Leveraging data for early detection of food and nutrition crises in Yemen

Yemen is currently experiencing one of the world's most severe and prolonged humanitarian crises, marked by conflict, economic collapse, and natural disasters. In 2022, an estimated 17 million people faced food crises or worse situations in Yemen, unable to meet their minimum dietary needs without resorting to irreversible coping strategies.

Conventional food insecurity assessment methods typically involve extensive data collection, consensus-building processes, and expert evaluations.  However, these approaches can be prone to access limitations, data collection issues, and differences in subjective analysis. The new modelling behind the Joint Monitoring Report—a bi-monthly food and nutrition security report delivered in collaboration between The World Bank, FAO, WFP, UNICEF, WHO and ACAPS—introduces an innovative data-driven methodology aimed at the early detection of food and nutrition security emergencies in Yemen. As outlined in the recent World Bank working paper and related reproducibility package, this innovative approach combines quantitative indicators with statistically optimized thresholds to generate reliable food and nutrition security alerts.

This data-driven approach focuses on simplicity and transparency by utilizing a core set of indicators to regularly monitor food and nutrition security crises. These efforts leverage and complement existing systems such as the Integrated Food Security Phase Classification (IPC) and Famine Early Warning Systems Network (FEWS NET) and promote greater simplicity and transparency in food and nutrition security modelling.

The analysis utilizes several key indicators, each optimized to capture critical dimensions of food and nutrition security:

1. Food Prices: Average prices of top-performing food items, with deviations from exponential moving averages signaling potential emergencies.
   
   
2. Fuel Prices: Prices of petrol and diesel, critical for transportation and food distribution, analyzed in a manner similar to food prices
   .
3. Exchange Rate: Volatility in the Yemeni rial (YER) to USD exchange rate, influencing purchasing power and food affordability.
   
   
4. Drought: Standardized Precipitation Index (SPI) indicating drought conditions impacting agricultural productivity.
   
   
5. Conflict: Average conflict fatalities in districts and neighboring areas, reflecting security and access issues.
   
   
6. Displacement: Summed displacements to and from districts, indicating severe disruptions to livelihoods.

By utilizing IPC-compatible data, the team developed a target variable model to establish optimized thresholds used to identify historical transitions into more severe states of acute food and nutrition insecurity. The model strikes a balance between false positives (that is, incorrectly identifying a period as a period of food security deterioration) and false negatives (that is, incorrectly missing a period of food security deterioration) to enhance the predictive power of each indicator. As a result, historic validation shows that using only these 6 core indicators can reliably identify major deteriorations in food and nutrition security approximately 80% of the time when compared to historical time series.

The graphs below show the combination of food security alerts over time for each indicator, providing a visual representation of historical food security crises in Yemen. The graphs can provide a breakdown of each food security risk alert over time as well as by geography, enabling a nuanced understanding of food security risks at a subnational level.

Generating such strong results with only a small number of indicators represents a significant innovation for modern food and nutrition security analysis. Such information is useful to practitioners and decision-makers leading crisis responses in the country by providing teams with high frequency updates of an emerging crisis.

This methodology offers a scalable solution for other regions facing similar data challenges such as Somalia, the Sahel region in Africa, and can support food and nutrition security monitoring in countries developing Food Security Crisis Preparedness Plans (FSCPPs). The analysis shows that simple, data-driven models can reliably detect impending food and nutrition security emergencies, providing a crucial lead time for anticipatory and early action—critical interventions for saving lives, livelihoods, and making the most of limited humanitarian and development resources.