Large earthquakes are more predictable than smaller ones

Large earthquakes have been viewed as highly chaotic events regardless of their magnitude, making their prediction intrinsically challenging. Here, we develop a mathematical tool to incorporate multiscale physics, capable of describing both deterministic and chaotic systems, to model earthquake rupture. Our findings suggest that the chaotic behavior of seismic dynamics, that is, its sensitivity to initial and boundary conditions, is inversely related to its magnitude. To validate this hypothesis, we performed numerical simulations with heterogeneous fault conditions. Our results indicate that large earthquakes, usually occurring in regions with higher residual energy and lower b-value (i.e., the exponent of the Gutenberg-Richter law), are less susceptible to be affected by perturbations. This suggests that a higher variability in earthquake magnitudes (larger b-values) may be indicative of structural complexity of the fault network and heterogeneous stress conditions. We compare our theoretical predictions with the statistical properties of seismicity in Southern California; specifically, we show that our model agrees with the observed relationship between the b-value and the fractal dimension of hypocenters. The similarities observed between simulated and natural earthquakes support the hypothesis that large events may be less chaotic than smaller ones; hence, more predictable.