Seismic data acquired directly over near-surface limestone formations are c
ommonly observed to be of inferior quality. For a dataset from West Texas,
we found that data acquired over mesas are badly contaminated by wave-energ
y scattered at shallow depths. Despite intensive data processing, we enhanc
ed the data only marginally. One 'effective' method was a double dip-filter
ing of common-source as well as common-receiver gathers (Larner et al., 198
3) which removed all events propagating at speeds less than the P-wave velo
city at the surface. Another 'effective' procedure was to remove all cohere
nt wave-energy propagating toward the zero-offset point by triple FK-filter
ing common-source, common-receiver, and common-midpoint gathers.
Possible mechanisms for this scattering in the near-subsurface are the weat
hering layer, rough free-surface topography, or volume heterogeneities such
as clusters of vugs, joints, or cavities. We applied different numerical s
cattering schemes to study the effects of each of these scattering mechanis
ms. For our West Texas dataset, we find that the weathering layer is the do
minant cause of source-generated noise on records acquired in valleys. Howe
ver on the mesas, we find that topography alone cannot cause the observed s
cattering. We believe that near-subsurface volume heterogeneity is a major
cause of scattering. Using energy-density and energy-flux vectors to study
the frequency dependence of the different scattering models, we observe tha
t near-surface heterogeneities form waveguides which efficiently trap seism
ic energy near the surface. In addition, the wavefields penetrating the het
erogeneous surface region are rendered incoherent masking reflections from
deeper targets.