Seismotectonics-driven estimation of b-value; implications for seismic hazard assessment

Taking full advantage of good-quality historical earthquake and seismogenic source data available for Italy (Catalogo Parametrico dei Terremoti Italiani [CPTI15] and Database of Individual Seismogenic Sources [DISS] databases), we tested the regional variability of the b-value of the Gutenberg-Richter law, focusing on the dominantly extensional, nearly 1200-km-long seismogenic corridor straddling the Apennines chain; an 18,200 km (super 2) area generating about 45% of the seismic moment released countrywide. We carefully chose and tested the most appropriate completeness interval and completeness magnitude (M (sub c) ), and used the Lilliefors method, the Utsu test, and a bootstrap technique to verify the quality of our results and inferences. The 0.65 b-value we obtained is substantially lower than b-values used in Italian seismic hazard models, often close to 1.0; yet it predicts more accurately the observed Apennines earthquake record, suggesting that most currently adopted magnitude-frequency distributions underpredict events in the M (sub w) range 6.0-7.1 and overpredict those in the range 5.2-5.5. We also highlight that the time, space, and magnitude distribution of the largest Italian earthquakes hardly follows the Gutenberg-Richter law: it rather favors a characteristic behavior. We contend that the b-value is too critical of a parameter for being calculated using statistically weak data sets, such as those resulting from overly detailed area-source models: one may end up characterizing low-hazard areas more accurately than high-hazard areas. Conversely, we advocate the use of fewer, larger area sources that fully exploit the current understanding of regional seismotectonics, as a way of obtaining more realistic regional hazard models.