Joint Inversion for 3D Velocity Structure of the Broader Africa-Eurasia Collision Region

Suzan van der Lee (Northwestern University), Heather Bedle (Northwestern University), Megan Flaganan (LLNL), Eric Matzel (LLNL), Michael Pasyanos (LLNL), Federica Marone, Barbara Romanowicz, Christian Schmid (ETH Zurich), Arthur Rodgers (LLNL),


The need to monitor broader areas and an increasing number of nations with nascent nuclear weapons programs has lead to major challenges to nuclear explosion monitoring research. Agencies must, in fact, be prepared to detect, identify and locate nuclear explosions in wide regions, often aseismic and lacking previous seismic observations. Since the 1980's, the importance of monitoring at regional distances has been well established. However, such monitoring is complicated by the passage of seismic waves through the structurally complex crust and uppermost mantle. As a consequence, traveltimes and amplitudes of regional phases show great variability leading to large uncertainties in event locations and decreased performance of regional discriminants. A major requirement for the accurate modeling of regional seismic data, and therefore improved event locations and regional discriminant performance, are 3D regional velocity models characterized by high resolution from the crust down to the transition zone.

Our aim is a 3D velocity model of the crust and upper mantle for the geographic region extending from the western Mediterranean to Pakistan, including the aseismic region of North Africa. The joint inversion of different types of seismic data with diverse sensitivity to the crust and mantle is essential to achieve a high resolution image of the structure in this tectonically complex area, where six major tectonic plates and several microplates interact with each other. We expect predictions for seismogram characteristics (phase arrival times, amplitudes, dispersion) based on this new model to match most observations and be useful for event discrimination. Simultaneously, the new model will refine our understanding of the structure and tectonics in the study region.

Technical approach and dataset

Our 3D S-velocity model will be derived from the joint inversion of regional waveform fits, surface wave group velocity measurements, teleseismic arrival times of S and P waves, receiver functions and published results from active source experiments. The strength of jointly using various datasets lies in their redundancy (increase in the results accuracy) and complementarity (resolving power increase and trade-offs reduction).

The fitting of regional fundamental and higher mode Rayleigh waveforms has been accomplished using the Partitioned Waveform Inversion method (Nolet, 1990) for thousands of paths covering the study area (Figure 32.1). The modeling of both fundamental and higher mode surface waveforms ensures resolution of the entire upper mantle structure down to the transition zone. The inclusion in the inversion of teleseismic arrival times will further boost the resolving power at mid and deep upper mantle levels, while group velocity measurements and constraints on crustal thickness from active-source literature and receiver function analysis will ensure high resolution in the shallow upper mantle.

The seismograms used in this work have been recently recorded by a variety of different stations and networks, both permanent and temporary, operating in the study region: MIDSEA deployment, Kuwait National Seismic Network (KNSN), the United Arab Emirates (UAE) Broadband deployment, the Jordan deployment, the Eastern Turkey Seismic Experiment (ETSE), the Caspian Broadband deployment, the Global Seismic Network (GSN), the International Monitoring System (IMS), MedNet and Geofon. While each of these waveform datasets is valuable on its own, their combination is unique and key to this study.

Figure 32.1: Great circle wave paths for the vertical and radial component seismograms used in this study to date.

Preliminary results

The broad consistency between seismic velocity anomalies inferred from existing and performed measurements of teleseismic arrival times and Rayleigh wave group velocities as well as from regional waveform fits implies that these different types of dataset are at least in part redundant. The consistency further shows that the datasets record the same structural phenomena, despite differences in size and character between typical sensitivity kernels for each dataset.

The preliminary model (Figure 32.2) obtained by jointly inverting linear constraints provided by the performed waveform fits and collected point constraints on crustal thickness from active-source seismic studies and receiver function analysis shows that the uppermost mantle in the study area is strongly heterogeneous, reflecting the complex tectonics of this region. Particularly low velocities are observed at 100km depth beneath the Mid-Atlantic Ridge, the East African Rift and the western Mediterranean basin. Lower velocities than average are also present deeper (150-200 km) beneath Turkey and western Arabia. High velocity mantle material has been imaged at 100 km depth in correspondence with the Russian Platform and the Ukraine Shield. Linear high velocity features, possibly representing subducted material, are observed in the central Mediterranean basin (Italy and Greece) and beneath the Zagros Mountains Range in Iran (e.g. at 150 km depth).

The observed large scale features of our preliminary model are in agreement with existing regional models (e.g. Marone et al., 2004; Maggi and Priestley, 2005).

Figure 32.2: Horizontal slices at different depths through our preliminary upper mantle model. Anomalies are relative to a 1D average model for the Mediterranean region.

Future of the project

Our final 3D S-velocity model will be derived from the joint inversion of regional waveform fits, surface wave group velocity measurements, teleseismic arrival times of S and P waves, receiver functions and published results from active source experiments. This model will be converted to a 3D P-velocity model, using both published data on elastic properties (and their partial derivatives with temperature and pressure) of mantle rocks and empirical information provided by measured arrival times of teleseismic P and Pms waves. This new P-wave model will provide an improved ability to locate seismic events.

The prediction and calibration of regional traveltimes and waveforms depend strongly on the methodology used to compute synthetic traveltimes and waveforms from a 3D velocity model. Our goal is to test the obtained S- and P-wave models' ability to predict regional P and S traveltimes, deflect wave paths and deform waveforms using different approximations (e.g. path average vs. exact numerical approaches). We will assess the effects of 3D heterogeneities first on the studied seismograms (traveltimes and waveforms) and subsequently on the 3D models derived from these data.


This work has been financially supported by the National Nuclear Security Administration, Office of Nonproliferation Research and Engineering and Office of Defense Nuclear Nonproliferation (DE-FC52-04NA25542).


Maggi, A. and K. Priestly, Surface waveform tomography of the Turkish-Iranian Plateau, Geophys. J. Int., 160, 1068-1080, 2005.

Marone, F., S. van der Lee and D. Giardini, 3-D upper mantle S-velocity model for the Eurasia-Africa plate boundary region, Geophys. J. Int., 158, 109-130, 2004.

Nolet, G., Partitioned waveform inversion and two-dimensional structure under the network of autonomously recording seismographs, J. Geophys. Res., 95, 8499-8512, 1997.

Berkeley Seismological Laboratory
215 McCone Hall, UC Berkeley, Berkeley, CA 94720-4760
Questions or comments? Send e-mail:
© 2006, The Regents of the University of California