Tests and Applications of 3-D Geophysical Model Assembly

Transcription

1 John N. Louie, Aasha Pancha, Glenn P. Biasi, Weston Thelen, James B. Scott Nevada Seismological Lab 174, University of Nevada, Reno, NV 89557, Mark F. Coolbaugh Great Basin Center for Geothermal Energy, University of Nevada, Reno, Nev. Shawn Larsen Center for Complex Distributed Systems, Lawrence Livermore National Laboratory, Livermore, Calif. Abstract Assembling 3-d geophysical models of the western Great Basin region is a critical task for regional geothermal and earthquake hazard assessments. Diverse data sets having highly variable spatial sampling demand we assemble trial models testing needs for new data. Unlike the SCEC effort, we describe our models with point-profile values rather than with 3-d surfaces. Testing of data-selection and spatial interpolation criteria is needed for each application. Crustal-scale, basin-scale, geologic map-scale, and geotechnical-scale data sets are merged to produce grids for regional seismic-wave modeling. Our assembled models have helped separate magmatic from extensional types of geothermal systems, and have advanced our understanding of the geophysical parameters most important to predicting earthquake shaking. 1

2 Introduction The lateral heterogeneity of the crust has a great effect on assessments of resources and hazards. For the regional-scale problems of assessing both geothermal resource potential and earthquake hazard, progress can be limited by the completeness and quality of the 3-d geophysical models that can be built. The Southern California Earthquake Center (SCEC) has assembled a 3-d Community Velocity Model, the SCEC-CVM of Magistrale et al. [2000], that describes details of crustal structure in their region. Derived mostly from petroleum exploration data in the Los Angeles and Ventura basins, regional academic refraction experiments, and USGS geotechnical borings, the SCEC-CVM is a critical database for prediction of wave propagation and seismic shaking in that urban area. We are developing a CVM for the western Great Basin, to contribute toward geothermal power evaluation of this large region, and to assess the potential for damaging earthquake ground shaking in the Reno and Las Vegas, Nevada, urban areas. Our region, aside from being much larger than that covered by the SCEC-CVM, also offers fewer data constraints overall, as well as extreme variations in data quality and density. Sensitivity studies in which we assemble and compute models help determine where new data are most needed. Fig. 1 shows a model assembled to examine shaking in the Reno, Nevada area from an earthquake 60 km to the west. Geological maps and a local geophysical study of the Reno area basin [Abbott and Louie, 2000] define a number of basins up to 4 km deep surrounding the urban assessment area. The 3-d grid we assembled, fed into a finite-difference solution of the elastic 2

3 wave equation [Larsen and Grieger, 1998; Larsen et al., 2001] for the earthquake, shows the complex influence on seismic shaking of basin geometry, and regional basin distribution. Model Assembly Methods Although many of our purposes are similar, we cannot use the same model-assembly methods as the SCEC-CVM. Overall, our density of geophysical control points is poor. As well, this large region has many diverse data sets, with many areas not covered by any data. Without abundant borehole data, we are not able to construct time-stratigraphic or constant-velocity surfaces. So we interpolate between possibly isolated control points at depth below often widely-spaced geophysical depth-profiles. Further, although SCEC has put together one consensus model for southern California, a consensus model for the western the Great Basin is not practical or even perhaps very useful at this time. Before a model is assembled, we must first select and prioritize the input data sets. Where no other data are available, we default to consensus models such as CRUST5 [Mooney et al., 1998] or CRUST2. Then we prioritize datasets according to their level of local detail, and their applicability to the problem at hand. Fig. 2 shows a data-selection problem in the northwest Nevada region. In the background is the consensus crustal thickness [Mooney et al., 1998], which has clearly been selected from some refraction and some receiver-function control points. New refraction results [, 2004] disagree with the consensus thicknesses but do agree with many of the surrounding refraction 3

4 points. In building models of geothermal resources it is crucial to be able to test alternative data selections. We hold to several general principles in model assembly: higher-priority, detailed data sets substitute for background and default data sets in areas of overlap, with no mixing; data files should be plain text for easy editing and archiving; data locations are not assumed to be in any order and are given with geographic coordinates; the code should be open-source for the same archival and review reasons; velocity averaging and interpolations should be done by slowness, to preserve travel times; and for now our methods of controlling model assembly are integrated with the control methods and file formats of the E3D finite-difference package [Larsen and Grieger, 1998; Larsen et al., 2001]. The assembly code and some datasets are available at Visualizations of the 3-d models are provided by the model assembler in the form of a certain type of seismic amplification map [after Field et al., 2000], and can be constructed from the grids by an open-source seismic graphics package ( Fig. 3 illustrates some of the difficulties with interpolation between widely and irregularly spaced data points. The model assembler constructs a trial seismic amplification map based on Reno basin depth [from Abbott and Louie, 2000] and geotechnical measurements of shallow shear velocity [in the manner of Field et al., 2000]. The geotechnical measurements are widely scattered except for 50 locations along a transect [Scott et al., 2004]. A nearest-neighbor spatial interpolation honors the transect s measured variance between adjacent measurements, but the boundaries are controlled purely by the data distribution. Distance-weighted averaging, though, 4

5 is too smooth and doesn t honor the measured spatial variance. A nearest-neighbor, distanceweighted average among the closest points in each of four directional quadrants appears to be a compromise. Applications of Regional 3-D Models Our assembled 3-d models have been useful so far in regional geothermal resource exploration, and to model regional earthquake-wave propagation for urban hazard assessment. Because of the scale independence of our model assembler, we expect it to be useful down to the prospect level where detailed geophysical data are available. Coolbaugh et al. [2003] have been able to use the crustal results of fig. 2 to associate the apparent thin-crust anomaly with geochemical and geodetic results to propose a target for new geothermal exploration in northern Nevada. This association motivates the collection of new crustal refraction data, where existing crustal thicknesses are unknown or in dispute. Since existing geothermal resources are rarely associated with crustal measurements, the 3-d interpolations have gained us statistics on correlations between crustal parameters and resources. Geothermal fields associated with extensional tectonics appear correlated with thin crust. Fig. 3, trial seismic amplification maps for the Reno area basin, demonstrates a hazardassessment application of our assembled models. Basin-depth data from a local gravity survey [Abbott and Louie, 2000] is combined with scattered shallow geotechnical measurements according to regressions for 1994 Northridge earthquake motions by Field et al. [2000]. All of 5

6 the trial maps point to some parts of the basin having more, and some less, shaking hazard. These assessments will be combined with scenario shaking predictions such as in fig. 4 to test how accurately we can plan for the effects of shaking in the event of an earthquake. Current ShakeMaps [Wald et al., 1999] use assumed background geotechnical velocities that our models can refine with measured data, as from fig. 3. More accurate scenarios will also aid in planning emergency-services response following an earthquake. Similar map-evaluation efforts are underway in Las Vegas. Since we assemble complete 3-d grids, 3-d finite-difference modeling of earthquake-wave propagation is now possible, from both local and regional events. Fig. 1 shows the modeled shaking from an earthquake in year 2000 across the Reno basin and surrounding regions. The many large basins in the region (such as Tahoe, at the bottom of the images) may be focusing, screening, and converting seismic energy. Pancha et al. [2004] uses such models to reproduce the long durations and some of the amplification of earthquake seismograms from the Reno basin. Laterally varying, 3-d models with regional basins are required; 2-d models do not fit earthquake records as well. Conclusions A 3-d model assembler allowing selection of local data sets to layer atop background models is a powerful tool for exploring the sensitivity of resource evaluations and hazard assessments to geophysical constraints. These assessments demand datasets at all scales, from crustal to shallow geotechnical. Interpolation between control points uses our scattered data more effectively than 6

7 attempting to construct surfaces. Assembled models have helped target areas for new geothermal-power exploration, and will enable realistic mapping of earthquake shaking hazards. References Abbott, R.E. and Louie, J.N. (2000), Case history: depth to bedrock using gravimetry in the Reno and Carson City, Nevada, area basins, Geophysics, 65(2), Braile, L.W., Hinze, W.J., von Frese, R.R.B., and Keller, G.R. (1989), Seismic properties of the crust and uppermost mantle of the conterminous United States and adjacent Canada, in Pakiser, L. C., and Mooney, W. D., Geophysical Framework of the Continental United States, Boulder, Colorado, Geological Soc. Amer. Memoir 172, Catchings, R.D., and Mooney, W.D. (1991), Basin and Range crustal and upper mantle structure, northwest to central Nevada, Jour. Geophys. Res., 96, CDMG: California Division of Mines and Geology (various dates ), Geologic Atlas of California, 1:200,000 scale. Coolbaugh, M., Sawatzky, D., Oppliger, G., Minor, T., Raines. G., Shevenell, L., Blewitt, G., and Louie, J. (2003), Geothermal GIS coverage of the Great Basin, USA: Defining regional controls and favorable exploration terrains: Transactions Geothermal Resources Council 27,

8 Field, E.H. and the SCEC Phase III Working Group (2000), Accounting for site effects in probabilistic seismic hazard analyses of southern California: Overview of the SCEC Phase III Report, Bull. Seis. Soc. Amer., 90(6B), S1-S31. Larsen, S., and Grieger, J. (1998), Elastic modeling initiative, Part III: 3-D computational modeling, Soc. Explor. Geophys. Ann. Internat. Mtg., Expanded Abstracts, Larsen, S., Wiley, R., Roberts, P., and House, L. (2001), Next-generation numerical modeling: incorporating elasticity, anisotropy and attenuation: Soc. Explor. Geophys. Ann. Internat. Mtg., Expanded Abstracts, Louie, J.N. (2002), Assembly of a crustal seismic velocity database for the western Great Basin, presented at the Geothermal Resources Council Annual Meeting, 25 September, Reno, Nev. Louie, J.N., Thelen, W., Smith, S.B., Scott, J.B., and Clark, M. (2004), The Northern Walker Lane refraction experiment: Pn arrivals and the northern Sierra Nevada root, Tectonophysics, in press. Magistrale, H., Day, S., Clayton, R.W., and Graves, R. (2000), The SCEC Southern California reference three-dimensional velocity model version 2, Bull. Seis. Soc. Amer., 90(6B), S65-S76. 8

9 Mooney, W.D., Laske, G., and Masters, G.T. (1998), CRUST 5.1; a global crustal model at 5 degrees X 5 degrees, Jour. Geophys. Res., 103, Pancha, A., Anderson, J.G., Louie, J.N., Anooshehpoor, A., and Biasi, G., 2004, Data and simulation of ground motion for Reno, Nevada, to be presented at 13th World Conf. on Earthquake Engineering, Vancouver, B.C., 1-6 August, paper no Scott, J.B., Clark, M., Rennie, T., Pancha, A., Park H., and Louie, J. N. (2004), A shallow shearvelocity transect across the Reno, Nevada area basin, submitted to Bull. Seismol. Soc. Amer., Oct. 7, Wald, D.J., Quitoriano, V., Heaton, T.H., Kanamori, H. Scrivner, C.W., and Worden, C.B. (1999), TriNet ``ShakeMaps'': Rapid generation of instrumental ground motion and intensity maps for earthquakes in Southern California, Earthquake Spectra, 15, Acknowledgements This work was performed under the auspices of the U.S. Dept. of Energy by the University of California Lawrence Livermore National Laboratory under contract No. W-7405-Eng-48. 9

10 Figure Captions Figure 1: (left) Upper-crustal velocity map assembled for the Reno, NV Truckee, CA region from geological maps (e.g., CDMG) and a detailed gravity study of the Reno area basin (Abbott and Louie, 2000). (right) Maximum ground motions computed through the assembled model from an earthquake in 2000, using E3D (Larsen and Grieger, 1998; Larsen et al., 2001). Figure 2: Test map of crustal thickness in nortwestern Nevada integrating new crustal refraction results (, 2004) into a consensus model (Braile et al., 1989; Mooney et al., 1998). The consensus model had to select among adjacent data points that varied significantly in crustal thckness (labeled in km). Many of the points in this area are from Catchings and Mooney (1991). Figure 3: Test maps of lateral interpolation methods between unevenly spaced data points. Shallow shear velocity is extrapolated from measurements in the the Reno area and used with basin depth (Abbott and Louie, 2000) to estimate seismic amplification according to the regressions of Field et al. (2000). (Upper) A nearest-neighbor extrapolation resembles a geologic map but completely depends on data distribution. More than 3 km from any measurement, shear velocity reverts to a default basin Vs of 350 m/s. (Middle) A distance-weighted average of all data within a 3 km radius smooths over known sharp variations. (Lower) Including only the closest 4 points in each of 4 directions seems a reasonable compromise, preserving detail near measurements. Figure 4: Planning-scenario ShakeMap (as by Wald et al., 1999) projected for a severe earthquake in urban western Nevada. Basin depths, shallow and crustal shear velocities, and 10

11 topography are factors that need to be assembled into 3-d models that could improve the shaking predictions for this scenario. 11

12 Fig. 3 Figure 2: (left) Upper-crustal velocity map assembled for the Reno, NV Truckee, CA region from geological maps (e.g., CDMG) and a detailed gravity study of the Reno area basin (Abbott and Louie, 2000). (right) Maximum ground motions computed through the assembled model from an earthquake in 2000, using E3D (Larsen and Grieger, 1998; Larsen et al., 2001). 12

13 Fig. 1 Figure 2: Test map of crustal thickness in nortwestern Nevada integrating new crustal refraction results (, 2004) into a consensus model (Braile et al., 1989; Mooney et al., 1998). The consensus model had to select among adjacent data points that varied significantly in crustal thckness (labeled in km). Many of the points in this area are from Catchings and Mooney (1991). 13

14 Figure 3: Test maps of lateral interpolation methods between unevenly spaced data points. Shallow shear velocity is extrapolated from measurements in the the Reno area and used with basin depth (Abbott and Louie, 2000) to estimate seismic amplification according to the regressions of Field et al. (2000). (Upper) A nearest-neighbor extrapolation resembles a geologic map but completely depends on data distribution. More than 3 km from any measurement, shear velocity reverts to a default basin Vs of 350 m/s. (Middle) A distance-weighted average of all data within a 3 km radius smooths over known sharp variations. (Lower) Including only the closest 4 points in each of 4 directions seems a reasonable compromise, preserving detail near measurements. 14

15 Fig. 3 Figure 4: Planning-scenario ShakeMap (as by Wald et al., 1999) projected for a severe earthquake in urban western Nevada. Basin depths, shallow and crustal shear velocities, and topography are factors that need to be assembled into 3-d models that could improve the shaking predictions for this scenario. 15