Seismic Well Tie Guide 2010

Interpreters Guide for Seismic to Well Ties

Petrel 2010

About Petrel* Development on Petrel seismic-to-simulation software began in 1996 in an attempt to combat the growing trend of increasingly specialized geoscientists working in increasing isolation. The result was an integrated workflow tool that allows E&P companies to think critically and creatively about their reservoir modeling procedures and enables specialized geoscientists to work together seamlessly. With the enhanced geophysical tools and the integration of ECLIPSE* reservoir simulation software and streamline simulation, Petrel is now a complete seismic-to-simulation application for

3D visualization 3D mapping 3D and 2D seismic interpretation well correlation 3D grid design for geology and reservoir simulation depth conversion 3D reservoir modeling 3D well design upscaling volume calculation plotting post processing streamline simulation ECLIPSE

Copyright Notice 1998-2010 Schlumberger. All rights reserved.

No part of this manual may be reproduced, stored in a retrieval system, or translated in any form or by any means, electronic or mechanical, including photocopying and recording, without the prior written permission of Schlumberger Information Solutions, 5599 San Felipe, Suite 1700, Houston, TX 77056-2722.

DisclaimerUse of this product is governed by the License Agreement. Schlumberger makes no warranties, express, implied, or statutory, with respect to the product described herein and disclaims without limitation any warranties of merchantability or fitness for a particular purpose. Schlumberger reserves the right to revise the information in this manual at any time without notice.

Trademark Information*Mark of Schlumberger. Certain other products and product names are trademarks or registered trademarks of their respective companies or organizations.

Authors

About this Guide:

The Interpreters Guide for seismic to well ties aims to give an interpreter understanding of seismic to well ties in the Petrel suite of software.

This guide provides the theory of seismic to well tying as well, what makes a good seismic to well ties and the workflows involved in creating such.

In addition this guide aims to provide questions that every interpreter should ask themselves when performing a seismic to well tie and workflows that can be used for synthetic creation.

However we cannot cover every possible scenario; the workflows in this guide are written to provide a guide for an interpreter and will not cover every scenario for seismic to well tying.

Ralph DaberSenior Product AnalystPetrel GeophysicsStavanger, Norway

Charles WagnerPrincipal GeophysicistReservoir Characterization

Houston, USA

Interpreters Guide to Seismic Attributes Table of Contents 5

Table of ContentsAbout Petrel* ...........................................................................2Copyright Notice .......................................................................3Disclaimer .................................................................................3Trademark Information .............................................................3

Module 1 - Background ............................................................9General Well Log Considerations .....................................11

Wavelet Extraction .................................................................13Statistical Wavelet Extraction ..........................................13Deterministic Wavelet Extraction - Seismic Well Tie ......16

Uses for Deterministic And Statistical Wavelets ..................17Velocities: Relating Depth Domain and Time Domain Data ..18Datums and Replacement Velocities .....................................19

Seismic Reference Datum ..........................................20Marine Seismic ...........................................................20Land Seismic (and VSPs) .............................................21

Sonic log calibration basics ...................................................23

Module 2 - Wavelets, seismic conventions and auxiliary logs ............................................................................25

Wavelets.................................................................................25Wavelet Descriptions .......................................................25

Polarity Conventions ...............................................................30Phase display convention .................................................30Phase Rotation ..................................................................31Seismic Polarity ................................................................32

Auxiliary Logs .........................................................................32Gamma (GR) ......................................................................32Spontaneous Potential (SP) ..............................................34Caliper (CALI) ....................................................................34

Washout response ......................................................35Density response ........................................................36

Overall display ..................................................................37

Module 3 - Seismic Well Tie (SWT) - workflow ................39SWT Workflow .......................................................................39

Sonic Calibration ..............................................................39Wavelet (2 options Statistical and Extracted) ..............40

Statistical (Wavelet Builder) .......................................40

6 Table of Contents Interpreters Guide to Seismic Attributes

Extracted (Wavelet Extraction) ...................................40Wavelet Viewer ................................................................42LogSets explained ............................................................42

Module 4 - SWT Synthetic workflows.................................45Generate accurate time-depth relationships ..................45Tie geologic markers to seismic horizons ........................48Understand the seismic response of lithologies and fluids at the well location ................................................51Understand the phase characteristics of theseismic data .....................................................................58Understand the frequency characteristics of the seismic data .....................................................................60

Notes on the Predictability Window ......................................63Predictability and Correlation ...........................................66How to interpret Predictability results .............................66

Where is the well (truly) located .......................... 66Well Head location .....................................................67Deviated well location ................................................67Selected Wavelet .......................................................68Where is the seismic data (truly) located...................68How good are the logs ................................................68How good is the wavelet ..........................................69How consistent is the geology ..................................69

Using Predictability to Optimize Well Location ................70Notes on Estimating an Initial Bulk Time Shift ......................72Notes on Stretching and Squeezing .......................................73

Validity ..............................................................................74Measured data .................................................................74Important points to consider ............................................75

Seismic Processing .....................................................75Phase ...........................................................................75Datum Management ...................................................76VSP ..............................................................................77

Miscellaneous well tie points ................................................77Velocity Log.......................................................................77

Well Tops ....................................................................77Auxiliary Logs ..............................................................77Cross correlation .........................................................78

Interpreters Guide to Seismic Attributes Table of Contents 7

Module 5 - Workflows ............................................................79Additional workflows to help an interpreter ..........................79

Using Seismic Well Tie in an exploration environment .79Using continuous output of the T/D Curve .......................82Using SWT in conjunction with other Petrel windows ....83Using auxiliary log tracks and posting horizons workflow ...........................................................................83Interactive use of wavelets for quick synthetic analysis .............................................................................84

References ...........................................................................85

Interpreters Guide to Seismic for Well Ties Background 9

Module 1 - Background

We regard the response to seismic signal energy (the wavelet, part F of Figure 1) of each interface in the lithologic log (part A of Figure 1) as generating a reflection pulse (part C) whose amplitude and polarity are determined by the acoustic contrast across that interface. The property of the lithology that defines the acoustic contrast is the acoustic impedance, which is the product of density and seismic velocity. The appropriate measure of the acoustic contrast is the reflection coefficient, which is the difference of the two acoustic impedances divided by their sum (Figure 2). The ability of the lithologic sequence (part A ) to generate a reflection is therefore shown by the reflection-coefficient log (parts B and E). The final synthetic (part D) is the sum of the individual reflection pulses, each of which is the product of the centered signal wavelet and the reflection coefficient..

Synthetic principles

10 Background Interpreters Guide to Seismic for Well Ties

Reflection coefficient definition

The preparation of a synthetic therefore has these basic stages:

the calculation of the velocity log (the reciprocal of the edited and calibrated acoustic log);

the calculation of the edited density log; the multiplication of the two to obtain the acoustic-impedance log; the calculation of the reflection coefficient log (or reflectivity series); the conversion of the reflection coefficient log from depth to two-way time; the convolution of the reflectivity series with a waveform representing the seismic pulse;

and the display of the resulting synthetic in juxtaposition with the seismic section through the

well, so that the match between synthetic and surface seismic can be confirmed, and so that each reflection can be positively tied to an interface observed in the well.


General Well Log ConsiderationsGood well logs are a crucial element of the well-seismic tie and wavelet extraction process. Whenever possible, logs should be prepared and edited by a petrophysics professional who can assure that the appropriate environmental corrections and editing procedures are applied uniformly on all the wells used in a study.

Bad logs can cause impedance computations and/or trends to be wrong, which will affect reflectivity strength. Washouts can cause reflection interfaces to appear to be located incorrectly, or worse, to be entirely absent

Some common conditions that can cause logs to contribute to incorrect and/or misleading well-seismic ties include:

Washout: Defined in the Schlumberger Oilfield Glossary as: An enlarged region of a wellbore. A washout in an open hole section is larger than the original hole size or size of the drill bit. Washout enlargement can be caused by excessive bit jet velocity, soft or unconsolidated formations, in-situ rock stresses, mechanical damage by BHA components, chemical attack and swelling or weakening of shale as it contacts fresh water. Generally speaking, washouts become more severe with time. Appropriate mud types, mud additives and increased mud density can minimize washouts.

The caliper tool is useful to identify washout intervals. Care must be taken to ensure use of the correct caliper (that which was recorded on the same tool string as the logs under study), as borehole conditions often vary over time and caliper readings can change. In washout intervals, sonic logs are typically more reliable (see Cycle Skip, below) than density logs because the sampling is made over a larger volume and penetrates further into the rock The density is a shallow reading tool and is very sensitive to borehole irregularities. It reports anomalously low values when washouts cause mud or mudcake to be incorrectly incorporated into the density reading.

Cycle skip: Occurs when sonic signal arrival is delayed due to attenuation by borehole fluids, and is typically associated with more severe washouts or gas in drilling mud. In this case, the transmit/receive cycle time is shorter than sonic pulse travel time and erroneously long travel times are reported. This results in short sharp spikes of lower than expected slowness.


Cable stretch/rebound: Stuck tools that become unstuck can cause a rubber band phenomenon that affects any tool using a timed sensing gate, or that requires a constant tool velocity, for an accurate reading. This effect can be mitigated with a stretch or squeeze of the logs in cases where correct interval thicknesses can be determined from other properly recorded logs. The readings over such modified intervals may t hen be better positioned but are still not rigorously correct.

Bad clock calibration: When identified, a stretch or squeeze over the whole log may mitigate or eliminate this problem.

Tool out of calibration: Often related to temperature or pressure issues, a tool that drifts out of calibration can result in unreasonable or misleading computations and depth trends. Consistency between wells in the same formation is a good QC for identifying logs with incorrect calibration baselines. Log calibration tails and logger remarks can also provide insight into non-geologic causes of odd tool response.

Derived density and/or sonic log(s): There are several methods in common use within the industry that can estimate a missing sonic or density log from the response of some other log(s) acquired over (or adjacent to) the interval of interest (Gardner, Faust, etc). Each method has assumptions that restrict the validity of the method, so any such pseudo-log generation should be performed by someone who is aware of the limits of the method and the intended use of the pseudo-log.

Mixed tool technology: Well log technology has changed dramatically over time. When working on fields with wells of varying age, one must keep in mind that tool responses may need to be normalized or re-calibrated before the logs can be used together.

Fluid Substitution: Well logs respond to the materials in close proximity to the borehole. Most often, borehole fluids have invaded permeable rock, and well logs are affected accordingly. Better ties to seismic data will be achieved if fluid substitution computations are applied to the well logs (sonic and density) so that the responses match the rock and fluid properties measured by the seismic method.


Wavelet ExtractionWavelet Extraction provides the capability to extract an embedded wavelet from the seismic data. The extracted wavelet can be convolved with the RC series to create a synthetic seismogram to provide a vital link between the borehole and the seismic data.

Statistical Wavelet ExtractionThe statistical extraction method assumes that the embedded wavelet is the same as the truncated autocorrelation of the seismic trace. The average autocorrelation from many traces is used to provide a more representative estimate.

It is possible to access statistical extraction even if no sonic log exists for the borehole. The statistical method transforms the autocorrelation of all the input seismic traces into the frequency domain, averages the spectra, and then inversely transforms the averaged spectra into the phase specified. Such wavelets are zero-phase by definition.

A statistical wavelet can approximate the correlated source signature used during Vibroseis (non-minimum phase) seismic acquisition. It is particularly useful where the quality of seismic data is high, and the quality of the RC series is low.

Theory:

A seismic trace can be represented by the convolutional model:

)()()()( tntrtwts +=

Where:

s(t): observed seismic trace

w(t): wavelet we are trying to find

r(t): reflection coefficient series

n(t): noise assuming noise does not correlate and therefore is additive.


This equation can be written in terms of power:

)()().()( 2222 fNfRfWfS +=

For a single trace, all elements that compose the seismic trace are present: wavelet, reflectivity and the noise.

By averaging many traces:

)()().()(2222

fNfRfWfS +=

It is a reasonable assumption that, over a sufficiently large sum, the following may occur

)(2

fR : Roughly a white spectrum due to the random occurrence of r(t) in space.

)(2

fN : Average noise spectrum. In good areas, noise will have a minor additive effect on the average and can be ignored

Given these condition )(2

fS is a good representation of the average seismic wavelet )(2

fW .


How it works:

The autocorrelation s(t) of the truncated seismic trace s(t) is computed using FFT, squaring the amplitude spectrum, zeroing the phase spectrum and inverse transforming.

0% means no taper A value of 60% leaves the center 60% of the autocorrelation window with a weight of 1

and applies the selected taper over the outer 40% of the window (Statistical default) A value of 100% (deterministic default) results in the tapers starting at 0 and being

applied to the entire correlation window.

Windowed autocorrelation function *s(t) is transformed to the frequency domain. Repeated for all the traces. The extracted wavelet is then computed as the inverse FFT of the square root of the spectral average. Optional user defined wavelet phase is then applied to the extracted wavelet.


Deterministic Wavelet Extraction - Seismic Well TieDeterministic wavelet extraction derives the wavelet from the seismic trace but utilizes the RC data to produce the wavelet. The process derives time shift and phase in two separate steps utilizing envelope cross-correlation methods. The optimal seismic trace and time position (lag) are automatically selected.

1. Theory

Wavelet can be determined in the frequency domain as:

NfCrr

fCrsfW

+=

)(

)()(

W(f): waveletCrs(f): cross correlation of the reflectivity and the seismic trace in the frequency domainCrr(f): autocorrelation of the reflectivityN: White noise factor

To identify the best estimate, we define the Signal to Noise (S/N) as the power in the smooth synthetic over the power in the residual of the match with the seismic data:

P

PN

S

=1

)().(

)().(

fCssfCrr

fCsrfCrsP =

P energy ratio called predictability

2. How it works a brief summary of the implementation1. Truncate well log reflectivity log to the specified time range. Compute zero lag

autocorrelation of the reflectivity envelope.2. Select seismic trace and compute zero lag autocorrelation of reflectivity (amplitude)

envelope.3. Cross-correlate lag slipped envelope autocorrelations, seeking maximum cross-

correlation to determine optimal bulk time shift between log and seismic..4. Following time shift, rotate wavelet phase to maximize cross-correlation, resulting in

determination of optimal phase.


Uses for Deterministic And Statistical WaveletsA simple 1D synthetic seismogram is computed by convolving a reflectivity series with a wavelet. A wavelet is defined by amplitude and phase spectra in the frequency domain. A deterministic wavelet extracted using both seismic and log data contains information on the frequency content and phase characteristics of the seismic. It is our link from the reflectivity series to a synthetic that matches the seismic. Wavelet estimation is something of an art form. It requires careful thought and a good strategy driven by the availability and quality of the data.

1. Analytical wavelets are simply standard model wavelets; the user supplies a few basic parameters such as the central frequency and the phase. The amplitude spectrum is pre-defined and the phase set by the user

2. Statistical wavelet techniques derives the amplitude spectrum from the seismic trace. The user has to specify the phase and length of the wavelet.

Deterministic wavelet methods derive both amplitude and phase spectra AND wavelet timing information by cross-correlating the reflectivity series derived from sonic and density logs with the seismic data. This technique should in theory generate a wavelet that produces a synthetic closely matching the seismic data.

This is often not the case, as there are many reasons why a synthetic match is difficult to obtain. These include:

Poorly edited log data Missing or insufficient log data (logging run too short to be useful for synthetics) Inherently noisy seismic data (acquisition noise, gas effect, diffractions from fault planes,

scattering of energy) Poorly imaged or processed seismic data (velocities and/or multiples) Missing or inadequate depth-time control points

The preferred approach is to extract deterministic wavelets wherever possible. The aim is to extract a series of wavelets with reasonably consistent amplitude and phase spectra.

Achieving this serves three main purposes:1. It provides an understanding of the true phase of the seismic data. This is crucial

information for an interpreter. 2. It provides a series of wavelets that can be used on other wells to generate synthetics.

Common practice would involve averaging of wavelets.3. The averaged wavelet can be used to generate an inverse filter to apply to the seismic

data in order to zero phase it. Zero phase seismic data is a requirement for processes like seismic inversion and AVO analysis.


Once the phase is understood, statistical or analytical wavelets may be sufficient to produce good synthetic ties on other wells where deterministic wavelets cannot be extracted.

Velocities: Relating Depth Domain and Time Domain DataBecause surface seismic data is acquired using both the downgoing and upgoing wave paths (the sources and receivers are at the surface), the convention is to use Two Way Time (TWT).

Assuming:1. a strictly vertical travel path2. time domain is Two Way Time (TWT)3. depth domain is True Vertical Depth (TVD4. homogeneous rock with vertical compressional velocity: 5000m/sec (roughly 16404 ft/s),

equal to a slowness of 0.0002 sec/m (roughly 60.9 us/ft)5. and seismic reference datum elevation (the 0.0 time reference point) is 0.0 m MSL

then, the depth equivalent of 1 sec (1000ms) TWT is exactly 2500m (16404 ft) TVD.

VSP and checkshot surveys use surface sources, but the receivers are beneath the ground surface at some point in the borehole. In this case, we measure One Way Time (OWT). Using the same velocity, a TWT of 1 sec would be the equivalent of 5000m (32808 ft) TVD.

The typical frequency bandwidth for seismic acquisition is 4-150 Hz. Sonic logs, however, are typically recorded at 7000Hz and higher. Velocity in rock materials is often frequency dependent, and sonic log velocities tend to be higher (the slowness is lower) than seismic velocities over the same interval.

This velocity disparity results in inconsistencies between time/depth determinations performed by integrating log travel times (slowness) acquired over constant time intervals, and the time/depths determined by integrating the inverse of seismic velocity (also, slowness).

The desire to align finely sampled well log data in the depth domain with features in time domain seismic data brings us to the requirement for sonic log calibration, for it is here that sonic log slowness is brought into agreement with seismic acquisition velocities.


Datums and Replacement VelocitiesThe process of bringing sonic log velocities into agreement with seismic velocities does not, by itself, guarantee a correct depth-time relationship. The logs must also be placed in the correct position relative to the seismic data prior to integration. This is established using datums and replacement velocities.

In the context of sonic log calibration and well-seismic tie, a datum is defined as the elevation at which some property is set to (typically) 0.0, in either the depth or time domain. The location the datum is either coaxial with, or directly above, the wellbore trajectory. In most cases, a datum consists of a single scalar value. However, in some cases, a datum may be derived from a surface with either a constant or variable elevation, which can be important when dealing with deviated wells (such as ground surface or water bottom elevation).

Replacement velocities, when utilized, cause static (vertical) shifts to seismic and other time domain data (checkshots, VSPs, etc.) that bring a specific point in time (or some time domain surface) into alignment with some common elevation feature, which may be real (ground level) or arbitrary (3000m above MSL). These velocities may be constant (air -ground surface replacement velocity) or space-variant (ground surface -base weathering layer replacement velocity).

Datum errors are a common cause of confusion when working with well and seismic data. A clear understanding of the datums is required to ensure that all necessary shifts are applied and that all data objects are properly aligned. Alignment errors can result in misties or erroneous phase estimations. This is particularly true when working with VSP data on land. The VSP acquisition and processing reports are crucial to understanding misties and inconsistencies.

As described in more detail below, a number of vertical shifts and artificial datums are used when processing seismic and VSP data, and if the data is not put back in the right place, a residual (and often difficult to identify) shift will be embedded in the data.


Seismic Reference DatumInitial seismic acquisition and processing takes place in the time domain, as well as many post-processing analytical activities such as wavelet based inversion. A convention in the seismic industry is that processed seismic records begin at a time of 0.0 sec. This 0.0 time reference point has to be associated with an elevation reference point, which is called the Seismic Reference Datum (SRD). In almost all marine and many land cases, SRD is set at a constant value a flat surface. However, there are cases where SRD is a non-flat surface with different values at each seismic trace (see Land Seismic, below).

Marine SeismicIn the marine seismic case, this 0.0 time is typically adjusted to mean sea level (MSL), and SRD is most often set at 0.0 feet/meters above MSL. During processing, however, data is often realigned (static shifts are applied) to alternative datums such as water bottom or gun depth. If marine data appears to be 25 feet shallower than it should be, go ask the processors if they remembered to move back to MSL after processing at gun depth. If they did not, the data may be zero datumed at 25 ft below MSL. Similarly, data acquired in land-marine transition areas, or very shallow water, may have been processed using datums related to tidal fluctuations during acquisition.

Some other common datum related causes of alignment ambiguity between seismic and well data for marine cases include:

1. Use of a constant (or linear trend) water column velocity (not valid in deepwater Gulf of

Mexico, proximal offshore Brazil, and other areas where water column velocity is affected by salinity changes)

2. Inconsistencies between log and seismic shallow velocity profiles. Often, no logs are available for a significant distance below mudline. Linear trends are often substituted and are often incorrect and typically too fast.

3. Assumption that water velocity represents the lowest possible velocity. For example, the presence of biogenic gas in near-shore shallow deltaic or marsh sediments can result in velocities significantly lower than water velocity.


Land Seismic (and VSPs)The land seismic case can be more complex because of surface topography and near surface phenomena that are not present in the marine case, or more severe in the land case. In many areas, the irregular thickness and rapid lateral variation of near surface layers, water table anomalies, karsting, etc., cause velocity anomalies that cannot be adequately characterized in the seismic velocity model.

As a result, time shifts (static shifts) are often present from one trace to the next, causing timing perturbations that affect deep reflectors. It is common practice to try to align the traces by making minor up/down shifts so that a reflector that should appear continuous and smooth does take on this appearance. Such shifts shift some traces such that the first sample(s) are moved up above the 0.0 time point, into negative time.

Similar, larger, shifts occur when data is shot beneath a water body (lake, river, etc) or other features with velocities that depart significantly from the seismic velocity model.

Now, some geophysicists (and some software systems) do not really like the idea of negative time very much, and the solution is to move SRD quite far up above the point that any trace might reasonably be expected to be shifted. Often, this puts SRD somewhere up in the sky, especially in mountainous areas where surveys are shot over large changes in elevation. In West Texas, SRD is often set at 3000 ft MSL, although ground level is often closer to 2300 ft MSL.

When this is the case, the thickness of the air gap between SRD and ground surface is used to compute a replacement velocity such that the first live sample on the seismic trace will be located


at the surface. Any subsequent static shifts will be small enough that the data remains in positive time.

A similar situation occurs with what is called the weathering layer, which is the region between the ground surface and the portion of the seismic that the geophysicist can understand. The time thickness of weathering layer velocity anomalies are estimated and weathering statics are applied that compensate for the associated timing shifts. These weathering static shifts must be taken into account when trying to align land seismic data with VSP traces that typically do not have weathering statics applied.

In most cases, land VSP and checkshot data is datumed (e.g. 0.0 time set) at the ground surface. In areas with elevations relatively close to sea level, such data may actually be datumed beneath the ground surface at MSL, or in the case of 3rd party contractor data libraries datumed at some arbitrary subsurface elevation; again, in West Texas this is 1000 ft MSL..

This means that 0.0 time for VSPs and checkshots is often at a different elevation than 0.0 time for the seismic data. In cases where this datum elevation is below ground level, a VSP/checkshot replacement velocity is often computed and applied so that 0.0 time is shifted to ground elevation but the data is still positioned correctly in the subsurface.


Sonic log calibration basicsThe primary function of a checkshot velocity survey is to establish a set of measured depth-time tie points along a wellbore. These points are then used to constrain sonic log velocities so that the integrated sonic log depth-time relationship is comparable to measurements at seismic frequencies.

Checkshot information is recorded in the field using a geophone clamped at a series of target levels in the well and a surface seismic source. Time-depth pairs equivalent to checkshot points are also derived from the first break travel times of VSP surveys.

The sonic log is recorded in the wellbore, typically at a frequency of 5-15Khz. The sonic log can be integrated to give a cumulative travel time, but over thousands of feet small errors at individual sample levels can accumulate and produce more substantial errors.

The sonic log may be acquired in several runs, sections of it may be influenced by casing, it may be prone to cycle skipping and so on; all these complications lead to cumulative errors in the uncalibrated sonic log. In addition the logging tools operates at frequencies of 5-15Khz, three orders of magnitude higher than typical seismic data. Velocity measured in a rock at 30 Hz will be slightly different from that measured at 10Khz. This frequency dependence is termed the dispersion effect, and needs to be accounted for.

Finally, the effects of anisotropy need to be considered since it is well known that vertical and horizontal velocities are commonly 5-15% different when shales or clays are present. As a well becomes progressively more deviated the sonic log will see a larger and larger horizontal component of the velocity field. We require vertical velocities for seismic interpretation and depth conversion purposes.

Checkshot data is corrected to vertical datumed travel time and can therefore be used to adjust the sonic log recorded in a deviated well. One must be aware that anisotropy can affect checkshot results if the source is not directly over the downhole sensor.

Calibration is done by slightly increasing or decreasing the sonic slowness values over sections of the log until the integrated sonic log travel times match those derived from the checkshot survey. The mechanism used to control this process is called the Drift Curve. Drift can be computed at every depth level and is defined as: Drift = checkshot time - integrated sonic time (Tcheckshot - Tlog) If the drift curve is positive the checkshot travel time is longer than the integrated sonic travel time, meaning that the sonic log is too fast. If the drift curve is negative the sonic log is too slow.

The aim of calibration is to define a drift curve through the checkshot data that will correct the sonic log to the known travel time values derived from the checkshot data. The result is a


calibrated sonic log. There are two ways of applying the necessary corrections.

The differential correction method uses multiplication to shift the sonic log to higher velocities. The correction is applied as a velocity dependent percentage so that larger corrections are applied to the lowest velocity sections. This is based on the assumption that low velocity sections of the log are prevalent in poor borehole conditions and contribute heavily to transit time errors.

The linear correction method uses simple addition to shift the sonic log to lower velocities. This is based on the assumption that high velocity sections of the log will be in better borehole conditions and contribute relatively little to the overall transit time errors.

Interpreters Guide to Seismic for Well Ties Wavelets, seismic conventions and auxiliary logs 25

Module 2 - Wavelets, seismic conventions and auxiliary logs

WaveletsWavelets used for SWT in Petrel include 3 analytical types:

1. Ricker 2. Ormsby 3. Tapered Sinc

Other wavelets that can be used in generating synthetics;4. Butterworth 5. Klauder

Wavelet DescriptionsThe most important decision for a user of Petrel is whether to use a minimum or zero phase wavelet. Dynamite and airgun sources produce impulse (minimum phase) source signatures, and the resultant data may not necessarily be reprocessed to zero phase. Be careful, as the convention for a zero phase wavelet for the USA and Europe are completely opposite in phase (see figure below).

USA Zero Phase Wavelet European Zero Phase Wavelet

26 Wavelets, seismic conventions and auxiliary logs Interpreters Guide to Seismic for Well Ties

In addition, some of the other wavelets used by other software products include Hanning, Hamming and reverberation wavelets.

Petrel shows the phase spectrum in the form of Radians, whereas other software products may show this spectrum in the form of degrees.

The characteristics of each are as follows;

Ricker A Ricker filter is a filter that requires only 1 input, the peak frequency as seen in the Petrel screenshot below. This is commonly used for synthetic modeling. No bandpass filter is involved and the frequency and phase spectrums are purely a function of the peak frequency input.

In the below example, a 40Hz peak frequency was chosen.

With a Ricker wavelet no side-lobes are seen in the amplitude vs. time spectrum, whereas the Butterworth, Ormsby and Klauder and tapered sinc filters all have associated side-lobes.

Ormsby The bandpass (see figure below) of an Ormsby filter can be described by using up to 4 corner frequencies as in the figure below. In Petrel;

1. is the low cut frequency, where all lower frequencies are filtered out and not used: 2. is the low pass frequency where after this frequency, 100% of all higher frequencies will

be used:


3. is the high pass frequency where frequencies higher than this will be linearly tapered until point 4.

4. is the high cut frequency where any frequencies higher than this will be filtered out and not used.

Figure showing the bandpass for an Ormsby Filter.


Tapered Sinc - A tapered sinc filter is defined by a low and a high cutoff similar to a Butterworth filter but applies no further filters whereas a Butterworth filter applies two slopes (see Butterworth description)

Butterworth The Butterworth bandpass (see following figure) consists of 2 cutoff frequencies taken at 3dB down from maximum power or approx. half power (~50% on the amplitude scale on the figure below). For the example below they are at 10Hz and 50Hz. The Butterworth filter also requires 2 slopes. The slopes are defined as decibels/octave, where an octave is a doubling in the frequency (i.e. 10 to 20Hz). A decibel is a unit of measure for acoustics defined by the formula;

dB = 20log(X/Y)

where X and Y is the ratio of amplitudes.

If the ratio of X/Y is ratio of 2 then it will be 6dB,

and if the ratio is 10 that translates to 20dB.


Figure showing the bandpass of a Butterworth filter

Klauder An analytical approximation of the Klauder wavelet computed via autocorrelation of an actual Vibroseis sweep signal. A Klauder is defined by 2 frequency cutoff values, a low and a high cutoff which in the example below are set at 10Hz and 70Hz. The contributing frequencies are represented by a box-car that assigns the same constant amplitude for all frequency components. Because of sudden discontinuities in the amplitude of frequencies at the beginning and end, the wavelet has some undesirable side-lobe oscillations


Polarity ConventionsPhase display convention

The phase convention for wavelets discussed previously implies that for an impedance increase (hard kick) the display convention is shown to below..

And subsequently the convention for a high impedance layer is shown below with examples of geological events for checking polarity.


Phase RotationThe rotation of a zero phase wavelet is shown below for European polarity.


Seismic PolarityHistorically in the North Sea for zero-phase data a hard event (hard kick) or impedance increase, such as the top chalk, is represented as a trough.

The opposite polarity convention is used in the USA.

By the SEG standards defined for a zero-phase wavelet the North Sea convention is reverse standard polarity.

The best composite definition for the North Sea is something like positive reflection coefficient is represented by a negative number on tape and as a white trough in display.

In the following figure, red represents negative values blue represents positive values.

Auxiliary LogsGamma (GR)Gamma logs are designed to detect the natural gamma radiation of the materials within a short distance of the borehole wall. Some count total gamma radiation per unit time and others have sophisticated spectroscopy capability and can report on concentrations of individual radiating elements per unit time, such as specific ions of potassium, uranium, and thorium.

Gamma logs are often assumed to be indicators of clay and/or shale content. The most common sedimentary rock framework elements (silicon, calcium, carbon) do not emit meaningful quantities of gamma radiation, whereas clay minerals often contain potassium or other gamma emitters.

Lets investigate the GR (Gamma log) in the below example. In the 4th track from the left, it shows a decreasing then increasing response just above marker horizon 3 indicating a shale to sand


lithology transition. In this case a thin bed of sand which also shows a SP response and a RHOB /NPHI crossover. As an initial inspective interpretation this could indicate a fluid filled thin sand, but as the resistivity response is very small, most likely water filled with a small portion of hydrocarbons.

Gamma based shale volume estimates may be misleading in continental clastic sediments because of the possible presence of gamma emitting potassium feldspars and zeolites (and their weathering products). The arkosic Granite Wash sands in Oklahoma are a classic example.

Gamma logs are often used for formation and/or bed boundary determination. Marker tops are often picked using only gamma logs. When trying to tie seismic reflectors to marker tops, one must keep in mind that acoustic impedance boundaries (identified from sonic & density logs) may not fall at the same location as boundaries selected based on gamma response.

Interpreting gamma logs can become complex in circumstances where neutron-emitting tools (such as pulsed neutron tools) have been run repeatedly over the same interval prior to the gamma logging run, as these tools can warm up metal ions of iron, aluminum, manganese, sodium, and potassium which can then emit induced gamma radiation.

Shales can absorb liquid from drilling fluids, causing rocks to soften and clay minerals to swell, which can lead to borehole breakout beyond that already induced by the physical drilling process. As a result, it is common to see calipers indicating washouts over intervals that also have elevated gamma readings.

Gamma log response must be corrected for borehole diameter and mud weight before well-to-well correlation or quantitative analysis.


Spontaneous Potential (SP)SP tools measure the electrostatic potential that develops via selective ionic diffusion when the electrolyte content of the mud filtrate is different from that of the formation water. SP log deflections are typically interpreted as indicators of permeability and as a result are often used as sand/shale indicators. Sand counts and bed boundary determinations are often performed using SP logs.

In the example below, the SP curve, in black, in the third track from the left is seen to have an easily discernable response around the marker horizon 1. In the second track from the left a RHOB / NPHI crossover response at the same depth indicates a changing of lithologies between shale and sand. The SP event and the RHOB / NPHI crossover coincide in depth and are both powerful tools when working with auxiliary logs.

One must keep in mind that SP log responses over any particular interval may not be comparable if the drilling fluid was changed between logging runs. In addition, deflection response in sands can be damped if there is dispersed clay in the sand body, or if the electrical characteristics of the formation fluids are closer to that of the mud filtrate than in another otherwise equivalent sand body.

Caliper (CALI)Single arm caliper logs measure the distance from the borehole wall to the side of the tool string using a spring loaded lever arm that presses against the borehole wall. measurements are corrected for tool diameter and standoff (when the tool is not centralized) to report borehole diameter.

Multi-arm calipers use similar technology and report diameter across several azimuths. Acoustic calipers exists and are used in special circumstances; they operate on a principle similar to sonar and have significant limitations.


One must be aware of the fact that two conditions can lead to a featureless caliper reading: a very smooth borehole wall, and, a bore wall that is so badly washed out that the caliper is at maximum extension and cannot accurately measure borehole diameter. For this reason, it is good to plot the bit size as well as the caliper log.

One should also be aware that tools typically rotate when pulled uphole, and that boreholes are often somewhat elliptical. As a result, single arm caliper readings are not always definitive with respect to borehole diameter.

In addition to geologic causes, washouts often occur at bit size change points and casing string landing points - due to increased drilling fluid flow and tool impact. This can introduce undesired features in sonic and density logs, and the caliper is often the best indicator of problems.

The example below illustrates the importance of auxiliary logs in determination of whether reflection coefficients are the result of a true impedance contrast or if the contrast under study is due to borehole washout.

Washout responseThis example shows the reflection coefficient response due to washout which can be clearly seen from the Caliper log.


Density responseThis example of a reflection coefficient response can be seen to be a true response as there is no washout indicated shown on the caliper log.


Overall displayThe overall display, with auxiliary logs, can provide the interpreter many pieces of additional information. From the following example, one can identify lithologic changes from the gamma log, washout from the caliper and velocity abnormalities from the interval velocity log - in addition to the residual drift.

Interpreters Guide to Seismic for Well Ties Seismic Well Tie Workflow 39

Module 3 - Seismic Well Tie (SWT) - workflow

SWT WorkflowAn overall Seismic Well Tie workflow is described below to give an idea on the use of the process within Petrel. The workflow utilizes two software components:

1. Sonic Calibration2. Wavelet Extraction

The below workflow is a basic outline for generating a Synthetic and is not meant to represent all variations that are possible but represent a viable way of generating a well to seismic tie via a resultant synthetic seismogram.

There are four basic elements of the workflow:1. Use sonic calibration to calibrate the wells to the seismic and save a time/depth

relationship.2. Use wavelet extraction to develop a deterministic wavelet from the seismic and logs

using the initial time depth relationship from sonic calibration3. Use wavelet extraction to convolve wavelet and RC and compare to the seismic.4. Perform any required depth-time relationship adjustments to match the synthetic to the

seismic (bulk shift, stretch & squeeze) and output a final time/depth relationship as well as a synthetic.

Sonic Calibration1. Add or import well data into project (if it doesnt exist). Either via loading or RPT import.

Key log information required for sonic calibration is a sonic curve (most commonly DT) and a depth-time relationship (commonly Checkshot data). If no checkshot is available, the sonic log can be integrated to establish a depth-time profile.

2. Create a Log set for the well by RMB clicking on the Well Data item and choosing the Insert new Log set option. Open the settings for the Log set and select the log in the bottom of the window and use the add option to add it to the logset - at a minimum, a sonic and density curve. Or RMB on a log in the Petrel Input pane and add it to a logset.

40 Seismic Well Tie Workflow Interpreters Guide to Seismic for Well Ties s

3. Sonic Calibration: Open the Seismic Well tie > Sonic Calibration Window and drag and drop (or select the checkbox of) a Log set that contains at least a sonic and density curve from the Petrel tree into the window. If a checkshot is present in that well, then seismic well tie automatically picks up the checkshot and uses it. A selection dialog will open up if more than one checkshot is present. If no checkshot is located in the well, you can drag and drop in a checkshot from another well.

4. Calibrate the sonic curve. Use the variables and settings GUI or interactively edit the curve to your liking by using MB1 to insert or delete knees along the TD relationship, and by moving the knees using MB1. Save the Calibrated Sonic and Time/depth relationship using the Output tab of the variables and settings GUI. Here, it is possible to save the log and the time depth relationship. At any time, the variables and settings window can be brought up by using MB3 > Variables and Settings in the Sonic Calibration window.

Remember to set this saved time/depth relationship to be the default one for the well in the Petrel input tree.

Wavelet (2 options Statistical and Extracted)Statistical (Wavelet Builder)

1. Open the Wavelet Builder module and enter the required wavelet information, (Name, Wavelet Type, Length, sample rate, frequency, high/low cut filters). Click Compute and then save the wavelet. This can then be dropped in to the Wavelet extraction process as a pilot wavelet.

Extracted (Wavelet Extraction)1. Open the Wavelet Extraction GUI and drag your seismic cube into the window. Drop in

a Log set that contains a sonic and a density curve.

If a density log is not present in the logset a warning will pop up. A density curve is needed to generate the AI curve which in turn generated the RC curve. Seismic well tie generates these on the fly. At this time, seismic well tie cannot use an existing RC curve as input.

An AI curve can be used if present, in which case a sonic and density do NOT need to be present in the Logset.

Importantly, seismic well tie can use any log curve as a pseudo-AI curve for apparent reflectivity computation. Gamma logs often work well in this circumstance., Sonic logs also work well, but the polarity is typically reversed from that derived from true acoustic impedance.


2. In the Wavelet Extraction variables and settings dialog box, select the Position tab and enter the extraction position (by default, this is automatically set to the well location). If the well has a deviated trajectory, select the Interpolate Seismic Data to Follow Trajectory option.

3. Select the Method tab in the Wavelet Extraction variables and settings dialog box and edit the parameters

4. Select the Extract tab in the Wavelet Extraction variables and settings dialog box and edit the parameters. Click Extract to extract a wavelet. 2 new windows will appear, a Predictability window and an Extracted wavelet display window.

5. The Predictability window and the Extracted wavelet display window are interactive so clicking on predictability with MB1 will change the extracted wavelet display and the main wavelet extraction window.

6. Using the Tools options in the Extracted Wavelet display window edit the wavelet. Output the wavelet once it is edited via the Output tab.

Seismic Well tie wavelets are not saved unless an Output or Save action is performed.

7. Use the Generate Synthetic option in the Extracted wavelet display window.

8. Optional: MB3 on the synthetic in the wavelet extraction window and use the Synthetic along well attributes option to edit your display.

9. In the Output tab of the Wavelet Extraction variables and settings window, save the synthetic by using the Output Synthetic and save the time/depth relationship.

10. A new synthetic log set should appear in the Petrel Input pane under the well.

Other tools that can be used in this workflow:


Wavelet ViewerThe Wavelet Viewer is a viewing tool and is only able to be used as such. To view wavelets in this window, drag and drop the Petrel wavelet objects into the viewer window or toggle them on and off in the Petrel tree with the Wavelet Viewer window open.

LogSets explainedA Logset is an object that contains one or more logs from a well. The purpose of a logset is to act as a bounding object wherein logs which have different sample rates can be combined into a common sampling framework.

Any log added to a logset is resampled so a sample depth exists for it compared to every other log. This process does not shift any data but honors all data in all logs.

Petrel logs are not resampled by this process, the resampling is contained only in the logset object.

Using RMB on the well in the Petrel input tree a logset can be created using the

option.

Once a logset has been created, each log in that well can be added to be part of that logset (i.e. by highlighting the log it can be added to logset) and using RMB it can be added to either insert a new logset, or set the currently selected log (in this case GR) to be part of an existing logset.

A log can only ever be part of 1 logset

In the case where a check mark exists next to one of the logsets, this denotes that the log is already part of that logset. If it is desired to move this log from being part of the current logset to another, the new logset can be selected.

Using RMB on the logset object will bring up 4 options.


Opening the Settings will allow the user to edit the content of the logset.

In the screenshot above, there are 3 logs contained in the logset (as indicated in the top part of the dialog box). All other logs available under the well are listed in the bottom dialog. The user can add a log to the logset by highlighting it in the bottom part of the window, then using the

option to add it to the logset. To remove a log from the logset use the option after highlighting the log in the top dialog.

This only removes the log as being part of the logset. It does NOT delete the log from the well or Petrel project.

The option will only delete the Logset from the Petrel project. No individual logs will be deleted from the well.


BE CAREFUL, the will delete the logset and ALL logs that are contained in it from the well and Petrel project. The user will be asked to confirm this action before proceeding as indicated by a warning message.

Lastly, the option allows the user to get a spreadsheet outline of all values for the logs contained in that logset. In this case, the logset contained a DT and RHOB curve.

Interpreters Guide to Seismic for Well Ties SWT Synthetic Workflows 45

Module 4 - SWT Synthetic workflows

In Module 1, four main uses for seismic to well ties (synthetics) were identified.

Each workflow has a different final aim, although often the method of achieving that aim can overlap.

Each of the 4 uses is outlined below while working in the Seismic Well Tie process. Tie geologic markers to seismic horizons Generate accurate time-depth relationships Understand the seismic response of lithologies and fluids at well location Understand the phase characteristics of the seismic data

Generate accurate time-depth relationships Calibrating the sonic log is part of the synthetic modeling process. Calibrating a sonic log corrects the log velocities to checkshot data and accurately hangs the sonic log in time. A time-depth relationship can be generated from the calibrated sonic log and should be used as the preferred time-depth relationship for the well

1. Create a logset with a DT and display it in the sonic calibration window. The logset is described in Module 4 and the contents of an example LogSet is shown

below by opening the settings for the LogSet.

46 SWT Synthetic Workflows Interpreters Guide to Seismic for Well Ties

The sonic curve (DT) is shown in the left track while the checkshot is in the right track. Matching the checkshot to the drift (blue line in the right track) will calibrate the sonic log to the checkshot.

2. Create knee points either freehand by using the Add knees functionality or automatically using the Create knees at checkshots.

The aim is to match the drift to the checkshot and as a result create a new more accurate time to depth relationship for the well.


The options for adding and editing drift points are found in the Knee picking tab of the Variables and settings dialog shown to the right.

3. Once drift curve editing has been completed, the original DYT and calibrated sonic curve will likely be different as can be seen in the figure below.

The result of the calibration creates a new time depth relationship.

4. The new time depth relationship is saved in the Output tab and then set as the new time depth relationship for the well.


When saving the time depth relationship there are options to set the sample interval of the time depth relationship and for continuous output to be set (see Module 6 for a description of using continuous output)

Tie geologic markers to seismic horizons A successful synthetic match clearly establishes the relationship between a change visible on the log data (at which a marker can be interpreted) and the response to that change on the seismic data. Thus the interpreter can clearly establish what the events on the seismic data represent in geological terms.

As a precursor to matching geological markers to seismic horizons, an accurate time depth relationship is needed. The first workflow in this section describes how to create an accurate time depth relationship.

Using the calibrated sonic and a density curve, an acoustic impedance (AI) curve is generated. If no density curve exists then Gardners equation can be applied to create a derived AI log. Also, if an acquired AI log exists, this can be used.

From the acoustic impedance curve a reflection coefficient curve is calculated by SWT.

The layout below shows all these components and also some auxiliary logs (GR and CALI) in the second track from the right. Auxiliary logs are very important to provide additional information to the user to understand the geology and the log responses (for a detailed description about auxiliary logs refer to Module 2 and the next workflow in this section).


The reflection coefficient (RC) log shows the major acoustic impedance contrasts that exist (see Module 1 for a description of acoustic impedance).

Geological tops in the above layout show where a geologist has identified a lithology boundary of interest based on logs.

Matching tops to a seismic volume or line allows us to extend the geological boundary away from the well bore.

To do this, the creation of a synthetic seismogram is necessary by convolving a seismic wavelet with the RC response generated from the log data.

Generating the wavelet for the convolution can be analytical or deterministic (see Module 1 for a description of analytical and deterministic wavelets). The process of generating these wavelet and a description of each type of wavelet can be found in Modules 2 and 4.

In summary; Using an analytical wavelet can be a good way to see what the seismic should look like if

the seismic was created using a wavelet of defined frequency, sample rate and length (i.e. a zero phase wavelet) or in area where no seismic may exist to extract a wavelet from.


Using a deterministic wavelet requires the extraction of the wavelet from the seismic volume. A deterministic wavelet represents the true wavelet from the seismic rather than an idealized numerical wavelet.

Setting the area of extraction to extract a wavelet from is defined by an interval of interest (i.e., a reservoir interval). Keeping this interval defined is important as the wavelet frequency and shape will change with depth and changes in the extraction interval.

A synthetic should only be considered an valid representation within the interval that the wavelet was extracted from.

The below example shows an extracted wavelet showing the wavelets amplitude, frequency and phase spectrums and the predictability of the wavelet in vicinity of the well bore. (The well bore is located at the center of the predictability window).

For a description of predictability refer to the Notes on the predictability window in this Module.

Convolving the extracted wavelet with the RC produces the following synthetic which matches the seismic and shows clearly the seismic response matching the geologically interpreted top moving away from the well.


Understand the seismic response of lithologies and fluids at the well location The lithology and fluid types are often known at the well location. Generating a synthetic from the sonic and density logs will produce a characteristic response. Understanding this character allows the interpreter to accurately determine where to pick top and base reservoir on the seismic data. Only by knowing this can we expect to produce accurate volumetric computations.

As mentioned earlier, well logs modified by fluid substitution computations can often provide crucial insight into a well-seismic tie, as light liquid hydrocarbons and/or a small amount of gas can make a significant difference in impedance and reflectivity computations used to generate a synthetic.

In the following figure, the synthetic on the left was created by re-computing sonic and density logs to reflect significant gas saturation. The synthetic on the right was computed using 100% water saturation, but using formation water parameters rather than borehole fluid parameters. Substitution interval outlined in red:


Displaying logs in SWT is important to initially understand what information the logs are showing. As an interpreter it is important to understand the logs and the information they bring.

The below example shows a SWT display with the following logs displayed;

Track 1: DT and a calculated AI The DT (multiplied with the RHOB) produces the calculated AI curve. From the AI curve

the Reflection coefficient will be generated and then convolved with the wavelet to produce a synthetic.

Track 2: RHOB and NPHI The RHOB and NPHI in combination show where sand / shale lithology transitions occur.

Where RHOB decreases and NPHI increases a transition from shale to sand is generally indicated and vice versa.

Other features like gas-effect can be interpreted when seperation between the two logs is large.

Track 3: SP, ILD and CALI The Caliper (CALI) shows any areas where washout may have occurred (see Module 2).


Where this occurs other log values can be unreliable. From the SP indicators of permeability can be shown and as a result are often used as sand/shale indicators. (see Module 2)

Track 4: GR The GR indicates lithology changes between shaly or sandy formations.

Taking each of the section highlighted above (A,B & C) individually and assessing them in more detail;


Section A

Examining the section from approximately 1850 to 1863, the gamma ray response shows a transition from a shaly to sandy formation, and a cross-over on the RHOB / NPHI track starting at 1850 indicated the formation is fluid filled. The resistivity in the third track from the left shows no coinciding resistivity response however, indicating the most likely fluid in the formation is water and not hydrocarbons.

Hence this feature can be interpreted as a wet sand (water filled).

Section B

The section from 1952 shows a gamma ray response indicating a rapid transition from a shaly to sandy formation. A large RHOB / NPHI separation, commonly known as a gas-effect can be seen and together with a resistivity response shows a strong likelihood of hydrocarbons in place.


Also there is a large variance in the calculated AI log and hence a strong RC.

Section B can be interpreted as a gas filled sand.

Section C

Investigating the section from marker Paris to 2036, a small RHOB / NPHI cross-over can be seen. However, the separation is not large so this would not indicate a gas-effect. There is a small resistivity response so likely there is fluid present in the formation but mostly water with a small presence of hydrocarbons.

A good overall interpretation could be, a wet sand, mostly water filled with a small amount of hydrocarbons.

Section B is therefore an area of interest for hydrocarbons.


To understand the seismic response attributed to the gas filled sand of section B (at and away from the well) a wavelet needs to be extracted from the seismic and convolved with the RC to produce a synthetic.

Using a 44ms window around the area of interest (blue markers in the RC track in the above display), the following wavelet is extracted (for understanding the predictability window, see the predictability explanation later in this Module).

The wavelet is a 2 degree rotated wavelet with an approximate central frequency of 25Hz.


For comparison, increasing the extraction window, it can be seen that extracted wavelet changes.

68ms 100ms 200ms

This illustrates that when extracting a wavelet it is important to balance the length of the extraction window with the formation of interest.

The minimum RC window that should be used should cover 1 cycle as illustrated below for a 45Hz wavelet is 22ms.


Calculated by using the formula;

1/frequency *

Therefore for the 25Hz wavelet as extracted for section B, the minimum extraction window should be 40ms.

Convolving the extracted wavelet with the RC produces the following synthetic which matches the seismic and shows clearly which seismic response is of interest moving away from the well.

Understand the phase characteristics of the seismic data Understanding the phase characteristics allows events to be interpreted on the correct part of the seismic waveform. And only by knowing the phase of the data can we reasonably expect to zero phase it correctly. Zero phase seismic data is essential for many geophysical analysis techniques including Seismic Inversion.

To understand the phase of the seismic over a formation of interest, SWT provides comprehensive tools to examine the phase over the extraction interval.

When extracting a wavelet, the extracted wavelet phase is reported in the extraction window (as indicated in the below extraction window example.


Interpreting on zero phase data is extremely important to be assured that the RC (or impedance change) from the log interpretation matches a peak, trough or zero crossing wavelet character in the seismic.

Various tools are available to adjust the phase of the extracted wavelet, including the rotation to zero phase.

Once the phase of the seismic is clearly understood the seismic can be re-processed to apply the dephase operator generated from SWT to the seismic to create zero phase seismic over the formation of interest.

(For further information on the effect of phase rotation, see phase rotation in Module 2.)


Understand the frequency characteristics of the seismic data Sometimes, low frequency is your friend and sometimes it is your enemy. When high frequency features are obscured due to the presence of low frequency noise or natural attenuation of higher frequencies, evaluating well tie details can become problematic.

When evaluating a well tie particularly when using a short extracted wavelet or an analytical wavelet (Ex: Ricker) - one must keep in mind that the wavelet may be bandlimited relative to the seismic and not fully represent low frequency response.

To assist with frequency evaluation, there is a filter option available on the Tools tab that will apply band limiting to the wavelet and to the synthetic traces built using the wavelet.

One can use this feature to evaluate whether a wavelet extracted in a deep interval would still be applicable in a shallow interval by filtering back the lower frequencies and observing the well seismic tie at a shallower interval.

You may also use the filter facility to help determine optimal low cut filter parameters to be applied to the seismic data in a later filtering operation.

Additional insight into phase behavior may also be gained, as selective filtering will allow band constrained phase behavior to become visible in the wavelet display. This is particularly important when working with data acquired with an impulse source (minimum phase) because of the potential for residual minimum phase in lower frequencies after processing.

In the following three figures (full bandwidth, low cut, high cut), observe how certain reflectors in the synthetic are brought into focus by filtering out low frequencies, and observe the that the


phase of the wavelet after low cut is much closer to zero. The phase of the high cut wavelet is rotated significantly.


Notes on the Predictability WindowPredictability in SWT is used for determining the optimal wavelet.

There are 3 windows showing predictability displays; Maximum Predictability Predictability for an inline Time of maximum predictability (or) Wavelet phase

and a section showing statistics related to the exercise (lower right) .

Predictability is calculated using the following methodology An autocorrelation is computed of the Borehole Data and Seismic Data traces Acor1(t)

and Acor2(t), and a cross correlation is computed between them Xcor(t). The autocorrelations and cross correlations are tapered from Time zero with a cosine

taper up to max-lag samples using;

where n is the number of input samples in the window, k has been set from experience to be (n* SR / 100), where SR is the sample rate of the data in milliseconds.


The global predictability value is now calculated with the following equation based on the tapered autocorrelations and cross correlations;

Predictability values range from 0 to 100, where 100 represents perfectly matching data.

At any time, the maximum predictability map, predictability for a single inline, and either time of maximum predictability or phase of maximum wavelet are displayed. The latter two are interactively interchangeable.

The predictability maps themselves are interactive and linked to the parameters in the position tab of Wavelet Extraction. Users can compare predictabilities a maximum of 21 Xline traces either side of an inline by selecting a location. The wavelet for that location is then shown and editable.

The white x on the Maximum Predictability map indicates the position of the currently selected extracted wavelet, which is convolved with the log reflectivities to build the synthetic displayed in Wavelet Extraction. By clicking on a different location, the wavelet will be changed to the wavelet extracted at that map location and the Predictability for Inline display will move to the selected inline.

The seismic data display remains at the position where the extraction took place. The Selected Wavelet Position button on the Position tab can be used to move the displayed seismic to the selected wavelet location.


The Predictability for an inline window displays the time lag for the position selected in the maximum predictability window.

The time of maximum predictability shows the time lag for the best predictability in map view.

The Phase of maximum wavelet display shows the phase rotation for a maximum predictable wavelet for each trace. The phase is calculated by;

Calculating the envelope of the wavelet and from that the peak of the envelope.1. Use the envelope peak to work out the time lag.2. Calculate the instantaneous phase at the envelope peak. This is the reported wavelet

phase.


Predictability and CorrelationWhy do we use predictability? Predictability is a measure of the similarity of the underlying reflectivity and as such is independent of the wavelet on the seismic and fairly insensitive to amplitude scaling differences and wavelet phase uncertainty between the two time series. Simple cross correlation methods do not benefit from these features.

How to interpret Predictability resultsThe predictability and extracted wavelet displays are used together to determine which of many extracted wavelets is the best to use. Sometimes, the optimal predictability does not occur at the exact location of the well. A few questions to ask yourself when this happens are:

Where is the well (truly) located Was the well location surveyed correctly?

Is everyone using the same cartographic system? Were cartographic transforms done correctly? Are we near the edge of a cartographic zone, where errors are often large? Is the well deviated? If so, where is the downhole location relative to the surface

location? Is the well cataloged as a vertical hole, but actually slightly deviated ?

There are three extraction positioning methodologies for SWT; Well head location Deviated well location Selected wavelet


Well Head location

Deviated well location


Selected Wavelet

Where is the seismic data (truly) located, and how good is it Is there a lot of noise or disruption at the well location, but not at the high predictability location?

Are we in an area with a lot of structure? Migration velocity or geometry problems may be positioning features incorrectly, both spatially and temporally. The deeper the target, the broader the Fresnel zone, which increases positioning ambiguity.

What is the bin spacing of the seismic? Were large spatial window smoothers used during processing? Has the data been zero-phased? Issues with velocity anisotropy, poor illumination, or frequency attenuation ?

How good are the logs Are we using measured logs or pseudo logs derived by some formulation? Are there borehole washouts at key bed boundaries that could cause the synthetic to be

in error? Was the sonic velocity recalibrated with checkshots to get closer to seismic velocities?

Were there problems with tools sticking in the hole that would affect alignment of log responses in depth?


How good is the wavelet If the wavelet is too short it will not accurately reflect low frequency phenomenon, but if

it is too long it will capture multiples or repetitive geology. The intended use of the wavelet drives the extraction methodology. If the seismic or log data is noisy, this will also be reflected in the wavelet.

Is it a zero-centric wavelet? If so, and there is residual minimum phase in the data, the tie will get progressively worse with increasing depth. If the extraction window includes a big reflector then the reflector may dominate the wavelet statistics.

Does the seismic data in the extraction window have problems? If so, then the wavelet will have problems as well.

How consistent is the geology Is the geology highly variable? If so, we would not expect to find a good tie at a

significant distance from the well. If the geology varies quite slowly, or repeats itself from place to place, then several of the factors mentioned above could easily combine to cause the best well tie to be quite some distance from the well location. If the distance between reflectors is not constant (getting thicker, or thinner, away from the well), and migration velocities are not perfect, then the best match may actually be some distance from the well.

Of course, the purpose of the well tie also comes into question. If the goal is an AVO inversion feasibility study, then a very good tie at the exact well location is strongly desired.

If the goal is simply to identify a reflector for interpretation, the geology is spatially uncomplicated, and there are no large faults between the well location and the best tie, then the well could be some distance from its true location without causing a real problem.


Using Predictability to Optimize Well LocationWhen a well location is questionable, or when selecting a location in a 2D line for a well projection, the wavelet extraction facility and the predictability display can be used to help identify the optimal location. In the following examples, shallow geology is laterally consistent over long distances. Deeper, dipping, reflectors can be seen to change significantly over shorter distances.

Figures are constructed by stretching the predictability window laterally and positioning it behind the wavelet extraction window.

In the following figure (a 2D line where the well is roughly 1.5 km offline), the Wavelet extraction window (red arrows) does not include the strong dipping reflectors in the deeper portion of the seismic data. We observe that the predictability is high for much of the area to the left of the well location (synthetic traces, center). As a result, optimal well location is ambiguous, and definitely not indicated at the current well position.


In the next figure, the strong dipping reflectors are included in the extraction window (red arrows), and the picture changes significantly. The position of these reflectors relative to strong interfaces in the log reflectivities (e.g., strong peaks in the synthetic trace) causes the predictability to be quite sensitive to timing errors that occur when the well is moved laterally.

As a result, the predictability display indicates a very small range of traces where the well reflectivity would be consistent with the seismic data, and the projection point for the well is correctly localized.


Notes on Estimating an Initial Bulk Time ShiftHow do we choose a starting log and seismic alignment point when no checkshots are available ? When our log interval is short ? When the initial checkshot corrected alignment just looks wrong ?

It may be necessary to bulk shift the entire log reflectivity series down (or up, if shallow velocities are too slow) in order to properly align log data with the corresponding seismic reflectors. This may be true even after checkshot and integrated sonic log data are properly applied. Bulk shift is applied on the Time Shift tab on the Wavelet Extraction dialog.

Estimation of velocities just beneath the surface are often problematic due to lack of supporting measured data. As a result, migration velocity models are often incorrect at the shallowest intervals, and typically too fast. This results in downward displacement of seismic reflections.

Before any wavelet is extracted or applied, a quick visual inspection of the characteristics of the distribution of energy in both the log reflectivity and seismic amplitudes can often provide some insight into the appropriate initial shift required.

Total energy (amplitude envelope) is phase independent, i.e.strong peaks and strong troughs should be considered in terms of unsigned absolute value. If our seismic data is relatively noise free, there should be a strong correspondence between energy packets in the amplitude data and energy packets in the log reflectivity.

In the following diagram, red ellipses highlight areas of strong positive and negative energy in the log data that correspond to high energy in the amplitude data:

One can make a good initial bulk shift estimate using this approach. Of course, in complex geology such an approach can be complicated by features caused by faulting, such as repeated (or overturned) geologic sections.

Absence of checkshot data can cause confusion after such an exercise because the logs will not be stretched or squeezed to reflect true time/depth relationships. Even so, energy bundle alignment followed by manual stretch/squeeze can be effective, as this figure illustrates (initial


shift on the left, alignment after stretch/squeeze on right):

Other complicating factors are possible. One common problem is undesired enhancement of smaller seismic reflections due to application of excessively short AGC windows or other scaling operators.

Also, an apparent energy imbalance may be present when working with true amplitude data. The synthetic will have power normalized to a single reflector and will not decrease in amplitude with time, but true amplitude seismic is progressively weaker with increasing time.

Notes on Stretching and SqueezingStretching and squeezing is popular with some practitioners and unpopular with others. Exact rules defining when to perform such an action (or when not to) do not exist. The choice is dependent on the data available including the purpose of the well tie, well specifics, well logs, seismic processing details, and the quality the resulting seismic data.

A discussion on some of the key areas to consider when considering a stretch and squeeze operation should include

Validity / Practicality Measured Data Phase / Data Management / VSP ties Miscellaneous well ties points


Validity There is often a necessary balance between validity and practicality in our industry. Scientific validity is typically required in theory but approximated in engineering design (example: interpolated/extrapolated wireline environmental correction algorithms derived from very specific test borehole conditions), and these are infrequently supported with sufficient data in everyday practice.

In cases where circumstances dictate deviation from a best practice, the crucial element is good communication so that any such deviation is well understood, agreed by all parties, and properly documented. One cannot generally refuse to work on a project unless all the data is perfect, and a tool that accommodates an imperfect world is not inherently flawed - it is well designed.

One must consider the reason for tying a well before determining the level of rigor applied to the process, and whether unconstrained stretch and squeeze is appropriate. For example, VSP derived phase determination requires far more attention to detail than inserting a pseudo-well interpolation control well and stretching it into a regional interpretation framework. The Wavelet Extraction tool is suitable for both situations.

Measured dataThe use of measured data rather than interpreted (or edited) data is clearly preferable, but it is not always available. Data like log and checkshot data is frequently sourced from publicly available source libraries, and their seismic data from generically processed multi-client libraries. Checkshots are rarely co-located with the well of interest, and almost never have complete documentation. As a result, fundamental questions related to datuming are often unanswered, and tying logs to seismic becomes more art rather than science.

Even when available, measured is not always the same as useful, a comment which goes to the purpose of the specific exercise in question. In this case whether one stretches/squeezes often depends on what one is doing.

Date post:	24-Sep-2015
Category:	Documents
Upload:	fran
View:	1,042 times
Download:	400 times

Seismic Well Tie Guide 2010

Documents