The proposed approach addresses several challenges in integrated studies, specifically (1) interpolating missing well logs at wells that have incomplete well log suites and (2) providing a methodology for semiautomatically tying wells to seismic and validating the consistency of the ties. In our example, the proposed approach accurately computed a TDR at all well locations regardless of the completeness of the initial well log suite. Consequentially, integrated studies need not be constrained to pilot wells where full log suites are collected. The proposed approach should be particularly useful in onshore plays where the number of wells drilled is much higher compared to the number of sonic and density log acquired.

Our method involves interpolation techniques that assume that rock properties do not vary significantly laterally. We make several additional assumptions related to the interpolation of missing log data:

  1. Gamma logs are matched to estimate the alignment shifts; therefore, estimated section is limited to section in each well with available gamma log.
  2. All gamma logs are aligned with a single reference gamma log, estimated log section is limited to the stratigraphy found in this reference log. This reference well log can be thought of as a type log which contains the entire stratigraphic column observed in other well logs.
  3. We did not perform fluid substitution prior to solving Equation 8 for each well. The proposed approach is based on interpolation, and we assume fluid substitution to have negligible impact on the results. This assumption may present challenges in reservoirs where hydrocarbons impact the well log response within the same stratigraphic interval.
These assumptions may not be valid in geologically complex areas with significant stratagraphic variations such as unconformities or channels where entire stratigraphic units may be absent due to erosion. Additionally, the well log correlation approach, may meet similar challenges as those experienced by conventional, interpreter driven, workflows where rapid stratigraphic variability (e.g. slope deposits, clinoforms, etc.) may not correlate, or may correlate ambiguously in several places, between wells. Although we did not account for changes in the fluid content, the proposed approach provides a reasonable first-order approximation of the unknown well logs. The predicted velocity and density well logs provides the minimum required logs to forward model a synthetic seismogram and tie the well with real seismic data.

Additionally, well log data interpolation by predictive painting may result in errors when crossing faults. We observe this challenge when comparing the fault in Figure 17a to the log property volumes in Figure 18. The fault in the property volumes is not accounted for during interpolation. Interpolation schemes that account for discontinuities would further improve results. Recently suggested approaches by Xue et al. (2017) and Shi et al. (2017b) address predictive painting across faults.

Our methodology can also be directly impacted by errors in the seismic data. An incorrect migration velocity will improperly place reflectors which will result in an incorrect TDR estimated from seismic well ties. Inaccuracies in the migration velocity may be a reason for the need for a velocity log update shown in Figure 15a.

The assumptions we make and errors in migration can compound resulting in inaccuracies in our integrated study. However, by relating several sources of information (multiple well logs and 3D seismic) we provide an approach and validation technique to minimize the impact of these challenges and to provide a better characterization of the subsurface.