Earthquake Distributions Along Northern Caribbean Faults Are Sensitive to Assumptions of Plate Boundary Coupling and Segmentation

Abstract:

We simulate how earthquakes are distributed along the complex fault system of the northern Caribbean plate boundary using the integer programming (IP) method. Input to the simulations is a regional synthetic catalog of M≥6 earthquakes drawn from a Gutenberg-Richter (G-R) distribution spanning 10 kyr. The binary decision vector in the IP method is composed of every possible location for each earthquake in the G-R sample. Integer programming globally optimizes the spatial distribution of the earthquakes according to target slip rates along faults provided by previously published GPS block models for major faults, and by other data for secondary faults. All feasible IP solutions must fall within the uncertainty in these slip rates. In addition to fitting earthquake distributions to the target slip rate, the northern Caribbean study region has an extensive record of large magnitude earthquakes that can be used to evaluate different assumptions about fault coupling, segmentation, and inter-fault connectivity. Both the magnitude and location of historical earthquakes, with their attendant uncertainty, are used to determine which parameter combinations result in models that best explain the observations. Results are particularly sensitive to the seismic coupling coefficient (α) along the Puerto Rico Trench (PRT) plate boundary fault and to the assumed pattern of rupture segmentation of the largest faults. Results that conform to the historical observations are obtained if the PRT fault is very weakly coupled (α =0.1) and if the PRT fault is disconnected from the westward extension of the plate boundary interface north of Hispaniola. Optimal on-fault magnitude distributions vary significantly and cannot be simply described as either a G-R or characteristic distribution. This study provides estimates for maximum earthquake magnitudes along faults with unknown seismic potential and highlights the importance of determining the location and magnitude of historical earthquakes as well as understanding fault coupling and connectivity to constrain system-wide models of on-fault earthquake distributions.