IPA13-G-050

PROCEEDINGS, INDONESIAN PETROLEUM ASSOCIATION Thirty-Seventh Annual Convention & Exhibition, May 2013

A NEW APPROACH TO TIME DEPTH CONVERSION AND DEPTH ERROR ESTIMATION FOR 2D INTERPRETATION USING SEISMIC VELOCITIES

Rosidah Hardiani *

Sanisah Soed *

Lothar Schulte**

ABSTRACT

The area of investigation covers 15,785 square km in Sarawak Malaysia. It is covered by 128 2D lines oriented in NW-SE and NE-SW direction. For depth conversion of two key horizons, seismic velocities and one exploration well are available.

The seismic velocities were loaded as data-points, DIX converted and adjusted to the well velocities to address velocity anisotropy. Then the interval and average velocity points were derived for both seismic layers. These data points were visualized in a 3D display window that facilitated easy identification of true velocity and deletion of apparent outliers. The edited point sets were interpolated and smoothed in order to remove noise not addressed by the point editing. The optimum smoothing operator was determined thru maximizing the correlation coefficient between the time horizon and the filtered seismic velocities. The two time surfaces were depth converted based on the smoothed interval velocity surfaces (layer cake approach). Results demonstrate that this depth converted 2D interpretation is more accurate than depth converting the interpolated time horizons. In addition the lower horizon was depth converted using the average velocity set. Its depth surface showed differences of more than 100ft compared to the result from the layer cake approach.

In order to estimate depth discrepancy the velocity differences at intersecting 2D lines were calculated. For each 2D time horizon the resultant residual velocity standard deviation was converted into depth, delivering a distribution of the time dependent depth error standard deviation. The analysis shows that the layer cake approach delivers smaller depth errors for the lower horizon than the average velocity approach. Apparently, smoothing

of the interval velocities can address the velocity noise more precisely than smoothing of the average velocities.

INTRODUCTION

The time depth conversion of seismic horizons of an exploration area is typically challenging because of lack of well data. The area of investigation for this paper lies offshore Sarawak, Malaysia and has an area of 15,785 square km. It is covered by 128 2D lines oriented in NW-SE and NE-SW direction. For the depth conversion of two key horizons, seismic RMS velocities and one exploration well were available. Different approaches were used for the time depth conversion and their results compared. The reliability of the depth conversion result was estimated from the noise inherent in the velocity data.

DATA PREPARATION

The available seismic velocities were DIX converted and the interval velocities derived for each seismic layer. The velocities showed quite a high noise level as displayed in Figure 1, which displays the velocity points together with the interpolated velocity surface.

In a 3D display the velocity outliers were easily identified and deleted prior to velocity interpolation. The velocity surfaces were smoothed to address remaining anomalies. Typically the challenge is to find a criterion for optimum smoothing. For this area the interval velocity surfaces showed a high correlation with the time interpretation. As documented in Figure 2 the smoothing of the velocities had an impact on the correlation coefficient. It was assumed that the optimum filter length for the velocity smoothing could be

identified by the largest correlation coefficient.* PETRONAS Carigali Sdn Bhd,Malaysia** SCHLUMBERGER IS, Stavanger, Norway

After the smoothing the velocity data was adjusted to the well velocities.

TIME DEPTH CONVERSION

A number of different work flows can be used for the time depth conversion of the 2D seismic horizons. Figures 3 and 4 present two approaches: the first work flow converts the interpolated time surfaces into the depth domain with the help of interpolated 2D seismic velocities. The second work flow depth converts the 2D interpretation and interpolates the depth data. Along the 2D lines the two work flows give similar results. However between the 2D lines the depth surfaces can deviate by more than 100m (Figure 5). Obviously the proper work flow is of importance to reduce uncertainty of the depth calculations. The first work flow includes two interpolation processes, e.g. interpolation of the 2D interpretation and of the 2D velocities. The second work flow interpolates only the depth converted 2D interpretation. Generally the interpolation of data points is subject to uncertainty. Consequently the second work flow should be used because it includes only one interpolation.

For the lower horizon it is worthwhile comparing the depth conversion based on interval velocities with the conversion based on average velocities. Typically depth conversion using interval velocities is based on a layer cake model. This means that the time thickness of each layer is converted to depth and added to the depth surface of the layer top. Comparing the results of both approaches shows quite a large difference between the two depth surfaces. Typically the disadvantage of the layer cake approach is that the deeper surfaces are influenced by the time uncertainty of the upper layers that is a direct consequence of the layer cake approach. On the other hand this approach allows optimizing the smoothing of the velocity field: typically the noise in the seismic velocities increases with depth and consequently the interval velocities of the lower layers need a stronger smoothing operator compared to the upper layers.

VELOCITY AND DEPTH UNCERTAINTY CONSIDERATION

There are two major reasons for velocity uncertainty: noise and anisotropy. Stacking velocities are sensitive to horizontal velocities

which may deviate by several percent from vertical velocities. Typically the anisotropy factor is estimated from the well data. However in this case only one well is available. Consequently, lateral anisotropy changes and the depth error caused by anisotropy effects could not be estimated. In order to estimate the influence of the seismic noise on depth conversion error the velocity difference at the crossing 2D lines were measured. The standard deviation of this velocity error distribution was converted to a depth error standard deviation as shown in Figure 6. This work flow is only valid for the first layer or for depth conversion based on average velocities. For the velocity model with two layers the depth error standard deviation was estimated for each layer; then their variance was calculated and summed, resulting in the depth error variance of the bottom surface (depth error propagation). Note that the depth error standard deviation derived from a constant velocity error distribution is a function of time.

The standard deviation error maps were calculated for two depth surfaces of the lower horizon, one based on average velocities and one based on the layer cake approach (interval velocities). Figure 7 shows the two depth maps color coded with the error standard deviation. The histograms give the distribution of the depth error standard deviation. The depth conversion based on the layer cake approach shows a larger spread in the error standard deviation. However the mean error standard deviation is much smaller (250m) compared to the mean error standard deviation of the depth surface based on the average velocities (400m).

CONCLUSION

Careful editing and smoothing of 2D seismic velocities as well as a careful selection of the work flow for depth converting the seismic 2D interpretation can increase the reliability of the depth conversion considerably. The estimation of the depth error is a challenge in case of sparce well data. However 2D seismic velocities offer the possibility to estimate the velocity uncertainty thru comparing the velocities of crossing 2D lines. This uncertainty can be converted into depth uncertainty.

Figure 1 - 2D seismic velocity points and interpolated velocity surface for the top horizon.

Figure 2 - Work Flow I for depth converting 2D seismic interpretation. The work flow includes two interpolations: interpolation of the velocity points and interpolation of the time interpretation.

Figure 3 - Work Flow II for depth converting 2D seismic interpretation. This work flow only requires one interpolation: interpolation of the depth points. The velocity points are interpolated only to allow a smoothing of the velocities. The depth conversion of the time interpretation uses the velocities along the 2D lines and consequently the velocity interpolation has no influence on the depth conversion.

Figure 4 - Difference between the depth surfaces of the two work flows shown in figure 2 and 3.

Figure 5 - Depth difference of lower horizon converted with interval velocities and lower horizon converted with average velocities.

Figure 6 - Calculation of the depth error standard deviation map.

Figure 7 - Distribution of the depth error standard deviation. The depth surfaces of the lower horizons are shown based on the layer cake approach (left) and the average velocities (right). Color-coded is the time varying standard deviation of the depth error. The histograms show the distribution of the time dependant depth error standard deviation.