AN R PACKAGE TO IMPROVE USGS OPERATIONAL AFTERSHOCK FORECASTS

The USGS Earthquake Hazards Program provides information to the public when potentially damaging earthquakes happen. The Operational Aftershock Forecasting (OAF) system publishes the chance of more earthquakes occurring in the next day, week, month, and year after all M5+ earthquakes in the US, that are available on https://earthquake.usgs.gov/data/oaf/. OAF makes ~70 forecasts per sequence and produces ~5000 forecasts per year. Accuracy is important because the forecasts are used by emergency managers and infrastructure operators to make well-informed decisions, increasing the efficiency of emergency response. The forecasts also help increase public awareness of potential dangers, such as damaged buildings.

We are building an open-source R package to visualize aftershock sequences and determine the accuracy of the forecasts produced by OAF. OAF tunes its models using data from the Comprehensive Earthquake Catalog (ComCat), where it also stores its forecasts. The package starts with a flexible query function that allows the user to focus on the area and time period of their interest, query ComCat for earthquakes with forecasts, pull the forecast data, and retrieve seismicity to visualize and analyze the forecasts.

The first plot shows earthquake magnitude versus time, with emphasis on the seismic networks that provided data and the method used to determine the magnitude. The second plot maps the aftershocks with background layers providing context on the affected region. We will also collect all the forecasts to determine what fraction of the observations are within the forecasted range.

Improving the plots helps seismologists understand complexities in the data. During the 2021 M6.0 Antelope Valley sequence, the new magnitude-time plot showed us that ComCat used different seismic networks and inconsistent methods for different earthquake magnitude ranges. Improved maps help seismologists evaluate the spatial region used for the forecast, identify nearby active faults that could be involved, and improve communication with the public and emergency managers. Comparing past forecasts to observations will help improve the OAF system. Thus, our R package will help the OAF team maintain operational awareness of the ~5000 forecasts per year and provide accurate information to the public.