PUSH is the first globally consistent seismic hazard model. Full hazard curves at any return period are available in several key intensity measures corresponding to homes, moderate buildings, and tall buildings; these curves are corrected for local site amplification. PUSH is based on the independently tested Global Earthquake Activity Rate (GEAR) model. PUSH can be licensed, accessed through our API, and viewed in our Temblor Hazard Model web platform.


PUSH, Temblor’s globally consistent seismic hazard model, across Asia and the western Pacific nations.


All other probabilistic models are based on mapped fault locations and large historical earthquakes. These models assume that earthquakes are “characteristic,” with events of a similar size assumed to recur at some regularity. In these models, earthquake magnitudes are estimated from either mapped fault lengths or the largest historical quakes. Earthquake frequencies are based on inferred fault slip rates. In these models, the areas between mapped faults are populated by ‘area sources,’ with an assumed maximum magnitude and occurrence frequency.

Except in a handful of countries, fault inventories are woefully incomplete, slip rates are poorly known, and the largest ‘area source’ earthquake that can strike is unknown. Therefore, even in the best mapped regions of the world, these assumptions, upon which all other models are based have failed us. Examples of earthquakes that were much larger than expected for their mapped faults include 1992 M 7.3 Landers, CA; 2011 M 9.0 Tohoku, JPN; 2016 M 7.8 Kaikoura, NZ. Examples of large earthquakes striking where no fault was mapped include 2019 M 7.1 Ridgecrest, CA; 2010 M 7.1 Darfield, NZ; 2008 M 6.9 Iwate-Miyagi JPN; and the 1989 M 6.9 Loma Prieta shocks. Thus, there must be a better way.

PUSH is not a fault-based model and is instead based on GPS-derived contemporary strain rates (a proxy for the forces that build stress on faults) and seismicity from a single, uniform worldwide catalog of earthquakes (which release stress on faults), as captured by the GEAR model. Because these datasets are uniform worldwide, the model does not depend on local judgments, data, or biases. So, the hazard in different parts of the world—Los Angeles, Santiago, Istanbul, Tokyo—can be rigorously intercompared.


PUSH across Central America and the Caribbean.