2. Data Requirements

2.1. Seismic Data

SNP does not directly use seismic data however it is designed to operate on attributes extracted from time domain, coloured inversion (CI) datasets.  The SNP algorithm assumes a binary model such that pay and non-pay lithologies have constant and distinct impedance values.  Attributes must therefore be extracted from a dataset approximately consistent with this model.  Extended elastic impedance theory (EEI & AIGI) may be used to construct an appropriate dataset.  A fluid impedance volume should be used for calculating net pay and a lithological impedance volume for net rock volume

Three input horizon attributes are required:

  1. Vertical time thickness (isochron) of the reservoir interval

  2. Vertical depth thickness (isochore) of the reservoir interval

  3. Average amplitude across the reservoir interval

Horizon attributes are read directly from Petrel.

All three input horizon attributes must be consistent i.e. derived from the same initial time picks.  These time horizons must be picked on zero crossings on Coloured Inversion data.  If the reservoir is low impedance relative to non-pay then the top pick will be on a +/- crossing and the base pick will be a -/+ crossing.  The reservoir interval should be isolated from other reflectors above and below it.  The interval may contain multiple seismic loops but accuracy will decrease the thicker the interval.

SNP requires a wavelet specification in order to detune the amplitude response. The wavelet must be specified in trapezoidal format. This is consistent with coloured inversion, which should result in a dataset having a flat-topped wavelet spectrum.

The wavelet may be estimated using any suitable spectral estimation procedure and should be representative of the spectrum at the location and two-way-time of the reservoir. It is not recommended to use the CI bandwidth parameters to define the wavelet. Note that the detuning algorithm is particularly sensitive to the low-end wavelet parameters and relatively insensitive to the high-end parameters.

Net-to-gross in time will differ from net-to-gross in depth if the pay and non-pay velocities are different.  SNP can correct for this if the user supplies the ratio of the pay to the non-pay velocities.  Unless the ratio differs significantly from unity the correction is small.

Net pay estimates can be made without any well data based only on self-calibration.  This method assumes that the reservoir peaks at 100% net-to-gross somewhere within the study area.  This can be a good assumption for thinner, turbidite reservoirs but will be increasingly inaccurate for thicker reservoirs and other depositional settings.

Well data will provide further constraints.  The user must provide the inline and crossline location for each well and the net pay value.  The net pay value is the verticalised pay between the seismic picks.  The calibration method assumes vertical wells; for wells deviated across thicker reservoir intervals the calibration will necessarily average across the lateral extent of the well penetration.

2.6. Uncertainty Estimation

This Dialog is used to estimate the errors that can arise from calibration errors and inaccurate wavelet estimation. The results are displayed in several charts and in tables of standard deviation values as a function of apparent thickness and seismic net-to-gross.

SNP may optionally output six horizon attributes:

  1. Net-to-gross (time)

  2. Net-to-gross (depth)

  3. Net Pay

  4. QC horizon

  5. Bias Uncertainty

  6. Total Uncertainty

Horizon data is written back directly to Petrel.

 As previously discussed, net-to-gross in time and net-to-gross in depth will generally be very similar or identical if the velocity ratio is one.  Net pay is the product of net-to-gross in depth and the isochore.  The QC horizon provides a number at each trace location indicating how each data-point was treated by the application (i.e. whether it was valid data etc.)

SNP will also generate a Parameter Report which is a text file containing a summary of all the input data and parameterisation.