Published on October 9, 2007
New Empirical Relationships among Magnitude, Rupture Length, Rupture Width, Rupture Area, and Surface DisplacementDonald L. Wells and Kevin J. CoppersmithBSSA, 1994: New Empirical Relationships among Magnitude, Rupture Length, Rupture Width, Rupture Area, and Surface Displacement Donald L. Wells and Kevin J. Coppersmith BSSA, 1994 03/09/2007 Earthquake of the Week Introduction: Introduction A precise estimates of the size of the largest earthquake is necessary for seismic hazard analysis. It is rare during the historic period. Earthquake magnitude can be estimated from fault rupture parameters. The purpose is to present new and revised empirical relationships between various rupture parameters. Include data from recent earthquakes. New investigations of older earthquakes. Data Base: Data Base Source parameters of 421 historic earthquakes are used. Seismic moment, magnitude, focal mechanism, focal depth, slip type, surface and subsurface rupture length, maximum and average surface displacement, downdip rupture width, and rupture area. Only continental interplate or intraplate earthquakes (M > ~4.5, h < 40 km) No earthquakes associated with subduction zones. All data are evaluated. The most accurate value or average value is taken for further analysis. If there are many observations. 244 earthquakes are selected to develop empirical relationships. Ms vs. Mw: Ms vs. Mw No systematic difference between Ms and M within the range of 5.0 to 8.0 Ms is systematically smaller than M in the range of 4.7 to 5.0 The standard deviation of the difference between each pair of Ms and M is ~ 0.19. Surface/Subsurface rupture length: Surface/Subsurface rupture length Surface rupture length averaged about 75% of subsurface rupture length. The ratio appears to increase with magnitude. Mw vs. displacement: Mw vs. displacement The ratio of average to maximum displacement does not vary systematically as a function of magnitude (Figure 5) Ave. surface disp < Ave. Subsurface disp. < Max surface disp. Regression Model: Regression Model The distribution of residuals shows no obvious trends -> a linear regression model provides a satisfactory fit to the data. Regression Results and Statistical Significance: Regression Results and Statistical Significance Only the reverse-slip relationships for maximum and average displacement are not significant at a 95% probability level Analysis of Parameter Correlations (I): Analysis of Parameter Correlations (I) Strong Correlation (r = 0.89 to 0.95, s = 0.24 to 0.28) Mw vs. Surface Rupture Length, Subsurface Rupture Length and Rupture Area Analysis of Parameter Correlations (II): Analysis of Parameter Correlations (II) Weak Correlation (r = 0.75 to 0.78, s = 0.39 to 0.40) Mw vs. Maximum and Average Displacement Analysis of Parameter Correlations (III): Analysis of Parameter Correlations (III) Weakest Correlation (r = 0.71 to 0.75, s = 0.36 to 0.41) Maximum and Average Displacement vs. Surface Rupture Length Effect of Slip Type on Regressions: Effect of Slip Type on Regressions T test for the regression coefficients for individual slip-type data sets to the coefficients for the rest of the data. SS to N+R, N to R+SS, and R to SS+N Individual slip relationships to each other. SS to R, SS to N, R to N The difference between regression coefficients are negligible if they are not different at a 95% significance level. Effect of Slip Type on Regressions (II): Effect of Slip Type on Regressions (II) The difference is negligible Surface rupture length and subsurface rupture length vs. Mw Effect of Slip Type on Regressions (III): Effect of Slip Type on Regressions (III) The difference is not negligible for R to N+SS (left) R to SS (right) Rupture Area and Subsurface Rupture Width vs. Magnitude Effect of Slip Type on Regressions (IV): Effect of Slip Type on Regressions (IV) Mw vs. maximum and average displacement The difference is significant (SS to N+R) SS regression has the highest correlation and the lowest standard deviation Effect of Slip Type on Regressions (V): Effect of Slip Type on Regressions (V) Maximum and Average displacement vs. surface rupture length The difference is negligible (SS to N) SS regression has the highest correlation and the lowest standard deviation Effect of Data Selection: Effect of Data Selection Sensitivity test by removing two data points at random from each data set and recalculating the regression coefficients. More than approximately 14 data points are stable. Larger data sets typically have higher correlations and lower standard deviations. Although there are far more data points for subsurface rupture length and rupture area relationships than for surface rupture relationships, they have only slightly higher correlation coefficients and slightly lower standard deviations. These regressions are very stable and are unlikely to change significantly with additional data Effect of Tectonic Setting: Effect of Tectonic Setting The difference between the extensional and compressional coefficients is insignificant The rupture area regressions differ at a 95% significant level (SCR vs. non SCR) -> difference in expected magnitudes generally is small (< 0.2) Conclusion (I): Conclusion (I) Surface rupture length typically is equal to 75% of the subsurface rupture length. The average surface displacement typically is equal to one-half of the maximum surface displacement The ratio of surface rupture length to subsurface rupture length increases slightly as magnitude (M) increases There is no apparent relationship between the ratio of average displacement to maximum displacement and magnitude (M). The average subsurface displacement is more than the average surface displacement and less than the maximum surface displacement No systematic difference between Ms and M (Mw) over the range of magnitude 5.7 to 8.0. Ms is systematically smaller than M for magnitudes less than 5.7 Conclusion (II): Conclusion (II) The empirical regressions show a strong correlation between magnitude and various rupture parameters M vs. surface rupture length, subsurface rupture length, downdip rupture width, and rupture area are well determined (r = 0.84 – 0.95 and std = 0.24 to 0.41) Displacement vs. rupture length/magnitude are less well correlated (r = 0.71 – 0.78) The sense of slip doesn’t significantly change the regressions. Relationship between M and rupture area/rupture width are different at 95% significant level Regression coefficients are similar and differences in parameters estimated from these regressions are small. All-slip-type regression may be used for most situations, especially true for poorly known faults or blind faults. The regressions of displacement vs. magnitude show a mild dependency on the sense of slip Conclusion (III): Conclusion (III) Regressions containing approximately 14 or more data points are insensitive to change in data. Smaller data sets (less than 10 to 14 data points) generally are sensitive to changes in the data and correlations may not be significant. The relationships based on large data sets (> 50 earthquakes) are unlikely to change significantly with the addition of new data Extensional and compressional regressions are the same SCR and non-SCR regressions for rupture area differ at 95% significant level Subsurface regression do not differ at the same level Subdividing the data set according to various tectonic settings or geographic regions may provide slightly different results, but typically does not improve the statistical significance of the regressions. Reliable estimates of the maximum expected magnitude for faults should include consideration of multiple estimates of the expected magnitude derived from various rupture parameters.