Share

GEM released its new global earthquake hazard and risk products at the opening of the GEM Conference 2026, held in Zagreb from June 23 to 25. The discussions, participated in by more than 200 individuals from more than 40 countries, moved from faults, catalogues and assumptions behind the models to future exposure, practical applications and emergency response.

Three main messages emerged. 

1. The new work extends global earthquake risk assessment beyond ground shaking to include changing exposure, liquefaction, infrastructure disruption, embodied carbon and future risk. 

2. Global models need national data, local evidence and institutional engagement to support specific decisions. 

3. Scenario and rapid-impact products must represent uncertainty and support decisions both before and immediately after an earthquake.

Running in parallel during the second and third days, a one-and-a-half-day National Seismic Hazard Modelling Workshop focused on the development and use of national seismic hazard models. Discussions from this workshop are covered in a separate recap.

Expanding the global view of hazard and risk

Marco Pagani presented the Global Seismic Hazard Model, which combines 36 models and includes the oceans and polar regions for the first time. Looking at recorded seismicity over the completed map, he said it appeared that GEM was finally “connecting all the dots” that had been missing when the model covered only land.

The hazard session also covered the new global homongenised earthquake catalogue, fault mapping, harmonised ground-motion datasets and the new global stochastic event set (SEESAWS). Kirsty Bayliss introduced a global catalogue containing around 495,000 earthquakes. Richard Styron described fault maps as “paintings, not photographs”, since incomplete geological information still requires interpretation. Christopher Brooks explained the practical reason for bringing 22 ground-motion datasets into a common format. Manuela Villani then presented SEESAWS, a global stochastic event set spanning 100,000 years for use in hazard and risk calculations.

Vitor Silva presented the Global Seismic Risk Model. Compared with 2023, the exposure model now represents around 200 million more people, 48 million more buildings and an additional USD 8.5 trillion in residential, commercial and industrial built value. He described keeping exposure information current as almost a losing battle. Cities continue to grow, buildings are replaced and populations move rapidly.

Al Mouayed Bellah Nafeh noted that GEM’s 2023 Global Seismic Vulnerability Model had been used by researchers, insurers, government agencies and humanitarian organisations. His presentation focused on the vulnerability models used to translate shaking into damage and loss, and its limitations.

Catarina Costa used a global dataset containing around 17.5 million kilometres of major roads to examine their exposure to liquefaction, with particular attention to post-earthquake access, emergency operations and recovery.

Martina Caruso estimated the environmental cost of repair and reconstruction at around 17 million metric tonnes of carbon dioxide equivalent in average annual loss. In the example presented by Alejandro Calderon, only the pathway combining code-compliant construction and targeted retrofitting met both safety and sustainability goals.

Keeping models connected to a changing world

Exposure and vulnerability do not remain fixed after a model is published. Cities expand, buildings age, people move and construction practices vary between places. Carmine Galasso described the purpose of future modelling in practical terms: “We can explore many plausible futures and design decisions that hold up across all of them.” This included urban growth, migration, informal construction, climate-related displacement and whether building standards are enforced.

Studies from Indonesia and Zagreb examined locally calibrated fragility and retrofit choices, while Gerard O’Reilly showed how spatial and damage correlation can change the probability of many buildings being damaged in the same event.

Nicole Paul examined how destroyed housing relates to immediate shelter needs and longer displacement, while Karim Aljawhari showed that using repair cost as a proxy could overestimate carbon emissions by 30–40% at some shaking levels.

Turning models into decisions

In Aizawl, the earthquake risk assessment and related engagement with local authorities contributed to regulations on site development and slope modification, the hiring of three municipal geologists and the empanelment of 203 geologists. As one presentation concluded, “Sustained, deep engagement with local professionals, agencies and leaders is key.”

Making Nepal’s vulnerable schools safer through rebuilding and retrofitting remains the preferred solution, but at the current rate, the work could take around 200 years. Earthquake desks were presented as an interim measure, not as a replacement for safer school buildings. Local production could also reduce transport costs and build skills in rural communities.

In Vancouver, Canada’s national seismic risk model was adapted from census-area to individual-building scale for a city-level assessment. The work informed the Buildings Seismic Risk Reduction Action Plan, which was unanimously adopted by the City Council. The work also required stakeholder engagement and more detailed studies of vulnerable building types, retrofit options, costs and possible effects on occupants.

GEM data and the OpenQuake Engine are also being used in catastrophe models, underwriting scores, parametric insurance and post-event assessment. Insurers and brokers combine them with portfolio information, claims experience, sensors and financial assumptions. Gallagher Re described using OpenQuake as an independent benchmark for internal models and as a source of hazard information where no internal model was available.

In El Salvador, Luis Mixco described an integrated workflow linking hazard, exposure and practical applications. He said global tools become most valuable when they are adapted to local needs and connected to actual decision-making processes.

From future scenarios to real-time response

The final sessions examined tsunami risk, physics-based ground-motion simulations, aftershock losses and uncertainty in earthquake scenarios.

Naveen Ramalingam carried the discussion from tsunami hazard to risk. His opening line, “Rare does not mean small,” summarised the problem well. Tsunamis may not dominate annual losses, but they can produce long recovery periods and severe fiscal shocks.

Ömer Faruk Kalaycı focused on another part of earthquake risk that can disappear from a single-event scenario: what happens after the mainshock. His Istanbul study found that aftershocks increased losses in every scenario, but the amount depended on the location of the aftershock, its proximity to heavily exposed areas and the damage already caused by the mainshock. The median results told only part of the story. Across the simulations, the additional loss ranged from 3.4% to 89%, leading to one of the session’s clearest messages: tail values are critical.

The closing session focused on forecasting and the rapid information needed after an earthquake. A study from the Philippines compared a 2023 event-based risk assessment with the Mw 6.9 Northern Cebu earthquake of September 2025. Across the 100 stochastic event sets, the largest simulated event in the area was Mw 6.85. The Bogo Bay Fault was not included in the model and was represented only through the background seismicity zone. The earthquake had not been predicted, but the comparison showed that the model had captured the broader tectonic potential of the area.

David Wald proposed adding static ground displacement and dynamic ground strain to ShakeMap and ShakeCast products used for pipelines. Kishor Jaiswal presented the development of PAGER 2.0 for rapid global earthquake-impact and response information, while Finn Løvholt described a 24/7 tsunami-exposure service that can produce maps and estimates in minutes to one hour.

During the same session, news of the twin Venezuela earthquakes reached Zagreb. During Marzia Santini’s closing presentation, ShakeMaps, liquefaction information and GEM OpenQuake results were already being used to identify areas of interest around Morón and Puerto Cabello for emergency satellite mapping.

##