Current models are almost fully dependent on faults, with off-fault hazard very low. But large, damaging shocks have recently struck California (2019 M 7.1 Ridgecrest), New Zealand (2010 M 7.1 Canterbury) and Japan (2008 M 6.9 Iwate-Miyagi) on unknown faults. And these are three best mapped places on Earth, so elsewhere, the problem is much worse. Current models assign maximum magnitudes to faults, but four recent shocks were much larger than expected on those faults (2016 M 7.8 Kaikoura, New Zealand; 2011 M 9.0 Tohoku, Japan; and 2023 M 7.8 Kahramanmaraş, Turkey). 

These are just a few examples of the limitations of fault-based ‘characteristic earthquake’ models, in which operators use their judgment to assign maximum magnitudes from fault lengths or seismic histories, and estimate earthquake repeat times from fault slip rates. However convenient and widely used, earthquakes don’t obey these rules. Just as important, fault mapping is a highly skilled and intensive endeavor, and so most of the world’s faults are inadequately mapped and have highly uncertain slip rates. Finally, each country relies on its own local datasets to make these judgments. These regional biases render the model in one country incompatible with that of its neighbors, as is evident from border discontinuities.