Roads on the Brink: Mapping Compound Earthquake and Landslide Vulnerability in Ethiopia’s Rift

A new open-access study published in the International Journal of Disaster Risk Science has produced an integrated, three-dimensional vulnerability assessment of a national road network exposed to compound earthquake and landslide hazards, focusing on the federal road network of northeastern Ethiopia within the tectonically active East African Rift System. Led by Yohannes Sisay Zeleke, together with Chiara Scaini, Worku Asratie Wubet and Maurizio Tira, the research combines graph theory, geographic information systems and multi-criteria decision analysis into a single framework that scores every one of 124 federal road links, spanning 5,910.5 kilometres, for its susceptibility to simultaneous seismic and slope-failure disruption. The headline finding is stark: 43.3 percent of the total network length, or 2,562.7 kilometres, falls into the high or very high vulnerability classes, while only 25.4 percent of the network, 1,500.4 kilometres, sits in the low vulnerability class.

The study’s central innovation lies in refusing to treat vulnerability as a single number derived from a single lens. Most previous assessments of road networks have examined one dimension in isolation: topological studies measure how a segment’s failure fragments the network graph, physical studies measure structural exposure to hazard forces, and functional studies measure the traffic and accessibility consequences of disruption. By integrating all three, the researchers show that road segments with nearly identical overall vulnerability can arrive there by markedly different routes. The Kombolcha-Bati link, which records the network’s highest vulnerability index at 0.694, earns its rank through the concurrence of high seismic intensity, confirmed landslide susceptibility, the largest structural inventory in the network and strong topological centrality. By contrast, the Kombolcha-Dessie corridor reaches a comparable vulnerability level largely through exceptionally high functional importance and network criticality, despite lower physical exposure.

Technically, the framework models the road network as a mathematical graph of 166 nodes, the intersections, and 124 edges, the road segments, weighted by length. Topological vulnerability is quantified through edge betweenness centrality, the proportion of shortest paths between all origin-destination pairs that traverse a given link, computed with a weighted Dijkstra algorithm, and through link redundancy evaluated via the cyclomatic number, a graph-theoretic count of independent cycles. Each link was systematically removed and the resulting cyclomatic number compared against the baseline: no change confirmed the absence of any alternative route, scoring maximum vulnerability, while a decrease of one indicated redundant paths existed. The analysis revealed a pronounced spatial concentration of topological importance along two dominant corridors, Kombolcha-Dessie-Haik and Kombolcha-Bati-Mille, where a disproportionate share of shortest-path flows is channelled through a narrow set of links with edge betweenness values reaching 1.000 for Kombolcha-Bati, 0.989 for Bati-Eliwuha and 0.977 for Kombolcha-Dessie.

Physical vulnerability was assessed through four indicators: a composite hazard index, pavement type and condition, the number of bridge structures, and the number of lanes. The composite hazard index itself is a methodological highlight, built by combining a seismic hazard layer representing peak ground acceleration with a 10 percent probability of exceedance in 50 years, sourced from the Global Earthquake Model, with a landslide susceptibility layer derived from a 30-metre SRTM digital elevation model, in which 39 percent of territory is classified as moderately susceptible, 10 percent as high and 0.1 percent as very high. Because the two hazards are independently sourced yet capable of concurrent interaction, the researchers used a criteria importance through intercriteria correlation, or CRITIC, weighted additive-multiplicative aggregation, assigning weights of 0.358 to seismic intensity, 0.397 to landslide susceptibility and 0.241 to their interaction term, capturing both individual hazard contributions and amplified co-occurrence effects.

Functional vulnerability drew on three indicators: the Ethiopian Roads Administration’s road functional classification, annual average daily traffic, and travel time. Traffic volumes were log-transformed before normalisation to address the strong positive skew typical of mixed urban-rural distributions. The highest functional vulnerability values appeared on three road links in Alemgena District and one in Adigrat, reflecting high traffic volumes and strategic proximity to Addis Ababa. To weight all nine indicators across the three dimensions, the team turned to the Analytic Hierarchy Process, aggregating pairwise comparisons from a panel of 12 domain experts drawn from the Ethiopian Roads Administration, the Ministry of Transport and Logistics and Ethiopian universities, using Saaty’s 1-to-9 scale and geometric mean aggregation with consistency verification. Hazard intensity emerged as the most significant factor, receiving the highest global weight of 0.35, followed by the number of structures along each link, while travel time and functional importance received the lowest weights.

The composite results classify the 124 links into four vulnerability categories using natural breaks in the index distribution: low below 0.25, moderate from 0.25 to 0.36, high from 0.36 to 0.51, and very high at 0.51 or above. The network-wide index ranges from 0.09 to 0.69, with a mean of 0.361 and a median of 0.351. The two very high vulnerability links are Kombolcha-Bati, with 46 bridge and culvert structures, and Bati-Eliwuha, with 42 structures and a near-identical compound hazard profile. Among the high vulnerability class, trunk corridors dominate, including Kombolcha-Dessie, which carries the highest traffic load in the network, Chefa Robit-Harbu and Debre Berhan-Gudoberet. Perhaps the most policy-relevant discovery is that the Ataye-Mehalmeda main access road ranks 10th overall, demonstrating that hazard intensity can outweigh road hierarchy in determining composite vulnerability, a result with direct implications for resource allocation frameworks limited to trunk and link road hierarchies.

The spatial distribution of vulnerability varies sharply across the four administrative districts studied. Kombolcha District shows the highest concentration of critical links, with 12 very high and 15 high vulnerability segments, driven by pervasive seismic exposure and steep, landslide-prone terrain. Alemgena follows, where functional vulnerability on high-traffic trunk roads amplifies moderate physical exposure, producing 5 very high and 15 high vulnerability segments and a maximum index of 0.58. Adigrat shows lower overall vulnerability with only 1 very high and 2 high segments, while Dire Dawa’s predominantly lowland network displays mostly moderate vulnerability, as low to moderate landslide susceptibility offsets significant seismic exposure. That contrast carries a warning: Dire Dawa contains 11 links exposed to the highest seismic intensities in the network, yet none reach the very high composite category, meaning a single-hazard seismic assessment would have systematically misdirected investment toward Dire Dawa at the expense of more compound-hazard-exposed districts such as Kombolcha.

The authors tested the robustness of their rankings through two sensitivity analyses. Varying the dominant hazard intensity weight across a range of plus or minus 0.10 in five scenarios left the top-five road links in their exact rank order. A second analysis independently scaled each of the three dimension weights from baseline up to twice their original values across 15 scenarios, revealing markedly different sensitivities: the topological dimension proved the most stable, keeping the top-3 ranking invariant at every valid step, while scaling the functional weight caused Kombolcha-Dessie to overtake Kombolcha-Bati from 1.5 times onward, though top-3 membership was preserved throughout. Scaling the physical weight was the most disruptive, producing mid-tier reshuffling among links ranked 4 to 15 at 1.25 times, a result the authors flag as a caution for prioritisation decisions. At extreme physical upweighting the composite index saturated, revealing that physical vulnerability is sufficiently dominant that the framework cannot absorb unlimited upscaling, underscoring the expert-calibrated weights as a ceiling on that dimension’s influence.

Validation against historical hazard occurrences, while limited, offers encouraging concordance. Applying a 2-kilometre buffer around the 10 highest-ranked links and overlaying 50 years of recorded seismic events and 31 Ethiopian Roads Administration roadside landslide incidents, the team found that 67.7 percent of landslide incidents and 23 percent of seismic events coincide with the top-ranked links. The authors are careful to note that this spatial concordance does not directly validate the composite index or confirm observed road disruption, and that the lower seismic concordance is structurally expected since a buffer around 10 links captures only a fraction of the seismically active zone. The absence of systematic historical records documenting compound seismic-landslide impacts on roads remains a key limitation, and the researchers call for post-disaster damage assessments in future work to strengthen empirical validation.

The broader significance of the study extends well beyond Ethiopia. Compound seismic-landslide events are among the most destructive multi-hazard threats to mountain road networks, as illustrated by the 2008 Wenchuan earthquake, which triggered thousands of landslides causing road disruption far exceeding direct seismic damage, and by the 2016 Kaikōura earthquake in New Zealand, which produced widespread slope failures and prolonged highway closure. In developing countries, where road networks often suffer from limited redundancy, maintenance backlogs and constrained disaster management capacity, the consequences of misallocated risk priorities are proportionally greater. The framework’s reliance on national hazard datasets and road authority records, or open-data proxies where those are unavailable, makes it transferable to other hazard-prone low-income regions. The authors suggest future extensions incorporating population served, access to health facilities and regional economic exposure, and complementing national-scale hazard layers with site-specific physics-based simulations of earthquake, landslide and compound scenarios, offering planners a spatially explicit decision-support tool for directing slope stabilisation, drainage intervention and structural retrofitting toward the links whose failure would matter most.

Subject of Research: Integrated topological, physical and functional vulnerability assessment of road networks to compound seismic-landslide hazards in northeastern Ethiopia

Article Title: Integrated Topological, Physical, and Functional Vulnerability of Road Networks to Compound Seismic-Landslide Hazards in the East African Rift

Article References: Zeleke, Y. S., Scaini, C., Wubet, W. A., & Tira, M. (2026). Integrated Topological, Physical, and Functional Vulnerability of Road Networks to Compound Seismic-Landslide Hazards in the East African Rift. International Journal of Disaster Risk Science. https://doi.org/10.1007/s13753-026-00779-x

Image Credits: AI Generated

DOI: 10.1007/s13753-026-00779-x

Keywords: road network vulnerability, compound hazards, earthquake, landslide, East African Rift, Ethiopia, graph theory, multi-criteria decision analysis, Analytic Hierarchy Process, GIS, infrastructure resilience, disaster risk reduction