The frequency dependence of attenuation is typically far greater for subsurface radar waves than for seismic waves. Since this frequency dependence causes the subsurface radar pulse shape to change as it propagates through the earth, conventional wavelet deconvolution techniques are often inadequate to reconstruct the earth's reflectivity series. The application of a time-variant filter that undoes the effect of frequency-dependent attenuation can substantially improve the resolution of subsurface radar data. This filter is minimum-phase with amplitude characteristics equal to the inverse of the attenuation function of the medium investigated. Preliminary tests on data from surveys over a water tank, an underground mine stope, and fractured granite suggest that good results can also be achieved by assuming attenuation is linearly proportional to frequency, similar to a constant Q model.