Abstract

Seismic magnitude has been an indistinct concept since modern seismology was established. The author discusses it thoroughly , provides an general expression of definition that is only relevant to the source characteristics and explains the magnitude conversion problem physically. Based on the essential understanding of magnitude estimation problems, the author introduces a method called two-way fixed effects regression, which has been applied in seismology since Keiiti Aki studied the properties of coda waves. A series of check tests are carefully designed to examine the hypotheses of this method and then the author applies it to the error analysis of the peak ground acceleration magnitude. The source, path and site errors are estimated by this approach and the robustness of separation is checked. The acquired knowledge about the errors in magnitude estimation may help us to improve the precision of magnitude estimation during earthquake early warning.