Introduction
On 24 June 2026, northern Venezuela experienced a rare and damaging earthquake doublet near the Caribbean-South America plate-boundary system. The sequence comprised an M7.2 earthquake near Yumare, west of Caracas at 22:04:33 UTC, followed only 39 seconds later by an M7.5 earthquake 5 km to the west. Both events were shallow, with centroid depths of 11.5 km and 23.5 km, respectively. The first earthquake was caused by right-lateral horizontal strike-slip movement on the Bocono Fault, which then triggered the second earthquake on the San Sebastian Fault (coloured green), which is the boundary between the Caribbean plate (shown in purple in Figure 1) and the South American plate.
Much has been made in the media about the occurrence of notable earthquakes in other parts of the world that day, including a M6.9 earthquake off northeastern Japan around 25 minutes later, and a M5.6 earthquake in northern California had occurred earlier the same day. It is unlikely that there is any relationship between these events: the northeastern Japan earthquake occurred in a seismically active area that recently experienced a M 7.5 earthquake.
The 2026 doublet should be viewed within the longer seismic history of northern Venezuela and the southern Caribbean plate-boundary system. Figure 2 shows large earthquakes along the northern Venezuela coast and the adjacent Caribbean margin since 1900, including the 1900 event and the 2018 earthquake east of Caracas.
Early loss reporting indicates that the human and physical impacts may be substantially larger than initial official Venezuela government estimates. The Guardian reported that, by 1 July 2026, the earthquakes had killed at least 2,300 people, injured more than 11,267, and left fifty thousand missing. The article also reported that the UN migration agency estimated that up to 6.8 million people could be affected and require shelter, water, sanitation, healthcare and essential relief items [2].
Earthquake Rupture Model
The first event evidently triggered the second event, which proceeded to rupture over a 200 km distance eastward to Caracas (Figure 3), generating a maximum fault slip of 2m. It is evident that rupture propagation generated strong pulses of ground motion near the fault due to the rupture directivity effect (Somerville et al., 1997) [3], which is a seismic shockwave equivalent to a sonic boom, illustrated further below in Figure 5. The centre panel shows rupture of 200 km over 90 seconds for a rupture velocity of 2.2 km/sec, which is quite low, and the average fault displacement of about 1 metre was also quite low. However, the rupture directivity pulse could still be quite strong because it represents the summation of the ground motions generated from the whole fault rupture up to that point.
The EERI report [5] notes that FUNVISIS recorded 138 aftershocks in Venezuela between 24 and 26 June 2026, dominated by small-magnitude events: 133 were M<4, three were M4–6 and two were M>6. Most were shallow, with 129 events at depths less than 30 km, and the highest concentrations were along northern Venezuela near Naiquatá, La Guaira, San Felipe, Maracay and the Caracas–Valencia region [5]. Aftershocks were concentrated along the rupture, especially toward the eastern half of the rupture and near Caracas [6].
Significance of the Double Event
A scientifically important feature of the Venezuela event is not only its magnitude, but also its sequence structure. A doublet is not simply a mainshock followed by a much smaller aftershock. It refers to two earthquakes of broadly comparable size occurring close together in time and space. In engineering and catastrophe-loss terms, this distinction matters because the first event can weaken buildings, slopes, lifelines, emergency services and public response systems, while the second event loads an already damaged environment before inspection, evacuation, cordoning or repair can occur [7].
For earthquake risk modelling, this raises a practical question: should the sequence be treated as one catastrophe event, two separate events, or a dependent rupture sequence? That distinction matters for loss aggregation, hours clauses, reinsurance attachment, business interruption, post-event demand surge and emergency-response assumptions. From a practical insurance point of view, the Venezuela earthquakes occurred so close together in space and time that they will be treated as a single event.
The event is also a reminder that shallow strike-slip earthquakes near major urban corridors can generate high intensities even when the epicentre is not directly beneath the capital because of the great length of the fault and the forward rupture directivity effect. The largest event occurred west of Caracas, but the affected region included major urban, coastal, transport and economic exposures. The event therefore provides a useful real-world stress test for catastrophe models, especially where accumulated damage, vulnerability and event definition interact [4,8].
Android earthquake early warning of about 10 seconds in Caracas (detection, not prediction)
One striking aspect of the Venezuela event was the public reporting of Google Android earthquake alerts. This alert does not predict the occurrence of the earthquake; it detects the occurrence of the earthquake and warns people of the imminent occurrence of strong shaking from the earthquake. This system detected the onset of the Venezuela earthquake and sent warnings to 11.4 million people in Venezuela before the strongest shaking reached them. People in Caracas reported receiving a warning about 10 seconds before the onset of strong shaking.
Google’s Android Earthquake Alerts System uses two approaches. In regions with established seismic early-warning infrastructure, it can distribute alerts from official networks. In many other regions, Android phones themselves can act as a distributed sensing network using built-in accelerometers. When enough phones detect earthquake-like motion, alerts can be issued to nearby users [9,10].
For an Asia-Pacific audience, the lesson is important. Countries such as Japan already have sophisticated earthquake early-warning systems, but many earthquake-prone regions do not. Smartphone-based warning systems cannot replace dense national seismic networks, but they may provide a useful additional warning layer, particularly in countries where instrumentation is sparse or unevenly distributed.
Exposure and vulnerability
USGS PAGER issued red alerts for both events. For the M7.5 mainshock, PAGER states that high casualties and extensive damage were probable, that past red alerts have required national or international response, and that economic losses were estimated at 2-10% of Venezuela’s GDP [4]. For the M7.2 foreshock, the PAGER economic-loss estimate was 0-7% of GDP [8]. These are modelled estimates and explicitly consider only shaking-related structural losses, so they are not a final economic loss assessment.
The built-environment evidence points to several vulnerability classes that are important for catastrophe modelling. The EERI (2026) report notes that Venezuelan seismic design is governed by NORMA COVENIN 1756-1:2019, with the previous NORMA COVENIN 1756-98 / 2001 edition applying to many buildings constructed before the newer code came into use. It also notes uncertainty about code compliance and that reinforced-concrete moment-frame buildings are common [11]. For housing, the report highlights the prevalence of low-rise, non-engineered reinforced-concrete frame buildings with hollow clay tile masonry infill walls, while media coverage has focused on mid- and high-rise reinforced-concrete frame buildings with unreinforced masonry infill. Reported damage mechanisms include non-ductile frame behaviour, column failures, excessive drift and collapse of masonry infill walls [12].
Observed loss reporting is still ongoing. Early post-event reporting indicates severe impacts in La Guaira, Catia La Mar and other coastal urban districts, including damage to transport and airport infrastructure [13]. These figures and damage descriptions should be treated as provisional until official consolidated assessments and engineering reconnaissance reports are published. Satellite and remote-sensing evidence is becoming central to both the damage assessment and the source interpretation of the Venezuela doublet. The Guardian reported that preliminary satellite and radar-based analysis by NASA (2026) [14] suggested approximately 58,870 buildings may have been damaged or destroyed across the affected region, substantially exceeding early official counts [2,15]. There is a very strong concentration of damage along a 40 km long coastal region north of the city of Caracas (Figure 4). The city spans a zone that lies between 10 and 20 km south of the coast. The concentration of extreme damage along the coast is probably due to the presence of the fault there, as shown in Figure 3.
This estimate of the number of damaged buildings should be treated as preliminary, but it highlights why satellite-derived damage mapping can become a critical early input to loss estimation when field access is limited [2]. Remote sensing also contributes to the earthquake source interpretation. The Earthquake Insights material reports early satellite-imaging evidence for more than 3 m of near-surface offset near Caracas and uses the spatial distribution of aftershocks to interpret how the rupture and subsequent stress redistribution evolved [6]. The before-and-after satellite imagery is shown in Figure 6.
Lessons for Insurance
From an insurance and reinsurance perspective, the event is a reminder that near-urban earthquake loss is a tail risk even where direct insurance penetration is limited. Capital cities, ports, airports, coastal apartment stock, public housing, financial centres and transport corridors concentrate correlated property, business-interruption, infrastructure, casualty, relief-cost and macroeconomic risk. For catastrophe modellers, the event stresses event-definition assumptions as much as vulnerability functions: a doublet can turn pre-existing damage, non-structural failure, inspection delay and aftershock risk into a single accumulated-loss problem.
The EERI report also notes that a large protection gap is expected because of low insurance penetration, Venezuela’s weakened economy and an insurance sector already under pressure. It states that residential and public infrastructure almost certainly do not carry insurance, although industrial and energy assets may have some international coverage [18].
The emerging vulnerability impacts are consistent with known earthquake-engineering concerns. USGS PAGER identifies unreinforced brick masonry and adobe block construction as predominant vulnerable building types in the region [4,8]. Early engineering-focused reporting also points to older concrete buildings, tall towers on soft soils, heavy masonry infill or non-structural walls, and soft-storey configurations as important contributors to severe damage and collapse [19]. Older reinforced-concrete apartment and public-housing blocks, soft-storey/tall-tower exposures, non-ductile concrete and heavy masonry infill, pending formal reconnaissance may have been especially vulnerable.
Human Impacts
The humanitarian impacts are at an extreme level. The Guardian reported that health services were operating beyond capacity, with concerns about disease outbreaks, overcrowding, surgical backlogs, gaps in obstetric care in La Guaira, and difficulties registering casualties and tracking missing people. These factors reinforce the need to interpret the event not only as a physical earthquake sequence, but also as a compound urban disaster involving damaged buildings, disrupted health systems, delayed rescue access, and prolonged shelter and recovery needs [2].
Asia-Pacific takeaways
The Asia-Pacific relevance is strong because the Venezuela doublet sits directly on one of the most important modelling questions in earthquake catastrophe risk: when is earthquake occurrence treated as independent, and when should clustering, triggering or rupture interconnection be represented explicitly?
In Australia, the National Seismic Hazard Assessment 2023 uses earthquake catalogues to support distributed seismicity and other source-model components. For probabilistic seismic hazard analysis, catalogues are commonly declustered so that long-term rates are not dominated by aftershock sequences. The aim is not to deny physical triggering, but to estimate a rate of earthquakes that is appropriate for long-term hazard calculations and engineering design [20,21].
This is where truncated and untruncated catalogue treatments become important. An untruncated treatment retains more of the observed sequence history, while a truncated declustering time window limits how long an earthquake remains classified as dependent for rate-estimation purposes. This is a pragmatic modelling choice in stable continental regions such as Australia, where aftershock sequences can be long lived and where strict declustering may remove substantial seismic moment from sparse regional catalogues [21]. This difference is illustrated in Figure 7, in which the left panel shows the effect of the long aftershock sequence of the 1970 Mw 6.0 Lake Mackay earthquake on the earthquake recurrence model for a source zone in southwestern NT and eastern WA.
The Venezuela doublet is a useful reminder that catalogue independence is not a physical truth; it is a modelling decision made for a specific question and timescale. A long-term building-code model, an event-based catastrophe model, a post-event loss estimate, and a reinsurance-hours-clause analysis may each require a different treatment of the same sequence.
New Zealand presents the complementary problem: a highly active plate-boundary system where ruptures can be long, complex and multi-fault, and where distributed seismicity is not always well represented by a simple stationary Poisson process. The 2022 New Zealand National Seismic Hazard Model uses a seismicity-rate model with an inversion fault model for known faults and a gridded distributed seismicity model, and explicitly allows complex single-fault, segment and multi-fault ruptures [22,23,24]. It also includes non-Poissonian or time-dependent rate behaviour in the distributed seismicity component through logic-tree branches, recognising that earthquake occurrence can be clustered and overdispersed rather than strictly memoryless [25].
This makes the model more complex, but the complexity is purposeful. Events such as the 2016 Kaikoura earthquake demonstrated that large ruptures can link multiple faults and segments rather than behaving as isolated sources. New Zealand’s NSHM therefore moves closer to modelling earthquakes as an interconnected rupture and rate-forecast problem, rather than as a collection of fully independent point events. For highly active plate-boundary regions, that added complexity is often a necessary representation of the physics and the uncertainty [23,24,25].
For insurers and reinsurers, the Venezuela event is a reminder that earthquake loss is controlled by multiple factors including earthquake rupture sequence, exposure concentration, construction vulnerability, infrastructure interdependence and the modelling assumptions used to define an “event”.
References
References are numbered according to the citation scheme used in the draft. Web sources should be rechecked before external publication because post-event loss and damage information can change quickly.
[1] Alataristarion. 2015. CaribbeanPlate.png. Wikimedia Commons. A map of the Caribbean Plate, derived from Tectonic plates boundaries detailed-en.svg by Eric Gaba / Wikimedia Commons user Sting. Licensed under Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0). Source: https://commons.wikimedia.org/wiki/File:CaribbeanPlate.png. Original source file: https://commons.wikimedia.org/wiki/File:Tectonic_plates_boundaries_detailed-en.svg. Image used as a cropped derivative for inclusion in Figure 1; modifications include cropping and resizing. Accessed 6 July 2026.
[2] Rangel, C., Jones, S. and agencies. 2026. “Tonnes of rubble”: 58,000 buildings estimated destroyed in Venezuela earthquakes. The Guardian, 1 July 2026. https://www.theguardian.com/world/2026/jun/30/tonnes-and-tonnes-of-rubble-more-than-58000-buildings-estimated-to-have-been-destroyed-in-venezuela-earthquakes. Accessed 3 July 2026.
[3] Somerville, P.G., N.F. Smith, R.W. Graves, and N.A. Abrahamson (1997). Modification of empirical strong ground motion attenuation relations to include the amplitude and duration effects of rupture directivity, Seismological Research Letters, 68, 180-203.
[4] U.S. Geological Survey. 2026. M 7.5 – 28 km SE of Yumare, Venezuela. Earthquake Event Page, including event summary, tectonic setting, ShakeMap, PAGER and finite-fault / rupture products where available. Event ID: us6000t7zp. https://earthquake.usgs.gov/earthquakes/eventpage/us6000t7zp. Accessed 3 July 2026.
[5] Earthquake Engineering Research Institute. 2026. Learning From Earthquakes Virtual Earthquake Reconnaissance Team (VERT): Phase 1 Response to M7.2 & M7.5 Venezuela Earthquakes, June 24, 2026. Version 1.0, last updated 29 June 2026.
[6] Hubbard, J. A. and Bradley, K. 2026. Aftershocks and satellite imaging of the Venezuela earthquake. Earthquake Insights, 1 July 2026.
[7] Phys.org. Was Venezuela struck by an earthquake doublet? Here’s what we know so far. 26 June 2026.
[8] U.S. Geological Survey. PAGER one-page product for the M7.5 Yumare, Venezuela earthquake. https://earthquake.usgs.gov/pdl/products/urn%3Ausgs-product%3Aus%3Alosspager%3Aus6000t7zp%3A1782429423257/contents/onepager.pdf
[9] Google Research. Android Earthquake Alerts: A global system for early warning. https://research.google/blog/android-earthquake-alerts-a-global-system-for-early-warning/
[10] Google Crisis Response. Android Earthquake Alerts System public information. https://crisisresponse.google/android-early-earthquake-warnings
[11] Earthquake Engineering Research Institute. 2026. Learning From Earthquakes VERT Phase 1 Response to M7.2 & M7.5 Venezuela Earthquakes, Section 5.1: Built Environment.
[12] Earthquake Engineering Research Institute. 2026. Learning From Earthquakes VERT Phase 1 Response to M7.2 & M7.5 Venezuela Earthquakes, Section 7.3: Housing and Section 7.4: Other Building Damage.
[13] Associated Press. Frustration grows in Venezuela as earthquake death toll reaches 1,430. 28 June 2026.
[14] NASA Earthdata GIS / NASA Disasters Mapping Portal. 2026. Sentinel-1 Likelihood of Damaged Structures (Experimental) for the Venezuela Earthquake, June 2026. ArcGIS web map / Map Viewer. https://gis.earthdata.nasa.gov/portal/apps/mapviewer/index.html?webmap=0c3d77dd5aae46e4829d9a282477615c. Screenshot/crop used in Figure 4; modifications include cropping and resizing for inclusion in the briefing note. Accessed 3 July 2026.
[15] Scher, C. and Van Den Hoek, J. 2026. Sentinel-1 likely damaged areas for the Venezuela earthquake, June 2026. Preliminary radar-based damage assessment using Copernicus Sentinel-1 imagery and building-footprint data, distributed through NASA disaster-mapping / ArcGIS infrastructure.
[16] Vantor / Maxar. 2026. Venezuela Earthquake June 2026 Open Data imagery collection. Vantor Open Data Program, before-and-after satellite imagery for the Playa Grande / Caraballeda area, La Guaira, Venezuela. Licence: CC BY-NC 4.0 for non-commercial use with attribution. Source: https://radiantearth.github.io/stac-browser/#/external/vantor-opendata.s3.amazonaws.com/events/Venezuela-Earthquake-Jun-2026/collection.json. Licensed under Creative Commons Attribution-NonCommercial 4.0 International: https://creativecommons.org/licenses/by-nc/4.0/. Image used as a cropped derivative of the original satellite imagery; modifications include cropping for inclusion in the briefing note.
[17] Google Maps. 2026. Location inset for Playa Grande / Caraballeda, near Avenida José María España, La Guaira, Venezuela. Google Maps screenshot used as the location inset in Figure 6. Map data and imagery © Google and data providers; Google attribution retained on the image. https://www.google.com/maps. Accessed 6 July 2026.
[18] Earthquake Engineering Research Institute. 2026. Learning From Earthquakes VERT Phase 1 Response to M7.2 & M7.5 Venezuela Earthquakes, Section 5.3: Governance / Insurance Coverage.
[19] Associated Press. Older buildings, substandard construction left Venezuela vulnerable to earthquakes. 27 June 2026.
[20] Allen, T. I., Griffin, J. D., Clark, D. J., Cummins, P. R., Ghasemi, H. and Ebrahimi, R. 2024. The 2023 National Seismic Hazard Assessment for Australia: Model Overview. Geoscience Australia Record 2023/053. Geoscience Australia, Canberra. https://doi.org/10.26186/148969. © Commonwealth of Australia (Geoscience Australia) 2023. Licensed under Creative Commons Attribution 4.0 International (CC BY 4.0). Figure 7 reproduced or adapted from this source.
[21] Risk Frontiers. The 2023 Update of the Australian National Seismic Hazard Assessment (NSHA23). https://riskfrontiers.com/insights/the-2023-update-australian-national-seismic-hazard-assessment/
[22] GNS Science / Earth Sciences New Zealand. National Seismic Hazard Model 2022. https://www.gns.cri.nz/data-and-resources/national-seismic-hazard-model/
[23] Natural Hazards Commission Toka Tu Ake. National Seismic Hazard Model. https://www.naturalhazards.govt.nz/resilience-and-research/data-and-modelling/national-seismic-hazard-model/
[24] NZSEE. The 2022 New Zealand National Seismic Hazard Model revision: source rupture model and multi-fault rupture representation. https://repo.nzsee.org.nz/handle/nzsee/2323
[25] Iturrieta, P. C., Gerstenberger, M. C., Rollins, C., Van Dissen, R., Wang, T. and Schorlemmer, D. 2024. Implementing Non-Poissonian Forecasts of Distributed Seismicity into the 2022 Aotearoa New Zealand National Seismic Hazard Model. Bulletin of the Seismological Society of America 114(1), 244-257. https://doi.org/10.1785/0120230168
[26] U.S. Geological Survey. 2026. ANSS Comprehensive Earthquake Catalog (ComCat), FDSN Event Web Service. Earthquake catalogue data used to generate Figure 2, including events with M ≥ 3.5 between 1900 and 2026 in northern Venezuela and the adjacent Caribbean region. https://earthquake.usgs.gov/fdsnws/event/1/. Accessed 6 July 2026.
[27] Natural Earth. 2026. Natural Earth vector map data. Public-domain basemap data used for map features in Figure 2. “Made with Natural Earth. Free vector and raster map data @ naturalearthdata.com.” https://www.naturalearthdata.com/. Accessed 6 July 2026.
