In this report, we propose a method of computing the traveltimes at the points of a regular grid by interpolating between adjacent rays. This procedure is based on the physical continuity of the traveltime field and uses rays generated by paraxial ray tracing Beydoun and Keho (1987); Cervený (1987) or Huygens wavefront tracing Sava and Fomel (1997). The paraxial ray tracing method, though not very robust when applied to models with big velocity contrasts, is very accurate in estimating times, and therefore it is expected to produce reliable traveltime maps.

The difficulty of interpolating traveltimes in media with complex velocities is that the rays do not follow smooth and uniform paths. In areas of high velocity variation, they bend and cross each other making the interpolation extremely difficult, if possible at all.

To avoid such a situation, we use additional parameters to separate the rays
so that they no longer intersect. For example, in the 2-D case, we can
associate with each point of a ray the value of the take-off angle.
We convert the 2-D problem, in which the points of the rays are described by
their *x* and *z* coordinates, into a 3-D one, in which the points of the rays
are described by their *x*, *z* and p parameters. The rays then appear to be
``stacked'' in the p dimension in equally spaced planes, defining a continuous
surface whose bending can represent the multiple, successive arrivals at a
given location (Figure 1).

Figure 1

A similar approach is possible in the 3-D case, where we describe the points
on the rays by their *x*, *y*, *z*, p, and q coordinates, where p
and q are the two take-off angles (azimuth and inclination).

10/9/1997