Using recursive inversion as input for gross-rock volume extraction from lithology prediction volumes: How bad can it be?

Abstract

Summary Reserves and Resources can be directly estimated by using seismic lithology prediction volumes generated from AVO inversion, calibrated to wells, to estimate sand rock volumes within a stratigraphic interval. However, high quality AVO inversion data and studies are not always available. This case study utilises recursive inversion to generate prediction cubes for input to geobody based gross rock volume estimates. Here, relative recursive inversions of the near and far seismic stacks are generated. EEI rotation theory was applied to calculate relative AI and Vp/Vs volumes. These have then been converted to band-limited absolute inversion volumes by adding a low frequency model built using seismic horizons, well logs and rock physics trends. Finally, probability density functions calibrated to wells were estimated to calculate lithology prediction volumes. A sand probability volume is then calculated using these probability density functions. A relative approach has not been applied due to poor separability of different lithologies in cross-plot space. This method is applied to an area where a single multiazimuth PSDM seismic survey covers two gas fields and several deep exploration prospects. However, previous inversion studies were limited to the individual fields (each inverted separately), incorporated fluid contact information and did not cover the deeper exploration, so were therefore considered sub-optimal. Although there is potential for results from this method to be “bad”, this case study was successful. The inversion volumes generated as part of this study enabled a wholistic view of the fields and exploration prospectivity, which had not been previously possible with the available QI volumes. The seismic data used for input was exceptionally good, and there was abundant well control to provide control and for use in blind testing. The amount of validation and quality-control applied to this project cannot be under-stated. It is critically important to be mindful of the limitations and broad assumptions that are applied as part of this work-flow. These include the addition of the low-frequency model, wavelet affects not being taken into account and depth decay of the lithology predictions due to the application of a single PDF.