to agricultural production that propagate across the global food production network. To understand and predict such effects and their consequences, it is key to take supply interdependencies into account. Network science provides for this the most suited framework of mathematical and computational models, as recently demonstrated in the case of the Russia–Ukraine conflict6.

The computational field that has undeniably attracted the most attention in recent years, in this and many other sectors, is artificial intelligence (AI) and machine learning. In the context of food and water security, this has been driven by the rapid combined improvements in satellite technology, with worldwide satellite imagery now freely available to the public (Sentinel, Landsat and MODIS), and in deep learning algorithms, which have presented an important opportunity to improve the automated analysis of agricultural lands and water bodies. Examples include monitoring and predicting soil fertility, early detection of crops diseases, monitoring of crop’s growth, and predicting yield7. Some solutions are instead based on imagery that can be captured directly by farmers’ standard smartphone cameras and analyzed in real-time through dedicated apps. For instance, PlantVillage Nuru is a mobile AI assistant piloted in Kenyan and Tanzanian communities that diagnoses cassava diseases, even without an Internet connection. AI-based water quality monitoring solutions have also been proposed. For example, Clean Water AI uses a convolutional neural network to detect dangerous bacteria and harmful particles in water. This system could potentially be used by cities through the installation of IoT devices across different water sources to monitor water quality and contamination in real time.

Predicting how many people lack access to quality food and water and where, and how these numbers are likely to evolve in the future, is key for governments and humanitarian organizations to make informed and timely decisions on relevant policies and programs to achieve the aforementioned goals. In 2020, the UN World Food Programme launched the HungerMap LIVE, a platform that combines near real-time data collection and machine learning8 to estimate the number of people that are currently consuming insufficient quality food or have to resort to coping behaviors because of constrained access to it. The Famine Action Mechanism (FAM), a global partnership dedicated to anticipatory action to prevent and mitigate food insecurity, has explored the use of statistical models to help predict the outbreak of food crises and to simulate crisis scenarios over longer time horizons9. Finally, forecasting water demand has also been the focus of a vast body of literature10, and the operationalization of one such model was recently piloted in Moulton Niguel Water District in California, an area severely affected by droughts.

Challenges and opportunities

Despite the many relevant examples of how computational tools can help monitor and improve access to quality food and water, important challenges and limitations also come with these approaches, which, if not seriously addressed, might hinder progress towards the aforementioned targets. For instance, low-income farmers might not be able to afford to integrate these technological innovations in their agricultural practices, and become even more vulnerable because they are excluded from new technology adopted by wealthier competitors. Government and development organizations will have to develop accurately designed funding and credit schema in order to overcome this challenge, which goes beyond economic access. Since farmers in many countries are illiterate, or non-English speakers, these solutions need to communicate results in local languages and provide as much information as possible through visual aids11.

Using a consensus-based expert elicitation process, Vinuesa et al. found that AI can currently enable the accomplishment of 134 targets across all of the SDGs (including six of the eight SDG 2 targets and all eight SDG 6 targets), but it may also inhibit 59 targets (including two SDG 2 and five SDG 6 targets)12. As discussed, AI-enhanced equipment might, for example, be unaffordable or inappropriate for small-scale farming, inhibiting the increase of small-scale food producers’ agricultural productive capacity and incomes (targets 2.3 and 2.a). Moreover, AI could, via automated decision making, hinder the participation of local communities in water- and sanitation-related programs (targets 6.a and 6.b). The authors suggest that, to avoid these inhibiting effects, additional research is needed to assess the long-term impact of such algorithms on equity and fairness. As already widely discussed for other AI applications, a common source of bias is the data that these algorithms are trained on. For instance, it was observed that agricultural tasks such as monitoring soil health and crop damage lack appropriate datasets for training, and this entails that deep learning in this area is restricted to those who have the means to run large data collection campaigns4. Hence, making data and code publicly available by sharing them on open repositories is key for the research community to capitalize on one another’s work and thus make more rapid and effective progress towards solutions for all.

As climate change’s impact on food and water security accelerates, future work will increasingly focus on developing computational models to forecast extreme weather events that can affect clean water sources and food production. These models will support preparedness action through early-warning and scenario exploration systems.

Moreover, beyond food production, an area where we should see more developments in the future is food waste. The United Nations Environment Programme (UNEP)’s Food Waste Index Report 2021 estimates that 17% of total global food production may be wasted13. Machine-learning-enabled food sharing and redistribution mobile applications have been developed to address this issue on the retailer and consumer side14, however additional work is needed to develop more systematic solutions along the whole food chain, from harvest to retail.

The computational tools discussed so far are of two kinds: either pilot projects developed in and for a specific setting, whose scale-up proves difficult, or across-the-board solutions, which are however then challenging to adjust to work effectively in local contexts. Bridging these two approaches to build solutions that are developed with the community and yet scalable is the main challenge we should try to solve. This will require capacity-building and close collaboration among all relevant stakeholders: local communities, local and international governments and organizations, academic researchers and industry partners. It is time to join forces to accelerate progress towards access to safe drinking water and nutritious food for everyone, everywhere.