MODELING REGIONAL SEISMIC PHASES AND CODA FROM EARTHQUAKES AND EXPLOSIONS USING GENERALIZED FOURIER METHODS IN 2D/3D Jessie L. Bonner 1 , David G. Harkrider 1 , James Britton 1 , and Jeffrey L. Orrey 2 Weston Geophysical Corp. 1 and GeoVisual Technologies, Inc. 2 Sponsored by the National Nuclear Security Administration Contract No. DE-AC52-05NA26610 Proposal No. BAA05-59 ABSTRACT We continue to develop a numerical code for modeling regional seismic phases from earthquakes and explosions in 2D/3D media using Generalized Fourier Methods (GFM). During the past year, we have modified the GFM code by implementing efficient and accurate Q parameterization and variable gridding in order to simulate topography. The objective of our current modeling is to better understand Lg and Lg coda propagation using synthetics generated with both earthquake and explosion source models. We have studied five velocity models as described in Yang (2002), including two Tibet models, a Chinese model, and a Tarim Basin model. We parameterized the models for 2D GFM with a grid spacing of 0.25 km, horizontal dimension of 962 km, and depth dimension of 110 km. We used a double couple source with and without attenuation to generate the synthetics. We estimated the geometrical spreading for Lg using the GFM synthetics without attenuation. We determined that Lg decays at a rate of ∆ -0.52 and ∆ -1.01 for spectral versus time domain measurements, which are similar to Yang’s (2002) results for wavenumber-integration synthetics. We are currently examining the 3D effects on geometric spreading. Next, we applied a Q β =200 and Q α =2.25*Q β throughout the entire model and generated synthetics using the 2D GFM for source depths of 1, 5, 10, and 30 km. We then used the two-station technique (Xie et al., 2006) to estimate power-law Q o and η from the synthetics. For all models and depths, the Q o s estimated from the synthetics were within 5% of the input Q β . With the exception of the two Tibetan models at a source depth of 1 km (η>0.4), all frequency-dependence η results were ~0. The estimated Q o s are all reduced when stochastic variations are added to the model. For example, when we add perterbations with von Karman distributions, 10% amplitudes, and 1 km correlation lengths, the Q o is decreased from the input Q by 6-25%. The amount of reduction caused by the heterogeneities is both model and depth dependent. In the process of porting source-time functions with our synthetics, we reviewed the available explosion source theories. We have included the Haskell (1961) source theory. Haskell suggested that simple analytical functions could fit the calculated reduced displacement potentials from near-source measurements of nuclear tests detonated in different lithologies. He showed that displacement scaling is inversely proportional to the cube root of yield at high frequencies and proportional to yield at low frequencies. The theory is based on continuum mechanics, which allows for either realistic plastic or fractured emplacement media. Additionally, it allows for different pressure and gas porosity considerations both above and below the water table. The theory also allows for different cavity and vapor radii, which could be used to model decoupled explosions. These and additional features of the Haskell source theory result in accurate predictions of the observed characteristics (e.g., corner frequencies, ψ ∞ , etc) of Nevada Test Site (NTS) explosions such as Cowboy and Rainier. We modeled small chemical explosions and found that the Haskell-predicted M w s are typically within 7% of the observed estimates based on moment tensor inversions. Haskell's source performed better at estimating the moment magnitudes than Denny and Johnson (1991), which was based on NTS data. Our next step is to convolve the Haskell source with the GFM synthetics to examine source effects on regional phase partitioning from explosions. 2009 Monitoring Research Review: Ground-Based Nuclear Explosion Monitoring Technologies 32