Hundreds of millions of people worldwide suffer from acute food insecurity. This emergency calls for timely preventive and response actions, which depend on our ability to anticipate food crises. Since food security results from complex combinations of societal, economic, political, and environmental factors, most existing models require extensive data and are based on many assumptions, limiting their realism and applicability across time and space. Here, we show that accurate models predicting the yearly onset of food crises can be generated using only temperature and precipitation data. When combined with demographic and poverty projections under different socioeconomic pathways, these models allow us to explore future global scenarios. Conflict and inequality pathways (SSP3 and SSP4) could expose more than 1.1 billion people—mostly in Africa and Asia—to at least one severe food crisis by century’s end. Of these, more than 600 millions would be under five years old at first exposure, and more than 230 millions would face a crisis in their first year of life. In contrast, a shift toward environmental and social sustainability (SSP1) could more than halve current yearly exposure and reduce worst-case cumulative exposure by 69%. These findings highlight that today’s policy decisions may lead to radically different food security futures.
| Authors | |
| Publisher | Nature |
| Geographic coverage | Global |
| Originally published | 22 Jan 2026 |
| Knowledge service | Metadata | Global Food and Nutrition Security | Food crises and food and nutrition security |
| Digital Europa Thesaurus (DET) | ForecastingModellingmachine learningrisk managementdisaster risk reductionpolicymaking |