Magnitude of the 23 January 2018 M7.9 Alaska Earthquake Estimated from Local Dense Seismic Records in Alaska

0. INTRODUCTION

The Gulf of Alaska is located in the north end of Pacific Plate, which subducts beneath the North American Plate at longitude west of 146°W along Aleutian trench, giving rise to extremely large megathrust earthquakes, like the M9.2 Great Alaska earthquake of March 1964, which was located about 550 km to the north of the 23 January 2018 M7.9 earthquake and many intermediate earthquakes, including two large (M7.9 and M7.8) strike slip earthquakes occurred in November 1987 and March 1988, respectively (Freymueller et al., 2008). The plate motion of the Pacific Plate relative to the North American Plate is ~5.6 cm/y in the direction of 350° at the eastern end of the Aleutian trench (Pegler and Das, 1996), showing complicated geodynamic environment along the trench.

On 23 January 2018, the M_w 7.9 earthquake occurred in the offshore region of the Kodiak, south Alaska. No foreshocks were recorded according to the earthquake catalogue issued by the United States Geological Survey (USGS). Up to 10 April 2018, there were 14 M5-6, 211 M4-5, and ~2 000 M3-5 aftershocks (Fig. 1).

Figure  1.  Locations of aftershocks (red circles) that occurred within 2.5 months following the mainshock and previous seismicity (gray circles) according to the USGS. Bold and thin straight black lines indicate fracture zones, and magnetic lineations, respectively (Naugler and Wageman, 1973; Pitman and Hayes, 1968). Focal mechanisms are determined by the GCMT.

DownLoad: Full-Size Img PowerPoint

Magnetic survey showed clear magnetic lineations between 152° W and 142° W, indicating a series of EW trending fracture zones including the Aja fracture zone and three other linear fractures with similar orientation (Naugler and Wageman, 1973; Pitman and Hayes, 1968). Magnetic anomalies along both sides of the fractures suggest right-lateral strike-slip offsets on the order of a few tens of kilometers to ~200 km. The earthquake sequence seems ruptured several conjugate faults including the EW trending fractures and the ones parallel to the magnetic lineations, as indicated by the complex locations of aftershocks.

The focal mechanism derived from Global Centroid Moment Tensor database (GCMT, or Global CMT) shows a strike-slip faulting, agreeing well with the fracture behavior in the source region.

In this work, we apply a new approach to determine the magnitude for the earthquake, using stations recorded at local to regional dense seismic stations (Wang et al., 2017). In this method, the magnitude is estimated by considering both the source duration and maximum displacement of the direct P wave, rather than solely using the amplitude of P waves (e.g., ML, Ms, and Mjma). Different from previous studies (Li et al., 2017; Rao et al., 2017; Convers and Newman, 2013; Hara, 2011; 2007a, b), the source duration here is evaluated by performing back-projection on seismic data recorded at dense stations. Dense seismic stations in Alaska facilitate a better estimation of the source duration since the coda waves are cancelled out in the procedure of stack during the back-projection.

The first estimate of the magnitude is 8.03 in 3-4 min after the origin time. Four minutes after the origin time, the magnitude is estimated as 7.87. The final estimate of the magnitude is 7.86 using all stations in and around Alaska at epicentral distances up to 15 degrees. Our method shows quite stable and rapid estimate of the magnitude for the earthquake, indicating a promising aid for source parameterization in tsunami warning and hazard assessment.

1. DATA AND METHOD

We use seismic data recorded in and around the Gulf of Alaska, US. There are ~258 broadband stations in the Gulf of Alaska, western Canada, and along the Aleutian trench. Stations with azimuths of -40° to 35° and epicentral distances of 8° to 15° are used for back-projection (Fig. 2). All of the stations at distances of 8° to 15° are used to calculate the maximum displacements of the direct P waves (Fig. 3). In this study, we want to investigate how well the local to regional stations work for determining magnitude in the new approach, therefore the farther stations that are available in the Incorporated Research Institutions for Seismology (IRIS) are not used.

Figure  2.  (a) Locations of broadband stations in Alaska used for back-projection (green inverted triangle), solid and dashed lines show the epicentral distances and strikes of the fault planes determined by the GCMT, large black star shows the epicenter determined by the USGS. (b) Waveforms recorded in stations shown in (a).

DownLoad: Full-Size Img PowerPoint

Figure  3.  (a) Locations of global stations used for calculating the maximum displacements of the P waves, the red star indicates the epicenter for the 2018 M7.9 Alaska earthquake according to the USGS. (b) Displacement waveforms of the P waves recorded at the seismic stations shown in (a).

DownLoad: Full-Size Img PowerPoint

Following Wang et al. (2017), we first determine the source duration by back-projecting the seismic data recorded in the Gulf of Alaska. The hypocenter (-149.166°, 56.004°, 14.1 km) determined by the USGS is used in this work. We first set up a horizontal grid of 20×20 points covering the possible source area at depth of 14.1 km. The interval between grids is 15 km. The back-projection method doesn't have good resolution in depth, so the assumed horizontal grid doesn't significantly affect the resolution in horizontal plane and of the source duration. We use the software TauP (Crotwell et al., 1999) and the velocity structure IASPEI 1991 (Kennett and Engdahl, 1991) to calculate the travel time between grid points and the seismic stations. In order to overcome the structure heterogeneities beneath the Alaska stations that are not taken into account in the IASPEI 1991 velocity model, we apply a station correction procedure (Wang et al., 2016; Fan and Shearer, 2015; Yao et al., 2013; Satriano et al., 2012; Zhang et al., 2011; Ishii et al., 2005) to further improve the travel time calculations. We align the first 60 and 10 s after P arrivals of the waveforms to the epicenter with filters 0.05 to 0.5, and 0.1 to 0.5 Hz, respectively, to constrain the initial source location to the epicenter determined by the USGS. In this step, we eliminate noisy waveforms with correlation coefficients less than 0.6 to the model waveforms (station COLD, longitude -150.201, latitude 67.227 8) for the other waveforms filtered in the frequency band 0.1 to 0.5 Hz.

To obtain the source migration, we use time windows that are offset by 2 s. By comparing the stacked energies for the tested grid points, we trace the rupture energy release in space and time, therefore estimate the source duration (Krüger and Ohrnberger, 2005). Following Wang et al. (2017), we determine the source duration based on the energy-time plot (proxy of source time function) derived from the back-projection. We calculated source duration, in which 90% of the total energy has been included, and another duration, in which the amplitude of the total is smaller than 0.1 time of the maximum stacked energy and 80% of the total energy has been included. We define the shorter one as the source duration.

For the seismic data recorded at distance of 8° to15° and azimuth of -97° to 98°, we remove the instrument responses, and calculate the maximum displacement for the waveforms from theoretically estimated P arrival to S arrival. We use the equations following Wang et al. (2017)

where A_i is the maximum vertical displacement of P wave recorded at the i_th station, and Δ_i is the epicenter distance (km). N₁ and N₂ are numbers of the stations with epicenter distances of up to 40°, and 40° to 85°, respectively. The operator log denotes the decimal logarithm.

2. MAGNITUDE ESTIMATION

After having the source duration and maximum amplitudes of direct P waves, we use Eq. 1 to estimate the M_dt. Using the stations distributed in the distance range of 8° to 15° and at azimuths of -97°-98°, and the source duration of 67 s estimated for the back-projection results, M_dt is estimated as 7.86.

We test how the selected frequency bands and length of stacking widows affect the resolution of the back-projections and therefore the uncertainties in source duration, we choose another frequency band of 0.08-0.4 Hz, and window length of 10, 15, and 20 s, and perform back-projection to evaluate the variations for the determined source durations (Fig. 4). The source durations show little variations between 62 and 68 s, around 0.95-1.05 times of the source duration (65 s) estimated from the entire stations, which affect the M_dt varying -0.02 to 0.02.

Figure  4.  The energy-time plots derived from back-projections using stacking length of 10, 15, and 20 s at frequency bands 0.08-0.4 and 0.1-0.5 Hz, respectively. The red lines represent the determined ends of the source according to Wang et al. (2017).

DownLoad: Full-Size Img PowerPoint

We mimic a real time environment to estimate the M_dt in real time, and evaluate the accuracy of the estimated magnitude over time (Fig. 5). When the direct P wave arrived at 33 stations that are 10° from the epicenter in the Gulf of Alaska (3 min after the origin time), we start to perform back-projection. The estimated duration is 91 s. Together with 68 maximum amplitudes of the direct P wave recorded in global stations, we estimate the M_dt 8.03. Four to five minutes after the origin time (P wave traveled 11 degrees away), with 73 stations in back-projection, and 94 global stations, the M_dt is estimated as 7.94. The final estimate of M_dt, using the entire stations in and around the Gulf of Alaska in azimuth of -69°-71° and distance of 10-15 degree, is 7.86. The magnitude fluctuation of 7.86-8.03 is very accurate, and therefore can be used for the purpose of tsunami warning and hazard evaluation.

Figure  5.  Magnitude determination for the 2018 M7.9 Alaska earthquake. (a) Maximum displacement versus elapsed time plot for all P-wave recorded at stations shown in Fig. 3a. (b) Energy-time plots derived from back-projections using waveforms recorded at 8°-10°, 8°-11°, 8°-12°, 10°-13°, 10°-14°, and 10°-15°, and filtered at 0.1-0.5 Hz. (c) Time evolution of M_dt. The determined magnitude varies from 7.87 to 8.03, consistent with the well-recognized magnitude M_w 7.9. The preliminary magnitude determined by the NOAA is 8.0, which was soon updated as M8.2, and M7.9 in a few hours (http://earthquake.alaska.edu/magnitude-79-offshore-kodiak-evolving-content-page).

DownLoad: Full-Size Img PowerPoint

Here we combine the source duration and maximum displacements of direct P waves to calculate magnitude (M_dt). Magnitude can be estimated from the source duration following Ekström et al. (1992) and Hanks and Kanamori (1979), although the complexities of frictional prosperities in fault planes distort the resolutions sometimes. Given a source duration of 62-68 for this earthquake, the estimated magnitude would be 8.25-8.33, which shows a large variation to the well accepted magnitude (M_w 7.9). M_w, estimated by W-phase inversion in USGS, shows a magnitude of 7.9, which is similar to the value of 7.87-7.95 determined from Global CMT method which utilizes the full waveforms. The M_dt, M_w from W-phase, and M_w from Global CMT for this earthquake resemble, indicating the effectiveness of this newly developed magnitude scale for rapid determination of large shallow earthquakes.

3. DISCUSSION AND CONCLUSION

At 12 : 31 : 43 am, 23 January 2018, the earthquake rupture began. At 12 : 35 am, National Oceanic and Atmospheric Administration (NOAA) issued the first tsunami warning for BC, southeast Alaska, and Aleutians based on M=8 (M_s) and offshore location. At 13 : 05 pm, NOAA issued the second tsunami warning based on M=8.2 and offshore location. At 14 : 18, largest tsunami wave was observed in Sitka at 0.4 ft height. At 14 : 19 pm, NOAA issued the third tsunami message based on M=7.9 and offshore location. In our method, the first release of the magnitude (M_dt=8.03) can be determined in 3-4 min. The M_dt is updated as 7.87 in 5 min after the origin time.

There are a number of magnitude scales that can estimate earthquake size in a few seconds, minutes, and hours. Fastest estimate of earthquake size would be from some empirical approaches frequently used in earthquake early warning. Usually those approaches regard the source as point source, and suffered from saturations. The most accurate estimate of size of earthquakes would be moment magnitude (M_w). M_w can be calculated by the direct P waves, W-phase waveforms, and full waveforms, among which W-phase is utilized as a standard approach running at PTWC, ERI, and many other institutions in an automated fashion. Regular W-phase inversion utilizes seismic waveforms recorded at regional to global stations, and output moment magnitude ~30 min after the origin time, although a few studies imply that the response time can be reduced to as short as 5-10 min if local and regional stations are available.

For the purpose of tsunami warning, resolving accurate magnitude of earthquakes in 5-10 min might be acceptable for a timely tsunami warning. For example, for the 2006 M7.7 Java earthquake, the damaging tsunami waves reached the coast areas at ~30 min after the origin time. For the 2011 M9.0 Tohoku, Japan earthquake, large tsunami waves first reached Miyako at 20-30 min after the origin time.

Earthquake early warning algorithms probably offer the fastest results for earthquake sizes, however with great uncertainty. The release of incorrect warning information would greatly destroy the confidence of the public on the warning system. Considering that more accurate estimate of magnitude can be obtained in a few minutes after origin time and ~30 min delay for the damaging tsunami waves reaching coast areas, we suggest that the first release of the tsunami hazard is evaluated by the more accurately determined magnitude and source parameters such as W-phase inversion based on local observations and our new approach.

DATA RESOURCES

Seismic data used in this work are downloaded from the Incorporated Research Institutions for Seismology (IRIS), and the Chinese seismic data center (http://www.ceic.ac.cn/). Focal mechanisms are downloaded from the USGS (https://earthquake.usgs.gov/, last accessed April 3, 2018) and GCMT (www.globalcmt.org, last accessed April 3, 2018). All the figures were created using the Generic Mapping Tools (GMT) of Wessel and Smith (1991).

ACKNOWLEDGMENTS

This work was supported by the National Natural Science Foundation of China (No. 41474050), the Fundamental Research Funds for the Central Universities, the China University of Geosciences (Wuhan) (No. CUG170602), and the National Programme on Global Change and Air-Sea Interaction (No. GASI-GEOGE-02). Comments from Alex Hulko, Chengli Liu, and two anonymous reviewers have greatly improved the manuscript. The final publication is available at Springer via https://doi.org/10.1007/s12583-019-1215-z.