In an isotropic media, the 21 anisotropic elastic rock parameters reduce to only 2 elastic rock parameters (bulk and shear moduli), each of which are simply a function of density, and P-wave and S-wave velocities. Detailed volumes of work using measured crustal velocities (e.g. Christensen & Mooney, 1995) are now leading to accurate constraints upon the petrology and mineralogy of crustal and upper-mantle rocks. Former ``outsiders'' such as geochemists and petrologists are now able to integrate their skills with the results of seismic-based studies. A particular project of interest involves the attempted identification of the deep source regions of xenoliths; lower-crustal and mantle rocks that have been rafted to the surface within magma uprisings. Having identified the physical properties of such rocks found at the surface, the challenge is to identify such parameters at great depths.
The use of AVO phenomena in deep seismic studies has been historically hindered by a lack of confidence in the validity of applying the concept to sparsely recorded, poor-quality data. In recent years however, AVO studies have been successfully applied to infer intra-crustal ``melt zones'' (e.g. Makovsky et al. 1996), transitions in the petrology of igneous intrusions and the presence of intra-crustal water.
Obviously at the crustal scale, anisotropy can be profound. Rather than being considered a hindrance however, crustal studies often rely upon the presence of anisotropy to infer useful observations. Historically, three of the most important mineralogies in the lower-crust/upper-mantle (to crustal geophysicists) have been olivine-rich rocks such as harzburgite, peridotite and dunite (orthorhombic mineral structure and hence anisotropic), eclogites (highly metamorphosed anisotropic rocks) and garnets (cubic mineral structure and hence isotropic). In extending oceanic crust near diverging plate boundaries, strain-related alignment of the olivine-rich rocks perpendicular to the ridge axis imposes a strong anisotropy. All of these examples relate to gross-scale applications. As the quality of recorded reflection seismic data improves, the use of anisotropy measurements at a more local scale will conceivably increase in application. Furthermore, the establishment of correct velocity models and structural geometry depends upon the accurate incorporation of anisotropy parameters, so it is expected that many detailed seismic profiling efforts will look to the developments in anisotropic data processing.
Geophysicists and geologists these days want to resolve both the shallow and deep geology together, such that the deep controls upon shallow structures can be established. This quest then demands the acquisition of high frequencies as well as the lower frequencies that have historically sufficed. Even at the most regional scale (using teleseismic data), the recording of higher frequencies should be of great use, conceivably using receivers anchored in boreholes that have been drilled into consolidated rock. Such acquisition of higher frequencies would allow the recording of lower magnitude earthquakes than is presently possible (at present, it is difficult to remotely record earthquakes with M < 5). As the number of global earthquakes varies roughly as the exponential inverse of magnitude, such recordings would create a wealth of new information.
At a simpler level, much of the challenge in using deep reflection seismic data is an interpreter's problem. Often the data quality is so poor that the interpreters find it less ambiguous to produce ``line drawings'', in which the dominant seismic events are traced by hand, effectively removing the contribution of the weaker signal and noise. This simplistic ``coherency'' filter is an indication that there exists great scope for the development of more powerful and automated coherency filtering techniques. Application of techniques such as ``skeletonization'' (Cook & Vasudevan, 1996) and fractal-based data dissemination already hint at the usefulness of such approaches. Coherency filtering is still in its relative infancy in commercial exploration data processing (e.g. Schwab et al. 1996), so it appears that this is one area in which the application of new industry technology will quickly follow in the deep seismic profiling domain.
What are the processing developments that hold most promise for the
future? Historically, due to the prohibitive logistical and financial
obstacles, 3D acquisition has never been undertaken. At present,
Scripps Institution of Oceanography (University of California, San Diego)
is acquiring the first 3D survey over the overlapping spreading center
of the 
N segment of the East Pacific Rise, Pacific Ocean, which will
obviously incur processing problems never before encountered in
crustal studies. These problems will open the way for the design and
application of routines such as DMO and pre-stack migration.
Certainly, the imaging of such a complex structure
is going to be a challenge. As consortium experiments between academia
and industry focus more upon natural resource regions, the associated
larger budgets will allow a previously unattained standard of
acquisition: close receiver spacing (of the order of 40 m), high data fold,
accurate surveying and thus refraction statics computations, and source
parameters designed to generate high-frequency signals. Hence, such
datasets will more closely resemble conventional exploration datasets
than has been the case in the past. In fairness to academia,
the industry will never acquire data in geologically interesting areas like
the Himalayas. Nevertheless,
there
would appear to be little reason to deny that industry surveys will conceivably
provide many high-quality images of the
deep geological features that have beguiled crustal geophysicists for so long.
In turn then, academia will look to industry to lead the way.