The prediction of the directivity function of a GPR antenna still remains a
partially unsolved problem because of the subject complexity. First, the f
ar-field conditions are often not satisfied and second, the antenna design
has little in common with the Hertzian dipole for which an analytical appro
ach can be used. Our contribution is both theoretical and experimental. On
one side, we solve the (electromagnetic) EM integral equations numerically
to derive the wavefield components from near to far-field distances. On the
other side, we experiment with two novel techniques for measuring the dire
ctivity functions in the near to far-field range on dry and saturated sand.
Theoretical and experimental results show that neither the analytic approx
imation of far-field directivity nor the numerical integration of near-fiel
d directivity can perfectly match the measured functions, although near-fie
ld solutions are generally more consistent. The mismatch should be attribut
ed to the present-day GPR antenna design that includes absorbers and shield
s. Although the effects of these elements are not included in the present n
umerical near-field solutions, we believe the approach to be of practical v
alue to predict an average directivity function. A smoothed version of the
analytic far-field solution can also be used in the range of the near to fa
r-field transition but near-field solutions are really recommended when TX-
RX distances are shorter than five wavelengths.