Global seismic tomography.

Scientists use seismic energy to map Earth's subsurface structure; for example, the Earth model at right was constructed using global seismic tomography (GST) data. The triangular cut into the globe shows deviations of shear wave speed in the planet's mantle from the spherically symmetric reference Earth model. The surface is the top of the mantle (the Mohorovicic discontinuity, see page 63), and the bottom is the core-mantle boundary (CMB). Green to blue colors indicate higher-than-average seismic wave speeds, and slower-than-average speeds are yellow to red. Seismic velocities decrease with increasing temperature: The inference is that the red areas are hotter than average and the blue, colder. Seismic wave speeds vary also with chemical composition, but there are strong indications that the thermal effect is dominant.

Density is also a function of temperature. Material hotter than average is lighter and, in a viscous Earth, tends to float to the surface; colder material is denser and tends to sink. Thus our picture can be thought of as a snapshot of the temperature pattern in Earth's convecting mantle. In particular, the picture implies a downwelling under North America and an upwelling originating at the core-mantle boundary under northwest Africa. Other cross sections of the model passing through this anomaly indicate that this upwelling may continue to the surface and be connected to the northern Mid-Atlantic Ridge.

Solving an "inverse problem" provides a three-dimensional model of seismic velocities. Like in optics, seismic energy propagates between two points (the earthquake and the seismograph) along the minimum traveltime path; because of the analogy with optics we often talk about "seismic rays." The travel time anomaly due to the small deviations from the reference-earth model is accumulated along the ray path. These are our data. What we need are the parameters describing the locations and sizes of wave-speed anomalies. To solve this, we must represent the properties of the medium using a finite number of parameters. The mathematics can get complicated. In the example shown above, basis functions (spherical harmonics in geographical coordinates and orthogonal polynomials in radius, both truncated at some degree) were used. Such basis functions vary smoothly with position in a way similar to that of the sine or cosine functions. Fortunately, there is evidence that the seismic spectrum of Earth's lateral heterogeneity is strongly dominated by very large wavelength features, so that truncating these basis functions does not pose a serious source of error. The alternative, frequently used, approach is to solve for velocity perturbation in a three-dimensional array of cells, assuming that the velocity deviation within each element is constant.

There must be sufficient coverage of the mantle volume with crisscrossing paths to locate the source of the observed travel-time anomalies. If the data sampling a particular region exist for only one azimuth, the observed anomaly could arise anywhere along the ray path.

There is an important distinction between GST and medical tomography, for example, in which the location of wave sources and receivers can be chosen according to the objective. In GST we are limited by the distribution of globally detected earthquakes for wave sources and by the locations of seismographic stations for receivers. There is not much we can do about the distribution of seismicity, except that now and then an earthquake occurs in an unexpected place, so coverage is expected to improve with time. Generally, earthquake distribution is more even in the Northern Hemisphere. At this time, there are no permanent ocean-bottom seismographic stations for receiving the broadband frequency signals we need, so the locations of the receivers are further limited to continents and islands. Significant progress has been made in recent years in the global deployment of state-of-the-art seismographic instrumentation. This has been accomplished through programs such as the Global Seismographic Network initiative in the US, and GEOSCOPE in France. Complementary programs were established in Canada, China, Germany, and Japan. These digital seismographic networks were federated in 1986 to set common instrumentation goals and coordinate data exchange. However, without deployment of permanent ocean-bottom observatories, the distribution of receivers will remain inadequate, particularly in the Southern Hemisphere.

A workshop held at the Woods Hole Oceanographic Institution in April 1988 outlined the steps required to determine the feasibility of an Ocean Seismographic Network (OSN). Presently, the development of the necessary instrumentation and testing of background-noise levels at the ocean floor and in Ocean Drilling Program holes proceed in Japan, France, and the US. Even though the deployment and operational costs of a 15- to 20-station OSN would be high, its data are critical for improving the resolution of GST models and for other applications such as earthquake studies. Since 1977, when the first large-scale GST study was published, scientists have believed that three-dimensional images of lateral heterogeneity in the mantle will be essential to solve some of the fundamental geodynamics problems. Results accumulated since then confirm this belief. Here are some examples of the GST application to various problems in Earth sciences.

Mantle convection. The distribution of seismic anomalies represents the current configuration of thermal and compositional heterogeneity, and it imposes a complex set of constraints on possible modes of mantle convection. The figure at right on page 71 shows the equatorial cross section through the same Earth model as the figure to its left. Clearly, the dominating feature from the CMB to mid-mantle are the two slow and two fast regions indicative of a very large wavelength convection pattern.

Mineral physics. The in situ ratio of the perturbations in shear velocity and compressional velocity inferred for the lower mantle from GST in 1986 was much higher than determined, at relatively low pressures, in the laboratory. Only recently the mineral physicists were able to reproduce the tomographic value.

Gravity. Under the assumption that seismic anomalies are proportional to density perturbations, they provide constraints on the modeling of large wavelength gravity anomalies, the viscosity distribution in the mantle, and the ratio of density perturbations to velocity perturbations.

Petrology and geochemistry. The GST models have the potential to provide integral constraints on petrological and thermal models of the ridge systems. The top image on the facing page shows a perspective view of -- 0.75 percent contour of shear velocity anomaly in the upper mantle. The trough associated with the mid-oceanic ridges is deeply depressed, and on occasion extends into the lower mantle. On the other hand, the negative anomalies under the back-arc basins are relatively shallow. The negative velocity anomaly under the East Pacific Rise is much wider and deeper than those under the north and south segments of the Mid-Atlantic Ridge. Velocity anomalies associated with the continental shields confirm the hypothesis of "continental roots." The cross sections through the upper mantle of North America, Africa, and Europe show that positive velocity anomalies extend to 400 kilometers or so. Correlation has been noted between low velocity anomalies near the core-mantle boundary, such as seen below, and the occurrence of large-scale isotopic anomalies.

Geomagnetism. Coincidence of regions with a high rate of secular variations and slow velocity anomalies (higher than average temperature) provides constraints on the thermal and mechanical coupling between the mantle and the core. Several recent papers point out that the virtual geomagnetic pole paths coincide with the two high-velocity regions circumscribing the Pacific and, effectively, connect the North and South poles.

Geodesy. GST models of velocity anomalies and topography of major discontinuities (CMB, 660 kilometers) can be compared with the data on Earth's rotation obtained by other techniques, such as very-long baseline inferometry.

In addition to the ever-present goal of improved resolution, the most urgent challenge in GST modeling is deriving comparable quality results for P- and S-wave heterogeneities, which is necessary to distinguish between the thermal and compositional variations.

Adam M. Dziewonski received his education in Poland at the University of Warsaw and the Academy of Mining and Metallurgy in Cracow. He arrived in the United States in 1965 and has been at Harvard since 1972. Even though his official title there is Professor of Geology, most of his work is in the area of global Earth structure, including one of the first tomographic studies in 1975. Having run out of data provided by the land-based stations, he started several years ago with G.M. Purdy of Woods Hole Oceanographic Institution and John Orcutt of Scripps Institution of Oceanography, among others, promoting the concept of the Ocean Seismographic Network. As an early warning of the possible consequences, he was appointed a co-chief scientist of ODP Leg 136, during which a hole off Hawaii was drilled for future experiments with a broad-band seismographic system.