Velocity modeling uncertainty; a challenge for regional

depth conversion in Tarakan Basin

1, a) 2, b)
Aan Hartono and Pintarwan Sasi Kirana
1
Medco E&P Indonesia
2
MedcoE&P Indonesia

a)
aan.harton [email protected]
b)
[email protected]

Abstract. Velocity modeling was conducted to provide the depth structural model which will be applied for further
Tarakan Basin characterization study. Uncertainty in velocity modeling was due to the limited data used in addition b
ecause of doubt about which the most suitable method for basin-scale of depth conversion was. Four velocity models with a
layer cake approach were generated for the following methods: interval velocity, V0+kZ, V0+k(Z-Z0), and 3D average
velocity property. The uncertainty should be reduced by choosing appropriate workaround and the smallest residual depth error
of four methods.

Layer cake velocity modeling was built by time depth relationship obtained from forty well -seismic ties and eight-time
structure maps correlating to actual well tops. The first method calculated interval velocity at the top of zones, where the
velocity i s constant through each zone. The second method defined vertical velocity changes by a factor of k, whereas V0
and Z represent the velocity at datum and the distance of the point observed from datum. The third method is similar to the
second method, but V0 was obtained at the top of the zone and (Z-Z0) represents the distance between point observed and the
top of the zone. The last method was generated from 3D average velocity property produced by up-scaled and statistically
populated the velocity of the key wells.

The residual depth error was calculated by differentiating actual well tops to estimated well tops in the blind wells. On
comparing the four methods it was observed that there was less residual depth error at 3D average velocity property method.
Based on th is result, regional depth conversion of Tarakan Basin was applied using this kind of method.

I. INTRODUCTION
The Tarakan Basin is situated at northeastern margin of Borneo Island and is located geographically in North
Kalimantan province, Indonesia. Tarakan Basin is one of prolific basins in Indonesia. It has been proved by many
discovery wells penetrated hydrocarbon zones within sedimentary sequence of the basin. Most of productive zones
were coming from Miocene to Pliocene of deltaic series.

Regionally as shown in Figure 1, The basin is bounded to the north by Mesozoic and older rocks of the Sampurna
High, to the west by strongly folded Mesozoic to Eocene melange of the Kucing High, to the south by
the Mangkalihat High separating the Tarakan Basin in the north from the Kutai Basin in the south, and
eastward the basin extends to unknown limit out across the continental shelf of deep Sulawesi Sea (Pertamina
BPPKA, 1995). Generally, Tarakan Basin is divided into 4 sub-basins, those are; Muara Sub-Basin, Berau Sub-
Basin, Tarakan Sub- Basin, and Tidung Sub-Basin. Two main structural trends are apparent in the Tarakan Basin,
NW to SE and NNE to SSW. These trends were initiated in the Eocene rifting, and were periodically reactivated
during generally compressive phase from the Middle Miocene to present.
 FIGURE 1. Location and regional geology of Tarakan Basin (Modified after Pertamina BPPKA, 1995).

Tarakan Basin is believed to still have plentiful hydrocarbon potential. Therefore the exploration activity is still
ongoing to find new resources. Medco E&P Indonesia as one of the operators in this area, Tarakan Sub-basin
particularly, has been conducting regional studies in order to minimize the further exploration drilling risk where
velocity modeling for regional depth conversion as paper topic is part of the studies.

Velocity modeling with layer cake approach aims to convert time structural model into depth structural model to
be used for the next stage of studies. To get a proper depth conversion, there are uncertainties in velocity
modeling related with limited data compared with coverage area, the vertical velocity variation and quality data in
study area. This paper attempts to discuss how the uncertainties should be diminished by using proper workaround
and selecting the most suitable method applied for basin-scale of depth conversion.

II. DATA AND METHOD

Layer cake approach of velocity modeling was made by integrating well data and seismic data. Well data used in
the study area consisted of 40 exploration wells which have logs and velocity, mud logs and biostratigraphy data,
either from the well itself or from previous regional studies. The wells were correlated by their similarity of age to
obtain eight following markers: Late Eocene 2 (LE-2), Middle Miocene 4 (MM-4), Late Miocene 6 (LM-6), Late
Miocene
9 (LM-9), Late Miocene 11 (LM-11), Early Pliocene 0 (EP-0), Early Pliocene 1 (EP-1) and Early Pliocene 3 (EP-3).
Location of reference wells and illustration of the well markers are shown by Figure 2a. The chosen markers
above are utilized for well-seismic tie process and are applied to the reference wells. The proper well-seismic tie is
very important to integrate vertical information in the wells into lateral information on seismic data. Well markers
tied to seismic markers and started seismic interpretation to produce eight time maps in order of age as shown
Figure 2b.
 (a) (b)
FIGURE 2. (a). Forty reference wells are used in the study with 4 blind wells highlighted by red and at the right shows
example of identified marker from B-1 well. (b). Eight regional time maps for surface input in velocity modeling with layer
cake approach.

One of uncertainties in the basin-scale of velocity modeling, especially in Tarakan Sub-basin, is the
vertical variation obtained from time depth function of wells. It requires selection of appropriate method to address
the challenge. Time depth relationship derived from forty well-seismic ties and eight time maps correlating to
actual well tops are used for input in velocity modeling with layer cake approach. In general, there are a lot of
methods that can be used, but for this paper will be discussing for the following methods: V0=Interval velocity,
V0+kZ, V0+k(Z- Z0), and 3D average velocity property.

The first method is defined by the following

formula:

V  V 0  V int .

Interval velocity of each marker in this method will be equal to V0 derived from well tops as shown at Figure
3. Layer cake approach is made by sorting time maps from the oldest (LE-2) to the youngest (EP-3) correcting
well tops as illustrated in Figure 4. Each interval layer or marker refers to zone so it is obtained seven zones
which will be calculated. Every zone has different velocity and constant value within the zones. After well tops
have calculated velocity respectively, then gridded for getting eight velocity map of each layer.

FIGURE 3. Simple illustration showing the velocities are calculated from identified well tops.
 FIGURE 4. Layer cake approach applied for the first method where the velocities are derived from correction (well
tops).

The second method is described with the following formula:

V  V 0  kZ

The formula is popularized by Marsden (1995). He stated that the velocity increases linearly with z with constant
V0 and k values as acceleration or compaction factor. Velocity and other parameters are calculated from check-
shot that have been modified in the well-seismic tie process. Unlike the first method calculated from well tops, V0
obtained from velocity at Z = 0 or datum so that the value V0 for each marker in the wells will be equal. Z
represents the distance of the point observed from datum and k is constant change of vertical velocity or
compaction factor (Figure
5 and 6).

FIGURE 5. Simple illustration showing the velocities are calculated from time-depth function of wells. V0 of each
marker’s obtained at datum, Z represents the distance of the point observed from datum and k reflects compaction factor.
 FIGURE 6. Layer cake approach applied for the second method where the velocities are calculated from time-depth
function.

The third method has similar formula with the second method, namely:

V  V 0  k (Z  Z
0)

The method has similarity with the second method in using the velocity wells. The difference is when calculating
the value of V0. V0 is obtained at the top of the zone so that it will have different V0 for each well marker
whereas (Z- Z0) represents the distance between the point observed and the top of zones (Figure 7 and 8).

FIGURE 7. Simple illustration showing the velocities are calculated from time-depth function of wells. V0 of each
marker’s obtained at top of markers, Z-Z0 represents the distance of the point observed to the top of markers and k reflects
compaction factor.

FIGURE 8. Layer cake approach applied for the third method where the velocities are calculated from time-depth function.
The last method utilized well data, time and composite surfaces time-depth function to create a simple grid
from time produced surfaces (Ismail et al, 2011). 3D property average velocity was produced by up-scaled
and statistically populated using moving average of time depth relationship (TDR) function of the key
wells. Furthermore, the value of average velocity is extracted from 3D property result to be set with layer cake
approach of velocity modeling as shown in Figure 9 and 10.

FIGURE 9. Simplified work flow of 3D model of average velocity

property.

FIGURE 10. Layer cake approach applied for the fourth method where the velocities are extracted from 3D model of
average velocity property.

Of the four methods will be selected the most appropriate method to be applied at the basin-scale depth
conversion in the study area. Selection is made by observing the smallest of residual depth error in the blind wells
meaning the wells were scripted no correcting to the actual depth upon depth conversion process. The errors are
calculated by
subtracting estimated depth to actual depth. The chosen method is expected to be one of ways to minimize the
uncertainties appearing in velocity modeling.

III. RESULT AND DISCUSSION

By comparison, the last method shows smallest residual depth error, as shown on Table 1. Average velocities
input at fourth method derived from modified check-shots in the well-seismic tie process. Other methods
generally provide larger residual depth error. This thing is possibly caused by the distance between wells
having velocity value and blind wells location is far apart. The addition of pseudo wells with certain values near to
the blind wells is predicted to reduce the error. In addition, it could be due to lack of lateral control at each zone,
unlike the surface layering process in the latter method.

TABLE 1. Residual depth error comparison of four methods at blind wells shows the smallest error in the last
method.

Residual depth Residual depth Residual depth Residual depth

Well Name Surface/Markers
error 1 (feet) error 2 (feet) error 3 (feet) error 4 (feet)

Blind Well-1 EP-0 -54 12 24 137

Blind Well-2 EP-0 438 221 15 31

Blind Well-2 EP-1 -925 501 -170 47

Blind Well-2 EP-3 608 561 561 72

Blind Well-1 LM-11 3133 -403 -369 117

Blind Well-3 LM-6 418 -213 27 61

Blind Well-4 LM-6 2383 3126 2592 -22

Blind Well-3 LM-9 646 282 331 100

Blind Well-1 LM-9 3413 928 951 155

Blind Well-4 LM-9 -41 1457 1113 -181

Blind Well-3 MM-4 -506 -1339 -1097 -218

Average absolute 1142 822 659 104

Comparison of velocity map of four methods at MM-4 shows different interpolation and extrapolation result
for each method (Figure 11). Velocity map of MM4 in the first and fourth methods look smooth er with
velocity variations are not as drastic as second and third methods, especially in locations away from the wells.
From several experiments, the use of gridding methods in velocity modeling formula also will result different
velocity maps. This suggests that there is another uncertainty arising during building the velocity modeling process
itself.
FIGURE 11. Comparison of velocity maps for top of MM-4 shows there is significant lateral velocity variation, first and
fourth methods seem to be smoother compared with others. (a). First method, (b).Second method, (c).Third method, (d). Fourth
method

Other challenges occur when building layer cake approach of velocity modeling like creating pseudo time maps
for shallow layer being outcropped or eroded to avoid error process. Pseudo surface is made by combining
with erosional surface so the entire areas will be covered by the data. Other related challenge is least wells
penetrating reference deep markers so it has no correcting to the actual depth in depth conversion process.
This sometimes causes the thickness in deeper becoming uncontrolled. The solution is by creating pseudo well
tops derived from the formula of check-shots. The last challenge is presence of fault shadows or artifacts in the
depth maps due to big faults in the shallower layers. Keho and Samsu (2002) brought the solution for the similar
problem by producing average velocity by dividing the depth map by time maps possessing fault shadow problem.
The average velocity maps should be smoothed to remove the artifacts, and then multiplying to the time maps to
compute the final depth maps.

IV. CONCLUSION
Based on explanation and discussion above, it can be concluded as
follows:

1. Layer cake approach of velocity modeling was built by integrating eight time maps and forty time depth
functions of well data.

2. Comparison of the four methods of velocity modeling shows the method using 3D model of average velocity
property has smallest value of residual depth error. Therefore the last method is the most appropriate for basin-
scale depth conversion in study area.
3. Need proper workaround to build layer cake approach of velocity modeling such as creating pseudo surfaces for
eroded areas, making pseudo well tops to maintain thicknesses and smoothing average velocity maps to eliminate
fault shadows in depth maps.

4. It was suggested to create pseudo well in between well reference to get lesser residual depth
errors.

ACKNOWLEDGMENTS
The authors wish to thank the management of Medco EP Indonesia and SKKMigas for permission to publish the
paper; to exploration east region team of Medco E&P Indonesia in particular Mr. Danang Priwastono as
the manager, Aswin Mustofa, Dedi Djunaedi, Rana Lesmana and Sri Mulyani for their support and discussion.

REFERENCES
1. D. Marsden, “V0-K Method of Depth Conversion,” in the Leading Edge of Exploration,
1992.
2. H. B. Ismail, M. I. Kassim, K. H. Khalid, S. Nurida, W. B. Nasifi, H.I. Darmawan, R. K. Pratama, M.
A.
th
Ismail, N. A. Abdullah, I. E. Amorita, “Conventional Velocity Model Seems To Be The Best,” in 35
Indonesian Petroleum Association Proceeding, 2011.
3. Pertamina BPPKA, “Volume 5 Tarakan Basin Northeast Kalimantan,” in Petroleum Geology of
Indonesia
Basins, 1995.
4. T. Keho and D. Samsu, “Depth Conversion of Tangguh Gas Fields,” in the Leading Edge,
2002.