GEOINT · Human security · Spatial communication
Sudan Food Security
Where Hunger Follows Conflict
A TCPED-structured GEOINT assessment and public StoryMap answering where food insecurity in Sudan is most acute, what drives it in each place, and where intervention can still change the outcome.
01Operational question
Fighting between the Sudanese Armed Forces and the Rapid Support Forces began in April 2023 in the capital region and spread to Darfur, North Kordofan, Gezira, and Sennar; by late 2025 the RSF had taken El Fasher. The consumers, humanitarian policy makers, strategic planners, and crisis-response coordinators, had to allocate scarce aid, access-negotiation effort, and logistics capacity across a country of 1.86 million square kilometres.
The brief posed five key intelligence questions.
- Which areas show the sharpest convergence of conflict, environmental stress, and humanitarian isolation?
- Which displaced populations face the greatest risk?
- Where are humanitarian corridors viable, and where are they blocked?
- Which agricultural zones have lost production, and why?
- Where will conditions escalate next, and with what lead time?
02Why the problem was difficult
- Attribution. Rain without harvest: separating conflict-driven crop loss from rainfall deficit requires a design that can contrast cases, not a correlation.
- Scale and cadence. Sources span 10 m to 25 km and update daily, weekly, every 16 days, monthly, or irregularly; harmonizing them without manufacturing precision is the central processing problem.
- Cloud season. Optical vegetation monitoring fails from June to September; Sentinel-1 SAR has to carry the season.
- Data-poor where it matters most. Active conflict zones have the least market, displacement, and conflict reporting, placing the greatest uncertainty where decisions carry the most consequence.
- Access modeling. A routable network is only as good as its damage and constraint layers; SAR-detected damage and OCHA constraints had to be integrated as barriers.
03My role
In a three-person team I co-authored the TCPED-structured brief across tasking, collection, processing, exploitation, and dissemination, and led the team's submission of the public StoryMap to the LA Geospatial Summit with the course instructor's endorsement.
04Data architecture
Five data pillars
| Pillar | Sources | Role in the analysis |
|---|---|---|
| Conflict and insecurity | ACLED weekly point events; territorial control | Violence hotspots, looting and siege patterns, corridor risk |
| Agriculture and food production | NASA Harvest imagery at 10–30 m with 5–16 day revisit; FAO GIEWS; GEOGLAM; FEWS NET livelihood zones | Crop stress, abandonment, yield anomalies linked to population |
| Population, displacement, vulnerability | WorldPop, HRSL, IOM DTM | Location and size of humanitarian demand |
| Environment and climate | CHIRPS rainfall, SPEI drought index, NDVI from MODIS and VIIRS | Separates environmental from conflict causes |
| Infrastructure and access | OpenStreetMap, UN Logistics Cluster, OCHA access constraints, WFP markets and prices, SAR-detected damage | Logistics modeling and corridor feasibility |
Processing protocols
Global datasets are held in EPSG:4326; regional analysis is reprojected to UTM Zone 36N, EPSG:32636. Continuous rasters resample with bilinear interpolation and categorical rasters with nearest neighbor to preserve class boundaries. Harmonization aggregates to monthly windows, interpolates MODIS NDVI to daily values for ±3-day event matching, and produces five analysis-ready layers: conflict stress from kernel density of ACLED events by type, agricultural stress from NDVI anomalies by livelihood zone, a population vulnerability composite, environmental stress from rainfall anomalies and SPEI, and an infrastructure-access layer.
05Methodology
| Approach | Method | Output | Decision served |
|---|---|---|---|
| Food production loss modeling | Crop masks and NDVI time series; anomalies against a 10-year median; CHIRPS downscaling with NDVI; overlay of conflict density and territorial control; spatial regression of crop loss on conflict intensity, rainfall anomaly, livelihood zone, and terrain | Maps separating conflict-driven, environment-driven, and compound crop loss | Livelihood support where environment dominates; conflict resolution as precondition where conflict dominates |
| Humanitarian accessibility modeling | Routable OpenStreetMap network; SAR-detected damaged segments removed; OCHA access constraints as barriers; least-cost paths from ports, border crossings, and airfields; time-distance to markets, aid sites, and hospitals | Settlements beyond 48 hours from supply; blackout zones; chokepoints | Corridor negotiation; ground versus air delivery |
| Vulnerability and exposure mapping | Population density, conflict density, livelihood zone, seasonal NDVI and rainfall anomalies, IDP sites; multiplicative composite index | Populations under simultaneous stress by livelihood type | Food aid versus livelihood support; cash-transfer targeting |
| Early warning and famine forecasting | IPC baseline; trigger variables: rainfall thresholds, conflict intensity, month-over-month price rises above 50 percent, displacement acceleration; lag analysis; scenario modeling | Thirty-day escalation zones; alert thresholds; scenario pathways | Pre-positioning, access negotiation, contingency planning |
The composite vulnerability index is multiplicative, so a population scores high only where exposure, sensitivity, stress, and displacement coincide.
V = Econflict × Slivelihood × Xenvironment × DdisplacementAttribution by contrast of cases
Where rainfall was adequate and violence high, crops failed; where rainfall was poor and violence absent, production continued. That comparison is the backbone of the claim that the geography of hunger follows the geography of conflict and access restriction, with rainfall explaining a minor share.
06Validation
- Cross-dataset validation matrix. Every source is paired with an independent observable that can confirm or contradict it.
- Stated confidence. Analytic confidence is assessed by region, each uncertainty paired with a mitigation, and every product carries its confidence level.
- Forecast-versus-outcome review. Escalation forecasts return to tasking, closing the cycle on the record of what happened.
- Design for attribution. The contrast-of-cases structure tests the conflict-versus-rainfall claim instead of assuming it.
Confidence falls with reporting density: Khartoum State carries strong conflict and market reporting; South Kordofan carries very low market reporting and low conflict-data quality.
Table view
07Output
- Intelligence brief on the TCPED cycle covering April 2023 to September 2025, with a decision matrix linking corridor negotiation to evidence and required confidence.
- Four dissemination tiers. Monthly executive briefings with a food-security priority map, famine-risk status, corridor status, and talking points; bi-weekly operational planning with accessibility, agricultural stress, market functionality, and displacement tracking; quarterly technical reports; continuous threshold alerts.
- Public StoryMap in six movements: the paradox of rain without harvest, deterioration from the September 2024 FEWS NET map to the September 2025 IPC map where Phase 4 expands and parts of Darfur reach Phase 5, El Fasher, attribution by contrast, access, and what can still change the outcome.
Conflict spikes precede market closures, market closures precede phase escalation, and vegetation stress precedes displacement acceleration; each interval is a planning window.
Table view
08Result
- Phase 3 and higher food insecurity concentrates where conflict, environmental stress, and isolation converge: central Khartoum State, parts of Gezira State, and North Darfur.
- Conflict is the primary driver in Khartoum and the central agricultural belt, where crop losses exceed what rainfall explains; pastoral zones in western Darfur and eastern Sudan show patterns consistent with rainfall deficit.
- Humanitarian isolation follows conflict geography; air delivery is the only option in some zones.
- Escalation follows a detectable sequence with two- to four-week lead times.
- Active conflict zones are data-poor, which places the greatest uncertainty where decisions carry the most consequence.
The StoryMap was submitted to the LA Geospatial Summit with faculty endorsement.
09Tradeoffs and limitations
- Proxies. NDVI anomaly stands in for crop condition; market prices stand in for access; both carry known lags.
- Network completeness. OpenStreetMap coverage and SAR damage detection bound the accuracy of accessibility modeling.
- Parameter sensitivity. Kernel-density bandwidths and index weights shape hotspot geometry; the multiplicative form was chosen to suppress single-factor artifacts.
- Open and commercial sources only. Confidence reflects what public reporting can support.