The reflected P- and S-waves in elastic displacement component data recorded at the earth's surface are separated by reverse-time (downward) extrapolation of the data in an elastic computational model, followed by calculations to give divergence (dilatation) and curl (rotation) at a selected reference depth. The surface data are then reconstructed by separate forward-time (upward) scalar extrapolations, from the reference depth, of the magnitude of the divergence and curl wavefields, and extraction of the separated P- and S-waves, respectively, at the top of the models. A P-wave amplitude will change by a factor that is inversely proportional to the P-velocity when it is transformed from displacement to divergence, and an S-wave amplitude will change by a factor that is inversely proportional to the S-velocity when it is transformed from displacement to curl. Consequently, the ratio of the P- to the S-wave amplitude (the P-S amplitude ratio) in the form of divergence and curl (postseparation) is different from that in the (preseparation) displacement form. This distortion can be eliminated by multiplying the separated S-wave (curl) by a relative balancing factor (which is the S- to P-velocity ratio); thus, the postseparation P-S amplitude ratio can be returned to that in the preseparation data. The absolute P- and S-wave amplitudes are also recoverable by multiplying them by a factor that depends on frequency, on the P-velocity α, and on the unit of α and is location-dependent if the near-surface P-velocity is not constant.