Does maximum earthquake size depend on focal depth?

While the Gutenberg-Richter "law" (GR) with b-value of 1.0 is an inexact description of the earthquake frequency-magnitude distribution, it does provide a convenient statistical basis for identifying when the largest earthquakes in a catalog are either anomalously large or anomalously small. When the largest earthquakes are as predicted by the GR distribution, it is inappropriate to infer that the largest historically known earthquake is the largest possible. Only when the largest observed earthquakes are significantly smaller than predicted can we infer that the largest possible earthquakes have occurred. Analysis of the global historical catalog demonstrates that at nearly all depths, the largest deep and intermediate earthquakes have sizes close to those predicted by a GR distribution with b of 1.0. Only between about 300 to 450 km and beneath 600 km are the largest known earthquakes somewhat larger than predicted. However, when geographically isolatable subgroups within catalogs are considered separately, largest earthquakes that are either anomalously large or small are quite common. Often this is because individual regions have b-values much different than 1.0; alternatively, sometimes there appear to be physical constraints limiting the size of the largest possible earthquake. As examples illustrating these assertions, this article evaluates two subgroups-deep-focus earthquakes occurring beneath Spain and intermediate-depth earthquakes in the Bucaramanga, Colombia, "nest.".