GEM Global Mosaic of Hazard ModelsThe GEM Global Seismic Hazard Mosaic
The Mosaic: A dynamic framework
The GEM Global Seismic Hazard Mosaic ("Mosaic") is a database of national and regional seismic hazard models all described using the OpenQuake (OQ) Engine data format. The mosaic consists of hazard models developed by various institutions, within collaborative projects, and by GEM Foundation scientists. It includes models originally developed using the OQ Engine, and models originally implemented in other software formats that have been converted into that of the OQ Engine.
The mosaic has a dynamic structure that allows for easily incorporating new models as they become available. The result is a dynamic database of hazard models representing the most recent and openly accessible information for every area. Meanwhile all previous versions of models are kept in the database and remain compatible with the current OQ Engine.
In December 2018, GEM achieved global coverage of inland areas (excepting the polar regions and small remote islands) with a total of 30 national and regional seismic hazard models. With global coverage, GEM produced its first global seismic hazard map - a milestone originally achieved by the Global Seismic Hazard Assessment Program (GSHAP; Giardini et al., 1999) - and the first fully transparent and open collection of seismic hazard models spanning the globe. In June 2026, GEM marked its latest version of the Mosaic - now 36 models that also cover offshore, remote, and polar areas - and maps with true global coverage.
Global Seismic Hazard Map
The GEM Global Seismic Hazard Map version 2026.1 (Johnson et al., 2026) depicts the geographic distribution of the Peak Ground Acceleration (PGA) with a 10% probability of being exceeded in 50 years, computed for reference rock conditions (shear wave velocity, Vs30 = 800 m/s).
Building the Global Seismic Hazard Maps
The mosaic strategy allows the use of the best publicly available model for each region, but does not ensure hazard continuity across model boundaries. To create the Global Seismic Hazard Maps (“global maps” herein), we use several strategies to minimise boundary effects and improve the transition between the hazard computed inland and in the oceans, including borrowing sources from neighbouring models and using joint rupture datasets computed by systematically combining the contributions of various models.
Past versions of the GEM Global Seismic Hazard Maps (“global maps” herein) have covered only onshore areas. The main challenge associated with collecting the per-model calculation results into global maps pertains to the border areas between two or more models, where each model may produce results that introduce a hazard discontinuity when viewed together. We addressed this by homogenising the hazard results of adjacent models. For each hazard model, we compute results not only for the official coverage, but also a buffer surrounding it. The homogenisation applies a weighted average at sites with multiple results; sites inside the official model coverage (the “inside model”) are weighted higher than those inside an exterior buffer (“outside model”), and the impact of the outside model decreases relative to the inside model based on the distance to the shared model border. The so-called homogenisation is done on the hazard curves - the most basic calculation output - and map values are derived from the homogenised curves. See Pagani et al., (2018) for a full illustration of the process.
The v2026.1 global maps differ from past versions in that they include the oceans (OAT, OIN, OPA) and polar regions (PAR, PAN). These models were developed so that the crustal sources directly abut the numerous land models that surround them. On the contrary, the sources within close proximity to any coastline are not included (e.g. subduction zone sources), despite their notable contribution to offshore seismic hazard. The main reason for this gap is to keep seismic hazard calculations manageable; for an ocean model to include the full range of subduction sources would prohibit the types of calculations typically performed as part of the global seismic hazard mosaic. The consequence of the ocean model construction is that a region of 100s of kilometers starting at the land model coastlines is not necessarily fully covered by any one model. Therefore, to cover this region, we must combine components of several models. We do this using two main calculation types.
For the first 300 km offshore, we “pad” the onshore models by adding sources from the oceans to expand their coverage by 300 km, a commonly used distance of influence for a given earthquake. Hazard for sites within a 300 km “transition zone” of the shoreline are then computed with the padded models, allowing for continuous coverage between the onshore and offshore sites. A second discontinuity exists between the transition zone and the sites fully covered by an ocean model. Here, the impact of subduction or other coastal sources abruptly ends despite that they may still contribute to hazard. Here, we use an event-based approach, e.g. the hazard is based on a collection of earthquake ruptures admitted by models in the Mosaic, covering a 10,000 year period. The ruptures are selected by pairing each geographic region with one Mosaic model. All the components needed to compute hazard are taken from the respective model, and the results are collected into the same results as from calculations of Type (1).
The main reason that Type 2 calculations cannot be used for the full transition zone results from the inherent mandate that larger sources are included in more than one model (i.e., the source impacts the geographic domain of multiple models, for example the Cascadia subduction sources affect both Canada (CAN) and the Conterminous United States (USA), while the South Solomon and New Britain subduction zones affect both Papua New Guinea (PNG) and the Pacific Islands (PAC). The event-based method currently only allows one representation of each source, binned geographically, and thus two adjacent sites (one offshore and one onshore) could have hazard computed using different characterisations (occurrence properties, ground motion attenuation) of a major seismic source; the significance of this inconsistency is much subdued at 100s km distance. Furthermore, the event-based approach admits only a possible rupture set for a 10,000 year period, while the full collection of possible ruptures are used in the Type 1 “classical” hazard calculations. We chose to prioritize the capability to use the classical hazard approach at sites near significant seismic sources, where infrequent earthquakes that may not be sampled by the event-based approach are more important to the computed results.
Like v2023.1 and earlier, the homogenisation of hazard curves is still completed for buffer zones surrounding the onshore models. Buffer zones of 300 km are used for offshore overlapping portions in the Type 1 calculations, while 100 km are used onshore.
References
Pagani, M, Garcia-Pelaez, J, Gee, R, Johnson, K, Poggi, V, Styron, R, Weatherill, G, Simionato, M, Viganò, D, Danciu, L, Monelli, D (2018). Global Earthquake Model (GEM) Seismic Hazard Map (version 2018.1 - December 2018), doi.org/10.13117/GEM-GLOBAL-SEISMIC-HAZARD-MAP-2018.1
Pagani, M, Garcia-Pelaez, J, Gee, R, Johnson, K, Silva, V, Simionato, M, Styron, R, Vigano, D, Danciu, L, Monelli, D, Poggi, V, Weatherill, G (2020). The 2018 version of the Global Earthquake Model: Hazard component. Earthquake Spectra. DOI: 10.1177/8755293020931866
Johnson, K, Villani V, Bayliss K, Brooks C, Chandrasekhar, S, Chartier, T, Chen Y,
Garcia-Pelaez, J, Gee, R, Styron, R, Rood, A, Simionato, M, Pagani, M (2023). Global Earthquake Model (GEM)
Seismic Hazard Map (version 2023.1 - June 2023), DOI: https://doi.org/10.5281/zenodo.8409534
Johnson, K, Villani, M, Bayliss, K, Brooks, C, Rood, A, Styron, R, Simionato, M, Pagani, M (2024) GEM's Global Seismic Hazard Mosaic: Updates since v2018, Applications, and Upcoming Activities. Proc. to the 18th World Conference on Earthquake Engineering.
Johnson K, Bayliss K, Villani M, Brooks C, Styron R, Simionato, M, Chen Y, Rood A, Pagani M (2026). Global Earthquake Model
(GEM) Seismic Hazard Map (version 2026.1), DOI: https://doi.org/10.5281/zenodo.20208536
