Surface waves provide key information about the lithospheric structure of the Earth. They are dominantly sensitive to shear velocity variations down to sub-lithospheric depths, and hence are a good data source for exploring regions of anomalously high or low velocities, or for mapping variations in crustal thickness (since the base of the seismological crust is in part defined by a large jump in shear velocity with depth). The technique used is called tomography, and involves applying inverse theory to infer the velocity structure given data about the surface wave group or phase velocities.

We have applied these methods to explore beneath the Tibetan Plateau and surrounding regions using inter-earthquake surface wave phase velocities, to map the velocity structure beneath the Eurasian continent, and our current work is devoted to producing global crustal thickness and average crustal shear velocity maps.

A key feature of almost all current earthquake tomography is that the inverse theory applied is linearised. This means that the relationships between the (surface wave) data that we measure, and the Earth model that we wish to estimate, are approximated by a linear function. This is an approximation, and particularly for geophysical problems which are all nonlinear in reality, it can be a very poor approximation causing bias and error in the Earth models found. As a result, all estimates of uncertainty in Earth models given to-date are likely to be in significant error.

*For the first time ever, we have produced a global tomographic model of the crust that employs no linearised methods.* The inverse theory used is fully nonlinear (it involves using neural networks to embody the full, nonlinear relationships) and produces full probability distributions that describe the uncertainty in the result of the inversion. This is a paradigm shift in the field of seismology as it shows that nonlinear methods *can* be made sufficiently efficient to be applied on to large scale problems (against popular opinion!)

- Tomography under Tibet and Eurasia using surface waves
*(Curtis and Woodhouse, 1997; Curtis et al., 1998; Devilee et al., 1999; Curtis and Snieder, 2002)* - Global tomography
*(Ueli Meier, Andrew Curtis, Jeannot Trampert. Geophys. J. Int., in press 2007; Ueli Meier, Andrew Curtis, Jeannot Trampert. Geophys. Res. Lett., submitted, 2007)* - Application of neural networks to invert surface wave velocities for Moho depth (crustal thickness) and other sub-surface discontinuities
*(Devilee et al., 1999; references above with Ueli Meier and Jeannot Trampert)* - Modelling and inverting for crustal deformation associated with large earthquakes
*(Curtis and England, 1997)*

**All citations are included in the reference list at the foot of my main page**.