Road-Closure Stress Testing of Essential-Service Accessibility in Yemen

Abstract Road disruption can change access to basic services even when the services themselves remain in place. This study develops a national network stress test for Yemen using OpenStreetMap road extract and WorldPop constrained population surface. Demand is represented by 8,460 population-weighted grid cells covering 41.18 million people and linked to mapped markets, health facilities, and fuel stations. The study then evaluates five stylized closure mechanisms at 1%, 5%, 10%, and 20% of simplified physical road links, including random removal, spatially clustered removal, major-road targeting, access-betweenness targeting, and exact-bridge targeting. At 1% random removal, the mean newly disconnected population share across services is 1.12%, indicating a shallow low-intensity response. The most disruptive mechanism depends on closure intensity. Exact bridges produce the largest disconnection at 1% closure (15.65% across services), access-betweenness targeting at 5% (35.28%), and major-road targeting at 10% and 20% (49.65% and 65.04%). Among people who remain connected, rerouting is highest under major-road targeting at 1% and 5%. At higher intensities, connected-only delay becomes non-monotonic because disconnected users no longer remain in that distribution. Spatial closures look modest in national averages but impose severe local losses. At 1% closure, 2.01% of the national population is newly disconnected, compared with 59.64% of the population within 25 km of the closure center and 93.95% within the realized closure footprint. These findings show that national averages should be complemented by local exposure measures when assessing road-network vulnerability and prioritizing access-sensitive recovery.