Machine-learning enhanced earthquake catalog for the Alaska Amphibious Community Seismic Experiment (AACSE)

This earthquake catalog was generated using seismic and hydrophone data recorded by the Alaska Amphibious Community Seismic Experiment (AACSE). Details of the experiment are documented in depth in Barcheck et al., (2020). Earthquakes were detected in all available broadband or short period seismic data 51°N to 60°N and 164°W to 148°W, from June 1, 2018 through August 31, 2019. This earthquake catalog contains 21,134 events total: 6,834 detected by the Alaska Earthquake Center following their standard detection methods (Ruppert et al., 2023), and 14,300 additional events detected using the machine-learning picker EQTransformer (Mousavi et al., 2020) trained on and applied to land and ocean-bottom seismic data from AACSE. Details of training and catalog generation can be found in Barcheck and Ruppert, 2026. This data set contains two comma-separated files: (1) File "AACSE_EQT_final_cat_arrivals.csv": This is a .csv file of the complete catalog including arrival information for each event. Events detected by both the Alaska Earthquake Center and by EQTransformer. Events are separated by blank rows. Columns are described below. (2) File "AACSE_EQT_final_cat_headers.csv": This is a .csv file of the complete catalog containing only event header information. Each row corresponds to the header of a single event. Columns are described below. Event headers contain the following 14 columns: Column 1: origin_id - A unique numerical ID for each event, ranging from 0 to 21133. Column 2: origin_time - A formatted string containing the event origin time in UTC time. Origin time is the estimated time that the earthquake occurred (not when seismic waves arrived at each seismometer). Columns 3-8: year,month,day,hour,minute,second - Separate columns containing origin time information. Columns 9-11: latitude,longitude,depth_km - Information about the location of the earthquake. Latitude and longitude are in degrees. Depth is in km. Column 12: std - The standard deviation of earthquake arrival times minus modeled arrival times. Arrival times are modeled using a 1-D velocity model developed for the Shumagin Islands region (Abers et al., 1994). Column 13: origin_catalog - Whether the event information came from the Alaska Earthquake Center's AACSE catalog (labeled "AEC"; Ruppert et al., 2023), or from the EQTransformer-based catalog (labeled "EQT"; this study). For the AEC catalog, arrival times and local magnitudes come from the original catalog, while locations are recalculated using the Abers (1994) velocity model to match the EQTransformer-detected events. Column 14: ml - local magnitude, calculated as described in the Supplementary Information from Barcheck and Ruppert, 2026. In the file "AACSE_EQT_final_cat_arrivals.csv", each event has a header formatted as above, and each also has arrival information, meaning information about the time that a P or S wave arrived at each seismometer (also often called "pick time"). Arrival times are either auto-detected and then human analyst-checked (AEC-detected events labeled "AEC") or detected by EQTransformer trained on either land or ocean-bottom data (events labeled "EQT"). Information about arrivals for each event is found in rows below each event header. Events are separated by a blank line. Arrival rows contain the following information in 9 columns. Column 1: station - The seismic station where the arrival was recorded. Column 2: phase - Whether the arrival was a P or S wave. Columns 3-8: year,month,day,hour,minute,second - The time of the arrival (the "pick time"). Column 9: julday - the Julian Day of the event origin time. The Alaska Amphibious Community Seismic Experiment (AACSE) was funded through NSF Award 1654568. This project is supported by U.S. Geological Survey Grant No. G22AP00040.