It is often very useful to be able to smooth velocity fields estimated
from exploration seismic data. For example seismic migration is most
successful when accurate but also smooth migration velocity fields are
used. Smoothing in one, two and three dimensions is examined using No
rth Sea velocity data. A number of ways for carrying out this smoothin
g are examined, and the technique of locally weighted regression (LOES
S) emerges as most satisfactory. In this method each smoothed value is
formed using a local regression on a neighbourhood of points downweig
hted according to their distance from the point of interest. In additi
on the method incorporates 'blending' which saves computations by usin
g function and derivative information, and 'weighting and robustness'
which allows the smooth to be biased towards reliable points, or away
from unreliable ones. A number of other important factors are also con
sidered: namely, the effect of changing the scales of axes, or of thin
ning the velocity field, prior to smoothing, as well as the problem of
smoothing on to irregular subsurfaces.