I have to simulate pulse wave propagation in soil using FDTD, my frequency of interest is 100MHz to 3GHZ (for different pulse widths, as of now, my source is in a form of double exponential signal).

I am searching for a good soil model, which considers dispersion effect, mixture, temperature(approx), soil texture. I need a model, which can be implemented in FDTD with less computational effort, as of now I have found much discussion on debye based model on FDTD implementation with improved effort in computational aspects. I found two types of soil modeling.

1. Experimental data interpolation (on cole-cole, cole-davidson..etc)

2. Mixture based model (much difficult to implement in FDTD), also I found some literature on circuit based model.

Even I try to find time domain model (at least for Gaussian input specific) of soil, where I can avoid convolution term in FDTD. Could you suggest to me a soil model, which may satisfy my requirements?

I know it is difficult to consider all the aspects of soil, and I have to make some approximations, but i am confused on what to approximate. I am not able to clearly conclude, what to neglect and what to consider (it is a irony of many soil models). Each paper suggests something different. Some say soil texture effect is negligible for above 1GHz whilst another piece of literature says the totally opposite. This even occurs with soil particle orientation, gravels, etc.