WO2014198347A1

WO2014198347A1 - Method and apparatus for determining rock properties

Info

Publication number: WO2014198347A1
Application number: PCT/EP2013/062446
Authority: WO
Inventors: Øyvind KJØSNES
Original assignee: Statoil Petroleum As
Priority date: 2013-06-14
Filing date: 2013-06-14
Publication date: 2014-12-18

Abstract

There is described a technique for determining a rock property. Embodiments include steps of providing geophysical data associated with a subterranean region, said region having a geological structure; providing at least one geological structure component; and processing the geophysical data to determine the rock property, said processing being guided by or based upon the geological structure component. This can provide improved rock property estimates.

Description

METHOD AND APPARATUS FOR DETERMINING ROCK PROPERTIES

Technical field The present invention relates to the field of determining properties of rocks. In particular, determining physical properties of a rock formation from geophysical data such as seismic or electromagnetic data, such as obtained from measured seismic or EM waves. Background

Geophysical data are commonly acquired and used for characterising the Earth's subsurface. In the oil and gas exploration and production industry, seismic data may be used to help detect geological structures in the subsurface and to locate hydrocarbon reservoirs. Electromagnetic data may also be obtained which may depend upon the resistivity of subsurface rocks. Such data may be used to determine physical properties of the subsurface rocks, which may typically comprise solid rock matrix with fluid contained therein. Geophysical data typically need to be processed in some way and interpreted in order to find hydrocarbons. In offshore reflection seismic data acquisition, where streamers are towed in the water behind a boat, acquisition may be limited to recording pressure wave (P-wave) reflection data. In seismic reflection acquisition, seismic waves are detected by a detector and amplitudes of the waves are recorded as data. The recorded data comprises amplitudes of reflected P-waves which have travelled through the subsurface from a seismic source at the surface and have reflected from a reflector in the subsurface and travelled to the receiver. Water does not possess a shear modulus which means that shear waves (S-waves) do not travel through water. Hence, there may be a limitation to P-wave recordal in offshore settings. However, it is possible to deduct S-wave properties of the subsurface through utilizing information available in the already recorded P-wave data. The S-waves are more sensitive to changes in rock properties and not so much the fluid content of the rock. The fact that quartz sand has a significantly higher shear modulus compared to shale makes the S- wave information better suited for lithology predictions than the P-wave information. More specifically, the deduction of S-wave information from P-waves is performed by observing how the P-wave reflection amplitudes from a Common Reflection Points behave as a function of increased offset (or incidence angle). This method is commonly referred to as Amplitude Versus Offset (AVO), the principle of which is shown in Figure 1 . The key concept is that portions of the seismic P-wave energy will convert to other kinds of energy. The amount of conversion is a function of the P- wave's incidence angle and the physical properties of the rocks on either side of a geological boundary.

The amplitude behaviour for a reflected P-wave (R_pp) as a function of incidence angle (6)and differences in rock properties (P-wave velocity (V_p) , S-wave velocity (V_s) and density (p)) across a geological boundary can be given by the Aki-Richards equation (Aki and Richards, 2002), as follows:

(Equation 1 )

The ray parameter p is given from Snell's law, p=sin Θι / V_pi . The study and mapping of a reflected seismic amplitude behaviour as a function of Incidence angle is commonly called AVO. Since equation 1 is given as a function of incidence angle it is most convenient that the input data to the equation is angle converted gathers or partial angle stacks. Both types of data can used directly with this equation. Equation 1 is one of many ways to generate seismic AVO and inversion volumes (i.e. of rock properties) from partial angle stacks or angle gathers.

It is desirable to make fluid and lithology predictions (LFPs) for example to detect a reservoir or monitor changes over time. The fundamental concept behind LFP using seismic amplitude versus offset data is that sands will have a much higher S-wave velocity and different density than shales, such that we can predict how the AVO behaviour of a sand-shale interface would be. This allows lithology prediction to be performed to determine for example the type or nature of solid rock. Similarly, the fluids contained in the rock (most commonly brine, oil and gas) have different density and P-wave velocity such that it can be seen from the Aki-Richards equation that a change in fluid, for example from brine to oil in a sandstone reservoir, will have an effect on the AVO behaviour of the rock formation when containing such a fluid. Hence, fluid prediction can be performed from seismic data.

In order to determine AVO behaviour and perform LFP typically, some initial processing steps or simplifications (e.g. stacking) are performed on the data to prepare the data as angle gathers or angle stacks. Thereafter, the angle gather or angle stack seismic amplitudes are used, directly and without further modification or filtering, as input from which an Intercept and Gradient can be calculated. In turn, from the calculated Intercept and Gradient, rock properties can be determined based on for example the Aki-Richards equation.

The "Intercept" is the zero offset reflection coefficient where no or very little wave conversion is occurring, giving an understanding of acoustic impedance contrasts at a geological boundary. It is the answer from the Aki-Richards equation for the angle 0=0 degrees. Figures 2A and 2B show this concept.

The "Gradient" is the derivative of the Aki-Richards equation for Θ (0-35 degrees). In other words it is the rate of change in amplitude as a function of offset (or incidence angle). This term is more related to differences in Vp/Vs (P-wave/S-wave velocity), and equally Poisson's ratio across a geological boundary. It has been proven mathematically that the Gradient equals the P-wave reflection amplitude minus twice the S-wave reflection amplitude. Different weighted stacking of Intercept and Gradient is what is typically used in AVO exploration to obtain a LFP prediction (a^*l + b^*G = different weighted stacks for all combinations of a and b (scalars)). This is used for producing an AVO rotated stack, Cartesian rotation for Rp, Rs or Producing AVO, etc. The Gradient is very often affected and sometimes even destroyed by noise in the data to the extent that AVO methods cannot be used for LFP.

Rock properties may also be determined from seismic data by performing an inversion of the data. Seismic inversion is a transform of the seismic data, by which reflections (i.e. amplitudes of received reflected waves) at geologic boundaries are converted into layer rock properties of Vp, Vs and Density. This method is also highly dependent and often limited by the quality and noise level of the seismic data. Seismic data may vary in quality, even within a dataset itself. This can be the case both in production and exploration settings. In fact, it has been demonstrated that reservoirs containing hydrocarbons are often associated with gas clouds above the reservoir. The presence of such gas clouds can significantly impact and lower the quality of seismic data, and in some cases render it useless with today's methods. This is particularly the case for sections of a reservoir below the gas cloud.

Low quality seismic data is often characterized by a low signal-to-noise ratio. The lack of signal compared to noise particularly limits the capability of correct estimation of the Gradient term and, where the data comprise low fold, angle stacks. Therefore, it might not be possible to perform an accurate LFP in such cases. A poor estimation of the gradient term can in turn lead to difficulties in estimating typical AVO. Estimating the 3^rd AVO term from equation 1 , the curvature of the reflected seismic amplitude across a geologic boundary as a function of incidence angle without a good estimate of gradient is impossible. It is then also impossible to estimate anisotropy properties (for example Epsilon and Delta) from AVO observations, which relies on curvature and higher order fit to that line with higher angles.

Noise in the seismic data limits the prediction capabilities and accuracy of prestack seismic inversions due to the fact that angle stacks are more affected by noise due to its lower fold. Angle stacks are angle transformed and partially stacked data based on the seismic prestack offset gather. The reason for performing the angle transform is to convert the data to the incidence angle domain for use in the AVO and Inversion equations, rather than in the acquired offset domain. The reason for doing the partial stacking in the angle domain is to enhance the reflected energy, and to reduce the effect of Gaussian and some coherent noise. A positive side effect is that the amount of data in megabytes is significantly reduced as a function of the partial stacking of the prestack seismic data. The partial stacking will hence improve the data quality and reduce the size of the data that will be input to the AVO or Inversion study. But still, even if the partial stacking is today recognized as one of the most powerful and commonly used methods to enhance the reflectivity and reduce the noise in the seismic data, the algorithm is not completely removing all noise in the data. The noise left in the seismic data even after angle stacking might therefore be so dominant that obtaining the correct prediction of the three fundamental properties Vp, Vs and density (eq. 1 ) from the seismic data might be impossible with today's methods. It has also been proven that the effect of anisotropy plays a significant part on amplitude and amplitude versus offset (or angle) responses. When not able to quantitatively predict the rock properties from Equation 1 , it can yet more difficult to quantitatively predict anisotropy properties.

Seismic data noise may be reduced by stacking amplitude time series (each which may define a seismic trace) recorded at different lateral locations but associated with a common midpoint. It is been commonly accepted in the seismic industry for many decades that stacking of traces is the most optimal way to remove non-coherent and dipping noise. The basic stacking principle is shown in Figures 3A and 3B, for non- noisy and noisy synthetic data respectively. Several attempts have been made in the industry to extend and refine this method into the post stack domain, by techniques such as often referred to as trace mixing, smoothing or median filtering, etc. Most of these techniques do not take into sufficient consideration the effects of topography of geological features upon the seismic data. Current methods assume that reflectors are flat, which is seldom representative of the true geological features. This is especially so and of significant importance in reservoir settings because reservoirs contain topographic, dipping geological structures that provide traps for accumulation of hydrocarbons.

This point is illustrated in Figures 3A and 3B. By applying a typical stacking technique, important dipping top reservoir reflectors may be lost. This may be significantly detrimental to any amplitude based techniques and in particular AVO and seismic inversion based techniques. Poor predictions of rock properties may therefore result. Indeed, it can upset the structural picture and the amplitude information for AVO or quantitative amplitude products such as seismic inversion. In this case where a dipping reflector is completely lost, a discovery could not be missed because the top reservoir could not be seen. The structural trap is of primary importance to map using LFP methods based on seismic data in order to optimally position the exploration well, drill and find hydrocarbons. It is also important during the production phase to characterize the reservoir from seismic in order to further improve the drainage strategy. Poor identification of reflectors at the trap can have significant implications. Summary of the invention

In a first aspect of the invention, there is provided a method of determining a rock property, the method comprising the steps of:

(a) providing geophysical data associated with a subterranean region, said region having a geological structure, said geophysical data being obtained from measured waves;

(b) providing at least one geological structure component; and

(c) processing the geophysical data to determine the rock property, said processing performed using the geological structure component.

In a second aspect of the invention, there is provided a geological structure component for use in processing the geophysical data of the first aspect to determine a rock property.

The geological structure component may preferably comprise a tensor or component thereof, e.g. a geological structure tensor.

In a third aspect of the invention, there is provided apparatus for performing the method of the first aspect of the invention.

In a fourth aspect of the invention, there is provided a computer program for use in performing the method of the first aspect. In a fifth aspect of the invention, there is provided a computer arranged to execute the computer program the fourth aspect to perform the method of the first aspect.

In a further aspect of the invention there is provided, a method of determining a rock property, the method comprising the steps of:

(a) providing geophysical data associated with a subterranean region, said region having a geological structure;

(b) providing at least one geological structure component; and

(c) processing the geophysical data to determine the rock property, said processing being guided by or based upon the geological structure component. In a yet further aspect there is provided apparatus for performing the method of the further aspect.

Each of the above aspects may include further features as set out in the claims or in the present description or the drawings in any combination. Features may be combined between any of the different aspects.

Each feature disclosed or illustrated in the present specification may be incorporated in the invention, whether alone or in any appropriate combination with any other feature disclosed or illustrated herein.

The invention provides numerous advantages as will be apparent from the description, drawings and claims. In particular, the embodiments of the invention provide an improved, more robust, way to determine rock properties from geophysical data such as seismic or electromagnetic data compared with prior art techniques.

Description and Drawings

There will now be described, by way of example only, embodiments of the invention with reference to the accompanying drawings of which.

Figure 1 is a representation showing the concepts of Seismic Acquisition for a Common Reflection Point (CRP), seismic data in the form of amplitude versus offset (AVO - equivalent to Amplitude versus Angle of incidence, AVA), and wave conversion at an reflection point across a subsurface reflection interface;

Figures 2A and 2B provide a representation of the concept of going from a seismic pre- stack angle gather as shown in Figure 2A to the AVO products of Intercept (+R/-R) and Gradient (Slope of the lines) as shown in Figure 2B, wherein the dots in Figure 2B are the peak and trough amplitudes taken from 2A, and R is the reflected vertical amplitude; in an AVO or inversion setting where the reflected properties are converted to rock properties rather than normalized differences in rock properties, R is similar to the reflection coefficient, whilst the Gradient is similar to the Vp/Vs ratio or Poisson's ratio; Figures 3A and 3B are representations showing the effects of stacking seismic data;

Figure 4 is a work flow according to an embodiment of the invention; Figure 5 is a representation of different quadratic shapes of a 3D quadratic tensor for identifying the location of amplitudes corresponding to a common geological boundary;

Figure 6 is a representation of a workflow according to an embodiment of the invention; Figure 7 is a representation of a computer device for use in performing the method of the invention;

Figures 8A shows synthetic noisy intercept data in an example embodiment of the invention;

Figure 8B shows dip-azimuth vectors identifying directions of coherence in the data of Figure 8A;

Figure 8C shows filtered amplitudes based on the data of Figure 8A, with amplitudes preserved and smoothed along the dip-azimuth vector;

Figure 8D is a graph of Intercept versus Gradient based on picks of amplitude from the noisy data of Figure 8A; and Figure 8E is a graph of Intercept versus Gradient based on picks of amplitude of the data filtered based on the dip-azimuth directions of Figure 8C.

Referring firstly to Figure 4, a flow chart is shown illustrating an example method for determining a subsurface rock property including steps S1 to S3. At S1 , seismic reflection data are obtained for a subsurface region of interest. The seismic reflection data comprises amplitudes obtained from seismic wave detectors that are arranged to detect seismic waves reflected from the subsurface in response to a seismic source event. Each amplitude record is associated with a lateral position, for example a mid point between a seismic source and detector, and a time relative to a defined source event. Seismic waves from the source are reflected where the elastic properties of the subsurface change, for example at boundaries between different rocks, and the amplitudes of the reflected waves depend upon the extent of contrast of the rock properties across such boundaries. Seismic waves that are detected at late times relative to the source event correspond to waves that have travelled the farthest and typically have reflected from a deep reflector. Thus, time can be an indicator of depth. Accordingly, seismic reflection data can be expected to contain information as to the nature of subsurface geological structures where such changes exist, both in terms of their location and the properties of rocks to either side. Such structures may be dipping or rugose, sometimes curvy and complex interfaces between rocks with different properties.

A geological structure component is also provided. In order to do so in this example step S1 also includes obtaining geological structure data, for example from a predetermined earth model, although it could be obtained from the reflection data as described further below. The geological structure data provides attributes of a geological structure known or hypothesised (e.g. simulated) to exist in the region of the seismic reflection data, and may have complex interfaces between rocks as indicated above. The attributes are for example the location and attitude of a part of the structure [e.g. interface or boundary with an adjacent rock type]. To define the attitude, the data can comprise dip angles of the structure (e.g. relative to horizontal) and azimuths in the direction of the dip (e.g. relative to North in the horizontal plane). The dip may preferably be the maximum dip for the structure at the relevant location. The location may be given by spatial coordinates.

In step S2, the attributes are used to identify records in the data that belong to or are associated with the geological feature. This step may include assigning the attributes to the seismic data. For example, one may specify that the seismic data in certain interval of travel time have a specific dip and azimuth attribute. The seismic data records in that interval may then be identified to be related to the same geological feature and to each other.

In step S3, the seismic amplitude data are processed and used to determine the rock property. This may involve picking or extracting the identified records from the data, and using the picked or extracted records to determine the rock property for example via the Aki & Richards equation.

Describing geological structure

The attributes, such as dip, location, curvature and attitude of the geological structure can be represented by a mathematical component, such as a structure tensor. The tensor can be given for each data point in the subsurface region in question. The tensor can indicate the gradients of the structure in defined, e.g. different, directions from the data point, and how these are spread.

The tensor for use in the present method can be provided in different ways. It may be obtained from the seismic data itself, or independently (e.g. from a non-seismic data source). In the case of deriving this independently, a simulated model of a geological fluid reservoir structure may be provided. Suppose such a structure has the form of a dome at a certain location in the subsurface region. A structure tensor can then be defined providing a determination of the gradient directions throughout the volume of data. Where there is no dipping structure, the tensor has no preferred gradient direction. Preferred (e.g. maximum) gradient directions will be defined for points on the structure, and in the case of a dome will differ between different points on the dome surface.

It is within the scope of the invention to simply give coordinates in the volume at which the structure is known or estimated to be present. From these coordinates, angular directions indicating where the structure is present may be readily derived.

The tensor or attributes of the structure may be found from seismic or other data by performing a scan or search of the data which are believed to relate to the structure. Numerous methods for estimating a Dip and Azimuth tensor from seismic data has been suggested in the literature (see for example Marfurt, K.J., 2006, Robust estimates of 3D reflector dip and azimuth, Geophysics, Vol. 71, No. 4, p29-40, and other articles listed in the References section below). Such methods may be used in the present invention to define a tensor that gives directions, where for example the amplitude at a given location in the data corresponds with amplitude at adjacent or different lateral locations. This may involve a determination of data coherence. These methods can be divided into three main competing methods for estimating Dip and Azimuth directly from seismic data: 1 ) by calculating temporal and spatial derivatives of the phase estimated using complex trace analysis, 2) by an explicit dip scan to find the most coherent reflector, or 3) by calculating the first eigenvector of the gradient structure tensor. These techniques may be extended to track curvature to suit more complex 3D quadratic shapes.

In the Marfurt technique, amplitudes are scanned along several lines of different dip and azimuth, and a measure of coherence of the data is determined for each line. The dip and azimuth providing greatest coherence in the data, e.g. semblance, is determined which indicates the boundary or the amplitudes in the data associated with a subsurface, e.g. dipping, reflector. In cases of only noise, the tensor will flip back and forth, reducing noise but not producing any signal if there is none. In cases of continuous reflections i.e. a geologic boundary in the sub surface, this tensor estimate will follow the topography of this geologic boundary. The tensor is composed by at least one set of parameters, such as Dip or Azimuth. The limitation in a seismic file is one value (parameter) per sample, such that a multi parameter tensor needs to be decomposed into multiple files, one file or "data volume" per parameter.

In general, the search technique to obtain can be applied with 3D data sets and to determine the attributes of 3D structures. Thus, a particular parameter of the tensor may be provided as a data set corresponding to the 3D data volume. Accordingly, the data can be analysed to assess the coherence in 3D so as to provide a 3D tensor. The tensor will relate to the structure responsible for the reflection amplitudes, i.e. represent structure, because data noise will cancel out. In some variants, the amplitudes may be analysed in 3D to identify quadratic structure and correspondence between amplitudes, such that complex 3D quadratic shapes can be identified. A 3D tensor with quadratic or curvature parameters may define the directions along which the structure is found in the relevant part of the data. Figure 5 shows examples of the sorts of 3D structures that could be identified. The Kpos=Kneg=0 is in this case in the middle of the figure. The tensor may optionally include additional parameter volumes for defining curvature of the 3D tensor for defining a 3D quadratic tensor shape.

The tensor provides in effect a mathematical representation of a boundary in 3D between the neighbouring rocks, representing the geologic interface in a small local 3D volume around a centre sample. The tensor may be considered to contain the 3D direction of maximum energy of the reflected waves of the seismic data locally in the local volume. The tensor is the mathematical description of the shape for a given geological boundary at a given X, Y, Z triple coordinate point. Z can be time, or depth, X, Y can be UTM or line numbers from a given seismic acquisition geometry. This tensor is variant both in the spatial and temporal domain, giving a high resolution, mathematical description of the shape of the geologic boundaries depicted in a set of data. The tensor can be calculated at centre sample positions throughout the whole 3D seismic data volume. In this way, the local small volume observations of geological boundaries are extended across the whole 3D dataset. This allows the tensor to follow and identify boundaries following complex variations in geology, depending on the size of the selected local small 3D volume. The 3D tensor for the complete seismic can then be broken down to dip and azimuth and stored as two cubes of seismic attributes. Thus, it is possible to represent the true shape and topography of the sub surface geology with the use of a tensor with a varying orientation according to the topographic shape. When analysing the seismic data in practice, a small volume of seismic data may be extracted around any centre sample for each calculation of the tensor for the centre sample. The amplitudes will have to be sine interpolated between samples so that exact values are used rather than the inherent sample interval amplitude resolution. A sine function may be applied to the amplitudes in the small data volume. The aim of the sine function is to perfectly reconstruct the analogue seismic signal from the digital representation given in the seismic data file. This is a typical way to get good accuracy out of the amplitudes present in the seismic data file, and also allow a greater resolution and accuracy in the estimation of the dip, azimuth and optional curvature tensor. Then the scan can then be performed on this smaller data volume that will find the tensor (dip and azimuth) of direction of maximum energy for the centre sample. A structural data volume comprising the tensor parameters / structure attributes can therefore be formed which corresponds with that of the seismic data, such that locations in the structural data volume map or correspond to locations in the seismic data. Accordingly, the tensor attributes indicate the where in the seismic data, data records may be located and identified that relate to the relevant structure.

Data work flow The seismic data for determining the rock property can be provided in any suitable form. For example, the seismic data may be amplitude versus angle gathers comprising amplitude time series each at different angles or offset locations with respect to a common midpoint. The data may also be intercept or stacked data comprising intercept or stacked amplitude time series each associated with a different lateral midpoint location. The intercept or stacked data may be derived from amplitude versus angle gathers. The identified data records in the seismic data may therefore comprise amplitudes of different kinds.

The processing of the seismic data to determine a rock property may be performed in different ways or involve a number of techniques. The processing technique adopted may depend on the kind of data used and data records identified.

The processing may comprise performing an inversion of the seismic data, wherein the identified data records are considered to be coupled data records. The coupled inversion may be performed using well known rock property inversion algorithms for inverting seismic or electromagnetic data with coupled data records. The data records identified from the tensor can be considered to be coupled.

The inversion core is arranged to use data observations from neighboring cells instead of only observations in a single cell. As discussed above, the tensor gives the information of how the geology is sloping (dip and Azimuth), and is used to let the inversion know how to couple observations from one cell to the neighboring cells. The inversion may take the mean, median or some weighted function of the coupled set of amplitudes, and run the actual core of the inversion to calculate the rock property. The key to producing a result that recognizes the structure is to provide a tensor that can couple data i.e. a seismic inversion in the temporal domain given that we know this tensor.

Coupled means that it is possible to relate observations in one cell (X, Y, Z) of the data to the observations in the neighboring cell (X+1 , Y+1 , Z) or (X-1 , Y-1 , Z) or (X±n1 , Y±n2, Z±n3) given that the tensor is representing the shape of geological structure or boundaries estimated from a volume of input observations with foot print size (±n1 , ±n2, ±n3). The tensor reveals in which directions the couplings must be done, in order to depict the same boundary of geology in the analysis of observations in the center cell.

In some variants, the seismic data may be processed to extract or select the identified data records from the data sets. The extracted or selected data records, may then be inverted to determine a rock property.

The seismic data associated with the identified records, e.g. amplitudes, may be processed by smoothing, weighting, and/or have a function fitted thereto to reduce noise. The data may be filtered to reduce amplitudes in data not related to the structure, or preserve or enhance amplitudes of the data records identified. Then, the processed or filtered data can be inverted using a seismic rock property inversion algorithm.

Intercept and gradients may be obtained from the processed data, and from the intercept gradient, rock properties determined using the Aki-Richards equations. Indeed, the most important rock properties for seismic amplitudes and seismic imaging are given by the Aki-Richards equation (Equation 1 ), that is, contrasts in Vp, Vs and density across a geologic boundary or interface. Once one has the Intercept and Gradient, it can be a matter of solving three simultaneous equations with three unknowns to solve Aki Richards for Vp, Vs and density. Another AVO product that could be used, in addition to Intercept and Gradient, is Curvature of the AVO data.

The robust filtering of the collection of amplitudes selected can for instance be a variety of edge-preserving mean, median and alpha-trimmed-mean filters. In other words, a filter of the above type can be applied to the data to, in effect, preserve amplitudes that are laterally coherent or geologically consistent based on the directions given by the vector or tensor, and to suppress amplitudes that are not laterally coherent or geologically consistent. The filtered amplitudes will then be less affected by noise such that cross plots, AVO products (Intercept, Gradient and any Cartesian rotation product from these), and Inversion will be more robust, and less affected by noise. The data quality will be enhanced such that the confidence in the final product (cross plot, AVO- or Inversion product) will be improved.

A spatial and temporal, robust filtering of the seismic amplitudes in a geologically consistent and topographically honouring manner makes the post-filtering seismic amplitudes more correct, and closer to revealing the true answer from any seismic amplitude study (cross plot, AVO, Inversion). The filtered amplitude can then be input into any of the previous mentioned amplitude studies. Since the post-filtered amplitude is a more correct representation of the sub surface geological boundary, and therefore a better, closer to a true depiction of the bounding/adjacent rocks and their properties, any amplitude study (cross plot, AVO- or Inversion study) will benefit from this as they will have greater prediction accuracy and be closer to reveal the true picture and properties of the sub surface. The increased prediction accuracy will also allow predicting more properties from the same seismic in later steps. The abovementioned tensor shape is considered to be sufficient to describe the topography of any geologic boundary within reasonable footprints of the given filter size. The footprint of the filter is a set of dynamic parameters that will have to be adjusted in every case by the user, as seismic acquisition layout, sparseness of sampling and geologic complexity are changing factors from case to case. The design of the filter will have to account for such variations in every case.

With reference to Figure 6, a representation of a work flow to produce improved AVO based rock property determination, with steps S1 to S4 as follows: S1 . Providing seismic data comprising multiple angle gathers or angle stacks;

S2. Producing dip and azimuth cubes. This provides the dip and azimuths for the whole data set with defined directions to amplitudes that replicate the geological structure, produced for example as a result of tensors determined at sample points throughout the data volume. S3 and S4. Using the azimuth cubes to filter the data and preserve amplitudes along surfaces defined by the cubes, and suppress data elsewhere, producing filtered data. Producing improved rock property estimations by using the filtered data as inputs.

With reference to Figure 7, there is shown an example of a computer device 10 for use in determining the rock property as described above. The computer device 10 has an In/Out device 12 used for reading in seismic data. The computer device 10 has a micro processor 13 used for processing the seismic data. In addition, the computer device 10 may be provided with a computer program, stored in a memory device 14, which comprises computer readable instructions for processing the seismic data, analysing the amplitudes, producing the filtered data and estimating the rock property based on the filtered data. The micro processor is connected to the In/Out device 12 and memory device 14, and is used for executing the program to determine the rock property. The computer device may include a display 15 which may be used for visualising the seismic amplitude data. Seismic amplitude data may be displayed for example as an image in the form of a 2D seismic depth section, 3D volume or time slices. The computer device may be provided in a compact unit or may be a distributed system in which for example each of the individual components 12-14 are provided separately. Any or each such component may be provided in separate location and may communicate with one another over a data communication network, for example by cable or wirelessly. The memory device 14 may comprise a portable storage medium which may contain the computer program or parts thereof. The medium may be an optical disk or memory stick or the like, which can be selectively connected or disconnected to the device 10 (for example through a wireless network) as required.

Results

Figure 8A shows seismic reflection amplitude data in the form of time series of intercept amplitudes. The time axis indicates a time of travel of a seismic wave from a seismic source through a region of the subsurface and reflected from a subsurface boundary or reflector and travelled back to a detector. Typically therefore, amplitudes recorded at the detector at late times correspond to waves that have travelled the farthest and typically have reflected from a deep reflector. High amplitude events in seismic reflection data are commonly accepted to represent boundaries between different rocks and rock properties in the sub surface. In Figure 8A, the positions 1 to 15 are lateral positions and time is plotted downwards to indicate depth. The data are synthetic noisy data for a subsurface structure having one flat reflector as seen by high amplitudes at 10 ms and one dipping reflector as seen by high amplitudes at 40-68 ms. It may be noted that in traditional stacking based techniques, the flat reflector in a data set such as this may be taken to be a primary reflector whilst the dipping reflector may be mistaken as an artefact such as a multiple reflection (noise), or linear noise (i.e. boat or nearby offshore installations) or a faulty velocity model (not noise). The data of Figure 8A are analysed and searched in order to find the attributes relating to the geological structure. The search is performed record-by-record to determine the coherency with neighbouring and/or nearby data records and identify the direction of maximum coherence. This direction can be given by a dip and azimuth angle. In order to find this angle, the amplitudes may be analysed along several lines with different dips and azimuth, and the direction identified along which there is greatest coherence.

In Figure 8B, the lines 1 , 2 indicate the maximum coherence or energy (consistent with the structure) from a 2D analysis of the data. These lines are consistent with the amplitudes due to the dipping reflector and the flat reflector. The data in processed form by filtering and smoothed based on these lines is shown.

This shows how a max energy tensor in 3D can be used as a steering filter for selecting or obtaining amplitudes that can then be input for the AVO Intercept and Gradient calculation or to the Seismic Inversion. Hence, the AVO Intercept / Gradient or the seismic inversion will be geologically consistent, and it will follow the contrasts and boundaries in the sub surface geology. As such, the quality and robustness of these AVO analyses and seismic inversion techniques which rely on subsurface contrasts are significantly enhanced. Figures 8B and 8C show traces of filtered amplitudes which are based on the traces of Figure 8A, but in which the Figure 8A amplitudes have been filtered according to the identified vectors 1 , 2. The amplitudes of the original data in proximity to the respective vectors are processed or modified to produce the filtered high amplitudes along the vector in Figure 8B. At data points further away from the vectors, the filtered amplitudes are suppressed versions of the original amplitudes.

As can be seen, the filtered data have a relatively good signal to noise ratio. The filtered data are then used for determining a rock property. For example, a seismic amplitude inversion of the filtered data may be performed for obtaining estimates of S- wave or P-wave velocity. The filtered amplitudes may also provide a better, more robust basis for gradient and intercept determinations according to the Aki-Richards equation. For example, amplitude values from each filtered trace may be selected along the direction of the vector 10 or vector 12 and used to calculate the intercept and gradient. This intercept and gradient can in turn be used to determine a rock property or make LFPs.

Results in the form of Intercept-Gradient cross-plots are provided in Figures 8D and 8E. In Figure 8D, it can be seen the results of the original noisy data, where amplitudes are picked out along the direction of the dipping vector 12. In Figure 8E, results are shown where amplitudes of the filtered data are used, showing much less scattering than in Figure 8D, and providing therefore a more reliable estimate. In Figure 8D, for data with no filter applied, the dot is the data after filter and the triangular is the correct model across the interface. In Figure 8E, the data points after filtering (square) are much closer to the correct model (triangle).

Advantages The method uses information contained in seismic amplitude data relating to geological reflectors for providing better estimates of geological properties.

The method depends solely on regular reflection seismic in the form of AVO angle gathers or angle stacks. Such data are practically always available in exploration and production settings such that new data is not required; i.e. existing data can be used. In the present method, more accurate predictions of rock properties can be obtained, leading to a better Lithology and Fluid prediction that will give faster model updates, and better well placements and drainage strategy. Greater value from existing data can be obtained. Thus, the determined rock properties may be provided in a drilling process and used to determine where to drill. The determined rock properties may be used to determine the location or character of the reservoir, or to monitor the condition of the reservoir over time, for example as production progresses.

This method will also help with statistical methods to perform uncertainty estimations, noise estimations, and data quality estimations. Greater confidence can be put in the seismic data, providing a better constrained input for data quality mapping, uncertainty mapping, Bayesian Inversions, or other seismic data based methods.

In prior art techniques, the amplitudes for use in AVO Intercept / Gradient methods are based on noisy data or data where important subsurface reflectors are not represented. However, by identifying the coherence in the data (corresponding to geological boundaries), and using the direction based on the coherency in the data as the basis for selecting amplitudes for the Intercept and Gradient, more accurate estimates are obtained. In particular, the technique can identify amplitudes resulting from dipping reflectors which may otherwise go undetected so that these can be applied in AVO Intercept / Gradient and seismic inversion methods. In prior art methods, such amplitude information has not been incorporated.

The use of a steering filter can be estimated and used from any supplied seismic volume, both 2D, 3D and 4D. The steering filter can be decomposed from a 3D tensor to two attributes namely dip and azimuth. 3D quadratic tensors can be used, including further attributes such as curvature, as a further improvement in the depicting of sub surface geologic topography in particular where geological boundaries are curved or complex.

The steering filter consistently guides a plane in 3D in a manner that is consistent with structures in a seismic image for picking the amplitudes for predicting rock properties from the seismic data. Then it becomes possible to apply statistic related methods to make a more robust and accurate estimate of the properties across the boundary particularly on noisy data. The improved accuracy in prediction of rock properties may allow further properties to be estimated from seismic data.

This method will also have a positive effect on highlighting or enhancing Direct Hydrocarbon Indicators (DHI). DHIs are amplitudes which are consistently changing with the geological relief, e.g. dimming of amplitudes across a certain depth. It could also be a flat spot indicating directly the position of a fluid contact. A direct hydrocarbon indicator is usually a bright spot, flatspot or amplitude that is conforming with the top reservoir structure. Different DHIs can also co-occur. Because this method automatically orients the plane for picking the amplitudes around the trap in a manner that is consistent with the reservoir structure, and because the generally accepted fact that the natural buoyancy of hydrocarbons make hydrocarbons accumulate inside a reservoir in a structurally conform manner, this method will - when median is used as statistical filter on the amplitudes - enhance the direct hydrocarbon indicators that are related to amplitude variations that confirm with structure. Thus, it will enhance flatspots as they have a reflection that the steering filter will pick up. Bright spots are close to the ideal reflector for the steering filter because they strongly contribute to the energy that the 3D tensor is searching for. The more relative energy in a reservoir setting (i.e. bright spot), the more energy the 3D tensor will find here relative to the surroundings, but contributing more strongly to the estimation of the tensor.

This method preserves the reservoir structure in the amplitude data, preserves true amplitudes, reduces the scatter effect from Gaussian noise present in the data, enhances DHI's using already available, standard data.

The method is applicable to all types of seismic data, including pre-stack and post stack. It provides robust mathematical estimation of rock parameters based on seismic data using the tensor, and in the estimation of rock parameters above and below the seismic reflector.

Getting a better estimate of the AVO gradient and curvature, or reducing the effect of noise on amplitudes in angle stacks enables us to give a better estimate of anisotropic parameters (epsilon/delta) or density, which in turn are important for seismic imaging, reservoir prediction and reservoir characterization. Various modifications may be made without departing from the scope of the invention herein described. In particular, the tensor or structural attributes could be applied to electromagnetic (EM) data sets for inversion in a similar way to the seismic data which is focused upon above. There could also be a step of acquisition of the seismic of electromagnetic data in order to provide such data, and the method may be considered a method of processing such data. The electromagnetic data can for example be obtained by transmitting an electromagnetic wave from a source into the subsurface region such that it interacts with the rock structure. The acquisition of electromagnetic data may be performed using receivers placed in appropriate locations so as to be able to detect and measure waves from the subsurface in response to the source transmission. The electromagnetic data may be controlled source EM (CSEM) data obtained from performing a CSEM survey. References

Aki, Keiiti, and Richards, Paul G., 2002, Quantitative Seismology, vol. I, 2^nd ed, , University Science Books, ISBN 0-935702-96-2

Fatti, J.L, Smith, G.C., Vail, P.J., Strauss, P.J., and Levitt, P.R. (1994) Detection of gas in sandstone reservoirs using AVO analysis: A 3-D seismic case history using the Geostack technique. Geophysics, 59, 1362-1376.

Walden, A. T., 1991 , Making AVO sections more robust: Geophys. Prosp., 39, 915-942 Buland, A. and Omre, H., 2003: Bayesian linearized AVO inversion.

Geophysics, 68, 185-198.

Buland, A., Kolbjornsen, O., Hauge, R., Skjaeveland, 0., and Duffaut, K.,

2008: Bayesian lithology and fluid prediction from prestack seismic data.

Geophysics, 73, C13-C21 .

Claims

CLAIMS:

1 . A method of determining a rock property, the method comprising the steps of:

(b) providing at least one geological structure component; and

2. A method as claimed in claim 1 , wherein the geological structure component comprises at least one tensor or vector.

3. A method as claimed in claim 1 or 2, wherein the geological structure component comprises at least one parameter comprising any one or more of: dip, azimuth, direction, curvature, position, and rates of change thereof.

4. A method as claimed in claim 1 or 2, which further comprises obtaining the structure component from analysing geophysical data or from a model of the geological structure.

5. A method as claimed in any preceding claim, wherein the geological structure component is arranged to associate data records with the geological structure.

6. A method as claimed in any preceding claim, which further comprises identifying data records associated with the structure using the structure component.

7. A method as claimed in claim 6, wherein the processing step comprises fitting a function to the data records associated with the geological structure; and weighting the geophysical data based on the function.

8. A method as claimed in claim 6 or 7, wherein the processing step comprises removing data records or removing noise from data records not associated with the geological structure.

9. A method as claimed in any of claims 6 to 8, wherein the processing step comprises filtering the geophysical data about a filter location determined by the data records associated with the structure.

10. A method as claimed in any preceding claim, wherein the processing step includes obtaining an Intercept and Gradient from the processed data, and using the Intercept and Gradient to determine the rock property.

1 1 . A method as claimed in claim 9, wherein the step of using the Intercept and Gradient is performed using the Aki-Richards equation.

12. A method as claimed in any preceding claim, wherein the processing step further comprises performing a rock property inversion of the data.

13. A method as claimed in any preceding claim, wherein said rock property comprises a rock lithology or fluid property.

14. A method as claimed in any preceding claim, wherein said structure comprises a dipping geological feature or boundary between rock types.

15. A method as claimed in any preceding claim, which comprises using the geological structure component to identify or determine data records, and/or locations thereof, within the data which relate to the geological structure.

16. A method as claimed in any preceding claim, wherein the geophysical data comprises either or both of electromagnetic data and seismic data.

17. A method as claimed in any preceding claim wherein said structure comprises a reservoir trap structure.

18. A geological structure component for use in processing geophysical data to determine a rock property.

19. A component as claimed in claim 18, the geological structure component comprising a tensor or component thereof.

20. Apparatus for performing the method of any of claims 1 to 17.

21 . A computer program for use in performing the method of any of claims 1 to 17.

22. A computer arranged to execute the computer program of claim 21 to perform said method.