← Back to Sessions

A Three-Year Update on the Performance of the USGS Aftershock Forecasts for the 2020 SW Puerto Rico Sequence

Description: 

The 7 January 2020 magnitude 6.4 Southwestern Puerto Rico earthquake caused significant damage and triggered a robust sequence of aftershocks. The U.S. Geological Survey has since issued regular forecasts for the number and probability of aftershocks in the following days, weeks, and years. These forecasts are released publicly on the USGS website, allowing for a truly prospective evaluation. The aftershock sequence has been unusually productive for an earthquake of this magnitude, with aftershocks still being experienced monthly. Here we provide an update and evaluation of the first three years of the forecasts.

Aftershock forecasts were generated using the USGS AftershockForecaster software, based on an Epidemic-Type Aftershock Sequence (ETAS) model. The initial forecast is trained on historical activity in similar tectonic environments, and ongoing forecasts use Bayesian updating to adjust to the sequence. Forecasts include both the expected number range and the probability of seeing at least one aftershock of a given magnitude within a given timeframe. Comprehensive tests of the number forecasts show the forecast model performing as expected, with close to 95% of the observations falling within the 95% confidence range of the forecasts. Tests of the probability forecasts, treating each forecast as a Bernoulli trial, also show success. The number and probability tests complement each other, with the number tests having high statistical power when probabilities and expected numbers are high, and the probability tests having power when the expected number may be small or zero. The analysis shows evidence for evolution of the magnitude-frequency distribution throughout the sequence, with fewer large events later in the sequence. This evolution is only partially captured by the existing forecast model, and longer-term forecasts could be improved by considering non-stationarity in the model parameters.

Session: The 2020-2021 Southwest Puerto Rico Seismic Sequence: Current State of Knowledge and Implications

Type: Oral

Date: 4/18/2023

Presentation Time: 10:15 AM (local time)

Presenting Author: Nicholas van der Elst

Student Presenter: No

Invited Presentation: 

Authors