Yan Y. Kagan Dept. Earth and Space Sciences, UCLA, Los Angeles, CA , Statistical properties of earthquake occurrence and their application for earthquake forecasting

Outline of the Talk Statistical analysis of earthquake occurrence – size, time, space, and focal mechanism distributions. Current global earthquake forecasts and their testing. Maximum size estimates for subduction zones relevant for Tohoku (Japan) M9 earthquake as well as long- and short-term seismicity rate forecasts in its region.

Modern earthquake catalogs include origin time, hypocenter location, and second-rank seismic moment tensor for each earthquake. The DC tensor is symmetric, traceless, with zero determinant: hence it has only four degrees of freedom -- one for the norm of the tensor and three for the 3-D orientation of the earthquake focal mechanism. An earthquake occurrence is considered to be a stochastic, tensor-valued, multidimensional, point process. Earthquake Phenomenology

GCMT catalog of shallow earthquakes

World seismicity: 1990 – 2000 (PDE)

Statistical studies of earthquake catalogs -- time, size, space, focal mechanism Catalogs are a major source of information on earthquake occurrence. Since late 19-th century certain statistical features were established: Omori (1894) studied temporal distribution; Gutenberg & Richter (1941; 1944) -- size distribution. Quantitative investigations of spatial patterns started late (Kagan & Knopoff, 1980). Focal mechanism investigations (Kagan, 1982; 1991; 2009)

Advantages of this distribution: Simple (only one more parameter than G-R); Has a finite integrated moment (unlike G-R) for  < 1; Fits global subcatalogs slightly better than the gamma distribution.

threshold magnitude 95%-confidence lower limit 95%-confidence lower limit not to be taken literally! (“a large number”) 95%-confidence upper limit

The maximum-likelihood method is used to determine the parameters of these tapered G-R distributions (and their uncertainties): An ideal case (both parameters determined) A typical case (corner magnitude unbounded from above)

Review of results on spectral slope,  Although there are variations, none is significant with 95%-confidence. Kagan’s [1999] hypothesis of uniform  still stands.

Gutenberg-Richter law For the last 20 years a paper has been published every 10 days which substantially analyses b-values. Theoretical analysis of earthquake occurrence (Vere- Jones, 1976, 1977) suggests that, given its branching nature, the exponent β of earthquake size distribution should be identical to 1/2. The same values of power- law exponents are derived for percolation and self- organized criticality (SOC) processes in a high- dimensional space (Kagan, 1991, p. 132). The best measurements of beta-value yields 0.63 (Kagan, 2002; Bird and Kagan, 2004), i.e. about 25% higher than 0.5.

Gutenberg-Richter law (cont.) We consider possible systematic and random errors in determining earthquake size, especially its seismic moment. These effects increase the estimate of the parameter β of the power-law distribution of earthquake sizes. Magnitude errors increase beta-value by 1-3% (Kagan, 2000, 2002, 2003), aftershocks increase it by 10-15%, focal mechanism incoherence by 2-7%. The centroid depth distribution also should influence the β-value by increasing it by 2–6%. Therefore, we conjecture that beta- (or b-) value variations are property of catalogs not of earthquakes.

Crystal plasticity Recent experimental and theoretical investigations have demonstrated that crystal plasticity is characterized by large intrinsic spatiotemporal fluctuations with scale invariant characteristics similar to Gutenberg-Richter law. In other words, deformation proceeds through intermittent bursts (micro-earthquakes) with power-law size distributions (Zaiser, 2006).

Earthquake size distribution CONJECTURE: If the hypothesis that the power-law exponent is a universal constant equal 1/2 and the corner moment is variable is correct, then it would provide a new theoretical approach to features of earthquake occurrence and account for the transition from brittle to plastic deformation (Kagan, TECTO, 2010).

Omori ’ s law (short-term time dependence)  C-value measurements yielded values ranging from seconds to days (Kagan, 2004). Its variation is believed to provide evidence of physical mechanism for the earthquake rupture process. However, Enescu et al. (2009) and other careful measurements suggest that C=0, therefore the non- zero C-value also is a property of catalogs, not of earthquakes.

 Most often measured value of P is around 1.0.  If the branching property of earthquake occurrence is taken into account, the P-value would increase from ~1.0 to ~1.5 (Kagan and Knopoff, 1981). P=1.5 is suggested by the Inverse Gaussian distribution (Brownian Passage Time) or at the short time intervals by the Levy distribution. Omori’s law (cont.)

Kagan, Y. Y., and Knopoff, L., Random stress and earthquake statistics: Time dependence, Geophys. J. R. astr. Soc., 88,

Kagan, Y. Y., Random stress and Omori's law, Geophys. J. Int., GJI , 186(3), , doi: /j X x

Spatial distribution of earthquakes We measure distances between pairs, triplets, and quadruplets of events. The distribution of distances, triangle areas, and tetrahedron volumes turns out to be fractal, i.e., power-law. The power-law exponent depends on catalog length, location errors, depth distribution of earthquakes. All this makes statistical analysis difficult.

Spatial moments: Two- Three- and Four-point functions: Distribution of distances (D), surface areas (S), and volumes (V) of point simplexes is studied. The probabilities are approximately 1/D, 1/S, and 1/V.

Spatial distribution of earthquakes (cont.) Pair distance distribution is fractal with the value of the fractal correlation dimension of 2.25 for shallow seismicity (Kagan, 2007). There is no known mechanism that would explain this value, the only limits are 3.0 > delta > 2.0.

Focal mechanism distribution It also seems to be controlled by a fractal type distribution, but because of high dimensionality and complex topological properties (non-commutative group of 3-D rotations) very little theoretical advance can be made. It seems possible that tectonic earthquakes are pure double-couples (Kagan, 2009).

Focal mechanisms distribution (cont.) Rotation between pairs of focal mechanisms could be evaluated using quaternion algebra: 3-D rotation is equivalent to multiplication of normalized quaternions (Kagan, 1991). Because of focal mechanism symmetry four rotations of less than 180 degrees exist. We usually select the minimal rotation. Distribution of rotation angles is well approximated by the rotational Cauchy law.

Kagan, Y. Y., Correlations of earthquake focal mechanisms, Geophys. J. Int., 110, Upper picture -- distance 0-50 km. Lower picture -- distance km.

Statistical analysis conclusions The major theoretical challenge in describing earthquake occurrence is to create scale-invariant models of stochastic processes, and to describe geometrical/topological and group-theoretical properties of stochastic fractal tensor-valued fields (stress/strain, earthquake focal mechanisms). It needs to be done in order to connect phenomenological statistical results and to attempt earthquake occurrence modeling with a non-linear theory appropriate for large deformations. The statistical results can also be used to evaluate seismic hazard and to reprocess earthquake catalog data in order to decrease their uncertainties.

Earthquake rate forecasting The fractal dimension of earthquake process is lower than the embedding dimension: Space – 2.2 in 3D,Time – 0.5 in 1D. This allows us to forecast rate of earthquake occurrence – specify regions of high probability and use temporal clustering for short-term forecast -- evaluating possibility of new event. Long-term forecast: Spatial smoothing kernel is optimized by using first temporal part of a catalog to forecast its second part.

Forecast: Long-term earthquake rate based on PDE catalog present. 0.1 x 0.1 degree, Magnitude M>=5.0

Forecast: Short-term earthquake rate based on PDE catalog 1969-present. 0.1 x 0.1 degree, Magnitude M>=5.0

Power-law kernel Fisher distribution kernel

Adaptive kernel smoothing improves our forecast in seismically quiet areas.

Earthquake forecast conclusions We present an earthquake forecast program which quantitatively predicts both long- and short-term earthquake probabilities. The program is numerically and rigorously testable both retrospectively and prospectively as done by CSEP worldwide, as well as in California, Italy, Japan, New Zealand, etc. It is ready to be implemented as a technological solution for earthquake hazard forecasting and early warning.

Tohoku M9 earthquake and tsunami

Flinn-Engdahl seismic regions: Why select them? Regions were defined before GCMT catalog started (no selection bias), and it is easier to replicate our results (programs and tables available).

Kagan, Seismic moment-frequency relation for shallow earthquakes: Regional comparison, J. Geophys. Res., 102, (1997).

DETERMINATION OF MAXIMUM (CORNER) MAGNITUDE Seismic moment rate depends on 3 variables -- 1.The number of earthquakes in a region (N), 2.The beta-value (b-value) of G-R relation, 3.The value of maximum (corner) magnitude. Tectonic moment rate depends on 3 variables Width of seismogenic zone (W – km), 2. Seismic efficiency coefficient (50-100%), 3. Value of shear modulus (30GPa - 49GPa)

McCaffrey, R., Global frequency of magnitude 9 earthquakes, Geology, 36(3),

END Thank you

Earthquake occurrence exhibits scale-invariant statistical properties: (a) Earthquake size distribution is a power-law (the Gutenberg-Richter relation for magnitudes or the Pareto distribution for seismic moment). Preservation of energy principle requires that the distribution should be limited on the high side; thus we use the generalized gamma or tapered Pareto distribution. The observational value of the distribution index is about However, it can be shown that empirical evaluation is upward biased, and the index of 1/2, predicted by theoretical arguments, is likely to be its proper value. The corner (maximum) moment has an universal value for shallow earthquakes occurring in subduction zones. We also determined the corner moment values for 8 other tectonic zones. (b) Earthquake occurrence has a power-law temporal decay of the rate of the aftershock and foreshock occurrence (Omori's law), with the index 0.5 for shallow earthquakes. The short-term clustering of large earthquakes is followed by a transition to the Poisson occurrence rate. In the subduction zones this transition occurs (depending on the deformation rate) after 7-15 years, whereas in active continents or plate-interiors the transition occurs after decades or even centuries. (c) The spatial distribution of earthquakes is fractal: the correlation dimension of earthquake hypocenters is equal to 2.2 for shallow earthquakes. (d) The stochastic 3-D disorientation of earthquake focal mechanisms is approximated by the rotational Cauchy distribution. Abstract

On the basis of our statistical studies, since 1977 we have developed statistical short- and long-term earthquake forecasts to predict earthquake rate per unit area, time, and magnitude. The forecasts are based on smoothed maps of past seismicity and assume spatial and temporal clustering. Our recent program forecasts earthquakes on a 0.1 degree grid for a global region 90N--90S latitude. For this purpose we use the PDE catalog that reports many smaller quakes (M>=5.0). For the long-term forecast we test two types of smoothing kernels based on the power-law and on the spherical Fisher distribution. We employ adaptive kernel smoothing which improves our forecast in seismically quiet areas. Our forecasts can be tested within a relatively short time period since smaller events occur with greater frequency. The forecast efficiency can be measured by likelihood scores expressed as the average probability gains per earthquake compared to the spatially or temporally uniform Poisson distribution. The other method uses the error diagram to display the forecasted point density and the point events. As an illustration of our methods, we are trying to answer a question: was the Tohoku mega-earthquake of 2011/3/11 a surprise? We consider three issues related to the 2011 Tohoku earthquake: (1) Why was the magnitude limit for the Tohoku region so badly underestimated, and how can we estimate realistic limits for subduction zones in general? (2) How frequently can such large events occur off Tohoku? (3) Could short-term forecasts have offered effective guidance for emergency preparation?

Two methods can be applied to estimate the maximum earthquake size in any region: statistical analysis of available earthquake records, and the moment conservation principle -- how earthquakes release tectonic deformation. We have developed both methods since For subduction zones, the seismic record is usually insufficient (in fact it failed badly for Tohoku), because the largest earthquakes are so rare. However, the moment conservation principle yields consistent estimates for all subduction zones. Various measurements imply maximum moment magnitudes of the order No statistically significant difference is found between different subduction zones in beta value and corner (maximum) magnitude. Historical magnitudes underestimate corner magnitude but discrepancy shrinks with time. A comparison of the inter-earthquake secular strain accumulation and its release by the coseismic slip implies a similar maximum earthquake size estimate. Beginning in 1999 we used our statistical short- and long-term earthquake forecasts, based on the GCMT catalog, for the western Pacific, including Japan. We have posted them on the web and included expected focal mechanisms as well. Long-term forecasts indicate that the average frequency for magnitude 9 earthquakes in the Tohoku area is about 1/400 years. We have archived several forecasts made before and after the Tohoku earthquake. As expected, the Tohoku mega-earthquake changed the forecasted long-term rate by just a few percent. However, the magnitude 7.5 foreshock increased the short term rate to about 100 times the long-term rate, and the magnitude 9 event increased it briefly to more than 1000 times the long-term rate. These results could well justify the development of an operational earthquake forecasting plan.