Extracting detailed lithology from seismic data

Abstract

The interpretation of seismic data in a traditional sense is limited by the bandlimited nature of the measured data. The frequencies of the data lead to a ?resolution limit? within which rock properties resulting from seismic inversion are inherently limited. However, the ?detection limit? of the seismic is another measure of its sensitivity to changes in rock properties and is typically much lower than the resolution limit. The only means of extracting information all the way to the detection limit is via a seismic inversion algorithm capable of exploiting the information derived from the clustering of elastic properties. Generally, such clusters are associated with different lithologies with distinctive rock properties. The presence of hydrocarbons is well known to affect the elastic properties of rocks. However, the large variation in reservoir rocks makes it nearly impossible to generalise on this effect. If multiple elastic properties can be measured the likelihood of identifying clusters is increased because some properties may cluster in one dimension more than in another. A cost effective way of obtaining multiple elastic properties is by inverting the AVO character of pre-stack seismic data. Using a series of synthetic and real data examples we will demonstrate how a geostatistical inversion algorithm can extract both detailed lithology and elastic properties from angle-stack seismic data. The inclusion of geostatistical parameters further constrains the results as the variograms impose constraints on the thicknesses and lateral continuity of the lithologies. The results are verified using a cross validation approach where pseudolog extractions from inversions not constrained to the well logs are compared to the known but unused logs. The close comparisons of the predicted and measured logs provide compelling evidence of the reliability of the predicted lithologies and elastic properties at and away from well control.