Drilling in frontier areas provides many challenges to drilling engineers in designing and planning for a well at a desired location. Traditionally, engineers have relied on information from an offset well to design and plan a new well. Unfortunately, demand for exploration in frontier areas and deep targets must contend with the lack of offset well data.

Modern seismic data and seismic data processing can offer high-quality velocity data that provide pore pressure prediction in such areas. The data also offer useful information to fill in the gaps between wells in areas of sparse well coverage. Surface seismic velocities from compressional (P) wave energy are influenced by the compaction of clastic sediments. Pore pressure affects the compaction; therefore, changes in formation velocity can be calibrated to changes in pore pressure, assuming there is no change in lithology. Velocities derived from surface seismic data can provide an indirect way to predict pressure in the subsurface before drilling.

Two scenarios applied this concept. In Area 1, five wells with logs of varying quality were available to calibrate the velocity from 3-D prestack depth migration; however, pressure prediction profiles were needed closer to these wells. Area 2, approximately 36 miles (60 km) northwest of Area 1, has no available well data.

It is common to calibrate the seismic velocities to velocities from sonic data recorded in the well bore. The calibration process can establish many parameters needed for the appropriate pore pressure calculation methodology.

The seismic velocity pressure prediction compares to the log predictions in the old wells in Area 1. (Images courtesy of Petronas)

Pressure prediction from well data

The initial step checks the behavior of the sonic log velocities encountering the pore pressure in existing wells. Pressure predicted from the sonic velocities should indicate that the seismic velocities would have similar behavior and predict the best possible model. All of the available well data of interest in Area 1 were collected for the initial pore pressure calculation. The first step was to establish the vertical stress or overburden gradient (OBG) from integrating the available density logs. Several wells were used to establish a composite density function because of missing sections or bad-quality density logs. The vertical stress for the first 1,640 ft (500 m) below the mudline was calculated using the Miller empirical formula. Pore pressure prediction for the initial input wells was calculated using calibrated pressure models employing the velocity-effective stress relation.

The matrix stress ratio is determined from the leak-off test measurements for calculating the fracture gradient. These values normally are estimated from the pressure-versus-time graph. This method makes calibration more difficult because of the calculation procedure. In this project, the sonic log data were used as an alternative.

Depth migration velocities

The final velocity model from prestack time processing was used as the initial model-building process for the tomography inversion of the prestack depth migration. The velocity model output of the last iteration from the tomography inversion was used as input to the residual move-out analysis.

A high-density simultaneous velocity analysis technique was used to pick a high-density VRMS and ? (effective eta) field. This helps flatten the events to a higher incidence angle than the output from a second-order stacking velocity correction. This automated velocity analysis was performed using a grid of eight inlines by eight crosslines at 300 ft by 300 ft (100 m by 100 m). An angle mute of 50 degrees was applied on the input gathers.

Automatic velocity-picking processes generally are noisy because of high-frequency picking, so geostatistical filtering is required for pore pressure prediction. To remove the noise in the velocity cube, velocity volume is decomposed into two components. The low-frequency component or trend cube should preserve the structural component of the velocity, and a residual cube should preserve the fine-scale variations and noise. Such decomposition has the following advantages:

  • The residuals are stationary, which is suitable for the condition to perform factorial kriging; and
  • The structural component of the velocity is preserved.

Factorial kriging was performed on the residuals to remove the noise. The final filtered velocity cube is the sum of the filtered residuals and the trend cube.

Seismic velocity pressure prediction is compared to the logs in the new wells in Area 2.

Interval velocity computation

The seismic velocity field (RMS) has to be converted to interval velocity to be used for pore pressure prediction. Normally, the regularly sampled velocity field (in time) is converted using the Dix approximation. This technique is known to produce some instability; to preserve as much information as possible, a small sample rate has to be used.

This new final interval velocity cube in time domain was considered as an attribute that preserves the velocity variations due to pressure influence. It was converted to depth using the same average velocity field from prestack depth migration that was used for the depth conversion of the newly imaged seismic data. In this way, the consistency of the depth information remains the same for the velocity attribute and the seismic interpretation.

Velocity calibration, pressure prediction

New seismic interval velocities were cross-plotted with the velocities from the sonic logs of the wells in Area 1. Velocities from the wells with stability problems deviate dramatically from the regression line. The regression was improved by removing the problem wells. To improve the correlation further, the residuals between the seismic and well velocities were computed and kriged in 3-D. These residuals then were added to the initial seismic velocity field. This method was tested using a kriging radius of three and six miles (5 and 10 km). One well was dropped from the kriging process to serve as a blind well test to give more confidence in the process. The good correlation with the blind test was the basis for applying the kriging with a radius of three miles.

This diagram shows the workflow to generate velocity attribute and accurate pore pressure prediction.

A calibrated Gardner relation was derived based on the well data to estimate the density from the P-wave velocity. This new calibrated interval velocity cube was converted to density after a first-order regression was applied in the correlation between the wells’ density logs and the derived density. The resulting equation then was applied to the “seismic” density cube to perform a residual calibration. Subsequently, the OBG was calculated. The correlation of the OBG from seismic and OBG from calibrated seismic density at the wells was examined. This final residual calibration improved the fit to the OBG from the original well densities as well as the OBG from the composite density function of the stability problem wells. This volume was used in the final pressure volume calculations.

The normal compaction curve (NCT) was computed on the same criteria observed from the well logs. The interval velocity cube was scanned to calibrate the NCT curves on a defined grid and interpolated to cover the whole final volume grid. This “scanning” process consists of extracting the velocity trace at the selected locations and defining the best lambda parameter for the Miller equation.

Good seismic data quality combined with good data processing and understanding of the geological model can achieve accurate pore pressure prediction in frontier areas. A robust and powerful workflow for velocity model building for pore pressure computation predicts pore pressure in such areas.