Marine controlled-source electromagnetic methods in the hydrocarbonindustry: A tutorial on method and practice

Lucy MacGregor1 and James Tomlinson1

Abstract

We provide a tutorial on the marine controlled-source electromagnetic (CSEM) method aimed at geoscient-ists and explorationists who are new to EMmethods. This tutorial highlights some of the issues to be consideredin planning and executing a CSEM survey, and interpreting the results. CSEM methods can add valuable in-formation on the resistivity structure of the subsurface, which complements measurements obtained from seis-mic or other geophysical methods. We also discuss the development of the method and provide an overview ofthe CSEM acquisition approaches applied today. Understanding the sensitivity of the CSEM method to seafloorresistivity structures is key to ensuring a successful survey. We illustrate this with simple examples, demon-strating the effect of reservoir and overburden properties and the effect of electrical anisotropy. It is also im-portant to understand how well a given resistivity structure can be recovered from realistic survey data. Weapply an inversion approach to illustrate this for 2D resistivity structures. Finally, we discuss the importance ofinterpreting CSEM in an integrated framework alongside seismic and well log data.

IntroductionMultiphysics approaches to geophysical analysis, in

which multiple geophysical measurements are inte-grated to provide a constrained earth model, are be-coming more and more popular. Controlled-sourceelectromagnetic (CSEM) survey methods provide avaluable complement to more conventional geophysicalapproaches. For example, whereas in many situationsthe seismic method provides high-resolution imagesof structure and stratigraphy, determining fluid proper-ties using seismic data alone can be difficult or, in somecases, impossible. The sensitivity of resistivity to hydro-carbon saturation and fluid properties such as temper-ature and salinity is well known, and widely exploited inwell log analysis and interpretation. The resistivityattribute derived from CSEM data, if properly inter-preted, can provide information which when integratedwith seismic data allows fluid properties to be deter-mined with more certainty. However, caution mustbe exercised: Resistivity is by no means a unique indi-cator of fluid fill. Variations in porosity and lithologycan also result in large resistivity variations, which mustbe distinguished from the fluid effects of interest.

The marine CSEM methodThe marine CSEM method in its most commonly

applied form uses a high-powered horizontal electric di-pole (HED) source to transmit a low-frequency electro-magnetic (EM) signal through the seafloor (Figure 1a).

The source is towed at about 30 m above the seafloor toensure good coupling of the signals with the earth,while avoiding the risk of entanglement with, or damageto seafloor infrastructure and obstructions. The trans-mitter itself comprises a 300 m neutrally buoyantstreamer supporting two conducting electrodes form-ing the two ends of a grounded dipole. The transmittercurrent is usually in the range 1000–1500 A, althoughcurrents up to 7500 A are reported for the deckmounted, shallow towed source system described inBarker et al. (2012). This provides source dipole mo-ments of around 250,000 Am in the lower harmonics.The source transmits a coded tristate waveform, de-signed to optimize the frequency content for sensitivityto the target of interest (Mittet and Schaug-Pettersen,2008; Myer et al., 2011). Frequencies in the range0.01–10 Hz are typically employed.

The receivers are deployed on the seafloor at the startof a survey and log autonomously. They detect and rec-ord up to three components of the electric and threecomponents of the magnetic field, although it is usualto record just the horizontal electric and magnetic fieldcomponents. At the end of the survey, the receivers arerecovered to deck and the data downloaded for process-ing and interpretation. By studying the variation in thereceived fields as a function of source-receiver separa-tion (range), geometry, and signal frequency, using acombination of forward modeling, inversion, and hy-pothesis testing approaches, the resistivity of the

1Rock Solid Images Inc., Houston, Texas, USA. E-mail: lucy.macgregor@rocksolidimages.com; james.tomlinson@rocksolidimages.com.Manuscript received by the Editor 15 October 2013; revised manuscript received 5 December 2013; published online 20 May 2014. This paper

appears in Interpretation, Vol. 2, No. 3 (August 2014); p. SH13–SH32, 18 FIGS.http://dx.doi.org/10.1190/INT-2013-0163.1. © 2014 Society of Exploration Geophysicists and American Association of Petroleum Geologists. All rights reserved.

t

Special section: Interpretation and integration of CSEM data

SH13SH13

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

seafloor can be determined at scales of a few tens ofmeters to depths of about 3 km below mudline, depend-ing on the structure and resistivity of the overburden. Arecent paper by Constable (2013) provides a review ofthis type of CSEM acquisition technology.

Alternative CSEM acquisition strategies have alsobeen developed. Yuan and Edwards (2000) describe asystem in which the source and receivers, linked in alinear array, are dragged across the seafloor. Sourcereceiver separations between 85 and 493 m are em-ployed. The resulting data, analyzed in the time domain,are used to quantify marine gas hydrate concentration inthe upper approximately 250 m of the seafloor. Consta-ble et al. (2012) describe a CSEM system that is also de-signed to map the shallow subseafloor. In this system,three-component electric field receivers, comprising in-line, crossline, and vertical electric dipoles, are towed atoffsets up to 1 km behind a horizontal electric dipolesource (Figure 1b). Although the system can be towedclose to the seafloor in deep water (in contrast to thesurface-towed system of Mattsson et al. [2012] discussedbelow) the short offsets make it primarily applicable tothe mapping of resistivity in the shallow subsurface, forexample mapping gas hydrates or shallow hazards.

Mattsson et al. (2012) describe a towed system de-signed to look deeper into the seafloor (Figure 1c).The source and receiver are towed at depths of 10–100 m below the sea surface in water depths from 5to 400 m. The source is again a horizontal electric di-pole, towed at the front of the array, transmitting intoa series of inline horizontal electric dipole receivers at

offsets of 500–8000 m. The effects of motional noise arereduced using data from auxiliary motion sensors onthe towed streamer. This acquisition method providessignificant operational efficiencies over seafloor de-ployed systems; however, it comes at the expense ofmultiazimuth coverage in the resulting data set. How-ever, good results have been achieved (e.g., Zhdanovet al., 2012; Bhuiyan et al., 2013).

Holten et al. (2009) describe a CSEM system in whichthe source and receivers are vertical electric dipoles(Figure 1d). The challenge with making this measure-ment is the small magnitude of the vertical electric fieldresponse in comparison to the horizontal electric field.As a result, only small tilt angles in the vertical sourceand receiver can be tolerated before the horizontalcomponent dominates the response. However, resistiveearth structure can be detected at significantly shorteroffset than in the horizontal electric dipole case (Cue-vas and Alumbaugh, 2011) with the result that the ver-tical dipole system has the potential to improve lateralresolution of subseafloor resistivity structure.

A brief historyThorough reviews of the history of the CSEMmethod

and its application to hydrocarbon exploration havebeen published by many authors (e.g., Chave et al.,1991; Edwards, 2005; Constable and Srnka, 2007; Con-stable, 2010), so only a short summary is provided here.The CSEM method was developed originally in the late1970s as a tool to map the resistivity structure ofthe deep ocean floor, and applied successfully to

this end in several surveys (Young andCox, 1981; Constable and Cox, 1996).Throughout the 1980s and 1990s, themethod evolved through surveys target-ing mid-ocean ridge hydrothermal andmagmatic systems (Evans et al., 1994;MacGregor et al., 1998, 2001). Even inthese early studies, the importance of in-tegrating seismic and CSEM was recog-nized. Seismic data were used to mapthe structure of the mid-ocean ridge sys-tem, with CSEM data providing comple-mentary resistivity information used toconstrain the fluid properties, primarilythe saturation and/or temperature ofmelt or hydrothermal fluid, within themagmatic and hydrothermal systemsunder study (Sinha et al., 1998).

Industry interest in CSEM began inthe mid-1980s with theoretical workon the use of CSEM to directly mapthe fluid content of reservoir structures(Srnka, 1986). Although the applica-tion looked promising in theory, theeconomics of the survey prevented itsapplication in practice. Throughout the1990s however, industry interest inCSEM continued, applied as a comple-

Figure 1. CSEM schematic diagrams. (a) Standard CSEM acquisition with ahorizontal towed source and seafloor deployed receivers. (b) A variation onthe standard in which a 1 km streamer of three-component receivers is deep-towed behind the source. (c) Streamer CSEM acquisition in which the sourceand receiver are towed behind the survey vessel. (d) Vertical electric dipole ac-quisition. See text for discussion.

SH14 Interpretation / August 2014

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

mentary method to the mapping of subbasalt sedimentsand structure (MacGregor and Sinha, 2000).

The first practical application of CSEM as a methodof mapping resistivity within a hydrocarbon reservoirtook place in late 2000 on a field offshore Angola. Dur-ing the six week survey exercise, a full 3D CSEM survey(albeit on a relatively sparse grid) was acquired, proc-essed, and interpreted. The results demonstrated inpractice for the first time that the CSEM method couldbe used to map a hydrocarbon reservoir (Ellingsrudet al., 2002; Eidesmo et al., 2002).

The history of the CSEM industry, born of that firstsurvey offshore Angola, has been somewhat checkered.Early successes were followed by a dramatic rise in thenumber and value of companies offering CSEM surveys,and bold statements as to the utility of CSEM across theexploration industry were made. However, in 2007,there was a dramatic crash in the CSEM market. Multi-ple reasons for this can be mooted. First, the strongmarketing and bold statements made in the early yearswere coupled with disappointing results where custom-ers failed to realize value from their CSEM data, or wereprovided misleading results. Second, despite the mar-keting hype, the CSEM method was still relativelyimmature, in terms of acquisition and particularly inter-pretation. The importance of careful integration withseismic and wells was not well understood, nor werethe effects of electrical anisotropy. Third, overly opti-mistic estimation of market size led to overcapacity. Fi-nally, it can be argued that aggressive patent battlesstifled market growth, making potential users wary ofadopting the technology.

However, resistivity remains a valuable attribute toinclude in an interpretation of rock and fluid properties,and CSEM, if acquired and interpreted carefully canprovide a robust estimate of this property. The purposeof this paper is to provide an overview of considera-tions when planning a CSEM survey or interpretingthe results, and to demonstrate some of the pitfallsand dangers in using the technology. It is an obviousstatement to make, but CSEM is not seismic, so particu-lar emphasis will be given throughout the paper tosimilarities and differences in the technologies andconsiderations when using them. An extensive bibliog-raphy is provided at the end of the paper for readerswishing to delve further into the details of the subjectsdiscussed.

Sensitivity analysisThe first stage in any CSEM exercise is to conduct a

thorough sensitivity analysis, and establish whetherCSEM, either alone or in combination with seismic,can address the geophysical question of interest. Pre-survey modeling should also include inversion analysisof how accurately, and with what resolution, structuresof interest can be recovered from the data. This “recov-erability” question is discussed later, and must also con-sider what constraints must (or can) be applied to the

inversion to optimize the results at the pre-survey mod-eling and post-survey interpretation stages.

In contrast to seismic sensitivity analysis that is oftenreservoir-focused, assessing sensitivity for CSEM sur-veys must consider not only the reservoir, but the com-plete structure, from seafloor to some distance beneaththe reservoir. This is because a CSEM response cannotbe said to be coming from, or caused by a single forma-tion: It is the cumulative result of the resistivity within avolume of the earth. This necessitates construction of acomprehensive background resistivity model prior toany sensitivity analysis. This model should includeany known resistors (for example, gas hydrates inthe shallow structure, carbonates, volcanics), and resis-tive electrical basement, if present, which may affectthe results even if it is some distance beneath the res-ervoir interval, and is often a source of uncertainty inthe analysis. The background model must also accountfor anisotropy in the overburden section. This latterpoint is also often a source of uncertainty in the analy-sis. Electrical anisotropy is seldom logged in wells, andwhere it is, such measurements are often confined tothe reservoir intervals. Only a few examples of electri-cal anisotropy measurements through the overburdenexist (e.g., Loseth et al., 2013). In most cases, overbur-den anisotropy must be estimated, either from prior ex-perience in the area, or from knowledge of anisotropyencountered in similar lithologies elsewhere. The un-certainty of such estimates can be high, and must befactored in to the sensitivity assessment.

To illustrate how sensitivity varies with the structureunder study, consider the simple 1D, isotropic modelshown in Figure 2a (the effect of anisotropy will be con-sidered later in the paper). The most basic question toask is whether a CSEM survey is likely to be sensitive tothe reservoir of interest. It is also important to establishthe optimum acquisition parameters, in terms of re-quired source-receiver separation (range) and transmis-sion frequencies.

The inline electric field is the component detectedalong a line passing through and parallel to the axisof the dipole, and is optimally sensitive to thin resistivelayers (MacGregor and Sinha, 2000; Ellingsrud et al.,2002). The basic sensitivity question can be answeredby contouring the percentage change in the measuredinline electric field as a function of source receiver sep-aration and source frequency (Figure 2b). The choice offrequency coverage to be used is a trade-off. At veryhigh frequency, the effect of the target reservoir onthe measured field is large; however, signals are attenu-ated rapidly and may not be detected above the noiselevel. At very low frequency, signal-to-noise ratio (S/N)is excellent, but the fields are unable to resolve small-scale structures. In practice, the usable frequency bandis therefore relatively narrow. Although an optimum fre-quency for sensitivity to a given target may be chosen inthis way, it is good practice to ensure that the trans-mission frequencies used cover a band around thisfrequency. As will be shown below, the choice of

Interpretation / August 2014 SH15

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

frequency depends critically on the background struc-ture, and therefore transmitting a range of frequenciesensures some contingency for uncertainty in this

parameter, as well as improving the quality of interpre-tation results.

It could be argued that the solution to this problemwould be to transmit a broadband signal and indeedsuch signals have been tested (Ziolkowski, 2007; Ziol-kowski et al., 2011). However, the CSEM source hasa finite energy budget. If this is spread across too manyharmonics, the magnitude of each is diminished, andhence S/N may be compromised. It has been shown thatbecause the variation in response with frequency issmooth and can be adequately represented by a fewcarefully selected frequencies, the most important fac-tor in controlling sensitivity and resolution is the overallfrequency bandwidth, whereas frequency samplingwithin the bandwidth is less critical (Key, 2009). Con-centrating power within certain harmonics betweenan upper and lower limit has the dual effect of ensur-ing good frequency coverage, while maintaining goodS/N.

The reservoir property to which the CSEMmethod isprimarily sensitive is the transverse resistance, definedas the vertically integrated resistivity. For a simple 1Dlayer, this corresponds to the resistivity-thickness prod-uct (Constable, 2010; MacGregor, 2012). The effect oftransverse resistance on sensitivity is shown in Figure 3.For ease of comparison between different models, per-centage anomalies for a range of frequencies are shownat a source-receiver separation of 5.5 km, correspond-ing to an extraction along the horizontal yellow dashedline in Figure 2b. As one would expect, the higher thetransverse resistance, the larger the effect of the reser-voir on the measured fields. However, an importantpoint to note here is that, although the magnitude ofthe anomaly changes with transverse resistance, thefrequency at which that anomaly is observed doesnot alter significantly. Optimum transmission frequencyis controlled primarily by the overburden structure, andmuch less by the properties of the reservoir to bedetected.

A frequency of 0.3 Hz provides a good trade-off be-tween S/N and anomaly size in this example. Figure 3bshows the variation in the anomaly associated with thetarget reservoir at this frequency, as a function ofsource receiver separation and reservoir transverse re-sistance. As the transverse resistance increases, so doesthe measured inline electric field anomaly. Taking aconservative limit on the level of 1D normalizedanomaly required for a response to be detected in realdata of 20% (other authors have suggested 15%, e.g.,Hesthammer et al., 2010), the minimum transverse re-sistance required for the CSEM survey to be sensitiveto the reservoir corresponds to the point at whichthe 20% contour crosses the noise floor, set to10−15 V∕Am2 in this case and marked by a white dia-mond in Figure 3b. This corresponds to a transverse re-sistance of 720 Ωm2, or a bulk average of 14 Ωmresistivity if the reservoir interval is 50 m thick.

It is clearly important to understand the sensitivity ofa CSEM survey to the reservoir properties of interest.

Figure 2. (a) Simple canonical model representing a hydro-carbon charged reservoir of resistivity 50 Ωm and thickness50 m embedded at 1500 m below seafloor in a gradually vary-ing background resistivity structure. The model is overlain bya water layer of depth 1500 m and resistivity 0.3 Ωm. (b) Sen-sitivity of a CSEM survey to the presence of the target reser-voir, expressed as the percentage change in the inline electricfield as a function of source-receiver separation (range) andsignal frequency. The solid white line shows the 10−15 V∕Am2

electric field strength contour, corresponding to a typicalnoise floor for this water depth. Signals above and to the rightof this contour lie below the noise floor. The dashed whitelines show the 20% and 50% percentage anomaly contours. Re-sponses in Figure 3 are extracted along the horizontal yellowdashed line.

SH16 Interpretation / August 2014

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

However, it is equally important to understand thebackground resistivity structure. This is illustrated inFigure 4, which again shows percentage anomaliesfor a range of frequencies at a source-receiver separa-tion of 5.5 km, corresponding to an extraction along thehorizontal yellow dashed line in Figure 2b, to allowcomparison between different models. In this case,the background resistivity structure of the baselinemodel shown in Figure 2a has been altered.

As the background structure becomes less resistive,the size of the anomaly associated with the reservoirincreases because the contrast between reservoir andbackground is greater; however, its effect shifts tolower frequency. This is because attenuation of the sig-nals in the more conductive overburden increases, withthe result that only low-frequency signals penetrate asfar as the reservoir. As the background resistivity in-creases, the size of the anomaly decreases as the

Figure 3. (a) Percentage anomaly in the inline electric fieldextracted at a source-receiver range of 5.5 km, correspondingto the horizontal yellow dashed line in Figure 2b. The line isdashed when the signal strength falls below a noise floor of10−15 V∕Am2. Although the magnitude of the anomaly in-creases with increasing transverse resistance, the frequencyat which the anomaly is observed is unaltered. (b) Effect oftransverse resistance on the measured inline electric fieldanomaly. Dashed white lines show the 20% and 50% anomalycontour. The solid white line shows a representative noisefloor of 10−15 V∕Am2. The minimum reservoir transverse re-sistance to which 0.3 Hz CSEM data will be sensitive is720 Ωm2 in this example, corresponding to the point at whichthe 20% anomaly contour falls beneath the noise floor and ismarked by the white diamond.

Figure 4. Effect of background resistivity on the sensitivityof a CSEM survey. (a) The background structure of the base-line model in Figure 2a has been made more conductive andmore resistive by factors of two and three. (b) The result is alarge change in the sensitivity of a CSEM survey. The percent-age anomaly is extracted at a source receiver separation of5.5 km for ease of comparison between models.

Interpretation / August 2014 SH17

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

contrast associated with the reservoir is smaller. In theexample shown, the more resistive overburden resultsin an anomaly that is unlikely to be detected.

Of course, in practice, variations in background re-sistivity structure are likely to be more complex than

the simple bulk shifts assumed here; however, this ex-ample does illustrate that variations in backgroundhave the potential to affect not only the transmissionfrequency chosen, but the overall feasibility of thesurvey. Knowledge of the background structure is asimportant as knowledge of the potential reservoir prop-erties: The background should be considered as muchof a target as the reservoir interval of interest.

In areas devoid of well log information, constructionof a background resistivity trend is difficult, and there-fore uncertainties in assessing survey feasibility and es-tablishing effective survey design parameters are large.In areas where standard well log-derived resistivity in-formation is available, this may be used to construct asuitable background model. However, the uncertaintyin this background model remains large. The reasonfor this is electrical anisotropy. In general, standard in-duction log measurements in vertical well bores mea-sure the horizontal component of the resistivity (e.g.,Luling, 2013). The vertical component of resistivity,to which the inline electric field is primarily sensitive(Ramananjaona et al., 2011), is often many timeshigher. Factors of two to three are common, andfactors as high as ten have been reported in some cases(Klein et al., 1997; Ellis et al., 2010; MacGregor et al.,2012; Colombo et al., 2013; Gabrielsen et al., 2013).

Anisotropy of this magnitude can significantly affectthe response of a CSEM survey and its sensitivity to areservoir target. Figure 5 shows a sensitivity plot of thesame form as Figure 2; however, in this example, theratio of anisotropy in the overburden has been in-creased to a factor of two. The reservoir is assumedto be isotropic because, although the anisotropy withinreservoir intervals can be large, the sensitivity of theCSEM method to anisotropy within the reservoir itselfis small (Brown et al., 2012). The response is dominatedby the anisotropy in the background structure. Compar-ing Figure 5 to Figure 2, it is clear that the anisotropy inthe overburden changes the sensitivity of a CSEM sur-vey to the reservoir significantly. The overall magnitudeof the anomaly has decreased and is now observed atlonger ranges than when an isotropic overburden isconsidered.

Correctly estimating or predicting anisotropy in thebackground structure is therefore important. It is alsoextremely difficult. Electrical anisotropy values mea-sured in three-component well logs, and CSEM surveysfrom around the globe show that this parameter isextremely variable, within a given section, and betweenareas. Even when three-component logs are available,they may fail to resolve the bulk anisotropy that affectsCSEM data (Gist et al., 2013). The controls on electricalanisotropy are not well-understood, and thereforebuilding robust models is a challenging task. Several au-thors (Ellis et al., 2010; Bachrach, 2011) have suggestedrock physics modeling approaches to address this; how-ever, these require calibration and are (at the moment)unlikely to be general enough to address all geologicsituations. Work continues to address this need.

Figure 5. (a) Simple canonical for which the horizontal resis-tivity is the same as that of Figure 2, but the vertical resistivityin the overburden is a factor of two higher. The reservoir tar-get is assumed isotropic. The model is overlain by a waterlayer of thickness 1500 m and resistivity 0.3 Ωm. (b) Sensitiv-ity of a CSEM survey to the presence of the target reservoir,expressed as the percentage change in the inline electric fieldas a function of source-receiver separation and signal fre-quency. The solid white line shows the 10−15 V∕Am2 electricfield strength contour, corresponding to a typical noise floorfor this water depth. The dashed white lines show the 20% and50% percentage anomaly contours. Comparing this to Figure 2,it is clear that the effect of the anisotropy is large.

SH18 Interpretation / August 2014

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

In the meantime, background models must be con-structed using best estimates of anisotropy derivedfrom previous CSEM experience in a given area,three-component well logs where available, and experi-ence in similar geologic domains. Recent work on theuse of isotropic seismic-resistivity transforms derivedfrom well log data to build background models basedon seismic velocity is also promising (Wertmuller etal., 2013). Background resistivity trends can also bederived from magnetotelluric (MT) data, which usesnaturally generated electromagnetic fields (Constableet al., 1998; Hoverston et al., 1998; Ellingsrud et al.,2002). However, such data provides a measure pri-marily of horizontal resistivity, and at a resolution scalelower than that of CSEM. Where significant uncertaintyremains, contingency should be built into survey plan-ning exercises to ensure that the resulting data willprovide the required sensitivity across a range of back-ground resistivity and anisotropy values.

Constraining electrical anisotropy usingCSEM data

It has been demonstrated that the presence of elec-trical anisotropy in the earth can have a significant im-pact on the choice of acquisition parameters. It alsoaffects the interpretation of CSEM survey data once ac-quired, and the constraint of anisotropy in the overbur-den section is a key component of any such analysis.Ramananjaona et al. (2011) have demonstrated that,for simple structures in deep water, the inline compo-nent of the electric field is primarily sensitive to the ver-

tical component of resistivity. In contrast, the broadsidecomponent, detected along a line passing through andperpendicular to the source dipole axis, is primarilysensitive to the horizontal resistivity. This leads to theobservation that multiazimuth data coverage is requiredto properly constrain anisotropy (e.g., Newman et al.,2010). However, the situation is considerably more com-plex in shallow water, and in more complex structures.

Figure 6 shows the electric field response of ananisotropic halfspace with a vertical resistivity of2 Ωm and a horizontal resistivity of 1 Ωm in polar dia-grams plotted at a source-receiver separation of 6 km,for water depths of 1500 and 100 m. The frequency is0.2 Hz. For the deep water case, along the axis of thedipole in the inline direction, the measured responseis close to that of a 2 Ωm isotropic halfspace suggestingas expected that the inline response is primarily sensi-tive to the vertical resistivity. In contrast, across theaxis of the dipole in the broadside direction, the re-sponse is much closer to that of a 1 Ωm isotropic half-space, suggesting that this geometry is more sensitive tothe horizontal resistivity. At azimuths between theseend members, the response is sensitive to horizontaland vertical components. This deep water responsehas been used to constrain anisotropic structures usingfield data (e.g., Newman et al., 2010).

The behavior in shallow water is different (Fig-ure 6b). The broadside response is still close to thatof a 1 Ωm halfspace; however, in the inline direction,the response now differs from the 1 and 2 Ωm isotropiccases, suggesting that, in shallower water, the responsehas sensitivity to the horizontal and vertical component.

Figure 6. Polar diagrams illustrating the effect of anisotropy on the measured response as a function of azimuth. The inlinedirection corresponds to an azimuth of 0° along the axis of the source dipole (represented by the double-headed arrow). Thebroadside geometry corresponds to an azimuth of 90°. The frequency is 0.2 Hz and the source-receiver separation is 6 km.(a) Water depth is 1500 m. (b) Water depth is 100 m.

Interpretation / August 2014 SH19

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

The response is governed by the complex interactionof signals interacting with the seafloor below, andthe resistive air layer above (Andreis and MacGregor,2008). The situation is further complicated if theearth has a more complex structure. Although multiazi-muth data are therefore useful, they do not provide sim-ple, independent constraints on the horizontal andvertical resistivity except in the simplest deep waterenvironments.

Given the sensitivity of the inline electric field com-ponent to horizontal and vertical resistivity, inline datacan be used alone to determine anisotropy. Sensitivityto anisotropy as a function of frequency is shown inFigure 7, where the quantity contoured is an anisotropicresponse similarity defined as

Anisotropic similarity ¼ E21 − E22

E11 − E22 ; (1)

where E21 is the inline electric field response of theanisotropic halfspace with vertical resistivity 2 Ωmand horizontal resistivity 1 Ωm, E11 is the responseof a 1 Ωm isotropic halfspace, and E22 is the responseof a 2 Ωm isotropic halfspace. This metric tends to 1 ifthe response of the anisotropic model tends toward theresponse of the isotropic halfspace equivalent to thevertical resistivity, and tends to zero in the oppositecase, where the response tends to that of the isotropichalfspace equivalent to the horizontal resistivity. Ittherefore provides a useful guide to the relative sensi-tivity to horizontal and vertical resistivity across a givenparameter range.

Figure 7 again highlights the large difference be-tween deep and shallow water cases. In deep water,at longer offsets and higher frequencies, the resultssuggest primary sensitivity to the vertical resistivity,whereas at the same offsets in shallow water, the re-sponse is primarily sensitive to the horizontal resistiv-ity, a result of signals interacting with the resistive air(Andreis and MacGregor, 2008). At shorter ranges andlower frequencies, the response is clearly sensitive tohorizontal and vertical resistivity in both cases. Thus,

although real situations are considerably more complexthan the example shown here, given a realistic multifre-quency, multioffset inline data set, it is possible to con-strain anisotropic resistivity, and indeed several casestudies bear this out (MacGregor et al., 2012; Mattssonet al., 2013).

Resolution and recoverabilityAlthough it is important to establish the sensitivity of

CSEM data to a target of interest prior to a survey, andchoose the optimum acquisition parameters, it is alsoimportant to understand how well the structure of inter-est can be recovered from a finite noisy CSEM data set,with what resolution and under what circumstancesgiven the inversion tools available. This “recoverability”question is addressed by Key (2009) and MacGregor(2012) for the case of a 1D structures. Here, simple iso-tropic 2D structures are considered. The goal of thissection is not to present a complex inversion study,but rather highlight some of the positives and negativesof, and controls on, resolution and recoverability.

In the first example (Figure 8), three resistive fea-tures of varying size are embedded at a depth of1 km below the seafloor in a 1D background structurein which the resistivity gradually increases with depth.The water depth is 900 m and the structure is invariantperpendicular to the plane of the page. Receivers aredeployed at the seafloor at intervals of between 400 mand 4 km. The source positions are 10 m above the sea-floor with a spacing of 200 m in all cases. A sensitivityplot (see Figure 2 for details) for this structure (Fig-ure 8b) demonstrates that there is extremely good sen-sitivity across a wide range of frequencies and source-receiver separations to the embedded resistive targetsin this case, at least if they are considered to be 1D.

Although in practice multiple frequencies are ac-quired and interpreted using either an amplitude/phase,or real/imaginary representation of the measured fields,in the interests of illustration, amplitude data at a singlefrequency of 0.75 Hz is chosen for the analysis below.Synthetic data were generated for each receiver andcontaminated with 2% Gaussian noise, assuming an ex-

ternal noise floor of 10−15 V∕Am2.The inversions use the Occam ap-

proach of Constable et al. (1987) to de-rive models that are smooth in the first-derivative sense, and are thus as close toa uniform halfspace as is compatiblewith the data. All inversion resultsshown have converged to a constantroot-mean-square (rms) misfit level justabove the fit of the true model to thesynthetic data to avoid overfitting thedata. Although the inversions may notbe precisely optimum for each modeland data set, this approach ensures aninternally consistent set of results thatmay be directly compared.

Figure 7. Anisotropic similarity metric (see text for discussion) for 1500 m ofwater (a) and 100 m of water (b). In each case, the response of an anisotropichalfspace with vertical resistivity 2 Ωm and horizontal resistivity 1 Ωm is com-pared to the responses of a 1 Ωm and a 2 Ωm isotropic halfspace.

SH20 Interpretation / August 2014

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

Figure 9 shows the results of unconstrained inver-sion of the synthetic data for receiver spacings of400 m to 4 km. In this example, only inline data are in-cluded in the data set: we assume a single tow line in theplane of the page, along the line of receivers. For the400 m receiver spacing, all three resistive featuresare resolved as discrete bodies. When the receiver spac-ing is 1–2 km, the larger two features are clearly re-

solved, but the central, smaller body becomes lessclear, and starts to merge with the deeper structure.At 4 km receiver spacing, none of the three resistive fea-tures are separated from the more resistive structureat depth.

To investigate the results further, Figure 10 showsthree pseudowells, extracted at across axis positionsof −3, 1, and 3 km through the true model and the four

Figure 8. (a) Simple 2D model (invariant perpendicular to the page) consisting of three resistive bodies embedded in a 1D back-ground. The model is overlain by a water layer of thickness 900 m. Receivers (white triangles) are spaced at between 400 m and4 km separation at the seafloor. The source spacing is a constant 200 m. (b) One-dimensional sensitivity plot for the model shownassuming the resistive bodies at 1 km depth are infinite in extent. There is good sensitivity to the resistive targets. A frequency of0.75 Hz (vertical dashed yellow line) is used in the analysis.

Figure 9. Unconstrained inversion of inline amplitude data generated from the model shown in Figure 8, and contaminated with2% Gaussian noise. The outline structure of the true model is shown by the white dashed lines, and results from receiver spacings of400 m, 1 km, 2 km, and 4 km are shown.

Interpretation / August 2014 SH21

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

inversion results shown in Figure 9. The pseudowell ex-tracted at 1 km across axis does not intersect any of theresistive features and so represents a backgroundtrend. Results from inversion of the data with 400 m,1 km, and 2 km receiver spacings accurately recovera smoothed version of this background trend, with littledifference in the results between cases. For the 4 kmresult, the background resistivity is significantly over-estimated, the result of merging the effect of the resis-tive bodies into the background structure.

Looking at the pseudowells at −3 km and 3 kmacross axis, that in both cases intersect one of the re-sistive features, the three smallest receiver spacings(400 m, 1 km, and 2 km) recover the resistive bodies

in each case as a localized increase in resistivity, albeitvertically smeared across a relatively large interval. Fora receiver spacing of 4 km, it is clear that in a verticalsense at least the resistive features have not been re-solved, but are merged into the background structure.

The pseudowells shown in Figure 10 highlight thepoor vertical resolution of the CSEM method: Featurestend to be smeared over a relatively large vertical inter-val and have lower resistivity than in the true model.However, although the absolute resistivity is not well-resolved, the transverse resistance is usually better con-strained by the inversion process (Constable, 2010;MacGregor, 2012). Figure 11 shows the transverse resis-tance calculated across an interval from 200 m above

the resistive features, to 200 m belowthem (800–1400 m below seafloor).The results for the two smaller receiverspacings reproduce the transverse resis-tance in the two wider bodies relativelywell. The results for the 2- and 4-kmreceiver spacing underestimate thetransverse resistance in these bodies.In all cases, the transverse resistancein the central narrower body is underes-timated. An important point to note inFigure 11 is that, even when the receiverspacing is 4 km and there is little verticalresolution, the transverse resistanceshows that the effect of the localized re-sistors is seen laterally. It is generallythe case that the lateral resolution ofthe CSEM method is considerably betterthan the vertical resolution.

The results in Figures 9–11 suggestthat resolution is compromised at re-ceiver spacings greater than 2 km whenonly inline data are considered. Theresolution of the relatively shallow resis-tive features considered here improves

as the receiver spacing decreases below this. Of course,a finer spacing also leads to significant data redun-dancy, that in itself can lead to improved data throughthe application of stacking or synthetic aperture ap-proaches to improve S/N and sensitivity (Fan et al.,2012; Mattsson et al., 2013).

Figure 12 shows the effect of including inline and off-line data in the inversion. Receivers are spaced at 4-kmintervals across the structure; however, in this case, thesource is towed over the line of receivers to give inlinedata, and in two parallel tows, offset by 4 and 8 km fromthe line of receivers (Figure 12a). The addition of themultiazimuth data for this sparse receiver array im-proves the result significantly, allowing the resistivebodies to be resolved as isolated resistive features,and giving a result that is comparable to the equivalentresults for finer receiver spacings when only inlinedata are considered. This is because inclusion of multi-azimuth data can resolve ambiguities between theresponse of localized resistive features and more

Figure 10. Pseudowells extracted from the inversion results shown in Figure 9.In each case, the true model result is shown by the gray dashed line. (a) Extrac-tion at −3 km across axis. (b) Extraction at 1 km across axis. (c) Extraction at3 km across axis.

Figure 11. Transverse resistance calculated across a depthinterval from 200 m above the resistive features to 200 m be-low. The dashed gray line shows the transverse resistance ofthe resistive features in the true model.

SH22 Interpretation / August 2014

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

gradual increases in resistivity with depth inherentin sparse inline data taken alone (Ellingsrud et al.,2002). This result should not be taken to imply that,in all situations, a sparse 3D survey layout is as goodor better than a finer inline only layout: The issue ofthe best survey layout is more complex in 3D struc-tures, compared to the 2D case considered here, andmust be assessed on a case-by-case basis with presur-vey modeling. However, it does serve toillustrate the effect of including multia-zimuth data in an interpretation process.

Figure 13 shows a slightly more com-plex model in which the resistive bodyof interest lies at approximately 2 km be-low the seafloor, and 1.5 km above a re-sistive basement fault block. The body isoverlain by a more resistive layer in theshallow background structure, repre-senting (for example) shallow biogeniccarbonates or hydrates. In this case, syn-thetic amplitude and phase data at threefrequencies, 0.08 Hz, 0.24 Hz, and0.56 Hz are inverted. Once again, thedata are contaminated with Gaussiannoise and are inverted to an rms misfitlevel just above that expected for thetrue model. Despite the overlying resis-tive layer and the deeper basementstructure, in this case, the presence ofthe localized resistive body is recoveredby the inversion: It lies far enough fromeach to be resolved as a discrete resis-tive body. The depth of the recoveredbody is slightly shallower than in reality,a consequence of the poor vertical res-olution of the CSEM method, madeworse by the depth of the target belowmudline in this example (Key, 2009;MacGregor, 2012).

Figure 14 illustrates the problem of resolving local-ized resistive features that are adjacent to more massiveresistive features. The model in Figure 14 is the same asthat in Figure 13; however, now the localized resistivetarget is slightly deeper in the section, lying less than500 m from the resistive basement fault block. In thiscase, the inversion is unable to resolve the localizedresistor, which is instead merged into the deeper

Figure 12. (a) Map view showing 3D survey geometry including inline and offline data. Receivers are placed at 4-km intervals, andthree tow lines simulated, the first across the line of receivers, second parallel to the receiver line and offset by 4 km, and the thirdoffset by 8 km. Synthetic amplitude data at 0.75 Hz were generated and contaminated with noise, to allow direct comparison withFigures 9–11. (b) Inversion of the resulting 3D data set. Resolution is considerably improved compared to the case of sparse inlineonly data (Figure 9d).

Figure 13. (a) Structure used to generate synthetic amplitude and phase data at0.08, 0.24, and 0.56 Hz. Receiver positions are shown by the white triangles. Datawere contaminated with Gaussian noise prior to inversion. The model contains aresistive target of resistivity 25 Ωm overlain by a shallow resistive layer repre-senting (for example) shallow hydrates or carbonates, and underlain be a resis-tive basement fault block. (b) Result of unconstrained inversion of this syntheticdata. The resistive target is recovered as a discrete body in this case because itlies far enough from overlying and underlying resistive features.

Interpretation / August 2014 SH23

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

basement structure. This does not mean the effect ofthe localized target is not felt. The transverse resistancein the basement is higher than the true model, theconsequence of the merged target structure, and ifthe basement resistivity were extremely well con-strained, one might be able to infer the presence ofthe localized resistor. Information on deeper basementresistivity can be obtained in some circumstances usingmagnetotelluric (MT) data (e.g., Constable et al., 1998;Hoverston et al., 1998), and can be interpreted sepa-rately or jointly inverted with the CSEM data. However,even with such MT-derived constraints, interpretationof the results in terms of a resistive prospect may bechallenging.

CSEM data analysis and inversionThe simple examples above serve to illustrate some

of the considerations in understanding the ability ofCSEM data to resolve subseafloor resistivity structure.Modern CSEM data sets acquired in the field in generalcontain a rich coverage of frequencies, source-receiverseparations and (for surveys using seafloor deployedreceivers) source receiver geometries. These data canbeused toconstrain resistivity andanisotropy in theback-ground structure, and the resistivity and lateral geometryof more localized resistors. Interpretation approaches ingeneral proceed in stages starting with simple 1D ap-proaches, moving through 2D analysis, and culminatingin 3D analysis and inversion if this is required.

In the first stage, 1D forward modeling and inversionapproaches are applied. These assume that the earth is

a 1D stack of layers, which can be either anisotropic orisotropic. The source fields are still those of a point di-pole (and are therefore 3D), and so the data analyzedcan include inline and multiazimuth fields, i.e., 3D datacan be analyzed using a 1D approach for constraint ofbulk resistivity structure across an area, and to examinethe presence and strength of anisotropy in the section.The assumption that the source can be modeled as apoint dipole is valid when the source receiver separa-tion is more than approximately four source dipolelengths (Chave and Cox, 1982). At shorter offsets, theeffect of a finite dipole can easily be calculated by in-tegrating the effect of point dipoles.

One-dimensional forward models or inversions de-rived locally at receiver locations can be stitchedtogether to derive resistivity pseudo sections or vol-umes (Chave and Cox, 1982; Andreis and MacGregor,2008). This process is similar to that used in seismic in-version where locally 1D models are used in the inver-sion and stitched together (with some cross traceregularization) to give an impedance section or volume.There is a major difference between the seismic and EMcases, however: In seismic, this approach is a fairlygood approximation. In marine CSEM, it is in generalnot, and resistivity sections/volumes derived in thisway should be used only for looking at bulk trendsand lateral variations in resistivity and anisotropy,and developing starting models for higher dimensionalinterpretation approaches.

Two-dimensional inversion approaches assume thatthe earth is invariant along one horizontal direction,

usually taken to be perpendicular tothe survey line (Unsworth et al., 1993;MacGregor, 1999; Key and Ovall, 2011).The source fields are again 3D, and soas in 1D approaches inline and multi-azimuth data can be included in theanalysis, although it is conventional toinclude only inline (transmission line-parallel) data. The results are verticalsections through the resistivity structureof the seafloor which can be stitched to-gether to provide a resistivity volume ifappropriate. In many situations, a 2D in-version approach, carefully constrainedwith seismic structural information, canprovide the information required tomeet survey goals without recourse tofull 3D inversion. If full 3D inversion isrequired, the 2D stage is critical in en-suring a robust starting model is con-structed, so that 3D run times can bereduced as far as possible. Such 2Dapproaches also have the advantageof being reasonably computationallyefficient, allowing multiple inversionscenarios, parameterisations, and con-straints to be applied to ensure thatthe result is robust. Hypothesis testing

Figure 14. (a) Synthetic model as in Figure 13; however, with the localized25 Ωm target now less than 500 m from the resistive basement fault block. Am-plitude and phase data at 0.08 Hz, 0.24 Hz, and 0.56 Hz were generated and con-taminated with Gaussian noise. Receiver positions are shown by white triangles.(b) Inversion of the synthetic data. In this case, the localized target cannot beresolved independently from the deeper basement structure.

SH24 Interpretation / August 2014

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

using synthetic forward modeling and inversion basedon geologic models of the area under study can also beused to further understand the inversion results.

Three-dimensional forward modeling and inversionapproaches (e.g., Maao, 2007; Commer and Newman,2008) provide the best approximation to the true resis-tivity structure because the earth and the source fieldsare assumed to be 3D. However, 3D inversion is a time-consuming process with run times on the order ofweeks to reach a converged solution. For this reason,the construction of robust starting models using 1D and2D approaches, and the constraint of the inversion itselfwith well log, seismic, or geologic information is criticalto ensure the 3D inversion process is as efficient as pos-sible. The result is an isotropic or anisotropic resistivityvolume.

As in the survey design process, technical consider-ations and constraints imposed by budget and timingmust be taken into account when considering the bestapproach to data analysis. It is not necessarily the casethat in all situations a 3D inversion approach will bemore effective at answering a given geophysical questionthan a carefully executed 1D or 2D inversion analysis:Workflows must be tailored for the problem at hand.

The result of the analysis is an understanding of thedistribution of resistivity in the earth. However, this isseldom the ultimate goal of a study: These variations inresistivity must be interpreted in terms of the underly-ing rock and fluid properties controlling them, and thisrequires careful integration of the results with seismicand well log data.

Integrated interpretation of CSEM dataWhen CSEM data are considered in isolation, struc-

tural resolution is poor and the results can be ambigu-ous because the effect of an increase in pore fluidresistivity cannot be distinguished from the effect ofa decrease in porosity. The presence of resistors inthe section; for example, cemented sandstones, tightcarbonates, volcanics, or salt will also complicate theinterpretation. Integration of CSEM-derived resistivitymeasurements with complementary seismic based mea-surements, under a rock physics framework can allowsome of the ambiguities inherent in each method to beresolved.

The term “integrated interpretation” can cover awide variety of different approaches and workflowsand is used in the literature to mean many differentthings; however, generally these fall into four broadclasses:

1) Recognizing the poor structural resolution of CSEMdata, seismic data is often used to condition the in-version of CSEM data by providing structural con-straints. Starting models can be constructed usingseismic stratigraphy as a guide, and seismic hori-zons included as smoothness breaks at known boun-daries in the structure; for example, top salt orbasalt, top reservoir or top basement. Such strate-

gies improve the resolution of the CSEM result(MacGregor and Sinha, 2000; Hansen and Mittet,2009; Lovatini et al., 2009).

2) Following inversion of the CSEM data, corenderingapproaches are perhaps the simplest category of in-tegration approaches (e.g., Lovatini et al., 2009; Mac-Gregor et al., 2012; Alcocer et al., 2013). Resistivitysections or volumes derived from inversion ofCSEM data are corendered with seismic data orattributes, and the correlation (or lack of it) be-tween the two is used to draw inferences on thegeology under study. This can be a very powerfulfirst look tool; however, it fails to address the lowervertical resolution of the CSEM method comparedto the seismic. Resistive bodies may be mappedat the wrong depth by unconstrained inversion ap-proaches, and if corendered and interpreted along-side seismic without care could lead to erroneousgeologic conclusions. Nor does it address the ques-tion of the underlying cause of the resistivity varia-tions observed. A zone of high resistivity could aseasily be caused by a lithological variation as by achange in hydrocarbon charge. Correlation betweenstructural closures and zones of high resistivity canperhaps provide confidence in an interpretation ofhydrocarbon; however, the underlying question re-mains: What is the cause of the resistivity variation?

3) Integrated interpretation approaches start to ad-dress this question, by using rock physics relation-ships to interpret the observed variations in seismicand CSEM-derived attributes in terms of the under-lying rock and fluid properties. Seismic data are in-verted first to provide acoustic and/or elasticimpedance, and derived properties such as Pois-son’s ratio. Similarly CSEM data are inverted toprovide a measure of resistivity and resistivityanisotropy. These are then coupled though rockphysics relations, either theoretical, or empiricallyderived and calibrated from well log information,and jointly interpreted to understand variations inrock and fluid properties (e.g., Harris et al., 2009;MacGregor, 2012; Morten et al., 2012).

4) Finally, joint inversion approaches seek to do bothsteps at once by inverting directly for a model thatsatisfies seismic and CSEM data sets simultane-ously. Various approaches to joint inversion havebeen investigated in the past. These fall broadly intotwo categories (de Stefano et al., 2011; Moorkampet al., 2011). In the first category, are methods wherethe coupling between electric and elastic propertiesis primarily structural, constraining variations in re-sistivity to be spatially coincident with variations inelastic properties. Approaches where the differentphysical properties are coupled through cross-gra-dient functions (e.g., Haber and Oldenburg, 1997;Gallardo and Meju, 2004) fall into this category.Such methods are extremely powerful in caseswhere the variations in properties are known tobe coincident, or where a direct relationship

Interpretation / August 2014 SH25

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

between electric and elastic properties is uncertainor hard to obtain. A second approach is to relate theelastic and electric properties of the earth directlyusing either deterministic (e.g., Hou et al., 2006;Hoversten et al., 2006; Gao et al., 2012) or statistical(e.g., Chen and Hoversten, 2012) rock physics rela-tionships. In this case, such rock physics relation-ships are used to relate elastic and electricproperties to a set of reservoir properties, such asporosity, fluid saturation, or shale volume. Althoughthis approach has a good physical grounding, thenonlinear nature of these relationships, uncertain-ties in them, and the possibility that the relationshipvaries (for example with lithology variations) overthe inversion domain, has the potential to introduceadditional nonuniqueness into the inversion.

Any workflow for the integration or joint inversion ofdifferent data types has to deal with a number of issues.Firstly, measurements made using very different physi-cal processes (electric and elastic in the case of CSEMand seismic) must be combined and linked to the under-lying rock and fluid properties in a consistent fashion.This requires a rock physics framework to be either nu-merically derived or empirically calibrated at well loca-tions. In both cases, such models are subject touncertainty, which in turn leads to uncertainty in theresulting interpretation.

Secondly, seismic, CSEM, and well log techniquessample the earth at very different scales, varying froma few centimeters in the case of well logs, to hundredsof meters for CSEM. For example, using seismic meth-ods, the details of layering within a reservoir intervalmay be resolved, whereas CSEM methods are likelyonly to be sensitive to the bulk properties of the reser-voir interval and not to the details of the fluid distribu-tion within it. These different scales must be reconciledin an integrated interpretation or joint inversion ap-proach.

Finally, in order for an integration approach to besuccessful, seismic and CSEM methods must be sensi-tive to the interval of interest and changes in propertieswithin it. Although this is perhaps an obvious statement,

it is nonetheless a key consideration in determiningwhere such approaches can be applied. For example,a reservoir that can be imaged and constrained seismi-cally may lie at too great a depth below mudline, or beembedded in too complex or resistive a backgroundstructure for CSEM methods to be effective. Similarly,low-saturation gas clouds above a reservoir may renderseismic method ineffective, while having little or no ef-fect on the CSEM response or interpretation.

Seismic and CSEM sensitivity to rock and fluidproperties

Detailed interpretation of CSEM-derived resistivityin terms of the underlying rock and fluid propertiesis fraught with ambiguity if considered in isolation. Thisis illustrated in Figure 15, which shows the variation inP-wave velocity, S-wave velocity, and resistivity withporosity and gas saturation of a clean sandstone reser-voir. Archie’s law (Archie, 1942) is used to calculate thevariation in resistivity

ρb ¼ aϕ−mS−nw ρw; (2)

where ρb is the bulk resistivity of the reservoir, ϕ is theporosity, Sw is the water saturation, and ρw is the inter-stitial water resistivity. The empirical parameters a, m,and n, respectively the tortuosity factor, cementationexponent, and saturation exponent, are set to 1, 2,and 2, respectively, in this example. The soft sandmodel of Dvorkin and Nur (1996) combined with Gass-man’s equation (Gassman, 1951; Mavko et al., 2009) areused to calculate the variation in seismic velocity.

Looking first at the resistivity, it is clear that there isa trade-off between porosity and saturation in the res-ervoir: A given value of resistivity may be explained byeither a higher porosity sandstone saturated with resis-tive hydrocarbon fluids, or by a much lower porositysandstone saturated with brine. These alternatives can-not be distinguished using resistivity alone. Velocity anddensity, in this example, are primarily controlled by theporosity of the reservoir. Significant variations with sat-uration are confined to low gas saturations (the area ofhigh Sw).

Figure 15. Variation in P-wave velocity (a), S-wave velocity (b), and resistivity (c) for a simple clean sand reservoir in which thesaturation of gas is varied.

SH26 Interpretation / August 2014

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

Figure 16 illustrates the effect that this sensitivity canhave on CSEM and seismic responses. The left-handpanel of Figure 16 shows the clean sand reservoir dis-cussed above embedded at 1570 m below seafloor in abackground structure in which resistivity increasesgradually with depth. The center panel of Figure 16shows the corresponding CSEM response at threefrequencies, calculated for an 80% gas saturated reser-voir and a water wet reservoir. As expected, the pres-ence of gas in the reservoir causes an increase in fieldstrength and an advance of the phase of the response.The response for a reservoir saturated with 20% gasoverlies that of the water wet case: In terms of CSEMresponse, a water wet reservoir cannot be distinguishedfrom one containing a low saturation of gas.

For comparison, the right-hand panel of Figure 16shows the amplitude versus angle (AVA) response forthe same reservoir scenarios, calculated using a two-term Shuey approximation to the Zoeppritz equations(Shuey, 1985). In contrast to the CSEM response, thereis now a large difference between water wet and 20%

gas saturated cases. However, when the seismic re-sponse is considered the difference between 20% and80% gas saturations is small, with both exhibiting a classIII AVO anomaly. The seismic response highlights thepresence of some gas, but the saturation cannot be con-strained well.

When interpreting geophysical data, such uncertain-ties have a large impact. Consider seismic and CSEMdata sets acquired over the reservoir above, chargedwith an 80% gas saturation. The job of the geophysicistis to constrain as accurately as possible the propertiesand fluid content of that reservoir. In the simple casewhere only the reservoir properties are to be examined(in a real example, the reservoir and overburden wouldhave to be constrained), the interpretation problem canbe illustrated by calculating the rms misfit between theacquired data and the geophysical responses of modelswith varying porosity and fluid saturation. These misfitsurfaces are plotted in Figure 17 for the case of seismicdata alone, CSEM data alone, and a joint data set con-sisting of seismic and CSEM data.

Figure 16. (a) Simple resistivity model containing a clean sand reservoir of porosity 28% at a depth of 1570 m below mudline. (b)CSEM receiver gather calculated from the model on the left at three frequencies, and for two reservoir scenarios: water wet and80% gas saturation. Note that the 20% gas saturation case overlies the water wet curve. (c) Seismic AVA response for this reservoir.

Figure 17. Root-mean-square misfit between geophysical data calculated from a 28% porosity, 80% gas saturated reservoir, andthe responses from models with a range of porosity and saturation values. In each case, the white star shows the misfit of the truemodel. (a) Seismic AVO data taken alone. (b) CSEM data taken alone. (c) Joint CSEM-AVO data misfit surface. Note the same colorscale is used in all plots.

Interpretation / August 2014 SH27

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

Consider first the misfit surface when only seismicAVO data are considered (Figure 17a). The true modellies at a misfit minimum; however, this minimum iselongated along the saturation axis, demonstrating that,although the porosity is well-constrained, the gas satu-ration could take almost any value between 100% andapproximately 20% and still be consistent with the data.This leaves a large uncertainty in saturation from anyseismic analysis.

Figure 17b shows the corresponding misfit surfacewhen only CSEM data are considered. Again, the truemodel lies within the zone of minimum misfit as itshould; however, the same misfit would be obtainedif the porosity were much lower, and the water satura-tion higher. The minimummisfit area is elongated alonga line of constant transverse resistance. Using CSEMdata alone, there is a tradeoff between porosity andsaturation.

Figure 17c shows the combined misfit surface whenseismic AVO and CSEM data are considered. In thiscase, the seismic constraint on the porosity resolvesthe saturation-porosity trade-off inherent in interpreta-tion of the CSEM data: The minimum in misfit isnow much better localized around the true model.Although this is a simple illustration, using perfect syn-thetic data, it demonstrates the advantages of combin-ing multiple data types so that the strengths of each areutilized.

Figure 18 shows a sensitivity analysis applied to welllog data from the North Sea. Given a well log throughthe intervals of interest, rock physics models linking re-sistivity to the underlying lithology and fluid propertiescan be built and calibrated. These models can then beused to alter the fluid content of the reservoir and pre-dict the change in resistivity. Potential ambiguities ininterpretation (for example, the presence of lithologicalor porosity changes) can also be identified at this stage.

This process is the same fluid substitution approach ap-plied routinely to understand the behavior of seismicproperties, such as impedance or Poisson’s ratio, tochanges in fluid content or other properties of a reser-voir interval.

Within the reservoir interval, which for this examplelies at approximately 2.2 km below the seafloor, Gass-man’s equation (Gassman, 1951; Mavko et al., 2009) hasbeen used to calculate the variation in acoustic andelastic parameters with fluid type and saturation, andArchie’s law (Archie, 1942) has been applied to calcu-late the variation in resistivity. In situ porosity and lith-ology are used in both cases.

Turning first to P-wave velocity-resistivity space(Figure 18a), it is clear that there is little variation inP-wave velocity with changes in the fluid propertiesor saturation. This is to be expected when only acousticparameters are considered. However, there is a largedifference in resistivity between the water saturatedand fizz gas cases, and the more resistive 80% gas or80% oil case. Although a measurement of resistivitycould not distinguish between gas and oil, it could beused to discriminate commercial from noncommercialhydrocarbon saturations, adding significantly to the in-formation available from P-wave velocity alone. How-ever, care must be exercised in the interpretation. Inthis example, a sequence of coals lying just beneaththe reservoir complicates the result. These coal mea-sures have a resistivity that overlaps with the resistivityof the 80% gas or oil cases, and their P-wave velocity isonly marginally lower than the reservoir case. In theP-wave-resistivity domain, these coals add significantambiguity to the interpretation.

If prestack seismic data are available for the analy-sis, elastic attributes may be added to the interpreta-tion. This is illustrated in Figure 18b, which showsthe same data in the acoustic impedance (AI)-Poisson’s

Figure 18. Fluid substitution example illustrating the sensitivity of resistivity, acoustic and elastic parameters based on a NorthSea well. (a) Shows a crossplot of resistivity against P-wave velocity. (b) Shows acoustic impedance against Poisson’s ratio. In bothcases, the in situ background trend is shown color-coded by clay content.

SH28 Interpretation / August 2014

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

ratio (PR) domain. There is significantly more sensitiv-ity to the properties of the reservoir fluid content in thisdomain, although the ambiguity between noncommer-cial and commercial hydrocarbon saturation remains.However, the subreservoir coal measures in this do-main are clearly separated from the fluid cases havingsignificantly higher PR. Resistivity and elastic parame-ters are therefore required in this example to provide arobust interpretation: the resistivity to assist in distin-guishing noncommercial from commercial hydrocar-bon saturations, and the elastic to characterize thecoal measures and ensure that high resistivitiesassociated with these are not misinterpreted.

DiscussionThe CSEM method has strengths and weaknesses

like all geophysical approaches. The strength of themethod is that it can provide valuable information onthe resistivity of the subsurface. This resistivity attrib-ute is complementary to more traditional attributes de-rived from seismic data, and when interpreted carefullyin a seismic framework is particularly helpful in deter-mining saturation within reservoir intervals.

However, the method also has weaknesses. Thestructural resolution achievable with CSEM data takenalone is poor, because CSEM makes a bulk measure-ment of a volume of the earth, using physics governedby the diffusion equation, rather than the wave equationas in the case of seismic data. Resolution in a verticalsense in particular is challenging: Often, the uncertaintyin vertical positioning of features can be severalhundred meters. Seismic data, however, providesextremely good structural and stratigraphic constraint,and so these weaknesses in the CSEM method can bemitigated through careful interpretation and integrationwith such seismic derived information.

As with any geophysically derived attribute, the re-sistivity attribute obtained from CSEM analysis mustalso be interpreted carefully. This is best achieved whenwell log calibration is available so that the cause of var-iations in resistivity, and indeed variations in seismicattributes or responses, can be understood in termsof the underlying rock and fluid properties.

This combination of strengths and limitations shouldperhaps guide the CSEMmethod to areas in oil field life-cycle where it can bring the most value to an analysis. Inrank exploration, in the absence of well log data andwith little or no seismic data, the CSEM interpretationproblem is poorly posed. An unconstrained inversion ofCSEM data will have poor resolution in terms of struc-ture, and even if resistivity variations are observed,there is no way to know the underlying cause: Is it alithology or fluid effect? Similarly, the absence ofhigh-resistivity features in an area may be the resultof poorly targeted acquisition parameters resultingfrom a lack of understanding of the background geol-ogy, or a lower-than-expected hydrocarbon chargedreservoir resistivity, and therefore cannot by itself betaken to rule out the presence of commercial hydrocar-

bons. Therefore, although it is of course possible to ap-ply CSEMmethods in frontier areas (e.g., Lovatini et al.,2009; Fanavoll, 2012), the interpretation risks resultingfrom the lack of constraint and calibration can beextremely high.

In more targeted exploration or prospect-ranking ex-ercises where seismic and (perhaps) well log data areavailable, along with detailed interpretations of struc-ture and stratigraphy, the CSEM interpretation processbecomes better constrained, and the answer achieved islikely to be more certain. Reservoir appraisal is betterstill, where seismic and well log data are available, andthe reservoir properties are known at the well location.Here, the question for the CSEM (and seismic) analysisto answer is how these properties vary away from thecalibration points.

Ultimately, CSEM methods may find a use in reser-voir monitoring where understanding changes in thefluids within a reservoir over time is the goal. Therehave been numerous studies of this application under-taken (e.g., Orange et al., 2009; Andreis and MacGregor,2011); however, the method has yet to be tried in thefield on an offshore reservoir.

ConclusionsThis paper has sought to provide an overview of

CSEM methods, the challenges and pitfalls that mustbe overcome when applying them, and considerationsfor designing a successful survey campaign. It is clearthat resistivity, which can be derived from CSEM mea-surements, can provide useful additional informationon subseafloor lithology and fluid properties in explo-ration, appraisal, and perhaps ultimately reservoir mon-itoring. Whatever the application, the best results areachieved when CSEM is analyzed and interpreted inan integrated framework in conjunction with welllog, seismic, and other available data types.

AcknowledgmentsThanks to many colleages at RSI who have helped

with this paper including Michelle Ellis, Slim Bou-chrara, and Kim Nichols. Thanks to BGP for permissionto use one of the synthetic case studies, and to RSI forpermission to publish this paper. Thanks also to JohnO’Brien, Antony Price, Mike Hoverston, Stefan Helwig,and an anonymous reviewer for providing helpful com-ments that greatly improved this paper.

ReferencesAlcocer, J. A. E., M. V. Garcia, H. S. Soto, D. Baltar, V. R.

Paramo, P. Gabrielson, and F. Roth, 2013, Reducing un-certainty by integrating 3D CSEM in the Mexican deep-water workflow: First Break, 31, 75–79.

Andreis, D., and L. MacGregor, 2008, Controlled sourceelectromagnetic sounding in shallow water: Principlesand applications: Geophysics, 73, no. 1, F21–F32, doi:10.1190/1.2815721.

Interpretation / August 2014 SH29

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

Andreis, D., and L. MacGregor, 2011, Using CSEM to mon-itor production from a complex 3D gas reservoir: A syn-thetic case study: The Leading Edge, 1070–1079, doi: 10.1190/1.3640531.

Archie, G. E., 1942, The electrical resistivity log as an aid indetermining some reservoir characteristics: Transac-tions of the American Institute of Mechanical Engi-neers, 146, 54–62.

Bachrach, R., 2011, Elastic and resistivity anisotropy ofshale during compaction and diagenesis: Joint effectivemedium modeling and field observations: Geophysics,76, no. 6, E175–E186, doi: 10.1190/geo2010-0381.1.

Barker, N. D., J. P. Morten, and D. V. Shantsev, 2012, Opti-mizing EM data acquisition for continental shelf explo-ration: The Leading Edge, 1276–1284, doi: 10.1190/tle31111276.1.

Brown, V., M. Hoversten, K. Key, and J. Chen, 2012, Res-olution of reservoir scale electrical anisotropy frommarine CSEM data: Geophysics, 77, no. 2, E147–E158, doi: 10.1190/geo2011-0159.1.

Bhuiyan, A., R. Sakariassen, Ã. Hallanger, and A. McKay,2013, Modelling and interpretation of CSEM data fromBressay, Bentley and Kraken area East Shetland plat-form, North Sea: 83rd Annual International Meeting,SEG, Expanded Abstracts.

Chave, A., S. Constable, and R. N. Edwards, 1991, Electri-cal exploration methods for the seafloor, in M. N. Nam-bighian, ed., Electromagnetic methods in appliedgeophysics: SEG, 931–996.

Chave, D., and C. Cox, 1982, Controlled electromagneticsources for measuring electrical conductivity beneaththe oceans, 1, forward problem and model study: Jour-nal of Geophysical Research, 87, 5327–5338, doi: 10.1029/JB087iB07p05327.

Chen, J., and G. M. Hoversten, 2012, Joint inversion ofmarine seismic AVA and CSEM data using statisticalrock physics and Markov random fields: Geophysics,77, no. 1, R65–R80, doi: 10.1190/geo2011-0219.1.

Colombo, D., G. MacNiece, E. S. Curiel, and A. Fox, 2013,Full tensor CSEM and MT for subsalt structural imagingin the Red Sea: Implications for seismic and electromag-netic integration: The Leading Edge, 436–449, doi: 10.1190/tle32040436.1.

Commer, M., and G. Newman, 2008, New advances in threedimensional controlled source electromagnetic inver-sion: Geophysical Journal International, 172, 513–535,doi: 10.1111/j.1365-246X.2007.03663.x.

Constable, S., 2010, Ten years of marine CSEM for hydro-carbon exploration: Geophysics, 75, no. 5, 75A67–75A81, doi: 10.1190/1.3483451.

Constable, S., 2013, Review paper: Instrumentation formarine magnetotelluric and controlled source electro-magnetic sounding: Geophysical Prospecting, 61,505–532, doi: 10.1111/j.1365-2478.2012.01117.x.

Constable, S., and C. S. Cox, 1996, Marine controlledsource electromagnetic sounding II: The PEGASUS

experiment: Journal of Geophysical Research, 101,5519–5530, doi: 10.1029/95JB03738.

Constable, S., P. Kannberg, K. Callaway, and D. RamirezMejia, 2012, Mapping shallow geological structurewith towed marine CSEM: 82nd Annual InternationalMeeting, SEG, Expanded Abstracts, doi: 10.1190/segam2012-0839.1.

Constable, S., A. Orange, G. M. Hoverston, and F. Morrison,1998, Marine magnetotellurics for petroleum explora-tion Part 1: A seafloor equipment system: Geophysics,63, 816–825, doi: 10.1190/1.1444393.

Constable, S., R. Parker, and C. Constable, 1987, Occam'sinversion: A practical algorithm for generating smoothmodels from EM sounding data: Geophysics, 52, 289–300, doi: 10.1190/1.1442303.

Constable, S., and L. Srnka, 2007, An introduction tomarine controlled source electromagnetic methodsfor hydrocarbon exploration: Geophysics, 72, no. 2,WA3–WA12, doi: 10.1190/1.2432483.

Cuevas, N. H., and D. Alumbaugh, 2011, Near source re-sponse of a resistive layer to a horizontal or verticalelectric dipole excitiation: Geophysics, 76, no. 6,F353–F371, doi: 10.1190/geo2010-0359.1.

De Stefano, M., F. G. Andreasi, S. Re, M. Virgilio, and F. F.Snyder, 2011, Multiple domain, simultaneous joint in-version of geophysical data with application to sub-saltimaging: Geophysics, 76, no. 3, R69–R80, doi: 10.1190/1.3554652.

Dvorkin, J., and A. Nur, 1996, Elasticity of high-porositysandstones: Theory for two North Sea datasets: Geo-physics, 61, 1363–1370, doi: 10.1190/1.1444059.

Edwards, R. N., 2005, Marine controlled source electro-magnetics: Principles, methodologies and future com-mercial applications: Surveys in Geophysics, 26, 675–700, doi: 10.1007/s10712-005-1830-3.

Eidesmo, T., S. Ellingsrud, L. M. MacGregor, S. Constable,M. C. Sinha, S. Johansen, F.-N. Kong, and H. Westerdahl,2002, Sea bed logging (SBL), a new method for remoteand direct identification of hydrocarbon filled layers indeepwater areas: First Break, 20, 144–152.

Ellingsrud, S., T. Eidesmo, S. Johansen, M. C. Sinha, L. M.MacGregor, and S. Constable, 2002, Remote sensing ofhydrocarbon layers using sea-bed logging (SBL): Re-sults of a cruise offshore West Africa: The LeadingEdge, 21, 972–982, doi: 10.1190/1.1518433.

Ellis, M. E., M. C. Sinha, and R. Parr, 2010, Role offine scale layering and grain alignment in the elec-trical anisotropy of marine sediments: First Break,28, 49–56.

Evans, R. L., M. C. Sinha, S. Constable, and M. J. Unsworth,1994, On the electrical nature of the axial melt zone at13° North on the East Pacific Rise: Journal of Geophysi-cal Research, 99, 577–588, doi: 10.1029/93JB02577.

Fan, Y., R. Snieder, E. Slob, J. Hunziker, J. Singer, J. Shei-man, and M. Rosenquist, 2012, Increasing the sensitivityof controlled-source electromagnetics with synthetic

SH30 Interpretation / August 2014

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

aperture: Geophysics, 77, no. 2, E135–E145, doi: 10.1190/geo2011-0102.1.

Fanavoll, S., S. Ellingsrud, P. T. Gabrielsen, R. Tharimela,and D. Ridyard, 2012, Exploration with the use of EMdata in the Barents Sea: The potential and the chal-lenges: First Break, 30, 89–96.

Gabrielsen, P. T., P. Abrahamson, M. Panzer, S. Fanavoll,and E. Ellingsrud, 2013, Exploring frontier areas using2D seismic and 3D CSEM data, as exemplified by multi-client data over the Skrugard and Havis discoveries inthe Barents Sea: First Break, 31, 63–71.

Gallardo, L. A., and M. A. Meju, 2004, Joint two-dimensional DC resistivity and seismic travel timeinversion with cross-gradient constraints: Journal ofGeophysical Research, 109, BO3311, doi: 10.1029/2003JB002716.

Gao, G., A. Abubakar, and T. M. Habashy, 2012, Joint pet-rophysical inversion of electromagnetic and full wave-form seismic data: Geophysics, 77, no. 3, WA3–WA18,doi: 10.1190/geo2011-0157.1.

Gassman, J., 1951, Uber die elastizitat poroser medien:Viertel. Naturforsch Gesellschaft Zurich, 96, 1–23.

Gist, G., A. Ciucivara, R. Houck, M. Rainwater, D. Willen,and J. J. Zhou, 2013, Case study of a CSEM false posi-tive, Orphan Basin, Canada: 83rd Annual InternationalMeeting, SEG, Extended Abstracts, 805–809, doi: 10.1190/segam2013-0307.1.

Haber, E., and D. Oldenburg, 1997, Joint inversion: A struc-tural approach: Inverse Problems, 13, 63–77, doi: 10.1088/0266-5611/13/1/006.

Hansen, K. R., and R. Mittet, 2009, Incorporating seismichorizons in inversion of CSEM data: 79th AnnualInternational Meeting, SEG, Expanded Abstracts,694–698.

Harris, P., Z. Du, L. MacGregor, W. Olsen, R. Shu, and R.Cooper, 2009, Joint interpretation of seismic and CSEMdata using well log constraints: An example from theLuva field: First Break, 27, 76–81.

Hesthammer, J., A. Stefatos, M. Boulenko, S. Fanavoll, andJ. Danielsen, 2010, CSEM performance in light of wellresults: The Leading Edge, 29, 34–41, doi: 10.1190/1.3284051.

Holten, T., E. G. Flekkoy, B. Singer, E. M. Blixt, A. Hanssen,and K. J. Maloy, 2009, Vertical source, vertical receiverelectromagnetic technique for offshore hydrocarbonexploration: First Break, 27, 89–93.

Hou, Z., Y. Rubin, G. M. Hoversten, D. Vasco, and J. Chen,2006, Reservoir-parameters identification using mini-mum-entropy based Bayesian inversion of seismicAVA and marine CSEM data: Geophysics, 71, no. 6,O77–O88, doi: 10.1190/1.2348770.

Hoversten, G. M., F. Cassassuce, E. Gasperikova, G. New-man, J. Chen, Y. Rubin, Z. Hou, and D. Vasco, 2006, Di-rect reservoir parameter estimation using jointinversion of marine seismic AVA and CSEM data: Geo-physics, 71, no. 3, C1–C13, doi: 10.1190/1.2194510.

Hoverston, G. M., F. Morrison, and S. Constable, 1998,Marine magnetotellurics for petroleum exploration PartII: Numerical analysis of sub-salt resolution: Geophys-ics, 63, 826–840, doi: 10.1190/1.1444394.

Key, K., 2009, 1D inversion of multicomponent, multifre-quency marine CSEM data: Methodology and syntheticstudies for resolving thin resistive layers: Geophysics,74, no. 2, F9–F20, doi: 10.1190/1.3058434.

Key, K., and J. Ovall, 2011, A parallel goal-oriented adap-tive finite element method for 2.5-D electromagneticmodelling: Geophysical Journal International, 186,137–154, doi: 10.1111/j.1365-246X.2011.05025.x.

Klein, J. D., P. R. Martin, and D. F. Allen, 1997, The petro-physics of electrically anisotropic reservoirs: The LogAnalyst, 38, 25–36.

Loseth, L. O., T. Wiik, P. A. Olsen, A. Becht, and J. O.Hansen, 2013, CSEM exploration in the Barents Sea,Part 1: Detecting Skrugard from CSEM: 75th AnnualConference and Exhibition, EAGE, Extended Abstracts.

Lovatini, A., K. Umpbach, and S. Patmore, 2009, 3D CSEMin a frontier basin offshore Greenland: First Break, 27,95–98.

Luling, M. G., 2013, The paradox of anisotropy in electriclogging: A simple proof and extensions to other physicsdomains: Geophysics, 78, no. 1, W1–W8, doi: 10.1190/geo2012-0123.1.

Maao, F., 2007, Fast finite-difference time-domain model-ling for marine-subsurface electromagnetic problems:Geophysics, 72, no. 2, A19–A23, doi: 10.1190/1.2434781.

MacGregor, L. M., 1999, Marine controlled source electro-magnetic sounding: Development of a regularised inver-sion for 2D resistivity structures: LITHOS sciencereport, 1, 103–109.

MacGregor, L. M., 2012, Integrating seismic, CSEM andwell log data for reservoir characterisation: The LeadingEdge, 268–277.

MacGregor, L. M., S. Bouchrara, J. T. Tomlinson, U.Strecker, J. Fan, X. Ran, and G. Yu, 2012, Integratedanalysis of CSEM, seismic and well log date for pros-pect appraisal: A case study from West Africa: FirstBreak, 30, 77–82.

MacGregor, L. M., S. C. Constable, and M. C. Sinha, 1998,The RAMESSES experiment III: Controlled sourceelectromagnetic sounding of the Reykjanes Ridge at57°45’N: Geophysical Journal International, 135, 773–789, doi: 10.1046/j.1365-246X.1998.00705.x.

MacGregor, L. M., and M. C. Sinha, 2000, Use of marinecontrolled source electromagnetic sounding for sub-basalt exploration: Geophysical Prospecting, 48,1091–1106, doi: 10.1046/j.1365-2478.2000.00227.x.

MacGregor, L. M., M. C. Sinha, and S. Constable, 2001,Electrical resistivity structure of the Valu Fa Ridge,Lau Basin, from marine controlled source electromag-netic sounding: Geophysical Journal International,146, 217–236, doi: 10.1046/j.1365-246X.2001.00440.x.

Interpretation / August 2014 SH31

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

Mattsson, J., F. Englemark, and C. Anderson, 2013, Towedstreamer EM: The challenges of sensitivity andanisotropy: First Break, 31, 155–159.

Mattsson, J., P. Lindqvist, R. Juhasz, and E. Bjornemo,2012, Noise reduction and error analysis for a towedEM system: 72nd Annual International Meeting, SEG,Expanded Abstracts, 1–5.

Mavko, G., T. Mukerji, and J. Dvorkin, 2009, The rock phys-ics handbook: Cambridge University Press, ISBN 978-0-521-86136-6.

Mittet, R., and T. Schaug-Pettersen, 2008, Shaping opti-mal transmitter waveforms for marine CSEM surveys:Geophysics, 73, no. 3, F97–F104, doi: 10.1190/1.2898410.

Moorkamp, M., H. Heincke, M. Jegen, A. W. Roberts, and R.W. Hobbs, 2011, A framework for 3D joint inversion ofMT, gravity and seismic refraction data: GeophysicalJournal International, 184, 477–493, doi: 10.1111/j.1365-246X.2010.04856.x.

Morten, J. P., F. Roth, S. A. Karlsen, D. Tampko, C. Pacurar,P. A. Olsen, A. K. Nguyen, and J. Gjengedal, 2012, Fieldappraisal and accurate resource estimation from 3Dquantitative interpretation of seismic and CSEM data:The Leading Edge, 448–456.

Myer, D., S. Constable, and K. Key, 2011, Broad band wave-forms and robust processing for marine CSEM surveys:Geophysical Journal International, 184, 689–698, doi: 10.1111/j.1365-246X.2010.04887.x.

Newman, G., M. Commer, and J. Carazzone, 2010, ImagingCSEM data in the presence of anisotropy: Geophysics,75, no. 2, F51–F61, doi: 10.1190/1.3295883.

Orange, A., K. Key, and S. Constable, 2009, The feasibilityof reservoir monitoring using time-lapse marine CSEM:Geophysics, 74, no. 2, F21–F29, doi: 10.1190/1.3059600.

Ramananjaona, C., L. MacGregor, and D. Andreis, 2011, In-version of marine electromagnetic data in a uniaxialanisotropic stratified earth: Geophysical Prospecting,59, 341–360, doi: 10.1111/j.1365-2478.2010.00919.x.

Shuey, R. T., 1985, A simplification of the Zoeperitz equa-tions: Geophysics, 50, 609–614, doi: 10.1190/1.1441936.

Sinha, M. C., S. C. Constable, C. Peirce, A. White, G. Hein-son, L. M. MacGregor, and D. A. Navin, 1998, Magmatic

processes at slow spreading ridges: Implications of theRAMESSES experiment at 57°45’ North on the Mid-Atlantic Ridge: Geophysical Journal International,135, 731–745, doi: 10.1046/j.1365-246X.1998.00704.x.

Srnka, L. J., 1986, Method and apparatus for offshoreelectromagnetic sounding utilizing wavelength effectsto determine optimum source and detector positions:U.S. Patent 4,617,518.

Unsworth, M., B. Travis, and D. Chave, 1993, Electromag-netic induction by a finite electric dipole over a 2Dearth: Geophysics, 58, 198–214, doi: 10.1190/1.1443406.

Wertmuller, D., A. Ziolkowski, and D. Wright, 2013, Back-ground resistivity model from seismic velocity: Geo-physics, 78, no. 4, E213–E223, doi: 10.1190/geo2012-0445.1.

Young, P. D., and C. S. Cox, 1981, Electromagnetic activesource sounding near the East Pacific Rise: Geophysi-cal Research Letters, 8, 1043–1046, doi: 10.1029/GL008i010p01043.

Yuan, J., and R. N. Edwards, 2000, The assessment ofmarine gas hydrates through electrical remote sound-ing: Hydrate without the BSR?: Geophysical ResearchLetters, 27, 2397–2400, doi: 10.1029/2000GL011585.

Zhdanov, M. S., M. Endo, M. Cuma, J. Linfoot, L. Cox, andG. Wilson, 2012, The first practical 3D inversion oftowed streamer EM data from the Troll field trial:82nd Annual International Meeting, SEG, Expanded Ab-stracts, doi: 10.1190/segam2012-0750.1.

Ziolkowski, A., 2007, Developments in the transientelectromagnetic method: First Break, 25, 99–106.

Ziolkowski, A., D. Wright, and J. Mattsson, 2011, Compari-son of pseudo random binary sequence and squarewave transient controlled source electromagnetic dataover the Peon gas discovery, Norway: Geophysical Pro-specting, 59, 1114–1131, doi: 10.1111/j.1365-2478.2011.01006.x.

Biographies and photographs of other authors are notavailable.

SH32 Interpretation / August 2014

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/