10
1 Statoil ASA, Stavanger, Norway. E-mail: [email protected]. Abstract Oil seeps, shallow gas, and surface features such as seabed pockmarks and mud volcanoes are historically be- lieved to be signs of deeper hydrocarbon accumulations. In the search for connections between shallow features and deeper hydrocarbon accumulations, gas chimneys and faults have been studied as possible routes for vertical mi- gration of gas and fluids from source rocks and hydrocar- bon-charged traps. Understanding these fluid migration pathways can help evaluate whether a trap is charged or has leaked. A method based on seismic attributes and use of neural networks has been developed to detect and display gas chimneys. This method makes it possible to detect and map gas chimneys in a consistent manner and to see the po- sition of chimneys relative to faults and traps. The detection of gas chimneys in seismic data has therefore been used as a tool in an effort to distinguish between hydrocarbon- charged traps and dry traps with associated chimneys. Based on such case studies, a model of trap classification has been proposed and tested on more than 100 drilled traps in the Norwegian North Sea with good results. Introduction To identify connections between the seabed or shal- low hydrocarbon indicators and oil and gas reservoirs or source rocks, seismic data have been used to infer migra- tion pathways (Heggland, 1997, 1998). Seabed features such as pockmarks and carbonate formations (Hovland and Judd, 1989; Roberts and Aharon, 1994; Hovland et al., 2010) are fluid escape features that may indicate deeper hydrocarbon accumulations. Buried features may evidence once-active hydrocarbon seepage. However, seabed and shallow features do not show the vertical fluid migration pathways that connect to a leaking trap or source. Gas chimneys and faults may provide connections to traps that can contain hydrocarbons as well as connec- tions to hydrocarbon sources. Method Gas chimneys are visible on seismic data as vertically oriented zones of low reflectivity and distorted reflector continuity. Gas chimneys are not clearly visible on time slices from 3D volumes nor on attribute maps. Therefore, a method was proposed to better detect chimneys by combin- ing seismic attributes in a manner that could distinguish gas chimneys from their surroundings (Heggland et al., 1999; Meldahl et al., 1999). Because of the difference in the seis- mic character inside and outside of a chimney, attributes such as amplitude, energy, trace correlations, and time-dip variance are well suited as input for chimney detection. A neural network (Steegs, 1997) was applied to the in- put seismic attributes and was trained on chimney and non- chimney locations. Based on this training, the network can make a classification of other samples in a seismic volume into chimney or nonchimney class by assigning a high probability to chimney samples and a low probability to nonchimney samples. Because noise and other features in the seismic data show a seismic character similar to a gas chimney (i.e., similar Chapter 14 Hydrocarbon Trap Classification Based on Associated Gas Chimneys Roar Heggland 1 221

Hydrocarbon Trap Classification Based on Associated … from 3D volumes nor on attribute maps. ... a simple model classifies traps using associated chimneys ... chimney and not the

  • Upload
    vuhuong

  • View
    225

  • Download
    1

Embed Size (px)

Citation preview

Page 1: Hydrocarbon Trap Classification Based on Associated … from 3D volumes nor on attribute maps. ... a simple model classifies traps using associated chimneys ... chimney and not the

1Statoil ASA, Stavanger, Norway. E-mail: [email protected].

AbstractOil seeps, shallow gas, and surface features such as

seabed pockmarks and mud volcanoes are historically be-lieved to be signs of deeper hydrocarbon accumulations. In the search for connections between shallow features and deeper hydrocarbon accumulations, gas chimneys and faults have been studied as possible routes for vertical mi-gration of gas and fluids from source rocks and hydrocar-bon-charged traps. Understanding these fluid migration pathways can help evaluate whether a trap is charged or has leaked. A method based on seismic attributes and use of neural networks has been developed to detect and display gas chimneys. This method makes it possible to detect and map gas chimneys in a consistent manner and to see the po-sition of chimneys relative to faults and traps. The detection of gas chimneys in seismic data has therefore been used as a tool in an effort to distinguish between hydrocarbon-charged traps and dry traps with associated chimneys. Based on such case studies, a model of trap classification has been proposed and tested on more than 100 drilled traps in the Norwegian North Sea with good results.

IntroductionTo identify connections between the seabed or shal-

low hydrocarbon indicators and oil and gas reservoirs or source rocks, seismic data have been used to infer migra-tion pathways (Heggland, 1997, 1998). Seabed features such as pockmarks and carbonate formations (Hovland and Judd, 1989; Roberts and Aharon, 1994; Hovland

et al., 2010) are fluid escape features that may indicate deeper hydrocarbon accumulations. Buried features may evidence once-active hydrocarbon seepage. However, seabed and shallow features do not show the vertical fluid migration pathways that connect to a leaking trap or source. Gas chimneys and faults may provide connections to traps that can contain hydrocarbons as well as connec-tions to hydrocarbon sources.

MethodGas chimneys are visible on seismic data as vertically

oriented zones of low reflectivity and distorted reflector continuity. Gas chimneys are not clearly visible on time slices from 3D volumes nor on attribute maps. Therefore, a method was proposed to better detect chimneys by combin-ing seismic attributes in a manner that could distinguish gas chimneys from their surroundings (Heggland et al., 1999; Meldahl et al., 1999). Because of the difference in the seis-mic character inside and outside of a chimney, attributes such as amplitude, energy, trace correlations, and time-dip variance are well suited as input for chimney detection.

A neural network (Steegs, 1997) was applied to the in-put seismic attributes and was trained on chimney and non-chimney locations. Based on this training, the network can make a classification of other samples in a seismic volume into chimney or nonchimney class by assigning a high probability to chimney samples and a low probability to nonchimney samples.

Because noise and other features in the seismic data show a seismic character similar to a gas chimney (i.e., similar

Chapter 14

Hydrocarbon Trap Classification Based on Associated Gas Chimneys

Roar Heggland1

221

Page 2: Hydrocarbon Trap Classification Based on Associated … from 3D volumes nor on attribute maps. ... a simple model classifies traps using associated chimneys ... chimney and not the

attribute values), the vertical extent of chimneys is extracted in order to discriminate between gas chimneys and features that have little vertical extent. This extraction is addressed by using multiple detection windows along the time axis. Still, features such as faults, channel edges, diapirs, and gas associ-ated with mud-volcano feeders will be detected because their seismic character is similar to chimneys and because they have large vertical extents. However, most of these features can be distinguished from chimneys by using other criteria, such as shape and appearance at different stratigraphic levels. A chimney can, for example, be present below a mud volcano as a result of gas invading the surrounding formation during eruption. Some mud volcanoes exhibit mud flows at different stratigraphic levels, which makes them easier to identify and hence distinguish the associated chimney from chimneys not associated with mud volcanoes. The final neural network out-put, called a chimney cube, can be regarded as a chimney probability volume.

Applications of the chimney-detection method have led to regional mapping of gas chimneys (Heggland et al., 2000; Meldahl et al., 2001), allowing comparisons of chim-neys and chimney distributions from different areas. The first chimney cube was created in 1998 (see Figure 1). Since then, chimney detection has been used to infer remigration routes from traps — useful information in predicting hydro-carbon-charged traps and dry traps.

ModelSome gas chimneys coincide with faults as well as with

fluid escape features, shallow gas accumulations, hydrocar-bon-charged reservoirs, and source rocks. The presence of

chimneys along a fault is interpreted to indicate that the fault is, or has been, open for vertical fluid migration. Chim-neys associated with faults have been referred to as type 1 chimneys (Heggland, 2005). Laterally extensive gas chim-neys that are not associated with faults are observed above hydrocarbon-charged structures as well as over deep parts of basins where source rocks are known to be present; such chimneys are referred to as type 2 chimneys.

Based on comparisons between dry and charged traps, a simple model classifies traps using associated chimneys (Heggland, 2005). The model, based on case studies from the Norwegian shelf, the Nigerian continental slope, the Gulf of Mexico, and the Caspian Sea, is comprised of three scenarios: class A, class B, and class C traps (Figure 2).

A class A trap has a leaking fault at the crest of the trap, indicated by one or more type 1 chimneys present at the fault, implying a risk that the trap has leaked and contains

Standard 3D Cube

Seismic Attributes

Neural network

Classification

Chimney Cube

Chimney

1 km

Figure 1. A 3D seismic volume (top) before and (bottom) after detecting gas chimneys.

Class A

Type 1 chimney

Leaked trap

Class B

Type 1 chimneys

Leakage from trap (left) orCharging of trap (right)

Class C

Type 2 chimney

Hydrocarbons in trap and seal

Figure 2. Trap classification model based on associated gas chimneys for classes A, B, and C.

222 Hydrocarbon Seepage: From Source to Surface

Page 3: Hydrocarbon Trap Classification Based on Associated … from 3D volumes nor on attribute maps. ... a simple model classifies traps using associated chimneys ... chimney and not the

only minor volumes of hydrocarbons. A small column can only be maintained by charge from a deeper source; if this is the case, hydrocarbons will continue to escape the struc-ture (i.e., remigration).

A class B trap leaks hydrocarbons vertically through a fault at the flank of the trap. Chimneys indicate leakage from the trap through the fault, but leaving a larger hydro-carbon column than in the class A case. Faults may act as trap leakage or charge routes and show associated chim-neys above or below class B traps, respectively. The latter case indicates fluid migration between a source rock and a reservoir (i.e., migration).

A class C trap has hydrocarbons in the trap as well as in the seal. In this class, a type 2 gas chimney is located within the map outline of the trap and extends from the top of the trap into the overburden. It can be widespread, covering part of or the whole surface of the trap. This kind of chimney is not associated with faulting of the cap rock. Similar models are presented in Connolly et al. (see Chapter 7, this volume).

Criteria for chimney recognitionNoise in seismic data can be created by features in the

overburden and can be misinterpreted as gas chimneys. Similarly, gas chimneys can be misinterpreted as noise cre-ated by shallower features. Some examples follow.

Shallow high-amplitude anomalies

Shallow high-amplitude anomalies or features exhibit-ing complex reflection patterns can create distortions below

them that look like gas chimneys. In many cases, high-am-plitude anomalies that are interpreted, or confirmed by drilling, as shallow gas accumulations are present above gas chimneys. These may have been charged through verti-cal migration pathways, as indicated by the gas chimneys.

Chimney detection creates an output in accordance with what the neural network has been trained to recognize. In the example in Figure 3a, the neural network is trained to detect features interpreted to be gas chimneys. The vertical noise on the right side of Figure 3a seems to be caused by the shallow high-amplitude anomalies and has not been se-lected for training the neural network for chimney detec-tion. As a result, the detection has picked up the interpreted chimney and not the noise below the shallow high-ampli-tude anomalies, as seen in Figure 3b.

Sand injections

Over large areas in the North Sea, V-shaped high-am-plitude anomalies, interpreted as sand injections (Cart-wright et al., 2007), are present. Below these anomalies are commonly zones of vertically extended noise that resemble gas chimneys (Figure 4). In Figure 5, these V-shaped high-amplitude anomalies are highlighted in red. The underlying noise has been detected as gas chimneys, highlighted in yel-low. At the base, a time slice from a trace correlation vol-ume is in blue and black to show how deeper faults at the Jurassic level compare with the locations of the V-shaped high-amplitude anomalies and the possible gas chimneys.

As can be seen in Figure 5, the amplitude anomalies (red) and the underlying noise (yellow) do not fully over-lap. This comparison indicates that the vertical noise is not

Amplitude anomalies

Noise

Gas chimney1

2

TWT s

2 km

b)a)

Min

Max

2 km

Gas chimney

BCU BCU

Figure 3. (a) Seismic section showing a gas chimney versus noise created by shallow amplitude anomalies. (b) A result of chim-ney detection.

Chapter 14: Hydrocarbon Trap Classification Based on Associated Gas Chimneys 223

Page 4: Hydrocarbon Trap Classification Based on Associated … from 3D volumes nor on attribute maps. ... a simple model classifies traps using associated chimneys ... chimney and not the

created by the V-shaped high-amplitude anomalies but like-ly represents gas chimneys. Both the anomalies and the gas chimneys are located above faults. The noise, or the gas chimneys, can be associated with the deeper faults, so they are interpreted to indicate continued vertical fl uid migration pathways. The faults have also probably infl uenced the in-jections of sand into shallower levels because they too are located almost directly above the faults. The injected sands probably contain gas because they exhibit high amplitudes and are located above the gas chimneys.

Pockmarks and associated chimneys

Gas chimneys are sometimes observed to be present below seabed pockmarks (e.g., Heggland, 1998). In an ex-ample from offshore Nigeria (Figure 6), a cluster of pock-marks can be seen on a seabed azimuth map derived from a 3D seismic volume. Vertically extended distortions are present below the pockmarks in the 3D volume (Figure 7), and the deep parts of the distortions are suspected to be arti-facts. To fi nd out, angle stacks with increasing offset be-tween the seismic source and the receiver are displayed to identify any changes in appearance with offset.

Sand injections

2 kmChimneys

2

1

0TWT s

Min

Max

Figure 4. Seismic section showing amplitude anomalies inter-preted as sand injections. Noise below the anomalies is inter-preted to be gas chimneys, based on the display in Figure 5.

Figure 5. A trace correlation time slice at 2344 ms TWT (blue) is used to display faults at the Jurassic level. Amplitude anomalies interpreted as sand injections are displayed in red. Noise below the anomalies is detected as gas chimneys, dis-played in yellow. The red and yellow spots do not fully over-lap, so the noise is interpreted as gas chimneys.

Figure 6. Seabed azimuth map derived from 3D seismic data, Nigerian continental slope, showing the presence of pock-marks.

2

TWT sPockmarks

Seismic distortions

3

1 kmMin

Max

Figure 7. Seismic section showing columnar distortions below pockmarks.

224 Hydrocarbon Seepage: From Source to Surface

Page 5: Hydrocarbon Trap Classification Based on Associated … from 3D volumes nor on attribute maps. ... a simple model classifies traps using associated chimneys ... chimney and not the

The result is shown in Figure 8. The deepest parts of the distortions look cone shaped; these features widen with offset, indicating they are artifacts. In addition, the reflec-tion sequences on the inner side of the cone-shaped distor-tions on the far offsets are preserved; they correlate with the reflection sequences on the outside and away from the cone shapes. The faults in the deep part of the seismic section, however, stay fixed in their respective positions throughout all offsets, whereas some of the deepest distortions move across faults when the offset increases. These comparisons suggest that only the upper part (e.g., 500 ms or less below seabed) of the distortions below the pockmarks can be

interpreted to be gas chimneys, creating artifacts farther down in the seismic section, which can be identified using angle stacks. In Løseth et al. (2011), these features are de-scribed as 1000-m-long blowout pipes. Based on the angle stacks, it seems that these features are less than half that length (i.e., down to only 2 s two-way time [TWT] in Fig-ure 8). This difference is significant because they are then not connected to the interpreted deep marine reservoir sands (at ~3 s TWT) described by Løseth et al. (2011).

Mud volcanoes

Vertically extended chimneys can also be seen on seis-mic sections below mud volcanoes. These chimneys may be detected in chimney cubes and are believed to be related to gas emitted during eruptions, in which case some of the gas invades the surrounding formation and is captured within the shales. This process can explain why the gas is visible on the seismic data as a chimney.

Figures 9–11 show a mud volcano on the Nigerian continental slope (Heggland et al., 1996; Heggland and Nygaard, 1998). Figure 9 illustrates a seismic section from 3D seismic data. A mud volcano and a wide zone of disturbance can be seen below the mud volcano. This disturbance may result from acoustic masking caused by the high amplitudes at the mud volcano; alternatively, it may be due to the presence of gas in the underlying sediments. In Figure 10, the mud volcano and two gen-erations of mud flows are made visible by an average ab-solute amplitude display, created by using a time interval parallel to and positioned below the seabed reflector. The time window used starts below the (black) seabed reflector and is large enough to capture the two mud-flow

TWT sa)

b)

c)

d)

2

3

2

3

2

3

2

3

1 kmMin

Max

Figure 8. Angle stacks of the section in Figure 7, showing that the distortions widen with increasing offsets and that the reflection sequences within the distorted zones are preserved in the far offset stacks, correlating with the reflection sequenc-es outside the distorted zones. Angle stacks are for (a) 2°–16°, (b) 14°–28°, (c) 26°–40°, and (d) 38°–52°.

Mud flow 1

1

TWT s

Mud flow 2

Mud volcano

1 km

Figure 9. Seismic section across a mud volcano at the Nige-rian continental slope. Two generations of mud flows are vis-ible as two reflectors terminating down the slope below the seabed reflector.

Chapter 14: Hydrocarbon Trap Classification Based on Associated Gas Chimneys 225

Page 6: Hydrocarbon Trap Classification Based on Associated … from 3D volumes nor on attribute maps. ... a simple model classifies traps using associated chimneys ... chimney and not the

reflectors seen in Figure 9. Figure 11 shows a side-scan sonar image of the same mud volcano. Two mud-expul-sion centers are clearly visible, indicating the mud feed-ers have a much smaller radius than their respective mud volcano and the seismic distortion below it. Seabed grav-ity cores taken at this mud volcano showed high content of oil and gas (Graue, 2000).

Mud volcanoes can bring large amounts of gas and sometimes oil to the surface during eruption, and they are known to seal themselves after eruption. Mud volcanoes are present in many areas where oil and gas discoveries have been made, e.g., in the south Caspian Sea (e.g., Azeri,

Chirag, Guneshli, and Shah Deniz oil and gas fields) and offshore Nigeria. An oil discovery was made close to the mud volcano shown in Figure 10. In Figures 12 and 13, oth-er mud volcanoes offshore Nigeria are indicated by arrows. These are located at the flank of a gas-filled trap (green structural closure in Figure 12).

Seismic section

Mud flow 2

Mud flow 1

1 km

Figure 10. Average absolute amplitude in a time interval par-allel to and positioned just below the seabed reflector, where high amplitudes are imaging the mud volcano in Figure 9 and its two mud flows.

Mudexpulsionfeatures

Samplelocations

500 m

Figure 11. Side-scan sonar image of the mud volcano in Figures 9 and 10. Two mud expulsion features can be seen. Gravity-core sample locations are noted.

Gas chimneys

Gas belowmud volvanoes

Leaked trap

Figure 12. A 3D visualization of a top reservoir horizon and de-tected gas chimneys (yellow) on the Nigerian continental slope.

Seabed

Leaking fault

Leaked trap

Gas chimneys

Mud volcanoes

Pockmarks

Figure 13. A 3D visualization of a top reservoir horizon and detected chimneys (yellow) on the Nigerian continental slope. A seabed azimuth map is displayed above, showing faults, mud volcanoes, and pockmarks.

226 Hydrocarbon Seepage: From Source to Surface

Page 7: Hydrocarbon Trap Classification Based on Associated … from 3D volumes nor on attribute maps. ... a simple model classifies traps using associated chimneys ... chimney and not the

Trap classificationThe following sections provide examples to illustrate

the three classes of traps.

Class A trap example

Figure 12 is an example from the Nigerian continental slope. As noted along the left side of Figure 12, detected chimneys are located at a fault oriented along the ridge of a structure. The fault extends to the seabed and can be identi-fied on the seabed azimuth map in Figure 13. A well drilled into this structure was dry, probably because hydrocarbons had leaked out through the fault. Many examples such as this one show chimneys located at faults at or close to the crest of traps that are dry or that may contain only small

amounts of hydrocarbons. In some class A traps, a small column may be maintained by persistent charge.

Figure 14 shows a case wherein a fault cuts through multiple stacked traps, causing partial leakage. A gas chim-ney indicates fluid migration through the fault from below the deepest trap to the uppermost trap. Small hydrocarbon columns may be present, as indicated by flat spots.

Class B trap example

Figures 15 and 16 show an example where subtle chim-neys are located at the flank of a Late Jurassic trap, the blue line in the figure being the base Cretaceous horizon. In Fig-ure 15a, the chimneys are located above faults present in the Late Jurassic at the flanks of the trap. Figure 15b shows the corresponding section with detected chimneys. In Fig-ure 16, the base Cretaceous horizon shows the structural closure of the trap; detected chimneys are displayed in white. The appearance of chimneys located at faults at the flank of the trap makes it a class B trap, according to the model. A gas discovery was made at the location shown by the red line in Figure 16.

Class C trap example

In class C traps, a trap is partially or fully covered by a gas chimney and no faults are associated with the chimney. Figure 17 displays an example of a class C trap. In the seis-mic section (Figure 17a), a widely extended gas chimney can be seen above a trap in the Jurassic, indicated by a base Cretaceous horizon shown as a blue line. A flat spot may

TWT s

2 Flat spots

Gas chimneyFault

Traps

1 km

3

Figure 14. Seismic section showing stacked traps with flat spots. A fault is cutting through the middle of the traps. A chimney that can be associated with the fault is visible be-tween the traps.

TWT s b)a) A B A B

Chimneys

Faults

2 km

Chimneys

2 km

Class Btrap

BCU BCU Max

Min

3

4

Figure 15. A seismic section showing a class B trap in the upper Jurassic, North Sea, where (a) chimneys and faults are present at the flanks and where (b) chimneys are detected in the corresponding section.

Chapter 14: Hydrocarbon Trap Classification Based on Associated Gas Chimneys 227

Page 8: Hydrocarbon Trap Classification Based on Associated … from 3D volumes nor on attribute maps. ... a simple model classifies traps using associated chimneys ... chimney and not the

indicate a fluid contact. In the 3D display (Figure 17b), the base Cretaceous horizon and detected chimneys (white) show that a large part of the trap is covered by the chimney. Figure 18 shows another class C trap example. Many known fields and discoveries in the North Sea are class C cases, e.g., Ekofisk, Valhall, and Tommeliten.

Results based on classification of drilled traps in the North Sea

A test of the trap classification has recently been made for more than 100 previously drilled traps in a region of the northern North Sea (well data taken from the Norwegian Petroleum Directorate’s Factpages). The actual number of discoveries in this region is high and represents 56% of all drilled traps regardless of chimney presence, bearing in mind that several well-known fields (Statfjord, Snorre, Vi-sund, Gullfaks, Troll) are present here. The following re-sults were obtained:

• The success rate for drilled traps with no chimneys is 46%.

• The success rate for drilled traps with chimneys is 78%.

• If the trap classification had been used, some of the leaked traps, class A, could have been avoided, increas-ing the success rate to above 90%.

• All drilled traps with associated chimneys (including dry traps) also contained a reservoir.

The recommendation is hence first to drill prospects with associated chimneys and then to drill traps without chimneys.

Discovery well

Seismic line

B

A

Figure 16. A 3D display of a base Cretaceous horizon, show-ing the structural closure of the class B trap in Figure 15, the location of a discovery well, and detected chimneys (white).

TWT s

2

Chimney

3

4

Max

Seism

ic lin

e

A

B

Min2 km

Flat spot

BCU

Aa) b)B

Figure 17. (a) A seismic section showing a class C trap in the Late Jurassic, North Sea, where a chimney covers a large area above the trap. A flat spot indicates a fluid contact. (b) A 3D display of a base Cretaceous horizon and detected chimneys (white), showing that a large part of the trap indicated in the seismic section is covered by a chimney.

228 Hydrocarbon Seepage: From Source to Surface

Page 9: Hydrocarbon Trap Classification Based on Associated … from 3D volumes nor on attribute maps. ... a simple model classifies traps using associated chimneys ... chimney and not the

ConclusionsCase studies show that it is possible to differentiate be-

tween hydrocarbon-charged traps and dry traps, based on the occurrences of associated gas chimneys (Heggland, 2005). A trap classification model that illustrates these dif-ferences has been proposed to predict hydrocarbon charge. Gas chimneys are believed to indicate fluid migration path-ways through faults, as for class A and class B traps. In some cases, chimneys may indicate sufficient seal capacity, as for class C traps. The trap classification, which is based on the presence of gas chimneys, has been tested using more than 100 previously drilled traps in a region of the North Sea, and it worked well.

In general, traps with associated gas chimneys seem to have a higher chance of containing hydrocarbons except for class A cases, which indicates traps have leaked. This obser-vation does not mean that traps without chimneys should be avoided; rather, traps with chimneys (classes B and C) are lower-risk traps that perhaps should be drilled first. Traps with associated gas chimneys also contained reservoirs, sug-gesting that chimney presence can be used as an indication of the presence of a reservoir. By using the trap classifica-tion model, the rate of discoveries can be increased.

AcknowledgementsStatoil ASA is acknowledged for giving permission

to publish this material. Many thanks go to my Statoil colleagues Nicholas Ashton, Benjamin Clements, Hallstein Lie, and Philip W. Mullis for supporting this work and of-fering valuable comments.

ReferencesCartwright, J., M. Huuse, and A. Aplin, 2007, Seal bypass

systems: AAPG Bulletin, 91, 1141–1166.Graue, K., 2000, Mud volcanoes in deep water Nigeria:

Marine and Petroleum Geology, 17, 959–974.Heggland, R., 1997, Detection of gas migration from a deep

source by the use of exploration 3D seismic data: Ma-rine Geology, 137, 41–47.

Heggland, R., 1998, Gas seepage as an indicator of deeper prospective reservoirs: A study based on exploration 3D seismic data: Marine and Petroleum Geology, 15, no. 1, 1–9.

Heggland, R., 2005, Using gas chimneys in seal integrity analysis: A discussion based on case histories, in P. Boult and J. Kaldi, eds., Evaluating fault and cap rock seals: AAPG, 237–245.

Heggland, R., P. Meldahl, A. H. Bril, and P. de Groot, 1999, The chimney cube, an example of semi-automated detection of seismic objects by directive attributes and neural networks — Part II: Interpretation: 69th Annu-al International Meeting, SEG, Expanded Abstracts, 935–937.

Heggland, R., P. Meldahl, P. de Groot, and F. Aminzadeh, 2000, Chimney cube unravels subsurface: The Ameri-can Oil & Gas Reporter: February, 78 –83.

Heggland, R., and E. Nygaard, 1998, Shale intrusions and associated surface expressions — Examples from Nigerian and Norwegian deepwater areas: Pro-ceedings of the Offshore Technology Conference, 1, 111–124.

Heggland, R., E. Nygaard, and J. W. Gallagher, 1996, Techniques and experiences using exploration 3D seis-mic data to map drilling hazards: Proceedings of the Offshore Technology Conference, 1, 119–127.

Hovland, M., R. Heggland, M. H. de Vries, and T. I. Tjelta, 2010, Unit-pockmarks and their potential significance for predicting fluid flow: Marine and Petroleum Geol-ogy, 27, 1190–1199.

Hovland, M., and A. G. Judd, 1989, Seabed pockmarks and seepages — Impact on geology, biology and the marine environment: Graham and Trotman.

Løseth, H., L. Wensaas, B. Arntsen, N. Hanken, C. Basire, and K. Graue, 2011, 1000 m long gas blow-out pipes: Marine and Petroleum Geology, 28, 1047–1060.

Meldahl, P., R. Heggland, A. H. Bril, and P. de Groot, 1999, The chimney cube, an example of semi-auto-mated detection of seismic objects by directive attri-butes and neural networks — Part I: Methodology: 69th Annual International Meeting, SEG, Expanded Abstracts, 931–934.

TWTs

1

2

Well

Chimney

1 km

Figure 18. Seismic section showing a broad gas chimney above the Tommeliten Gamma discovery, a class C trap.

Chapter 14: Hydrocarbon Trap Classification Based on Associated Gas Chimneys 229

Page 10: Hydrocarbon Trap Classification Based on Associated … from 3D volumes nor on attribute maps. ... a simple model classifies traps using associated chimneys ... chimney and not the

Meldahl, P., R. Heggland, A. H. Bril, and P. de Groot, 2001, Identifying faults and gas chimneys using multi attributes and neural networks: The Leading Edge, 20, 474–482.

Roberts, H. H., and P. Aharon, 1994, Hydrocarbon-derived carbonate buildups of the northern Gulf of Mexico con-

tinental slope: A review of submersible investigations: Geo-Marine Letters, 14, no. 2-3, 135–148.

Steegs, T. P. H., 1997, Local power spectra and seismic in-terpretation: Ph.D. dissertation, Delft University of Technology.

230 Hydrocarbon Seepage: From Source to Surface