Context

Conflict that escalated in April 2023 caused widespread destruction of housing, markets, public infrastructure, and livelihoods across Darfur, Kordofan, Khartoum, and Al Jazirah. By 2025, more than 14 million people will have been displaced, including approximately 9.9 million internally displaced persons, and 30.4 million people will require humanitarian assistance.

Since late 2024, approximately 2.6 million individuals have begun returning to their areas of origin. Many returnees encounter destroyed or damaged homes, non-functional water and sanitation infrastructure, damaged roads, unexploded ordnance, and a near-total absence of reliable information on the scale of infrastructure damage.

Due to insecurity and access constraints, comprehensive infrastructure damage assessments have been extremely limited. Humanitarian actors lack the reliable geospatial datasets needed to prioritize interventions, plan shelter responses, and coordinate recovery programming. This assessment directly addresses that critical information gap.

Our approach

HOT's approach uses remote mapping and geospatial analysis carried out entirely via satellite imagery, enabling rigorous assessment even where physical access is impossible due to ongoing insecurity.

The project comprises 3 phases of implementation:

• Phase 1: Baseline building footprint mapping
  • Establishing a comprehensive baseline dataset of building footprints within NRC priority areas using AI-assisted mapping and volunteer mapping platforms. The output will cover approximately 100,000 buildings with attribute tagging for structure types in NRC priority locality.
• Phase 2: Damage assessment
  • Visual interpretation of high-resolution satellite imagery combined with radar-based change detection where necessary. Buildings are classified as undamaged, minor damaged, or major damaged — aligned with internationally recognized standards including UNOSAT — with multi-step validation to ensure data quality.
• Phase 3: Methodology development and enhancement
  • Developing a replicable, multi-class damage classification framework integrating machine learning tools for damage detection, rubble detection, and accessibility analysis. The resulting methodology will be fully documented and designed for replication by humanitarian actors in other conflict contexts.

Deliverables

• Complete geospatial dataset of approximately 100,000 building footprints with attribute tagging for structure types across NRC priority operational areas.
• Geospatial dataset identifying damaged and undamaged structures, compatible with UNOSAT damage classification standards and structured for sharing via humanitarian data platforms.
• A complete methodology package including technical documentation, machine learning models and model weights, data processing workflows, training materials, and recommendations for future applications.
• Damage assessment maps, spatial analysis products highlighting damage concentrations, and a summary analytical report.
• Training and briefing sessions for NRC technical staff, with guidance for interpreting and using the generated datasets.

Outcome

The assessment will establish a baseline understanding of infrastructure and housing damage, enabling evidence-based prioritization of humanitarian interventions and strengthening coordination within the Shelter/NFI Cluster and the broader humanitarian community. Outcomes will be tracked through three indicators:

• A complete building footprint layer with attribute tagging for structure types, covering approximately 100,000 buildings in priority areas.
• A damage classification layer produced and shared, compatible with UNOSAT standards and structured for use across humanitarian data platforms.
• A replicable damage assessment methodology developed and documented, including machine learning workflows, for use by humanitarian actors in other contexts.