Imaging by forward propagating the data: theory and application

ABSTRACT The forward (modelled) wavefield for conventional reverse time migration (RTM) is computed by extrapolating the wavefield from an estimated source wavelet. In the typical case of a smooth subsurface velocity, this wavefield lacks the components, including surface reflections, necessary to image multiples in the observed data. We, instead, introduce the concept of forward propagating the recorded data, including direct arrivals, as part of RTM. We analyse the influence of the main components of the data on the imaging process, which include direct arrivals, primaries and surface‐related multiples. In our RTM methodology, this implies correlating the forward extrapolated recorded data wavefield with its reversely extrapolated version prior to applying the zero‐lag cross‐correlation imaging condition. The interaction of the data components with each other in the cross‐correlation process will image primaries and multiples, as well as introduce cross‐talk artefact terms. However, some of these artefacts are present in conventional RTM implementation and they tend to be relatively weak. In fact, for the surface seismic experiment, forward propagating the direct arrivals is almost equivalent to forward propagating a source and it tends to contribute the majority of the data imaging energy. In addition, primaries and multiples recorded in the data become multiples of one higher order. Forward propagating the recorded data to recreate the source will relieve us from the requirement of estimating the source function. It will also include near‐surface information necessary to improve the image in areas with near‐surface complexity. Data from a simple synthetic layered model, as well as the Marmousi model, are used to demonstrate some of these features.