New graph-based framework cuts disaster recovery time in simulation study

Scheduling repair crews, medical teams, transport units and shelter supplies after a major disaster is one of the hardest problems in emergency management, and most existing tools rely on deterministic models or static heuristics that struggle with the dynamic, uncertain and spatially uneven conditions of real disaster zones. A new study addresses this gap with STGR-DRS, a Spatiotemporal Graph-based Robust Disaster Recovery Scheduling framework that coordinates heterogeneous resources using large-scale, multi-source open data.

The framework operates in four stages. First, a data fusion pipeline builds a dynamic spatiotemporal recovery graph from OpenStreetMap, Humanitarian Data Exchange, Shelter Cluster and Copernicus datasets. Second, a gradient boosting module estimates demand urgency, with SHAP values providing interpretability of the predictions. Third, a multi-objective robust mixed-integer linear programming formulation jointly minimizes recovery time, transportation cost and unmet demand while enforcing fairness constraints. Finally, a rolling-horizon re-optimization mechanism updates schedules dynamically as new disruptions emerge, rather than committing to a single fixed plan.

The researchers evaluated the framework on a case study of the 2023 Türkiye–Syria earthquake. Under nominal conditions, STGR-DRS reduced total recovery time by 13.4% compared with the best-performing baseline, and under severe disruption scenarios it achieved a robustness advantage of 15.2%. Ablation studies showed the contribution of each framework component, and scalability analysis confirmed that the approach remains computationally tractable for networks with more than 5,000 nodes.

The authors are explicit about the limitations of their work. The framework depends on centralized coordination, which could be challenged by communication failures in an actual disaster, and the disruption scenarios used in testing are synthetic, so they may not fully replicate the complexity of real-time operations. Future directions include decentralized multi-agent scheduling, integration of predictive aftershock models, and field validation with humanitarian organizations.

Despite these caveats, the study offers a reproducible, open-data experimental pipeline for disaster recovery scheduling research and provides a decision-support tool that emergency management agencies could adapt. By combining open geospatial and humanitarian datasets with robust optimization and rolling-horizon replanning, the framework illustrates how data-driven methods can move post-disaster resource allocation beyond static heuristics while remaining transparent about where model-based schedules still fall short of field reality.

Subject of Research: Robust optimization and spatiotemporal graph scheduling for post-disaster resource recovery

Article Title: A spatiotemporal graph-based robust optimization framework for data-driven disaster recovery scheduling

Article References: Li, B., Tan, Z., Fu, G., Liu, Z., Jin, R., & Yang, C. (2026). A spatiotemporal graph-based robust optimization framework for data-driven disaster recovery scheduling. Frontiers in Earth Science, 14, 1872550. https://doi.org/10.3389/feart.2026.1872550

Image Credits: AI Generated

DOI: 10.3389/feart.2026.1872550

Keywords: disaster recovery, robust optimization, spatiotemporal graph, scheduling, Türkiye–Syria earthquake, rolling horizon, mixed-integer programming, open data, humanitarian logistics, gradient boosting, spatiotemporal, graph-based