Erratum to Earthquake Recurrence Estimates for Northern Caribbean Faults from Combinatorial Optimization

In the original article (ten Brink and Geist, 2025), we derived on-fault magnitude-frequency relationships with uncertainties from a combinatorial optimization method that assigns individual earthquakes from a regional catalog to the different faults.Our model used a magnitude versus area relationship from Shaw (2023).His relationship was published with an error, M f 1 , 2025).Here, we recalculate the magnitude-frequency relationship with uncertainties using the Shaw (2025) corrected formula.The remaining parameters of the model, the model itself, and the number of repeated optimization runs have not changed.Figures 1 and2 show the corrected results and replace figures 3 and 4 in the original article.Table 1 lists the corrected mean recurrence interval of the different faults and their one-sigma standard deviation.Colored values show values that are less than 85% of ten Brink and Geist (2025; thereafter, the original article) published results (in red) or higher than 115% (in blue) than the original article.The results marked "None" (in green) are faults that, in the corrected version, no longer predict M 8 earthquakes.Note that the main difference between the corrected mean recurrence intervals and the original intervals occurs in the largest faults (subduction segments and large strike-slip faults) and mainly for the maximum magnitudes.The reason is that the change in magnitude as a function of fault area is largest for the faults with the largest feasible slip areas.Note also that the convergent margin segments (Fig. 1) now accentuate the bimodal distribution of predicted earthquake frequency as a function of magnitude.In these segments, earthquake frequency decays logarithmically from M 6.5 to 7.5, and a second relatively high frequency is found around M 8-8.2. Data and ResourcesThe integer programming solver used was IBM ILOG CPLEX version 20.1.The bathymetric terrain model is available at