A 3-D cube requires four variables: 3 independent and 1 dependent. Independent variables can be spatial (longitudinal distance, ransverse distances and depth) or mixed (e.g. longitudinal/transverse distance and frequency). The dependent variable is the measured quantity (e.g. tilt angle, ellipticity, in-phase component etc.). The crucial question is: which parameter are you proposing to use for the third independent variable?
I presume that the depth is a result of modelling or inversion, or can be independently estimated... If not, I do not see how this is possible because VLF cannot yield depth as an observable. Then you simply have to generate a four column matrix, each row comprising the three independent and one dependent variables and use it as input to a 3-D interpolator. Which brings us to the second crucial question: what kind of interpopation (gridding) software do you have?
Surfer cannot generate 3-D volumes. Period. You can try Voxler though, which is also by Golden Software and is a very good program. I mainly use MATLAB and the interp3 routine. There's a respectable number of other programs, free-, shareware or prooprietary. Google the keyword "3D interpolation" and you'll come up with a lot of answers... Be carefull to choose a program that can cope with missing data (recall the trapezoidal shape of a Karous-Hjelt section) and go for algorithms based on Voronoi tessellation schemes (e.g. natural neighbor interpoplation and the likes.) Good luck...
Thank you very much Tzanis, my friend also told me about Voxler. I'll now try this one, and will post my experience with it.
I tried Matlab also, but was unable to generate the matrix as I dont have all the data points, so I need to extrapolate as we do in surfer for 2D sections.