“…For about half of the earthquakes in our database, the mean asperity depth along the fault rupture, R 0 , could be estimated from published seismological rupture inversions [Yoshida and Koketsu, 1990;Wald et al, 1991Wald et al, , 1996Zeng and Chen, 2001;Hernandez et al, 2004;Hikima and Koketsu, 2005;Pathier et al, 2006;Tan and Taymaz, 2006;Elliott et al, 2007;Cirella et al, 2009;Hashimoto et al, 2011;Wei et al, 2011;Cheloni et al, 2012;Fielding et al, 2013;Sun et al, 2013;Zhang et al, 2014] (Table S1). For less constrained cases, we set R 0 at the hypocenter [Given et al, 1982;Kawakatsu and Cadena, 1991;Kikuchi and Kanamori, 1991;Stein and Ekström, 1992;Anderson et al, 1994;Stevens et al, 1998;Doser et al, 1999;Abercrombie et al, 2000;McGinty et al, 2001;Januzakov et al, 2003;Berryman and Villamor, 2004;Hancox et al, 2004;Hamzehloo, 2005;Legrand et al, 2011;Alfaro et al, 2012;Has et al, 2012] or at half the hypocentral depth when surface rupture occurred or when the earthquake exceeded M w 7.5, accounting for the fact that for large earthquakes, rupture often propagates toward the surface, with large amount of slip occurring significantly shallower than the hypocenter. This holds for the large earthquakes for which we have the complete slip distribution but is not universal, the 1991 Limon (Costa Rica) case being an exception.…”