evidoria

← Back to browse

Good practice Imported

HungerMap LIVE's AI Nowcasting Tracks Sudan's Slide Toward Famine

Sudan · Port Sudan · See the Sudan profile

Evidence: Descriptive / self-reported Top 25% 73/100 · Ask Evidence Copilot about this practice

WFP's HungerMap LIVE uses machine-learning "nowcasts" to estimate food insecurity where ground surveys are impossible. In Sudan's civil war, it flagged 21.2 million people (45% of the population) in Crisis-level hunger by September 2025, with famine confirmed in multiple areas.

21.2 million (45%)
People in Crisis-level food insecurity, Sudan (Sep 2025)
375,000
People in Catastrophe (IPC Phase 5) (2025)
-25%
WFP field data footprint decline (past year)

Details

Maturity
Established
Promoter
World Food Programme (WFP)
Period
2023-2025
Keywords
food security, humanitarian response, disaster forecasting, public health

Context

Sudan's civil war, ongoing since April 2023, has displaced millions and cut off traditional door-to-door food security surveys in famine-hit areas such as Darfur and Kordofan.

Activities

WFP's HungerMap LIVE, rebuilt with Google.org and Gates Foundation support, combines over 300 food-security analysts with machine-learning nowcasting models fusing conflict, climate, market and population data across 16 UN-designated Hunger Hotspot countries, including Sudan.

Results

By September 2025 the IPC (a separate corroborating body) found 21.2 million people (45% of Sudan's population) in Crisis-level (IPC Phase 3+) food insecurity, including 375,000 in Catastrophe and 6.3 million in Emergency; famine was confirmed in at least five areas through May 2025.

Conclusions

WFP's own field data footprint shrank 25% over the past year, meaning the AI nowcasting layer is increasingly filling gaps field collection can no longer reach, but no independent accuracy evaluation of the model itself is published.

Implementation
Indicative cost
High (€500k–€5M)
Time to results
Long (> 3 years)
Staffing & skills
300+ WFP food-security analysts, machine-learning nowcasting model team

Conditions for success

  • fusion of conflict, climate, market and population data
  • corroboration against independent IPC analysis

Common failure modes

  • shrinking field data footprint in hard-to-reach areas
  • no independent accuracy audit of the nowcasting model itself

Commonly funded by

Philanthropic / foundation funding

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Similar practices you may find useful