Source signature variations versus repeatability - a study based on a zero-offset VSP experiment

In 1998 a zero-offset VSP survey was acquired in the North Sea. As a part o f this survey a dedicated VSP repeatability survey was acquired.100 test sh ots were fired in less than 3 hours, and these shots are the basis for the repeatability study that is described in this paper. Due to acquisition rel ated problems, only the z-component of one of the five geophone tools was u sable for this study. Stacking of neighbouring shots decreased the RMS (Root Mean Square) differe nce from approximately 16% (no stacking) to 3-4% (stack of 10 neighbouring traces). Increasing the number of traces going into the stack further, did not improve the repeatability significantly. We found a strong correlation between variations in the source signatures and lack of repeatability in th e VSP data. We found that a simple deterministic deconvolution using the ne ar-field recordings improved the repeatability considerably. Near-field rec ordings can therefore be valuable both for source monitoring and for later corrections of time-lapse seismic data in order to obtain increased repeata bility.