An integrated approach to geostatistical depth conversion and gross rock volume estimation

Abstract

Static, dynamic, and geomechanical models are typically built in depth because this is required when planning and drilling wells. These models are usually driven by seismic data via structural interpretation and seismic attributes, whose position is fixed in the time domain. To incorporate seismically derived data, or interpretations, time objects must be transformed to the depth domain using the process of depth conversion. The depth-conversion model should dynamically evolve during project life, from exploration to maturation. As more data become available and the geologic understanding improves, an evolving model can accommodate requirements for specific deliverables for various disciplines, e.g., gross rock volume (GRV) estimates, drilling uncertainty windows, and structural base case models. To meet the above specifications, we used a fully geostatistical method that enabled the incorporation of a large variety of data types and their associated uncertainties. We performed depth modeling simultaneously for all layers in the model, in which all layers were coupled. This imposed a dependency on depth surfaces, such that a depth surface was influenced by surfaces above and below, moving away from a classical top-down approach of depth conversion. A scriptable and mechanized method allowed for fast scenario modeling in which various geologic and geophysical concepts or model parameterization choices can be tested to provide discrete models and avoid anchoring at a single preferred model and associated probabilistic range. Depth results from multiple scenarios can be combined to provide common confidence intervals. The proposed method including scenario modeling has been successfully implemented in the UK Culzean HPHT gas field. The model successfully delivered input to the various disciplines in the asset team, providing depth structures for static model build, GRV estimates for each reservoir, and localized drilling uncertainty estimates.