Exploration Geophysics - Volume 26, Issue 2-3, 1995

A combined crosswell and VSP seismic reflection experiment was carried out in a metalliferous mine in an attempt to image in three-dimensions the stope tunnels and residual ore. The experiment provided initial field testing of a newly designed hardware - software system for high resolution in-mine seismic work.

High frequency (1-3 kHz) seismic signals were recorded on both hydrophones and triaxial geophones from detonator explosions. The major form of noise contamination was tube-waves. The data were migrated in 3-D by forming a partially coherent sum of the cross-products (semblance) between the components of the particle motion vector and the components of the wave vector, over all receivers and all sources, for every grid point in the investigated volume. Migration maps with the PP mode using the crosshole data were the most successful. The images correlated remarkably well with the known geology and mineworkings.

A genetic algorithm has been developed to tomographically construct the sub-surface velocity distribution using first-arrival seismic refraction data. The performance of the genetic algorithm is improved by the inclusion of a ‘zooming’ technique able to perform a staged subspace search in the higher dimensional solution space. This allows the complexity of the model defining the solution to be progressively increased during the inversion and results in an increase in the speed and accuracy with which a solution can be obtained.

The procedure has been tested with both synthetic and model data. The method proved able to correctly recover the geometry and velocity structure of a non-planar refractor and the overlying lower velocity layer without the need of any ‘a priori’ information.

In the petroleum and the mining industries, more detail about underground structures is often needed than can be determined from surface seismic surveys. Crosshole surveys and vertical seismic profiles are commonly used techniques for determining the velocity distribution between two wells by traveltime tomography. Seismic migration is also a powerful tool for imaging the structures between wells, but it requires velocity information. Neither seismic tomography nor migration alone can achieve an ideal underground image in both velocity and structure.

To get a high resolution image both in velocity and structure, we propose a two step inversion which is implemented by tomography and then pre-stack migration. The strategy uses the first arrival times to recover the low spatial frequency velocity components of the medium by tomography, then this velocity model is applied as a migration velocity to pre-stack migration to recover the higher spatial frequency components. As the result, high resolution images both of velocity and structure are obtained by this two step inversion process.

For the pre-stack migration, a general Kirchhoff integral is used, based on Huygens’ principle, which applies a Green’s function by traveltime and amplitude mapping, using the finite difference method. This method handles an arbitrary velocity model (including high velocity contrasts and shadow zones) represented by a grid of velocities. For the tomographic technique, either a constant velocity or a gradient velocity background can be used, depending on the particular case. This two step inversion offers a very useful tool for the petroleum and the mining industry for getting a high quality seismic image. A numerical example which uses a strongly contrasting velocity model for imaging a low velocity anomaly from crosshole survey data is given here using this two step approach.

Two stratigraphic bore-holes 335 m apart, drilled to the Bulli coal seam at a mine site in the Southern Sydney Basin, showed an average velocity discrepancy of 10 per cent throughout the entire lithological section. Subsequently, a comprehensive seismic experiment was performed by BHP Steel (AIS) Pty. Ltd. Collieries Division which included the surface seismic reflection, walk-away vertical seismic profile (VSP) and cross-hole recording over the area connecting the two bore-holes.

The surface seismic data indicated possible seam disturbances between the bore-holes. The VSP data were of variable quality, while cross-hole data were of the best quality. Strong transverse isotropy was measured through the sandstone unit. The top 100 m of the Bulgo sandstone showed weak P-wave anisotropy. However, shear wave splitting in the Bulgo sandstone suggested fracture induced anisotropy. Methane accumulations were inferred from the reflection and VSP data and Poisson’s ratio computed at the receiver borehole.

Analysis of these data revealed a high complexity of the zone between and away from the boreholes. The poor quality of reflection and VSP data resulted from the presence of faults, associated fractures and several methane accumulations in the Bulgo sandstone. The overall velocity difference between the boreholes is related to a change in stress field. The results of the experiment showed that where reflection seismic data was poor, the application of three component VSP and cross-hole seismic data can be used to define a structurally complex area which may contain faulting/fracturing and methane. In such areas special care has to be taken in data processing, analysis and time to depth conversion.

Conventional lithology interpretation from well log data becomes unreliable when the target bed is thinner than the vertical resolution limit of the logging tools. The thin bed log analysis method of Bateman (1990), using the concept of binary lithology and a filtering technique applied to conventional logging suites, is able to detect beds as thin as 15 cm. This paper will focus on the use of the Bateman method to improve log-derived lithology estimates in the Permian Patchawarra Formation sandstone gas reservoirs in the Toolachee Field of the southern Cooper Basin in South Australia. In three representative wells, conventional log analysis overlooks most of the thin sandstone beds and underestimates the gross sandstone by up to 36% when compared with core data. In contrast, the thin bed algorithm identifies the thin sandstone beds and gives a gross sandstone thickness which agrees with the core value to within about 5%. It therefore provides a more correct picture of reservoir rock distribution, and reveals more potential pay than conventional log analysis. The method also makes core to log depth matching easier and more accurate. Log character and flow test results imply the presence of relatively permeable gas-bearing zones within some of the thin sandstone beds, suggesting that overlooking thin bed reservoirs may significantly lower reserve estimates. Since the method does not require the use of any high resolution logging device, it is possible to re-evaluate old log data for which conventional log analysis is subject to such underestimation.

The economic constraints brought about by low oil prices make it necessary to maximise the structural and strati-graphic resolution of our seismic data while still being subject to these economic constraints. Thus, whenever any new technology or procedures become available, the cost impact of any parameter changes must be carefully considered against the potential technical benefits achieved.

This paper reviews some of the acquisition trials carried out during 1989 - 1994 and discusses the impact of those trials on the current acquisition parameters utilised in the Cooper and Eromanga Basins. The modification to parameters and procedures brought about by these trials has seen the productivity of seismic crews increase from 1.36 km/chg hr in 1988 to over 1.85 km/chg hr during 1992 - 1994. Thus, average monthly production in excess of 400 km/crew month in 1988 has increased to 550 - 600 km during the last few years, with the remarkable totals of 651 km and 665 km achieved in July 1993 and March 1995 respectively.

The main acquisition parameters analysed include group interval/receiver array length and source effort. The opportunity to use 60,000 lb vibrators is considered to have given a 10 -15 Hz increase in bandwidth at the high frequency end. This means that the previously perceived 65 Hz ceiling, which was considered to be caused by high transmission losses in the shallow section (Q = 25), is no longer restricting the bandwidth.

Further investigations into static control techniques have demonstrated that the refraction method has been unable to match the results gained from upholes, whereas subsequent upholes drilled to check residual static profiles generally add confidence to the use of residual statics tied to upholes.

The Tilbooroo seismic survey, recorded in early 1993, is the first 3-D survey acquired in the Otway Basin. It was preceded by an extensive study to optimise acquisition parameters in a cost-effective manner. This survey followed-up the minor oil recovery from fractured basement intersected by the exploration well Sawpit-1 and forms part of an integrated geological and geophysical analysis of the structural history of the Saw-pit area. Evaluation of this fracture play requires accurate prediction of the orientation and intensity of the basement fault and fracture pattern.

The 3-D data, sampled at a spacing of 15 m inline by 30 m crossline, are acquired over a 44 km² area. Land use is mainly pasture land, but also extends into the Coonawarra vineyards. There were numerous obstacles - swamps, fences, buildings, no-access areas, vineyards - that challenged Ingenuity in the planning, acquisition and processing stages of the survey. The 3-D seismic data are acquired at less than one quarter of the fold and twice the group interval of 2-D data, yet have a vastly superior signal-to-noise ratio. Confident interpretation of the basement fault pattern, combined with structural analysis of fault attributes and knowledge of the present day stress field, indicates the basement fracture play would be best evaluated by a deviated well with a southwest azimuth.

An ultrasonic pulse transmission method was used on a large block of gabbro (18×18×30cm), at low normal stresses (0.0115 MPa) and low frequencies (0.02-1 Mhz) in order to determine the transmission coefficient of a simulated fracture. The contact stiffness was measured directly and spectral ratios of the received signals calculated for comparison with predictions from the displacement discontinuity model. The agreement between the measured and derived spectral ratios was significantly better for the rough surface, indicating that the model provides a better approximation for natural fractures than for smooth interfaces. The model predictions also improved at lower frequencies, suggesting that spectral analysis of seismic data might yield useful information about the presence and stress state of fractures in the field.

The Pituri Prospect was a combined stratigraphic/structural trap on the Puffin Horst in the Vulcan Graben area of the Timor Sea. This paper describes a case study where the problem to be solved was one of mapping the lateral extent of a thin sand defined by well control. The sand is a 7m thick unit within the uppermost Puffin Formation claystones of Maastrichtian age. The interpreted environment of deposition is submarine which is biostratigraphically calibrated on a basin wide scale with water depths interpreted as Upper Bathyal to Outer Neritic. The fan units have been regionally mapped seismically to define the major feeder systems. The resulting regional sand distribution maps do not however provide a simple explanation for the thin sands which are seen in the Puffin-2, Grebe-1 and Prion-1 wells. Two models were proposed, a submarine fan unit at the very distal end of the Late Maastrichtian lowstand fan system or a shingled turbidite of the Highstand Systems Tract of the Prudhoe deltaics.

The sand distribution had to be defined by the seismic data with the calibration of the nearby Puffin and Pascal wells providing the possible northern limits of the sand unit. An innovative approach is called for when using seismic amplitude and attributes to try to describe facies distributions particularly when the units are at the limit of seismic resolution. The amplitude anomalies were analysed for tuning effects, inverted for acoustic impedance and mapped on timeslices datumed on the regional flooding surface at the base of the Palaeocene above the Puffin-2 sand. Modelling for seismic amplitude response was undertaken using nearby well data. The analyses carried out generated a model for the Puffin-2 sand where that unit is a crevasse splay/levee overbank unit which was associated with a major submarine channel system to the south of the well location. The channel system ended at a broad lobate feature which was interpreted to be a submarine fan lobe which was targeted by Pituri-1.

The well found no sand in the section interpreted to have the sand unit. What was present, was a 15 metre thick low density, low velocity claystone that is interpreted to be the thickening into the palaeo-low of a minor flooding surface seen at the Puffin-1 and Puffin-2 wells. This then highlights the difficulty in predicting sand distribution from seismic attributes alone. However the ability to predict reservoir distribution from seismic data is of critical importance in future exploration and further work which integrates all seismic attributes and geological/rock physics datasets could provide the means to better reservoir prediction.

In 1992, a 138 km² 3D seismic survey was recorded over the central eastern section of the Lake Hope Block within the South Australian PELs 5 and 6. The 3D data provided a marked increase in data quality and interpretability over the previously recorded 2D seismic data. The survey was recorded over the Sturt/Sturt East, Tantanna, Taloola and Malgoona Fields, and had specific exploration and development objectives.

The production history of the Cretaceous and Jurassic reservoirs of the Tantanna Field indicated that the recovery factors for some wells were anomalously high. Potential was recognised both in the Tantanna and Taloola Fields for pool extensions and unswept reserves behind some form of permeability barriers. The 3D dataset has allowed a complex fault pattern to be mapped over these fields. In the Sturt/Sturt East Fields, the objectives were to better define the strati-graphic edges of the Jurassic Poolowanna and Birkhead Formation reservoirs. In the Malgoona Field, the 3D dataset has more clearly defined the trapping geometry, which is a three way dip closure against a basement fault. The highly sampled amplitude data from the 3D may help describe the porosity distribution of the reservoir, a thin sand within the Late Carboniferous Merrimelia Formation. After drilling Hollows 1, an anticline, the remaining exploration potential within the 3D is now confined to stratigraphic, and lowside fault plays. The 3D data will allow these plays to be explored with reasonable risk.

The 3D seismic dataset has provided a clearer understanding of the structural style in both the Cretaceous and pre Jurassic sections. This will lead to direct economic benefit through the better definition of reservoir geometries. The depth conversion has been greatly improved through delineation of the high velocity calcite cemented sandstones in the Jurassic section, and the increased stability of the 3D velocity model.

The Lake Hope 3D seismic project has demonstrated that land 3D seismic data can be used to revitalise the exploration and development potential of a mature oil province.

Interactive interpretation of some 3000km of seismic section, combined with the integration of well, potential field, subcrop and surface geological data with satellite imagery have shown that the structure of this prospective basin is controlled by compressive tectonics.

Four stages of basin development between the late Proterozic and Carboniferous can be identified on calibrated seismic sections from the eastern Officer Basin.

Recent work suggests that the Officer Basin was originally part of an extensional Centralian Superbasin which included the present Amadeus, Ngalia, Savory and Georgina Basins. Justification for this hypothesis comes from the broadly similar stratigraphy in each of these basins. A model is proposed which explains the deformation of the Centralian Superbasin into those basins recognised today, separated by basement uplifted to shallow depth and exposure by compression.

The main deformation is believed to have commenced at or around the Proterozoic-Cambrian boundary, with a substantial pulse at the end of the Cambrian. As a result, the original broad platform depositional system was deformed into a series of foreland-style basins. Stacked thrust sheets of late Proterozic to Cambrian sediments, containing the best known source rocks in the basin, illustrate the intensity of the compressive deformation. Interbedded reservoir units have been folded to form traps suitable for hydrocarbon retention. This model will enhance the probability of prospect generation and the search for commercial hydrocarbons.

The economic constraints brought about by low oil prices make it necessary to maximise the structural and strati-graphic resolution of our seismic data while still being subject to these economic constraints. Thus, whenever any new technology or procedures become available, the cost impact of any parameter changes must be carefully considered against the potential technical benefits achieved.

This paper reviews some of the acquisition trials carried out during 1989 - 1994 and discusses the impact of those trials on the current acquisition parameters utilised in the Cooper and Eromanga Basins. The modification to parameters and procedures brought about by these trials has seen the productivity of seismic crews increase from 1.36 km/chg hr in 1988 to over 1.85 km/chg hr during 1992 - 1994. Thus, average monthly production in excess of 400 km/crew month in 1988 has increased to 550 - 600 km during the last few years, with the remarkable totals of 651 km and 665 km achieved in July 1993 and March 1995 respectively.

The main acquisition parameters analysed include group interval/receiver array length and source effort. The opportunity to use 60,000 lb vibrators is considered to have given a 10 -15 Hz increase in bandwidth at the high frequency end. This means that the previously perceived 65 Hz ceiling, which was considered to be caused by high transmission losses in the shallow section (Q = 25), is no longer restricting the bandwidth.

Further investigations into static control techniques have demonstrated that the refraction method has been unable to match the results gained from upholes, whereas subsequent upholes drilled to check residual static profiles generally add confidence to the use of residual statics tied to upholes.

Mapped seismic two-way-times to the top of the Cadna-owie and Toolachee Formations in part of Petroleum Exploration Licences 5 and 6 of the Cooper-Eromanga Basins were depth-converted using the interval velocity and velocity anomaly methods. Two types of input velocity data were used for each depth-conversion method: (i) based on mapped seismic isochrons combined with thicknesses at well locations, and (ii) based on drift-corrected checkshot/sonic log data and thicknesses at well locations.

The sequence was split into three layers for the depth-conversion: surface to top of the Mackunda Formation (layer 1), top of the Mackunda Formation to top of the Cadna-owie Formation (layer 2) and top of the Cadna-owie Formation to top of the Toolachee Formation (layer 3). The interval velocity method used constant layer interval velocities as predicted by interval velocity maps. In the velocity anomaly method, regional velocity/depth functions were fitted to the input layer velocities. Anomalies with respect to these regional functions were then mapped and used in conjunction with the regional functions in order to undertake the depth-conversion.

The accuracy of the depth-conversions was assessed with reference to test wells not used in the determination of the input velocity models. The 95.5% confidence limits of the error in depth-conversion for the velocity anomaly method are 7-16% closer to zero than for the interval velocity method at the top of the Cadna-owie Formation, and 21-41% closer to zero at the top of the Toolachee Formation. Hence velocity anomaly can be more accurately mapped and predicted away from the input wells than ‘raw’ interval velocity. There is no significant difference between the errors in depth-conversion based on seismic isochron and checkshot/sonic log input velocities. Given that the former are more widely available, and more easily determined than the latter, their use is preferred.

The Lake Hope 3D survey was acquired in 1992 and covers five producing oil fields in the Lake Hope Block of the South Australian sector of the Cooper/Eromanga Basins. A total of 38 wells have been drilled within the survey limits, of which 24 are currently producing from eight different reservoir levels. The bulk of the production comes from the Early Cretaceous/Namur Sandstone and Jurassic Birkhead Formation, Hutton Sandstone and Poolowanna Formation. The reservoirs occur between 1.2 and 1.5 seconds two way time (1300-1800 m below MSL).

The excellent quality of the Lake Hope 3D seismic data and the large number of wells in the area allowed extensive use of seismic forward modelling. The object was to better understand the quality and distribution of the various reservoirs in the area.

Integration of well data, seismic amplitude information and production history made it possible to define the presence and extent of producing Birkhead Formation sands. Two dimensional forward modelling was utilised to interactively alter well-logs and observe the corresponding change in seismic response. By using these procedures and horizon slices from the 3D seismic volume it was possible to prognose additional areas where similar sands are likely to be present.

The Hutton Sandstone reservoir contains over one quarter of the oil within the Lake Hope Fields and is also a very productive reservoir elsewhere in the Eromanga Basin. The Birkhead Formation is the top seal to this unit and the Birkhead/Hutton Interface marks the change from the high energy braided stream environment of the Hutton Sandstone to the lacustrine/low energy fluvial environment of the Birkhead Formation. This interface varies from an abrupt change to a transitional change over a 10 metre interval. Reservoir quality and pay thickness are heavily dependent on the nature of this transition. Well information was correlated with observed seismic response to establish the character of the Birkhead/Hutton Interface within the area of the 3D seismic survey.

The above studies demonstrate the importance of integrating geological information, well production history and observed variations in seismic response to fully utilise 3D seismic data.

Numerical studies have been conducted to analyse how the elastic parameter S* influences the shapes of the squared travel time-offset curves (T² versus X²) for surface seismic data and the subsequent effects on stacking velocities using reasonably large offsets typical of reflection seismology. The velocity functions of Green River shale and Mesaverde clayshale (Thomsen, 1986) used in this study represent opposite properties of P-wave propagation. As such they have been chosen to explain fully the role of δ in this analysis.

A one-layer transversely isotropic solid overlying a plane reflector represents the model geometry used in the numerical studies. The axis of symmetry is vertical and both horizontal and dipping reflector cases were considered. Travel times were computed for common mid-point gathers above the reflector. The stacking velocities were computed at different offsets by fitting a tangent line to the T²-X² plot at each X, making it possible to generate graphs of stacking velocity versus offset (X) using Green River shale and Mesaverde clayshale as representative examples.

The numerical modelling results indicate that the stacking velocities not only vary significantly with offset but also differ from the vertical ray velocity depending on the sign of the anisotropic parameter <5*. Hence, P-wave surface-seismic data cannot accurately predict depths to horizontal reflectors in the presence of anisotropy. This study further emphasises that interpretation of travel time graphs must consider the presence of anisotropy. The variation of the stacking velocity, with offset is expected to give rise to non-hyperbolic travel time curves, which if not taken into consideration may lead to deterioration of stacking itself.

Results of the physical modelling obtained using phenolite material which simulates a transversely isotropic medium, indicate that, if anisotropy is present, the stacking velocity used in aligning the primary reflection changes with offset and may not be representative of either the vertical or horizontal velocity.

Parigi Formation is a group of sedimentary rocks deposited during the Neogen time transgression in Northwest Java Basin. Its lithology comprises mostly limestone biostrome with 10-50 m thick, bioherm of 300-500 m thick. It was deposited in the shallow marine environment during the Late Miocene. Parigi Formation is one of the most important gas reservoirs in Northwest Java Basin.

In the limestone reservoir, the hydrocarbons were trapped within vughs of the formation. The vughy or porosity in the Parigi Formation is studied by seismic attribute and AVO analysis. In this formation the top of porous zone could be first identified by the reflection strength and instantaneous frequency from the seismic attribute analysis. After the spatial location of this top has been identified, the value of porosity is estimated by using AVO analysis. In Shuey‘s AVO equation, the reflection coefficient as a function of angle of incidence depends on P-wave velocity, density and Poisson‘s ratio. The Marquardt iterative algorithm was used to find the Poisson’s ratio which is directly related to the S-wave velocity.

For carbonate reservoir, an empirical relationship exists between the porosity and the Young modulus. Since the Young modulus is related to the density, the P-wave velocity, the S-wave velocity and the Poisson’s ratio, then the porosity of the reservoir can be computed.

The application of these two methods to one structure in the Parigi Formation shows a zone of high reflection strength and low instantaneous frequency in the reef limestone reservoir. Porosity calculation in that zone by AVO analysis give a value of porosity of 36.6 % with a Poisson‘s ratio 0.266. These results compared favourable with borehole data In the location (20 - 38 % porosity). The method has given a satisfactory result in estimating the porosity and delineating its lateral direction.

Seismic amplitudes are shown to provide useful information about rapid lateral velocity changes. Inversion of seismic amplitude data is used to constrain conventional traveltime tomographic inversion, producing an improved velocity field. Amplitude measurements recorded over a time gate, and normalised for each offset, can be assumed to be influenced solely by absorption in the overlying rocks. A normalised image of seismic velocity can be produced by using this simplification to invert seismic amplitudes. Lateral velocity gradients calculated from the amplitude inversion are used as a filter to constrain the traveltime solution. This approach results in sharper resolution of lateral velocity variations. This technique is applied to a model of a surface reef from the Northwest Shelf of Australia.

In 1988, Katnook-1 discovered gas in the basal Eumeralla Formation in the Penola Trough. All subsequent commercial discoveries have been within the underlying Crayfish Group. Amplitude-versus-offset (AVO) analysis performed on single gather records correctly predicted the subsequent discovery of gas within the Crayfish Group in the Katnook-2 and Ladbroke Grove-1 wells.

Seismic recorded since 1990 has generally better quality data which are more suitable for AVO analysis. Two lines were recorded in 1993 over the Katnook Field, tied to the three Katnook wells. Only a small AVO effect was predicted at some locations from modelling using the well data, which was probably due to the poor quality well logs. The seismic AVO analysis, however, using all gather records, produced clear anomalies at both reservoir levels which approximately matched the predicted extent of the gas in the field from structural mapping. The subsequent Katnook-4 appraisal well confirmed the presence of gas at these levels, but unfortunately was not commercial.

AVO and inversion were also run over the Haselgrove Gas Field prior to its discovery in 1994. AVO anomalies are seen at both Katnook reservoir levels, but the strong basal Eumeralla Formation anomaly is interpreted to be due to a clean, water saturated sand rather than gas. The anomalies at the top Pretty Hill Sandstone level are more consistent with structure although they extend below mapped closure. Subsequent reprocessing using pre-stack migration prior to AVO analysis produced anomalies which closely match the actual gas column height, and are within the zone of strong seismic amplitudes at this level.

AVO anomalies in the area appear to be caused by both clean, water filled sands and the presence of hydrocarbons. Seismic inversion can help the prediction of sand quality variations, and the mapped extent of structural closure can be used to identify those AVO anomalies more likely to be related to hydrocarbons. Pre-stack time migration prior to AVO processing is essential in most cases. Unfortunately, AVO anomalies also occur in reservoirs with low gas saturations or low reservoir permeability. Despite these drawbacks AVO with inversion has proven to be a valuable exploration tool in the Penola Trough. No gas discoveries have yet been made which are not associated with an AVO anomaly.

We present a geologic modeling system that can accept the large, irregular surface and volume data typical of exploration studies. The modelling system is based on topological data structures that are used in other areas of computer-aided geometric design. The explicit representation of the model’s topology in the data structures is crucial for carrying out the construction and manipulation of the model. The modelling system provides import functions to build 3-D velocity fields from isolated velocity points and to build 3-D surfaces from a collection of points on the surface. It also provides model visualisation and editing functions to merge multiple horizons and velocity definitions into a single model. We demonstrate the system capabilities with a complex salt body embedded in layered sediments and with a model of intersecting horizons and faults.