It is shown how surface impedance concepts can be incorporated into the FD-TD technique. To show the accuracy of the proposed approximation, the maximum percent error for any angle of incidence versus loss tangent p is graphed. It is demonstrated that the impulse responses can be approximated by a series of exponential functions, typically 5 to 10. The convolution of each exponential with H/sub x/(x,t) then requires only one storage location for a running sum variable. This is a great savings and makes the surface impedance concept a viable option for use in FD-TD codes. >