Alaska 2020 update for USGS G19AP00019: Initial Development of Alaska Community Seismic Velocity Models Eberhart-Phillips, Donna; Nayak, Avinash; Ruppert, Natalia; Thurber, Clifford Seismic velocity model AKEP2020 uses earthquake travel-time and ambient noise group velocity data to update Eberhart-Phillips et al. (2006: AKEP2006) The model is provided in a table: vlAKEP2020xyzltlnSFDRE.tbl.txt and in the simul output from velocity inversion: vlAKEP2020.out.txt Map plots of Vp and Vp/Vs are also provided, with lines denoting limits of adequate data. Velocity Inversion Procedure Notes for AKEP2020 model The 2006 AK model and the 2020 updated model both use Transverse Mercator coordinate transformation with central meridian= -150; and earth- flattening transformation. The depths are relative to sea-level and station elevations are used. For the group-velocity data, the surface is taken as the 30-km median filtered topography. Velocity with the 3D gridded model is defined by linearly interpolating between nodes. A gradational inversion approach was used with earthquake and shot travel- time data, and group velocity observations for periods 6-15 s. The table provides, from the computed resolution matrix, the diagonal resolution element (DRE), and the spread function (SF), for each Vp and Vp/Vs node. This material is based upon work supported by the U.S. Geological Survey under Grant No. G19AP00019. Note that an earlier model, AKEP2018, from the first year of this funded project was reported in Eberhart-Phillips et al. (2019). The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the opinions or policies of the U.S. Geological Survey. Mention of trade names or commercial products does not constitute their endorsement by the U.S. Geological Survey For this project, the model AKEP2020 was developed, expanding on the data and area of the AKEP2006 model. This includes data from stations throughout Alaska and adjacent western Canada. Earthquakes and active source data from AKEP2006 are used. Selected earthquakes were included from 2007-2008 MOOS array (Li et. al, 2013). Additional earthquakes from 2015/07 through 2019/03 were selected for spatial distribution and during periods of temporary arrays, primarily using data provided by the Alaska Earthquake Center (AEC). Additional Rest autopicks were obtained. Many had huge residuals and manual evaluation was not feasible in the project timeframe. Hence downweighting was applied after initial evaluation in the 3D inversion code. For initial residual > 2.7 s, data was eliminated. Picks had assigned quality codes of 0-4 (where 0 is best and 4 is unusable) and for initial residuals > 1.65 s, the pick quality code was increased by 3 to eliminate the most problematic observations. In contrast, for the earlier smaller central Alaska study (AKEP2006) arrival-time data was checked for each earthquake. The Rest data provided about 39,000 and 33,000 S observations. The expanded dataset included 4953 earthquakes, and 162 active sources from TACT and EDGE studies (Fig. EP1). This totalled 413,776 P and 118,732 S observations. During the velocity inversions, travel-time data are linearly downweighted for residuals 0.5-1.2 s, and for distances 75-600 km. The 3-D velocity model is oriented parallel to the central Alaska subducted slab and extends 1600-km in X and 2000-km Y (Fig. EP1), with coarser grids towards the periphery. A gradational approach is used for velocity inversions. This provides reasonable velocities throughout the region, and more detail where there is denser data coverage. The initial model used a coarse version of the AKEP2006, with some extrapolation down the Alaska Peninsula, and very coarse models with recent data for distant areas. A coarse inversion of the whole model area was done with ~50-km grid spacing. Then fine inversions were done with ~25-km grid-spacing and auto-linking in low resolution areas (Eberhart-Phillips et. al, 2014), obtaining the local earthquake travel-time model. Simultaneous inversion for hypocenters and 3-D Vp and Vp/Vs structure has been done with a gradational approach, using travel-time observations from the AKEP2006 study (Eberhart-Phillips et al., 2006) and recent data, and incorporating group velocity observations. Eberhart-Phillips (1990) developed the Simulps code to use P and S travel-times to solve for Vp and Vs, through modifying the code of Thurber (1983). All S travel-time ray- paths are calculated using the 3-D Vs structure, in the Eberhart-Phillips (1990) code and all subsequent modified codes such as Eberhart-Phillips and Reyners (1997, 2012) and Eberhart-Phillips and Fry (2017). For typical earthquake travel-time data, the Vs model is poorly constrained relative to the Vp model, less representative of crustal structure and difficult to use for interpreting Vp/Vs (Eberhart-Phillips, 1989). Thus, as described by Eberhart-Phillips and Reyners (1997), it is preferable to solve for Vp and Vp/Vs, when using local earthquake travel-time data. This parameterization is retained for group velocity, with the Herrmann (2013) Vp kernels related to Vp model inversion parameters and Vs kernels related to partial derivatives of Vp and Vp/Vs model parameters (Eberhart-Phillips and Fry, 2017). Model cartesian coordinates and distances are computed with Transverse Mercator conversion, and earth-flattening transformation is used. The model 0-km depth is at sea level, travel-time ray-tracing includes station elevations, and elevation for group velocity observations is taken from the 30-km median topography (Fig. EP2). Joint inversions of travel-time and group velocity observations were completed to enhance the earthquake travel-time model. Earthquake data has weak vertical resolution of the upper 6 km, although the velocity fits the travel-time to the underlying earthquakes. Group velocity provides information in these shallow depths for Vp and Vs, for the 6-15 s periods of the selected observations, although it has inherent broad horizontal smoothing compared to the earthquake travel-time data. The group velocity data incorporated 11,286 observations from group velocity maps, with quality assigned based on the DWS of the underlying Rayleigh-wave path distribution. A progressive series of joint inversions was done to promote improvement of the shallowest depths, with fixing the deeper portions of the model. The relative weight of the group velocity observations (wtU) is varied. This series comprised (a) 1 iteration for depth z = -1 km free and wtU=35; (b) 2 iterations for z = -1,2 km free and wtU=22; (c) 1 iteration for z = -1,2,6 km free and wtU=10; and (d) 2 iterations for z= -1 through 33 km free and wtU=5.5. The final model achieves good improvement in fitting the expanded data set. Compared to the initial model, the final model AKEP2020 provides 36.4 % decrease in P data variance, 24.6 % decrease in S-P data variance, and 96.7 % decrease in group velocity data variance. The results are shown in map-views for depths 2-140 km Vp and Vp/Vs in Fig. EP3-8. Cross-sections across the central Denali fault and across the Pacific-Yakutat slab boundary are shown in Fig. EP9-10, with comparison to AKEP2006. The cross-sections illustrate that Vp, Vp/Vs and Vs are all reasonable. More work could be done to improve the quality of the travel-time data, to assess the influence of specific group velocity observations on the results, and to test factors such as reducing the smoothing applied to group velocity maps during earlier processing of the dispersion data that provided group velocity. Varied approaches to the progressive inversions and the relative weight could be evaluated. Further evaluations are beyond the scope of this project. Such procedures would alter some specific values and shapes of features, but overall the results would be similar. FIGURE CAPTIONS Figure EP1 Earthquake and shot distribution Figure EP2 Topography 30-km median filter used for group velocity Figure EP3-5 Vp maps, with line for limit of adequate data for 3D grid model. Figure EP6-8 Vp/Vs maps, with line for limit of adequate data for 3D grid model. Figure EP9-10. Upper has AKEP2006, lower has AKEP2020. (A) Vp and Vp/Vs cross sections with hypocenters (plus)within 50 km of section. Image is masked or has line where there is low resolution. (A) Normal to central Denali fault. (D) Across the Yakutat slab. TF, transition fault system; Chug, Chugach Mountains; DF, Denali fault; edge, apparent edge of Yakutat slab. Spr, Spurr; Red, Redoubt; Aug, Augustine; Ilm, Iliamna; CRB, Copper River basin; P-Yak, Pacific-Yakutat boundary (STFS). REFERENCES Eberhart-Phillips, D., 1989. Investigations of crustal structures and active tectonic processes in the Coast Ranges, Central California, Geophysics. Stanford Univ., Stanford, CA, p. 209. Eberhart-Phillips, D., Three-dimensional P- and S-velocity structure in the Coalinga region, California, J. Geophys. Res., 95, 15343-15363, 1990. Eberhart-Phillips, D., and M. Reyners, Continental subduction and three- dimensional crustal structure: The northern South Island, New Zealand, J. Geophys. Res., 102, 11,843-11,861, 1997. Eberhart-Phillips, D., D. H. Christensen, T. M. Brocher, R. Hansen, N. A. Ruppert, P. J. Haeussler, and G. A. Abers (2006), Imaging the transition from Aleutian subduction to Yakutat collision in central Alaska, with local earthquakes and active source data, J. Geophys. Res., 111, B11303, doi:10.1029/2005JB004240. Eberhart-Phillips, D., Reyners, M., 2012. Imaging the Hikurangi plate interface region with improved local-earthquake tomography. Geophys. J. Int., 190, 1221-1242, DOI: 10.1111/j.1365-246X.2012.05553.x. Eberhart-Phillips, D., S. Bannister, and S. Ellis, Imaging P and S Attenuation in the Termination Region of the Hikurangi Subduction Zone, New Zealand, Geophys. J. Int., 198, 516-536, 2014. Eberhart-Philips, D. and Fry, B., 2017. A new scheme for joint surface wave and earthquake travel-time inversion and resulting 3-D velocity model for the western North Island, New Zealand, Phys. Earth and Plan. Int., 269, 98–111. Eberhart?Phillips, D., A. Nayak, N. Ruppert, and C. Thurber (2019), Alaska 2018 update for USGSG18AP00017: Initial Development of Alaska Community Seismic Velocity Models [Data set], Zenodo, doi.org/10.5281/zenodo.2544925. Herrmann, R. B. (2013), Computer programs in seismology: An evolving tool for instruction and research, Seis. Res. Lett., 84, 1081-1088, doi:10.1785/0220110096. Li, J., G. A. Abers, Y. H. Kim, and D. Christensen (2013), Alaska megathrust 1: Seismicity 43?years after the great 1964 Alaska megathrust earthquake, J. Geophys. Res., 118, 4861-4871, doi:10.1002/jgrb.50358. Thurber, C. H. (1983), Earthquake locations and three-dimensional crustal structure in the Coyote Lake area, central California, J. Geophys. Res., 88, 8226-8236.