Estimation of magnitudes in central and eastern North America using intensity and felt area

Over 480 earthquakes from central and eastern North America having instrumental mb magnitudes and maximum Modified Mercalli intensities ( I ) were examined using exploratory data analysis techniques and modeled with robust estimation methods (91 of those earthquakes also have felt area data). A previously undocumented distinct offset of magnitudes is observed between Modified Mercalli intensities VI and VII in both the central and eastern North America data set and a separate western North American catalog of earthquakes. The offset is most probably a characteristic of the Modified Mercalli scale and brings into question the often assumed linear relationship between magnitude and intensity. (In particular, linear regression models could not accurately estimate magnitudes for larger, I = VII or VIII, events.) Instead, robust estimates of center and spread for individual intensity interval distributions are recommended. In studies where conversion of intensity to magnitude for groups of earthquakes is required (e.g., studies involving the Gutenberg-Richter recurrence relation), the underlying distribution of magnitudes (or its approximation) for each intensity should be used. Linear regression models using I and/or felt area (and several transformations were tested. The robust linear regression models m b = 2.48 + 0.0769 log 2 ( F A ) and m b = 2.16 + 0.0219 l 0 2 + 0.0596 log 2 ( F A ) (where FA is felt area in square kilometers) proved to be the most accurate magnitude estimation models for central and eastern North America earthquakes. A comparison of regional felt area models indicates that regional differences in attenuation of seismic waves may exist between central and eastern North America that becomes apparent only for events of sufficient magnitude. Those differences appear as larger felt areas for earthquakes of central North America as compared to earthquakes of eastern North America.