ieee signal processing

IEEE SIGNAL PROCESSING MAGAZINE [36] JULY 2012

Digital Object Identifier 10.1109/MSP.2012.2192529

[Anton Ziolkowski and David Wright]

Date of publication: 12 June 2012

1053-5888/12/$31.00©2012IEEE

In this article, we show that the controlled source elec-tromagnetic (CSEM) meth-od is complementary to the seismic method. Land and marine CSEM

methods have developed almost independently. Both methods are discussed, but the focus is more on the applica-tion of the marine CSEM method, since this has had the most attention in the past decade. Active methods using man-made EM sources are used to investigate subsurface reservoirs and, in principle, are able to distinguish between those that are saturated with electrically resistive hydrocarbons and those that are saturated with electrically conductive brine. Therefore, they have the potential to rank the prospectivity of structures known from seismic data, but before drilling. Novel techniques for the processing of marine CSEM data include removal of the airwave that travels through the air at the speed of light and the suppression of magnetotelluric (MT) noise. For tran-sient pseudorandom binary sequence (PRBS) data, deconvolution is an important part of signal-to-noise ratio enhancement. EM

data have much lower resolu-tion than seismic data and therefore need to use the sub-surface structure obtained from seismic data plus rock physics relations to constrain resistivities in starting models for inversion.

INTRODUCTIONRocks are composed of miner-als that form a solid matrix containing pores. The fraction of rock volume occupied by pore space is the porosity f.The pores are full of fluids,

sometimes including air. The solid matrix is normally extremely resistive, as there are very few charged objects or ions free to move and conduct electricity. The fluid in the pores, on the other hand, contains ions that can move in the fluid and therefore con-duct electricity. Normally, the fluid is saltwater and the conduc-tivity of the rock depends on the concentration of salt in the water, the fraction of the pore space that contains saltwater, and the freedom of movement of the ions between pores.

Sometimes the pores may also contain hydrocarbons as liquid, gas, or both. When hydrocarbons are present, there are generally three fluid phases: saltwater, hydrocarbon liquid, and hydrocarbon gas—normally methane. The hydrocarbons are not ionized and so they are not conductors of electricity. It follows that the presence

[A powerful tool for hydrocarbon exploration,

appraisal, and reservoir characterization]

IMAGE COURTESY OF U.S. DEPARTMENT OF COMMERCE/NOAA/NESDIS/NATIONAL GEOPHYSICAL DATA CENTER


of hydrocarbons increases the resistivity of the rock. The greater the fluid fraction, or saturation, of hydrocarbons, the greater is the resistivity of the rock. As shown in Figure 1, the effect of replacing saltwater by hydrocarbons can increase the resistivity by orders of magnitude, whereas the effect on seismic P-wave veloci-ty is small. This is the reason the petroleum industry is becoming interested in EM methods in addition to very well-established seismic methods.

EM methods can be divided into passive and active methods. Passive methods use Earth’s naturally occurring EM field to inves-tigate its interior. Active methods use a man-made EM source and one or more receivers to investigate Earth’s interior. This distinc-tion is similar to that for seismic and acoustic methods: passive seismic and acoustic methods record and analyze events that cre-ate vibrations, such as naturally occurring earthquakes and rock bursts, man-made explosions, including underground nuclear tests, and movement of submarines. Active seismic methods are used in exploration especially in the search for potential traps for oil and gas.

Figure 2 illustrates the setup for conventional marine con-trolled source EM surveying (in [2]–[5]), which uses receiver nodes on the seafloor that measure all three components of both the electric and magnetic fields, and an electric dipole source, towed by a vessel, that transmits a continuous signal, typically a square wave or a similar periodic signal, with a spectrum con-taining discrete frequencies. The conventional marine CSEM method has been used mainly in deep water (deeper than 500 m),

but in the last few years it has also been used in shallower water (e.g., [6]). If the source is on the x-axis, a receiver also on the x-axis is in line with the source and is known as the in-line com-ponent. The in-line component of the electric field is used to detect subsurface resistors.

An alternative CSEM technique, which has also been used on land, also uses a horizontal electric dipole source and dipole electric receivers, but the source signal is transient: after a certain amount of time, the earth response reaches a steady state. After a steady state has been reached, the cycle

Fluid Saturation and Rock Properties100

10

10 20 40 60

Brine Saturation (%)

P-Wave Velocity

Resistivity

105

104

Velocity (m

/s)

1,00080 100 120

Res

istiv

ity (

Ω⋅m

)

[FIG1] Resistivity and P-wave velocity as a function of brine saturation for a porous sandstone. Figure redrawn from [1] and used with permission.

Air (Resistive)

Seawater (Very Conductive)

CSEM Transmitter

Seafloor (Variable Conductivity)

Electric and Magnetic Field Receivers

Magnetotelluric Source Fields

[FIG2] The acquisition setup for the marine CSEM method with a towed electric dipole source and nodal seafloor receivers. Figure used with permission from [2].


may be repeated. We describe the transient approach in detail later. One type of transient response is the response to a rever-sal in polarity of a direct current (DC) current (e.g., [7]–[11]). Another type of transient response is the response to a single period of a PRBS, demonstrated by [12] for land data and by [13] for marine data.

Since most of the articles in this issue of IEEE Signal Processing Magazine are concerned with seismic data process-ing, we thought it would be helpful to present the fundamen-tal principles of EM propagation in relation to seismic wave propagation. A plane wave analysis then leads to the concept of skin depth. In the conventional approach to the CSEM method, the source signal is continuous. Transient source signals have also been used and we discuss the issue of source control before discussing the multitran-sient EM method (MTEM). The discuss ion of CSEM data processing focuses on noise reduction and removal of the so-called airwave, although some CSEM experts prefer to inter-pret CSEM data without the use of airwave removal algorithms. The MT signal is a powerful source of noise for CSEM methods, especially in shallow water, and cultural noise associated with electrical equipment is especially important for the land CSEM method.

FUNDAMENTAL PRINCIPLES OF EM PROPAGATION IN CONDUCTING MEDIA

SEISMIC AND EM PROPAGATION: WAVE EQUATION AND DIFFUSION EQUATIONSeismic waves deform the media—fluids and solids—in which they propagate very slightly: the strains are usually much smaller than 1026. After a seismic wave has passed a particular point in the medium, the medium at the point returns to its original state. There is no change in the medium. The energy associated with the wave is contained in the wave. No energy is left behind in the medium after the wave has passed. It follows that there can be no DC component to seismic wave propagation. (This statement does not necessarily apply in a region close to the source, where the dis-placement may be large, the medium may be permanently deformed, and Hooke’s law itself may not apply.) If the recorded seismic data contain DC, this must be a mistake introduced by the recording electronics and is removed from the data in processing before subsequent processing.

EM waves in conducting media—fluids and solids—have both electric and magnetic fields. The electric fields are asso-ciated with currents according to Ohm’s law, the currents generate magnetic fields according to the Biot-Savart law, and changes in the electric and magnetic fields are related by Faraday’s law of EM induction. Maxwell developed his famous theory of electromagnetism by starting with the experimental evidence presented by Faraday. The important point for the

exploration geophysicist is that the electric and magnetic fields are related, and whenever there is current, as there must be in a conducting medium, there are losses. These losses are the principal difference between seismic and EM propagation. And because they are permanent, there must be a DC component to EM wave propagation.

Seismic waves propagate according to the wave equation that is derived from two more fundamental equations: Newton’s second law of mechanics (force equals mass times acceleration) and Hooke’s law of elasticity (stress is propor-tional to strain). In solids, there are two kinds of elastic waves:

longitudinal, or P-waves, in which the particle vibration is parallel to the direction of wave propagation; and shear, or S-waves, in which the particle vibration is perpendicular to the direction of propagation. Fluids have no shear strength and therefore do not support shear

waves. P-waves propagate in fluids and are known as acoustic waves. The wave equation in a fluid is

=2 p21

c2 '

2p

't2 5 0, (1)

where p 1x, y, z, t 2 is the pressure in the fluid at the point 1x, y, z 2 and time t, and c is the velocity of propagation, given by

c5ÅKh

, (2)

where K is the bulk modulus of the fluid and h is its density. For the simple case of a monopole source of acoustic energy in a fluid, the wave equation may be written as

'2 1rp 2'r2 2

1

c2 '2 1rp 2't2 5 0, (3)

where r is the distance from the source. The pressure function p 1r, t 2 is then

p 1r, t 2 5 1r qat2

rcb, (4)

in which q 1t 2 is the source time function with dimensions of pres-sure times distance. In exploration geophysics, the conventional units of q 1t 2 are bar-m (1 bar = 105 Pa).

Figure 3(b) shows the response to a monopole source in an infinite fluid. The arrival time at the receiver is proportional to the distance r from the source, while the amplitude decays as 1/r, as expressed in (4). In all this, we have ignored seismic attenuation, because we are assuming the deformations obey Hooke’s law, which is correct to first order. There is some seismic attenuation, which is a second-order effect, and is caused by rock physics effects that we ignore in this analysis. Figure 3(c) shows the corre-sponding response to an impulsive electric dipole source in an infinite conducting medium. The peak of the response arrives at a time proportional to r 2, while the amplitude decays as 1/r5, which

THE EFFECT OF REPLACING SALTWATER BY HYDROCARBONS CAN INCREASE THE RESISTIVITY

BY ORDERS OF MAGNITUDE, WHEREAS THE EFFECT ON SEISMIC P-WAVE

VELOCITY IS SMALL.


is dramatically different from the purely geometrical decay in the acoustic case.

The electric field at the point 1x, y, z 2 and time t is the vector E 1x, y, z, t 2 , which has units of Vm21 and obeys the following wave equation

= 3 = 3 E1me

'2E't2 1ms

'E't5 0, (5)

in which s is the electrical conductivity; m is the magnetic permeability, which, for most nonmagnetic materials, is normally taken to be equal to the magnetic permeability of free space m05 4p 3 1027 Hm21; and e is the electrical permittivity, normally taken to be the permittivity of free space e05 8.85 3 10212 C2/Nm2. Every component of the vector E field interacts with every other component as the field propagates.

It is important to understand the difference between the EM propagation in air and in the conducting earth. Air is an extremely poor conductor of electricity (3 to 8 3 10215 Sm21). For geophysical purposes, we may regard the conductivity of air as essentially zero. The last term in (5) is then quite negligible and EM wave propagation in air is therefore described by

= 3 = 3 E11

c02 '2E't2 5 0, (6)

in which c05 1/"m0e0 is the velocity of light. That is, it is the familiar EM wave propagation with all waves throughout the EM spectrum traveling at the speed of light.

To tackle the problem of EM propagation in the conducting earth, we transform (5) to the frequency domain using the deriva-tive theorem to yield

= 3 = 3 E2v2m0e0E2 ivm0sE5 0, (7)

where E is the Fourier transform of E and v is angular frequency. For exploration to depths greater than a few meters, low frequen-cies must be used and v V s/e0. Equation (7) then reduces to

= 3 = 3 E2 ivm0sE5 0, (8)

and (5) becomes

= 3 = 3 E1m0s'E't5 0. (9)

For uniform conducting media, where there are no free charg-es, (9) simplifies to the diffusion equation

=2E2m0s'E't5 0, (10)

in which the vector components are independent. The corre-sponding magnetic field intensity H 1x, y, z, t 2 , with units Am21, obeys a similar equation,

=2H2m0s'H't5 0. (11)

The electric and magnetic fields are related by Faraday’s law

= 3 E5 2m0'H't

. (12)

PLANE WAVE ANALYSIS AND SKIN DEPTHThe responses shown in Figure 3(b) and (c) are very different. One important difference is that the shape of the wavelet is indepen-dent of the source-receiver distance for the seismic case but is very dependent on distance for the EM case. Much of seismic data pro-cessing relies on this invariance of the wavelet with distance. These processes are not applicable to the EM data. To understand the reason for the differences, it is convenient to simplify the prop-agation to one dimension.

We choose the z-direction as positive downward and restrict the discussion to waves which propagate vertically. That is, the pressure and the EM fields are functions only of z and t. The wave equation for the pressure (1) can be written as

'2p

'z2 21

c2 '2p

't2 5 0, (13)

while the diffusion equation for the magnetic field (11) can be written as

2,000 m

Receivers 21

(a)

(b)

(c)

Receiver Number

0 5 10 15 20

r /c0.20.40.60.8

11.21.41.61.8

2

000.20.40.60.8

11.21.41.61.8

5 10 15 20T

ime

(s)

Tim

e (s

)

1,000 mr

1

x

z

Receiver Number

Source

[FIG3] (a) Configuration of source and receivers in water of velocity 1,500 m/s and resistivity 0.33 V -m; (b) pressure response at the receivers to acoustic monopole at the origin; and (c) x-component of electric field response to an x-directed current dipole source at the origin.


'2H'z2 2m0s

'H't5 0. (14)

Taking the Fourier transform of these two equations enables the fields to be expressed as functions of frequency v, instead of time t: p 1z, v 2 and H 1z, v 2 . The two equations now become

'2p

'z2 1v2

c2 p5 0, (15)

'2H

'z2 1 ivm0sH5 0. (16)

Comparing (15) and (16), we recognize that the quantity v /im0s is velocity-squared, so "v /im0s is the velocity of propagation of the electric field, which is frequency dependent and complex. Equation (15) has the well-known solution

p 1z, v 2 5 p1exp 1 ivz/c 2 1 p2exp 12ivz/c 2 , (17)

which is a wave of amplitude p1 propagating in the z-direc-tion at velocity c, and a wave of amplitude p2 propagating in the negative z-direction at velocity c. The amplitudes of these waves are p1 and p2 and are determined by the boundary con-ditions. Let’s now consider that there is only a downgoing wave; that is, p2 5 0, and

p 1z, v 2 5 p1exp 1 ivz/v 2 . (18)

Applying exactly the same reasoning to (16), the downgoing wave solution for the magnetic field intensity is

H 1z, v 2 5 H1exp 1"ivm0sz 2 , (19)

which can be written as

H 1z, v 2 5 H1exp 12z"vm0s/2 2exp 1 iz"vm0s/2 2 . (20)

This is a wave propagating in the positive z-direction with a fre-quency-dependent propagation velocity, but its amplitude decays exponentially with increasing z. The amplitude decreases by a fac-tor 1/e for a propagation distance

ds ; Å 2vm0s

. (21)

This distance is known as the skin depth. Conductivity of seawater is about 3.2 Sm21. Conventional CSEM surveys often use a source with fundamental frequency 0.25 Hz for which the skin depth in seawater is 563 m.

The physical significance of skin depth is that it is a seri-ous impediment to the resolution of small targets at depth. The smaller and the deeper the target, the harder it is to detect and resolve. Doubling the thickness of the overburden reduces the highest detectable frequency by a factor of four. Normally, a feasibility study should be done to ascertain the detectability of the target. It is important that this analysis be done with three-dimensional (3-D) not one-dimensional

(1-D) modeling to obtain reasonable estimates of the likely EM responses.

THE EVOLUTION OF THE MARINE CSEM METHODThe conventional marine CSEM method evolved from the marine MT method. Both techniques make measurements of orthogonal components of the electric and magnetic fields using autonomous receiver nodes. The key difference is that the MT method [14] uses the naturally occurring source field generated by solar interaction with the earth’s magnetic field, while the conventional CSEM technique uses a man-made or “controlled” horizontal electric dipole source towed close (50 m) to the seafloor. The source trans-mits a waveform with desired spectral properties into the earth. A continuous square wave with a fundamental frequency of between 0.1 and 1 Hz is often used.

The use of a controlled source was motivated by the need to provide information about the subsurface at frequencies above those available to marine MT in deep water (that is, above about 0.01 Hz). Frequencies in the range 0.01–1 Hz are sensitive to vari-ations in the top few kilometers of the subsurface where hydrocar-bon accumulations are found but are also required to constrain near-surface resistivities for the inversion of the lower frequency MT data and deeper earth resistivity. Another advantage of using a controlled source is that a horizontal dipole produces fields both tangential and orthogonal to the target whereas MT fields are purely tangential (in a 1-D earth). The component of the horizon-tal dipole source that produces a vertical electric field normal to horizontal layer boundaries, however, is particularly sensitive to thin resistive layers characteristic of hydrocarbon reservoirs. The MT source field is sensitive to conductive targets but has poor sen-sitivity to resistors.

The Scripps Institution of Oceanography developed techniques and instrumentation for marine MT [15] in the 1960s and 1970s as academic tools for investigation of the lithosphere and mantle. The use of a controlled source with the existing MT receivers fol-lowed on from the early work in MT [16], [17]. The node-based CSEM system used today is relatively unchanged from the original Scripps system.

A detailed description of the node-based CSEM equipment including the first towed source that was developed at Cambridge is given in [18]. The source is a horizontal electric dipole with cur-rent electrodes separated by 100–400 m and towed in-line by the vessel at about 1.5 kn. The data acquisition setup is illustrated in Figure 2. The source used in deep water CSEM applications trans-mits a high voltage and low current signal down a cable from the vessel to a transformer at the source. This signal is then trans-formed to a low-voltage, high-current signal, which is rectified, polarity reversed at prechosen times, and transmitted between two electrodes with the return current through the water. The high-voltage transmission from the vessel to the source minimizes loss-es suffered in the cable. Due to the conductive nature of the seawater, it is not necessary for the source electrodes to be in con-tact with the seafloor to obtain good coupling. The source is gen-erally towed approximately 50 m above the seafloor. This is close enough that the signal is not attenuated significantly in the water


column and large enough to avoid small topography variation of the seafloor.

An example of two different CSEM node receivers is shown in Figure 4 [19]. The electric field is measured by a pair of perpendicular dipoles each about 10 m in length and the magnetic field is measured by three orthog-onal magnetometers. A new com-pact CSEM receiver [Figure 4(b)] has recently been developed pri-vately in collaboration with Scripps [20] and uses capacitive electrodes and short electric field receivers (,1 m) within the body of the receiver module, which eliminates the need for extended electric field arms.

The commercialization of MT and CSEM methods for hydrocarbon exploration was largely due to exploration mov-ing into deep water where these techniques were traditionally applied. The increased noise level and airwave problem, which we discuss later, prevented these techniques from being suc-cessful in shallow water.

The first company to show interest in the CSEM method for hydrocarbon exploration was Exxon [21]. The recent development of node-based CSEM method for oil and gas exploration was a result of work carried out by Statoil, a Norwegian oil company. The first test survey took place offshore Angola in 2000 [4]. This was carried out in 1,200 m of water with the target 1,100 m below the seafloor using the academic equipment described above. From this initial survey, commercial hardware was developed with three

companies offering node-based CSEM services. The commercial-ization of the CSEM method coincided with the advent of deep water exploration in the Gulf of Mexico and offshore West Africa. With drilling moving into deep water and the associated higher

drilling costs, the CSEM tech-nique has become a key tool in derisking such wells [22].

A marine transient CSEM sys-tem was developed at the University of Toronto [23], [24] and first applied to shallow subsea

investigation of hydrothermal structures; source-receiver separa-tions, or offsets, were typically about 100 m and the source cur-rent was about 3 A. The system was then applied to the detection of gas hydrates with a source current of 50 A and offsets of up to 500 m [10], [25]. This system used a source and a two-channel receiver cable attached to a single boat with both source and receiver stationary during recording. The commercial develop-ment of a static transient marine EM system with an ocean bot-tom cable (OBC) receiver was developed from the land-based multitransient EM method [11]. A source current of 800 A and up to 30 in-line receivers each 200 m in length were used to record the in-line electric field at offsets of up to 8 km [13]. A fully towed streamer-based EM system is described in [26] and illustrated in Figure 5. With a cable-based system, data are transmitted to the acquisition boat in real time allowing real-time quality control.

In summary, there have been two separate marine CSEM developments. The conventional marine CSEM method uses

Stray Line Float withRadio Beacon and Strobe

Glass FlotationSpheres

–Ey Dipole

–Ex Dipole

+Ey Dipole

+Ex Dipole

Data LoggerPressure Case

Acoustic Release andNavigation Unit

Bx Coil

Electrode

AnchorRelease

By Coil

Concrete Anchor

(b)(a)

[FIG4] (a) The conventional marine CSEM node receiver (modified from [19] and used with permission) and (b) a new compact receiver.

THE PHYSICAL SIGNIFICANCE OF SKIN DEPTH IS THAT IT IS A SERIOUS

IMPEDIMENT TO THE RESOLUTION OF SMALL TARGETS AT DEPTH.


receiver nodes on the seafloor and a towed dipole electric source transmitting a continuous signal, as illustrated in Figure 2. The transient marine CSEM method uses receivers that are connected to a vessel, allowing real-time quality control and a dipole electric source that transmits a broad bandwidth signal with a beginning and an end. It is clear that source control and receiver deployment are separate issues. These two developments may not stay sepa-rate, however. In the following sections on data acquisition and processing, we maintain the separation but note that these tech-niques face the same problems in the subsequent data processing to attenuate MT noise and eliminate the airwave.

ACQUISITION AND PROCESSING OF CONVENTIONAL MARINE CONTROLLED SOURCE EM DATA

POSITIONINGThe position of the source above the seafloor is performed through real-time winch control based on altimeter data received from the source. The source is generally towed at about 1.5 kn (compared with 4–5 kn for seismic data), which is the minimum speed to have some control of the deep-towed source. The lower the speed, the longer the available stack time window for a given travel distance. The low towing speed results in a small towing angle to the vertical, which reduces the positional uncertainty of the source. Towing slower than 1.5 kn would result in the source being adversely affected by cross currents. Uncertainties in source positioning are still considered the biggest source of error in CSEM data [27] and result in errors in source-receiver offset and azimuth. The source electrodes may not be in the same horizontal plane (tilt), and may not be in the vertical plane of the sail line (yaw). The tilt and yaw must be mea-sured and included in the modeling. Errors associated with source electrode positioning are frequency dependent, as higher frequen-cies decay more frequently with offset than lower frequencies. The problems are greatest for high frequencies and short offsets. Short baseline acoustic positioning on both receiver electrodes and the source vessel are used in water depths of less than 3 km.

Positioning of receiver nodes is performed by acoustic survey-ing as the source vessel sails over the receivers, normally using two perpendicular lines. Seafloor orientation of the receivers is generally obtained from the data. Direct determination of the

receiver orientation from gyroscopes is expensive with the addi-tional drawback of added weight from the extra batteries required. Compasses are fitted to all receivers but are often inaccurate and require frequent recalibration [28]. Errors in positioning are cur-rently the main barrier to improving the quality of CSEM data. Reference [29] quote a positional accuracy of 5%, which equates to a maximum signal-to-noise ratio of 26 dB regardless of the source current. For this reason, improvements in positioning accuracy must be achieved before more powerful sources are required.

CLOCK DRIFTTiming on the source is controlled by the global positioning sys-tem (GPS) time from the boat. All the receivers contain internal clocks that are synchronized to GPS time before deployment and following recovery are again compared to GPS time. Drift can be on the order of a few milliseconds per day. Accurate phase requires that the source and receiver clocks are synchronized.

The present receiver clocks are accurate to within 10 ms over a period of a few days. By applying a temperature-dependent drift correction to the timing it is claimed that a phase accuracy of one degree is now possible [30], although the typical inversion misfit threshold for phase data is 5° [31]. The ability to use phase infor-mation in inversion is crucial in providing depth sensitivity.

ACQUISITION GEOMETRIESAcquisition of CSEM data generally involves placing the receivers on the seafloor in a grid and then towing the source over the array of receivers along two sets of perpendicular lines. When the

source is towed in-line with a receiver the measurement is called in-line or radial and the azi-muthal angle is zero. When the source is towed parallel to the receivers the measurement is called broadside or azimuthal and

the azimuthal angle is 90°. When the azimuthal angle is between these two extremes the received data is a mixture of these two components as shown in Figure 6 (a). The distribution of these three measurements for a source towed in a regular grid around a single receiver is shown in Figure 6 (b). The in-line (radial) and broadside (azimuthal) fields behave very differently in the pres-ence of a thin buried resistor: the in-line component is very sensi-tive to the resistor, while the broadside component is very insensitive to it. The reason for this is that the in-line component includes significant vertical current flow, which is strongly altered by the thin resistor, while the current flow in the broadside com-ponent is largely horizontal and so relatively unaltered by the resistive layer. MT data are similar to the controlled source broad-side data: the current flow is horizontal and therefore not sensi-tive to thin resistive layers.

Modern surveys often employ a large 2-D grid with over 50 receiver nodes placed approximately 1 km apart on the seafloor. As a result, a large proportion of the data acquired is neither purely in-line nor purely broadside. From Figure 6(b), it can be

10 m~100 m

Receiver V(t )Source I (t )

[FIG5] Acquisition setup of a fully towed EM system. Figure modified from [34] and used with permission.

THE ABILITY TO USE PHASE INFORMATION IN INVERSION

IS CRUCIAL IN PROVIDING DEPTH SENSITIVITY.


seen that almost 80% of the acquired data is a combination of the two components, which can be exploited only through a full 3-D inversion.

Following acquisition, the receivers release their concrete base weight and float to the sea surface for recovery and data transfer. Data are acquired continuously from the moment the receiver is deployed; during periods when the source is either not active or a long way from the survey site, the receivers acquire MT data that can be utilized in inversion, provided the water depth is not so deep that it filters out the MT source signal.

DETERMINATION OF AMPLITUDE AND PHASECSEM data require little processing before inversion. Most of the steps simply correct the data for orientation and amplitude to make them comparable with the synthetic data generated during inversion.

The amplitude and phase of frequencies transmitted in a con-tinuous source waveform are determined from time windows of the data known as “stack frames” [32]. A stack frame is defined as the length of time it takes the vessel to travel a distance of one source length. This is typically 120 s, but is restricted to consist of an integral number of fundamental source periods. The stack length can be increased to improve the signal-to-noise ratio, but this is limited by the towing speed. The amplitude and phase may be obtained simply through Fourier transform of the data within the stack window, or a best-fitting sinusoid in a least-squares sense at the various transmission frequencies may be found in the time domain and the amplitude and phase of this sinusoid then deter-mined. Amplitude and phase as a function of offset are the input to inversion.

SOURCE NORMALIZATIONMeasurement of the source waveform is made by a small data log-ger attached to the source. Source and receiver data for the same time window are transformed to the frequency domain, becoming complex. The complex receiver signal at a given frequency is then divided by the complex source signal at the same frequency, and finally the amplitude of the resulting complex number is scaled by the receiver electrode length to give an amplitude with units of V/Am2. The source signal normally has its energy confined to a small number of discrete frequencies and the frequency domain division is performed only at those frequencies.

DETERMINATION OF ROTATION ANGLEThe first step in CSEM data processing is to determine the orienta-tion of the receiver node on the seafloor. When the node lands on the seafloor its receivers are in an arbitrary position, with the Ex and Ey components orientated at unknown angles relative to the source line. In the in-line direction, the amplitude is a maximum;

in the cross-line direction, it is a minimum. Reference [28] describes a method for determining the receiver orientation by rotating the Ex and Ey components to find maximum (in-line) and minimum (cross-line) directions. Over a 1-D earth with source and receiver in-line, the cross-line signal amplitude is zero. However, 2-D and 3-D effects introduce a cross-line component and noise is always present. In addition, if the source towing angle is more than a few degrees from in-line, it adversely affects the determination of the rotation angle. Another approach to deter-mining the rotation angle is to find it with the inversion proce-dure itself; that is, allowing the rotation angle to be another unknown. Reference [33] describe a procedure to invert for the rotation angle and seafloor conductivity simultaneously, which appears to work on both in-line and out of line source-receiver geometries.

SOURCE CONTROL IN CSEM METHODSA variety of periodic continuous signals has been developed for use in marine CSEM methods, including a square wave and various waveforms that concentrate the energy in selected frequencies, as discussed by [34].

An alternative CSEM technique, also discussed by [34], uses a horizontal electric dipole source and dipole electric receivers, but the source signal is transient: after a certain time, the earth response reaches a steady state. After the steady state has been reached, the cycle may be repeated. One type of transient response is the response to a reversal in polarity of a DC current; another type of transient response is the response to a single period of a PRBS. For both cases the full response is measured over the whole cycle time. Reference [12] concluded that using a PRBS for the source current signal rather than a step in DC current allowed data with better signal-to-noise ratio to be obtained in a given time.

Let the source current be I 1t 2 and the resultant voltage at the receiver be V 1t 2 . Since Maxwell’s equations are linear, the response of the earth can be regarded as a causal linear filter with impulse response g 1t 2 that depends on the position and

In-LineReceiver

φ = 0°

BroadsideReceiverφ = 90°

Azimuth φ

In-Line CombinationBroadsideSource

Combination ofIn-Line andBroadside

In-LineReceiver

φ = 0°

Azimuth φ

Source

CombinaIn-LineBroad

(b)(a)

[FIG6] (a) The geometry of the in-line (radial) and broadside (azimuthal) components as well as measurements that are a mixture of both components. (b) The distribution of in-line, broadside, and source-receiver configurations that are a mixture of both for a source towed in a rectangular grid around a single receiver (black square).


direction of the injected current at the source and the position and orientation of the receiver electrodes. These three quanti-ties are related by the convolution

V 1t 2 5 3`

0

g 1t 2I 1t2t 2dt. (22)

The lower limit of the integral is zero because the earth is causal and cannot respond before there is an input. Since the flow of current in a conducting earth is a lossy process, the impulse response g 1t 2 must decay to zero as t S `. Within the precision of the measurements, therefore,

g 1t 2 5 0, for t . Tg , (23)

where Tg is a time greater than which the response is too small to detect. That is, the earth impulse response g 1t 2 is transient: it has a beginning and an end.

Given that g 1t 2 is of finite duration, what is the best func-tion for I 1t 2? This is the problem of source control. To obtain the response g 1t 2 , or its Fourier transform g 1 f 2 without bias to any particular frequency, I 1t 2 should be a function whose amplitude spectrum is constant over the known frequency range of interest.

Two time functions with flat amplitude spectra that are used extensively in the measure-ment of impulse responses are swept frequency sine waves (used in radar and exploration seismology in the Vibroseis technique) and PRBSs (used for many decades in electrical and electronic applications). The instantaneous power of a swept frequency signal is time variant, whereas the instantaneous power of a PRBS is con-stant for its duration. For Vibroseis, the implementation of PRBSs is difficult, because the inertia of the vibrating mass-es inhibits rapid reversal of the direction of motion. It is not difficult, however, to switch the direction of current flow in resistors, and PRBSs are therefore very suitable for EM applications.

A PRBS is a sequence of N5 2n2 1 samples that switches from one level to the other at pseudorandom multiples of a basic time interval Dt; n is known as the order of the sequence. The PRBS has an amplitude spectrum that is flat in the frequency interval

1

NDt# f #

12Dt

. (24)

The recorded data need to be sampled at a rate that is greater than, or equal to 1/Dt to obtain the full benefit of the source spectrum.

Using a current signal I 1t 2 such as a single period of a PRBS, of time duration Ts, which has the full bandwidth of the impulse response g 1t 2 , (22) can be written as

V 1t 2 5 3Tg1Ts

0

g 1t 2I 1t2t 2dt. (25)

This is a complete convolution of finite duration Tg1 Ts. The only issue remaining is to determine g 1t 2 , given I 1t 2 and V 1t 2 .

I 1t 2 , of duration Ts, must have the full bandwidth of the impulse response g 1t 2 . Because I 1t 2 has finite duration Ts, it is a transient source signal. The response V 1t 2 must be measured for the minimum duration Tg1 Ts, beginning at the start time of the source signal. Because V 1t 2 has a finite duration, with a known beginning and end, it is also a transient.

TRANSIENT CSEM METHODThe transient EM method has been used for many years for mineral exploration, but has not yet become a standard tool for hydrocarbon exploration and production. A standard work on the theory of transient EM was written by [35] who sum-marized 20 years of work done in Russia and North America. Reference [7] presented the state of the art of long offset tran-sient EM (LOTEM) surveying for land applications. The MTEM method was developed initially in the University of Edinburgh [9], and then by MTEM Limited and by PGS. It works both onshore and offshore.

In 1994 and 1996, a time-lapse transient EM data set was obtained along a line over a gas storage site at St. Illiers la Ville in

France using a long-period square wave dipole current source, in-line and cross-line electric dipole receivers, and horizontal loops to measure the rate of change of the vertical magnetic field [9], [36], as shown in Figure 7.

The transient responses obtained were essentially step responses. As described above, the propagation through the air travels at the speed of light, so the arrival of the step at the receiv-er is virtually instantaneous. The propagation through the earth is slower and frequency dependent, and arrives later. The step response and its time derivative, the impulse response, were mod-eled by [11] for a 1-D earth and a source-receiver separation of 1 km in a similar gas storage situation; the result is reproduced in Figure 8. The response without the resistive layer is shown in black. The response with the resistive layer is shown in red. The arrival of the step through the air is known as the airwave. When the step is differentiated it becomes an impulse, as shown in Figure 8(b). The impulse response of the earth, traveling more slowly, arrives after the impulsive airwave and is separated from it in time. There is a dramatic increase in amplitude when the resis-tive hydrocarbon layer is present.

Two different approaches were taken to the processing and inversion of the St. Illiers la Ville EM data. First, [36] used the classic modeling and inversion approach of [7], but with 3-D modeling. Second, [9] recovered the earth impulse responses from the data and processed them, taking special care to cor-rect timing errors, to obtain common-offset stacks, similar to seismic data processing. A comparison of the two approaches to the data processing is shown in Figure 9. Figure 9(a) shows the data input to the 3-D inversion by [36], displayed as common-midpoint gathers; Figure 9(b) shows the same data

THE FIRST STEP IN CSEM DATA PROCESSING IS TO DETERMINE

THE ORIENTATION OF THE RECEIVER NODE ON THE SEAFLOOR.


as Figure 9(a), but after careful adjustment for timing errors, and displayed as common-offset gathers. The resistive gas reser-voir can be seen clearly in Figure 9(b) at 4 ms on the verti-cal axis and between 3000 and 5000 m on the horizontal axis.

The MTEM method evolved from the result shown in Figure 9(b). The essence of the method is that both the volt-age at the receiver V 1t 2 and the input current I 1t 2 are mea-sured simultaneously and the earth impulse response is recovered from these two measurements by deconvolution. A diagrammatic plan view of one possible setup is shown in Figure 7(a).

A transient current, typically a step function or a finite-length signal such as a pseudorandom binary sequence (PRBS), is injected between two source electrodes and is measured and recorded. The time-varying voltage response between each pair of receiver electrodes is also measured simultaneously. If the response reaches a steady state before the next change in current is applied at the source, the full response has been measured and is the convolution

V 1t 2 5D xsD xr I 1t 2*g 1t 2 1 n 1t 2 , (26)

in which V 1t 2 is the measured voltage response at the receiver, I 1t 2 is the measured input current applied at the source, the aster-isk * denotes convolution, g 1t 2 is the unknown earth impulse response, and n 1t 2 is uncorrelated noise; D xs is the source dipole length, D xr is the receiver dipole length, and t is time. The source current and received voltage are recorded with identical devices,

whose effects cancel in the subse-quent deconvolution. The total circuit impedance is, in general, complex and therefore the source current is in general out of phase with the applied voltage. The response in (26) depends on the injected current, and that is what

we measure: the complex impedance effects are automatically taken into account.

For the land case, a schematic cross section of the setup is shown in Figure 7(a), and for the marine case, one possible configuration is shown in Figure 5. In both of these cases, the source and receiver electrodes are in a straight line. Since it is known from (5) that the different components of the electric field interact with each other in propagation, it is clear that more components and more azimuths should be measured. Onshore, the electrode positions, or pegs, are known from sur-veying. Offshore, acoustic transponders are attached to the cable at the electrode positions and are positioned using a commercial underwater acoustic positioning system. The whole setup can be moved along to continue the line, very similar to the 2-D seismic reflection method. It is necessary to have offsets up to four times the depth of the target to resolve both its top and bottom. Normally, about 40 receiver channels of equal spacing are used. This choice is somewhat arbitrary, but it has been found to give good lateral resolution equal to about half the receiver spacing. For a target at 1 km depth, the receiver bipole length would be 100 m and the receiver spread would be 4 km long. Onshore, a roll-along system similar to the seismic reflection method is used. Two special features of the method are precise timing and real-time quality control.

TransmitterOrientation

In-Line CurrentDipole Source Receiver Spread

(a)

(b)

2 kmY

X

0 500

N

1,000

Scale (m)

Reservoir Edge

1 1687.5

MTE

M

Pro

file

[FIG7] (a) The field layout for in-line configuration showing an in-line current dipole source and a 2-km receiver spread containing 16 in-line and eight cross-line 125-m electric dipole receivers and eight horizontal 50-m magnetic loops. (b) The position of the MTEM line relative to the gas reservoir [9]. Figure used with permission from [9].

AN IMPORTANT PART OF CSEM DATA PROCESSING IS THE REMOVAL

OF THE AIRWAVE BECAUSE ITS PRESENCE SIGNIFICANTLY REDUCES

THE SENSITIVITY TO SUBSEA RESISTORS IN SHALLOW WATER.


DECONVOLUTION TO RECOVER THE IMPULSE RESPONSEThe convolution in (26) applies because the earth system is linear. In the frequency domain, the convolution becomes a multiplication, so deconvolution becomes a division. A typi-cal land data example of the measured current input, mea-sured voltage output at one receiver, and the result of deconvolution to obtain the earth impulse response for the source-receiver pair are shown in Figure 10. Deconvolution in the presence of noise is described in [11].

In land data, the impulse response exhibits an initial impulse at the time break (which is the “airwave”); this is followed by the earth impulse response. The noise consists of

random noise, MT noise, and nonrandom cultural noise from power lines and railways that is normally orders of magni-tude greater than the MT component. The noise can be reduced by a variety of processes, including stacking. The fundamental frequency and harmonics of the cultural noise are often not constant and the phase of each of these fre-quencies also varies in an apparently random way.

IMPULSE RESPONSE OF A HALF-SPACEThe response of a half space of resistivity r ohm-m to a 1 A-m step applied at an electric dipole source on the surface was derived by [37]. The impulse response is obtained by differentiating the step response:

g 1r, r, t 2 5 r

8p"p expa2r 2

4c2tbt2

52 V/m2/s, (27)

in which r is source-receiver offset in meters, c25r/m, with mag-netic permeability m5 4p #1027 Hm21, and t is time. This func-tion has a peak at time

tpeak5mr2

10r s. (28)

Substituting t 5 t/tpeak into the expression in (27) gives the result

g 1r, r, t 2 5 5.65 # 106 r2

r5 expa2 52tbt25

2 . (29)

Equation (28) states that the peak of the earth impulse response arrives at a time proportional to r2, while (29) shows the ampli-tude decays as r25, as mentioned above in connection with Figure 3(c). The timing and amplitude of the peak can be used to estimate subsurface resistivities [11], [38].

The function described by (29) has the same shape as the black curve in Figure 8(b). It is very similar in shape to the real earth impulse response of Figure 10(c) and thus gives an analytical approximation to the real data.

PROCESSING OF MARINE CSEM DATA

THE AIRWAVEIn shallow marine CSEM methods, the response at the receivers is complicated by the interaction of the seawater/air interface. A large amount of energy that arrives at the receivers propagates through the air at the speed of light. This energy is often referred to as the “airwave.” Reference [39] gives the following analytic fre-quency-domain expression for the simple case of a double half-space of air and water

E5 L1 D1 I (30)

where,

L5Mr

2pr3 exp522kz6 (31)

D5Mr

2pr33111 kr 2exp522kr64 (32)

8×10–9

7

6

5

Ele

ctric

Fie

ld A

mpl

itude

(V

/m/A

-m)

4

3

2

1

0

1

0.8

0.6

0.4

Nor

mal

ized

Der

ivat

ive

(Ω/m

2 /s)

0.2

0

–0.02 0 0.02Time (s)

(a)

0.04 0.06

–0.02 0 0.02Time (s)

(b)

0.04 0.06

Uniform Half SpaceHydrocarbon Reservoir

Uniform Half SpaceHydrocarbon Reservoir

[FIG8] (a) Step response of a 20 V-m half-space at an offset of 1,000 m to a 1 A-m step at the source dipole (black curve), and with a 25-m-thick, 500-V-m resistive layer at a depth of 500 m (red curve). (b) Normalized impulse response, for same configuration as (a) with normalization factor 3.433E +6. The black vertical arrow represents the pure inductive effect of the impulse at the source. Figure used with permission from [11].


I 5Mr

2pr3 c2a131 3kr1 k2r2 2 12z 2 22r2 bexp52kR16d . (33)

In these equations, M is the source dipole moment, with units A-m, r is the water resistivity (V -m), z is the source and receiver depth below the sea surface (m), r is the horizontal source-receiver offset (m), R1 1m 2 is the distance between the receiver and image source given by R15"1r21 12z 2 2 2 , k is the wavenumber in the water (m-1), given by

k5"ivm0s, (34)

where v 5 2pf and f is frequency (Hz), m0 is magnetic permea-bility, taken to be 4p 3 1027 H/m, and s5 1/r is conductivity (S/m). The terms L, D, and I in (30) are known as the primary airwave, the direct wave and the source image, respectively. These three components are illustrated in Figure 11.

The primary airwave travels vertically through the water and then through the air at the speed of light with only geometrical spreading of 1/r3. It is a function of seawater conductivity, source and receiver depth, and source-receiv-er offset only. When a layered earth is considered, a second component of the water layer response is introduced, which is coupled to every resistivity boundary within the subsur-face. Reference [40] describe this energy as multiple rever-berations at the source and receiver side between every resistivity boundary and the sea surface. This is analogous to multiple reflections in seismic reflection data, though in this EM case the process is diffusive in the water vertically above and below the source and receiver. The reverbera-tions are most sensitive to the shallow resistivity structure and can be considerably larger in amplitude than the pri-mary airwave when the seafloor is resistive.

An important part of CSEM data processing is the remov-al of the airwave because its presence significantly reduces

0.5 02468

101214161820

1

1.5

2

2.5

1,000 2,000 3,000

Distance Along Profile (m)

Distance Along Profile (m)(a) (b)



NNE NNEGas

WaterGas

Water

SSW

7,0005,0003,0001,000 7,0005,0003,0001,000

100 m

SSW

100 m

4,000 5,000 6,000

1,000 2,000 3,000 4,000 5,000 6,000 7,000

–0.2–0.7

0.0

1.0

0

0.2

Log

Tim

e (m

s)

Tim

e A

fter

Tim

e B

reak

(m

s)[FIG9] Data from the St. Illiers la Ville gas storage site. (a) Input to 3-D inversion by [36] and (b) the same data set as (a) after different processing to give the first derivative of the impulse response displayed as 1,000-m common-offset section [9]. Figure used with permission from [9].

10 4e-4 6e-6

4e-6

2e-6

0–4e-4

0

5

0

–5

–100 0.1 0.2

Time (s)(a) (b) (c)

0 0.1 0.2Time (s)

0 0.01 0.02 0.03Time (s)

15 t Peak 15 t Peak30 t Peak

30 t Peak

Am

plitu

de (

A)

Am

plitu

de (

V)

Am

plitu

de (

Ω/m

2 /s) 4 t Peak

15 t Peak30 t Peak

[FIG10] (a) Current measurement for a PRBS binary sequence source time function; (b) receiver response for the input signal shown in (a); and (c) result of deconvolution, showing the impulse response of the earth, plus noise, for three different values of the listening time. (Note that (c) is plotted on a time scale different from (a) and (b) to enable the features of the impulse response to be seen clearly.)


the sensitivity to subsea resistors in shallow water. The effect of a thin conducting layer above the source and receivers provides a short path from the source to the receivers verti-cally through the water and then through the air at the speed o f l ight wi th only geometrical spreading of 1/r 3 where r is the horizontal source-receiver offset.

THE AIRWAVE IN TRANSIENT EM The impulse response of the earth recovered with a transient source exhibits partial temporal separation of the earth response and airwave in shallow water as shown in Figure 12. Reference [41] showed that there is an observation window in offset ver-sus arrival time space which grows as the water depth decreases. As a result the airwave is not considered a great problem in shallow water for transient EM.

A technique to remove the airwave in the time domain was proposed by [42]. It requires a long offset measurement of the

airwave on the order of 20 km, which is completely separated from the earth response and water layer response. This esti-mate of the airwave can then be used to estimate and remove the airwave in the received data. Limitations of this method are that it assumes that the water depth is the same for both measurements and that the near-surface resistivity structure is broadly the same in both locations to record the same mul-tiple reflections.

Another approach to the elimination of the airwave for a transient source is to position the source and receiver close to the sea surface such that the measured airwave is the same as that measured on land; that is, the airwave becomes a delta function band limited only by the acquisition system. The water layer is then included as the first layer in the inversion and becomes part of the subsurface response. This approach is particularly well suited to a towed system and shallow water so that the resolution of the target is not adversely affected by the water layer.

THE AIRWAVE AND CONVENTIONAL CONTINUOUS SOURCE EMIn the frequency domain, airwave contamination is recognized as a decrease in slope of the amplitude versus offset (AVO) curve to 1/r3 and a flattening of the phase versus offset (PVO) curve [27].

Many of the approaches to removing the airwave appear to provide an exact theoretical solution that is difficult to realize in practice. Receiver spacing in field data are not small enough to carry out up/down separation. Small devia-tion from the vertical when measuring Ez results in the air-wave being measured. Separation into TE and TM modes assumes a 1-D earth. Other techniques involve taking the difference of two large quantities to find a small quantity. When the airwave is a significant proportion of the total response, the required error on acquisition parameters is very small and the signal-to-noise ratio of the data must be very high. An important requirement of any technique is that it must be possible to model and invert the data after

airwave removal. As a result, any t echn ique tha t a l so removes the subsurface signal will make it difficult to accu-rately model the resultant response. For these reasons, the preferred approach of many in the CSEM community is to

include the airwave in the inversion process. The effect of the airwave is greatly reduced at frequencies in the range 0.01–0.1 Hz compared with the frequency range 0.1–1 Hz commonly employed in deep water.

NOISE IN SEAFLOOR EM MEASUREMENTSThe noise floor of seafloor measurements is a crucial factor in marine EM surveying. The noise characteristics are quite dif-ferent in deep and shallow water environments.

00

0.5

1

1.5

2

2.5

1 2Time (s)

Am

plitu

de (

V/A

m2 )

Earth Response

Airwave

Apeak

tpeak

×10–11

3 4 5

[FIG12] The impulse response of the earth in 100 m of water with source and receiver on the seafloor showing the peak amplitude and peak time of the earth response.

I (Source Image)

L (Primary Airwave )1/r 3

D (Direct Wave) Receiver

r

Source

Air

Water

z

z

R1e–z /δ e–z /δ

[FIG11] The three wave components for a double half-space of air and water.

IN WATER SEVERAL KILOMETERS DEEP, THE MAIN SOURCES

OF NOISE ARE FROM RECEIVER ELECTRONICS, ELECTRODE NOISE, AND

WATER MOTION.


In water several kilometers deep, the main sources of noise are from receiver electronics, electrode noise, and water motion. The movement of electrically conductive water at velocity v through the earth’s magnetic field B due to tidal currents and wave motion induces an electric field E at the receivers given by

E5 v 3 B, (35)

while microseisms, generated by sea surface gravity waves, dis-place the electrodes relative to each other at about 0.2 Hz [43]. There is currently no literature on the removal of induction noise; the effect of noise at a particular frequency such as that at 0.2 Hz caused by surface waves can be avoided by choosing source frequencies that avoid this frequency range. MT noise, which originates in the ionosphere, is low-pass filtered by the thick conductive seawater layer with little energy above 0.1 Hz reaching the deep seafloor.

In shallow water, the noise due to wave and tidal motion is far greater than in deep water and the amplitude of MT noise is considerably greater. Figure 13, modified from [44], shows the total noise (the lower solid line), including MT noise, measured in the Gulf of Mexico on the seafloor in 100 m water depth. The spectrum has a peak at about 0.1 Hz attributed to ocean swell noise in the shallow water. The background MT field can be seen to increase in amplitude towards low frequencies at approxi-mately 20 dB per decade.

The noise floor of CSEM receivers is the voltage measured between two electrodes and divided by the electrode separation to give units of volts per meter (V/m). This measurement may also be expressed as the square root of the power spectral densi-ty V/m/sqrt(hz). The quoted instrumental noise floor for deep sea measurements is 1029 Vm/"HZ [44] or 2190 dB in Figure 13 at 1 Hz.

A common practice in con-ventional CSEM literature is to quote a noise floor with units of V/Am2, where the noise voltage is normalized by both the receiver electrode separation and the source dipole moment (which is the product of the source length and the source cur-rent amplitude) as well as the bandwidth of the stacking win-dow. A reduction in the noise floor quoted for conventional CSEM methods in the past decade has in fact largely been due to the increase in the source dipole moment; the true noise levels are unchanged.

For node-based CSEM methods in shallow water during periods of high MT activity, the data can be degraded so much that lines must be retowed. Reference [45] describes a remote reference technique for the removal of MT signals from sea-floor nodes. The technique involves deploying two or three nodes outside the survey area and obtaining transfer func-

tions between the remote stations and stations in the survey area during a period when the source is not transmitting. These transfer functions are a function of frequency, subsur-face conductivity, and seawater conductivity but are indepen-dent of the MT signals. During periods when the source is active, these transfer functions can then be used to calculate the MT signals in data recorded in the survey area with the

source signal present, based on the source-free remote mea-surement of the MT signal.

Reference [46] demonstrated a technique for removing MT noise without a remote reference mea-surement by exploiting the prop-erties of the earth impulse response and the high spatial correlation of the MT signal.

Figure 14(a) shows a raw common-source gather, 250 s in length (vertical axis), with offsets (horizontal axis) increasing from 2,200 m on the left to 7,000 m on the right. The long-period noise is well correlated from trace to trace; the response to the PRBS input decays dramatically from near to far offsets. Figure 14(b) shows the result of deconvolution in a 20 s window containing the impulse response: the long signal responses to the pseudorandom input current have been compressed to impulse responses, but the noise remains. An estimate of the noise is obtained by subtracting the short impulse response from the nearest 250 s trace. This noise estimate is similar to the noise on the other traces. To determine the component of the noise on each subsequent trace that is corre-lated with this noise estimate, a Wiener filter is found for each trace that best estimates the correlated part—the noise—from this noise estimate. The noise estimated in this way on each

–110

–150

–190

–23010–3 10–2 10–1 100

Frequency (Hz)

Electrode Noise

Instrument Noise

Seafloor Signal,100 m

Seafloor Signal,1,000 m

101

Am

plitu

de in

dB

Rel

ativ

e to

V/m

/√H

z[FIG13] This figure shows electric field noise (lower solid line) that was collected on adjacent electrodes in 100 m of water in the Gulf of Mexico. (Figure modified from [44] and used with permission.)

REMOVING MT NOISE BRINGS THE NOISE LEVEL MUCH CLOSER

TO THAT OF DEEP WATER EM RECEIVERS AND SIGNIFICANTLY INCREASES THE OPERATIONAL

CAPABILITIES OF TRANSIENT CSEM IN SHALLOW WATER.


subsequent trace is then subtracted from the trace to reveal the impulse response, as shown in Figure 14(c). The increase in sig-nal-to-noise ratio from Figure 14(b) to (c) is about 20 dB. This is equivalent to increasing the source current from 700 A to 7,000 A.

Removing MT noise brings the noise level much closer to that of deep water EM receivers and significantly increases the operational capabilities of tran-sient CSEM in shallow water. This technique attenuates all types of spatially correlated noise, not just MT noise, and so may help to reduce noise asso-ciated with water and wave motion.

MODELING AND INVERSIONOne of the main barriers to the widespread use of EM in exploration has been the difficulty of making the data more understandable. Ray theory, which is so useful in under-standing seismic data, does not apply to diffusive EM data. The analysis of CSEM data is carried out through iterative forward modeling or inversion using a model of the subsur-face resistivity that generates synthetic data that fit the field

data to within a specified misfit based on errors present in the data. The for-ward modeler may discretize the earth in 1-D (plane horizontal layers), 2-D (resistivity variation in the x-z plane), or 3-D, with resistivity variation allowed in all directions. The 2-D and 3-D for-ward modelers can discretize the sub-surface using three different techniques: finite difference [47], finite element [48], and integral equation [49].

Inversion is very sensitive to the starting model: it pays to have the best possible starting model that is consistent with other known a priori information in the survey area, including especially seis-mic data and well logs. Since there are no common parameters in seismic and EM wave propagation, it is not straight-forward to connect this information to a background resistivity model. Normally, rock physics is used to connect seismic velocities to resistivities via porosity, although there is no agreed procedure for doing this and the steps are not always well explained. For example, a 3-D resistivity forward model shown in [2, p. WA9] was “guided by 3-D seismic data, well log data...”

In addition to the dimensionality of the earth, another complication (which is of particular importance in EM data) is

the effect of resistivity anisotropy, which is seen in induction log measurements in deviated wells. The vertical and horizon-tal resistivities are different. Normally, the vertical resistivity is greater than the horizontal.

Figure 15 from [50] shows the improvement that can be obtained by taking anisotropy into account in 3-D inversion. The isotropic inversion is sensi-tive to both vertical and hori-zontal resistivities and appears to include both these effects in

the same image resulting in low resistivity features above and below the target due to low horizontal resistivity. Although this result used only in-line data and two frequencies, there is still a clear improvement when anisotropy is included.

CONCLUSIONSThe CSEM method of exploration has the potential to dis-criminate between more conducting brine-saturated and more resistive hydrocarbon-saturated reservoirs. EM data are, however, very different from seismic data and normal seismic data processing cannot be applied. In particular, the

50

2 3

3 4 5 6 7

3 4 5 6 7

4 5 6 7Source-Receiver Offset (km)

Source-Receiver Offset (km)(a)

Source-Receiver Offset (km)(b)

(c)

100

150

200

0

10

20

Tim

e (s

)T

ime

(s)

0

10

20Tim

e (s

)

[FIG14] (a) Common-source gather, 250 s in length (vertical axis), with offsets (horizontal axis) increasing from 2,200 m on the left to 7,000 m on the right; (b) result of deconvolving source gather for measured current, showing only 20 s of data: signal has been compressed to impulse responses, but the noise remains; and (c) result of subtracting estimated noise from data in (b). Figure modified from [34] and used with permission.

FOR THESE REASONS, THE PREFERRED APPROACH OF MANY IN THE CSEM COMMUNITY IS TO INCLUDE THE AIRWAVE IN THE

INVERSION PROCESS.


EM wavelet shape varies as it propagates, high frequencies being attenuated faster than low frequencies, with the atten-uation characterized by the well-known skin effect. For this reason, seismic stacking techniques cannot be applied.

The conventional marine CSEM method evolved from marine MT and uses very similar multicomponent receiver nodes on the seafloor. Traditionally, conventional CSEM methods use a dipole electric source that transmits a continu-ous signal with energy concentrated in a few discrete frequen-

cies. The received data are divided into time windows of typically 120 s duration and transformed to the frequency domain for analysis.

Transient CSEM methods have had a separate evolution and use receivers that are connected to the vessel, permitting real-time quality control and data analysis. The source is also a current dipole, but the source signal is transient, having a beginning and an end, and normally containing a broad band-width of frequencies, enabling the complete impulse response of the earth to be recovered.

Presently, these two methods are separate. But since source control is unconnected with the measurement at the receiver, we see no technical reason to maintain the distinction between these techniques.

The CSEM method suffers from poor signal-to-noise ratio for targets deeper than about 2 km and a considerable effort is being devoted to noise reduction, especially for the airwave and MT noise. There is considerable scope for development. The three components of the electric field interact with each other in propa-gation, so it is clear that all three components should, therefore, be measured. It is known that 3-D effects are very important, so it is not sufficient to measure along a line.

At present, the interpretation product of the CSEM method is an inversion, which is essentially a resistivity model that yields synthetic data that match the real CSEM data within some error. The final model is obtained by iterative forward modeling, beginning with one or more starting models that are based on all available geophysical and well data and usually some rock phys-ics. For geophysicists accustomed to seismic data acquisition and processing, it is very unsatisfactory. The positive conclusion is that there is plenty of scope for improvement. The potential of the method to identify hydrocarbons before drilling is a terrific incentive to develop new acquisition and processing methods to make CSEM become a reliable geophysical tool for hydrocarbon exploration, appraisal, and reservoir characterization.

AUTHORSAnton Ziolkowki ([email protected]) has been a professor of petroleum geoscience at the University of Edinburgh since 1992 and has been sponsored by Petroleum GeoServices (PGS) since 2010. He was previously a professor of applied geophysics at Delft University of Technology (1982–1992). He cofounded MTEM Limited in 2004 and was MTEM’s technical director until 2007, when the company was bought by PGS, where he became a chief scientist of geoscience and engineering.

David Wright ([email protected]) is a PGS senior research fellow at the University of Edinburgh. He holds B.Sc. and Ph.D. degrees in geophysics from the University of Edinburgh and an M.Sc. degree in geophysics from the University of Durham. He was a cofounder of MTEM Limited, which was formed to commercialize research on CSEM for hydrocarbon exploration. He worked as senior geophysicist at MTEM from 2004 to 2007 and held the same position at PGS from 2007 to 2010.

0.1 10 1,000Resistivity (Ω⋅m)

0

0.4

Sediments

0 5 10 15 20Horizontal Distance (km)

(a)

(b)

(c)

(d)

Isotropic ρ

1 5 10 15

Twop

Twop31/5 31/6

20 24 31/3

TM

CP

0.8

Dep

th B

elow

Sea

floor

(km

)

1.2

1.6

0

1

2z (k

m)

3

0

1

2z (k

m)

3

0

1

2z (k

m)

3–10 –5 0 5

x (km)10 15 20

150.055.020.07.42.70.40.1

ρ (Ω ⋅m)

1.0

Anisotropic ρ (Horizontal)

Anisotropic ρ (Vertical)

Water 31/2

Trol Easiers

Seawater Layer

[FIG15] (a) Structural interpretation of the Troll gas field in the North Sea [51]. The three inversion results are (b) isotropic inversion, (c) horizontal resistivity, and (d) vertical resistivity [50]. Part (a) used with permission from [51]. Parts (b)–(d) used with permission from [50].


REFERENCES[1] M. Wilt and D. Alumbaugh, “Electromagnetic methods for development and production: State of the art,” The Lead. Edge, vol. 17, pp. 487–490, Apr. 1998.

[2] S. Constable and L. Srnka, “An introduction to marine controlled-source elec-tromagnetic methods for hydrocarbon exploration,” Geophysics, vol. 72, no. 2, pp. WA3–WA12, 2007.

[3] T. Eidismo, S. Ellingsrud, L. MacGregor, S. Constable, M. C. Sinha, S. E. Jo-hansen, F. N. Kong, and H. Westerdahl, “Sea bed logging (CSEM), a new method for remote and direct identification of hydrocarbon filled layers in deepwater ar-eas,” First Break, vol. 20, no. 3, pp. 144–152, 2002.

[4] S. Ellingsrud, T. Eidesmo, S. Johansen, M. Sinha, L. MacGregor, and S. Con-stable, “The meter reader-remote sensing of hydrocarbon layers by seabed logging (SBL): Results from a cruise offshore Angola,” The Lead. Edge, vol. 21, no. 10, pp. 972–982, 2002.

[5] R. N. Edwards, “Marine controlled source electromagnetic: Principles, meth-odologies, future commercial applications,” Surv. Geophys., vol. 26, no. 6, pp. 675–700, 2005.

[6] M. Darnet, P. V. D. Sman, F. Hindriks, G. Sandrin, P. Christian, L. Jensen, and A. Uldall, “The controlled-source electromagnetic (CSEM) method in shallow water: A calibration survey,” in Proc. SEG Expanded Abstracts, 2010, pp. 701–705.

[7] K. -M. Strack, Exploration with Deep Transient Electromagnetics. Amsterdam: Elsevier, 1992.

[8] R. N. Edwards, “On the resource evaluation of marine gas hydrate deposits using sea-floor transient electric dipole–dipole methods,” Geophysics, vol. 62, no. 1, pp. 63–74, 1997.

[9] D. Wright, A. Ziolkowski, and B. Hobbs, “Hydrocarbon detection and monitoring with a multichannel transient electromagnetic (MTEM) survey,” The Lead. Edge, vol. 21, no. 9, pp. 852–864, 2002.

[10] K. Schwalenberg, E. Willoughby, R. Mir, and R. N. Edwards, “Marine gas hy-drate electromagnetic signatures in Cascadia and their correlation with seismic blank zones,” First Break, vol. 23, no. 4, pp. 57–63, 2005.

[11] A. Ziolkowski, B. A. Hobbs, and D. Wright, “Multitransient electromagnetic dem-onstration survey in France,” Geophysics, vol. 72, no. 4, pp. F197–F209, 2007.

[12] D. Wright, A. Ziolkowski, and G. Hall, “Improving signal-to-noise ratio using pseudo random binary sequences in multi transient electromagnetic (MTEM) data,” in Proc. 68th Meeting, EAGE Expanded Abstracts, 2006, p. P065.

[13] A. Ziolkowski, D. Wright, G. Hall, and C. Clarke, “Successful transient EM survey in the North Sea at 100 m water depth,” in Proc. SEG Expanded Abstracts, 2008, vol. 27, pp. 667–671.

[14] L. Cagniard, “Basic theory of the Magneto-Telluric method of geophysical pros-pecting,” Geophysics, vol. 18, no. 3, pp. 605–635, 1953.

[15] J. H. Filloux, “Techniques and instrumentation for study of electromagnetic in-duction at sea,” Phys. Earth Planet. Interiors, vol. 7, no. 3, pp. 323–338, 1973.

[16] C. S. Cox, “On the electrical conductivity of the oceanic lithosphere,” Phys. Earth Planet. Interiors, vol. 25, no. 3, pp. 196–201, 1981.

[17] P. D. Young and C. S. Cox, “Electromagnetic active source sounding near the East Pacific Rise,” Geophys. Res. Lett., vol. 8, no. 10, pp. 1043–1046, 1981.

[18] M. C. Sinha, P. D. Patel, M. J. Unsworth, T. R. E. Owen, and M. G. R. MacCor-mack, “An active source electromagnetic sounding system for marine use,” Mar. Geo-phys. Res., vol. 12, no. 1–2, pp. 59–68, 1990.

[19] S. C. Constable, A. S. Orange, G. M. Hoversten, and H. F. Morrison, “Marine magnetotellurics for petroleum exploration Part I: A sea-floor equipment system,” Geophysics, vol. 63, no. 3, pp. 816–825, 1998.

[20] T. Nielsen, “Novel electromagnetic seafloor system as compared to conventional system,” in Proc.6th MARELEC Conf., 2009, p. 4.

[21] L. J. Srnka, “Method and apparatus for offshore electromagnetic sounding uti-lizing wavelength effects to determine optimum source and detector positions,” U.S. Patent 4 617 518, 1986.

[22] J. Hesthammer, A. Stefatos, M. Boulaenko, S. Fanavoll, and J. Danielsen, “CSEM performance in light of well results,” The Lead. Edge, vol. 29, no. 1, pp. 29–34, 2010.

[23] S. J. Cheesman, R. N. Edwards, and L. K. Law, “First results of a new short base-line seafloor transient EM system,” in Proc. 58th Annu. Int. Meeting, SEG Expanded Abstracts, 1988, pp. 259–261.

[24] G. W. Cairns, R. L. Evans, and R. N. Edwards, “A time domain electromagnetic survey of the TAG hydrothermal mound,” Geophys. Res. Lett., vol. 23, no. 23, pp. 3455–3458, 1996.

[25] J. Yuan and R. N. Edwards, “The assessment of marine hydrates through electri-cal remote sounding: Hydrate without a BSR?,” Geophys. Res. Lett., vol. 27, no. 16, pp. 2397–2400, 2000.

[26] C. Anderson and J. Mattsson, “An integrated approach to marine electromagnetic surveying using a towed streamer and source,” First Break, vol. 28, pp. 71–75, May 2010.

[27] S. Constable, “Ten years of marine CSEM for hydrocarbon exploration,” Geo-physics, vol. 75, no. 5, pp. A67–A81, 2010.

[28] R. Mittet, O. M. Aakervik, H. R. Jensen, S. Ellingsrud, and A. Stovas, “On the orientation and absolute phase of marine CSEM receivers,” Geophysics, vol 72, no. 4, pp. F145–F155, 2007.

[29] K. K. Zach, F. Roth, and H. Yuan, “Preprocessing of marine CSEM data and model preparation for frequency domain 3-D inversion,” in Proc. PIERS, 2008, pp. 144–148.

[30] J. J. Zach, M. A. Frenkel, D. Ridyard, J. Hincapie, B. Dubois, and J. P. Morten, “Marine CSEM time-lapse repeatability for hydrocarbon field monitoring,” in Proc. SEG Expanded Abstracts, 2009, vol. 28, pp. 820–824.

[31] D. Crider, M. Scherrer, T. Pham, J. J. Zach, and M. A. Frenkel, “Full-azimuth, anisotropic 3-D EM inversion applied to a low-resistivity pay reservoir with well con-trol,” in Proc. SEG Expanded Abstracts, 2010, vol. 29, pp. 690–695.

[32] J. Behrens, “The detection of electrical anisotropy in 35 ma Pacific Litho-sphere: Results from a marine controlled-source electromagnetic survey and implications for hydration of the upper mantle”, Ph.D. dissertation, Univ. of San Diego, CA, 2005.

[33] K. Key and A. Lockwood, “Determining the orientation of marine CSEM receiv-ers using orthogonal Procrustes rotation analysis,” Geophysics, vol. 75, no. 3, pp. F63–F70, 2010.

[34] A. Ziolkowski, D. Wright, and J. Mattsson, “Comparison of pseudo-random binary sequence and square-wave transient controlled-source electromagnetic data over the Peon gas discovery, Norway,” Geophys. Prospect., vol. 59, no. 6, pp. 1114–1131, 2011.

[35] A. A. Kaufman and G. V. Keller, Frequency and Transient Soundings. Amsterdam: Elsevier, 1983.

[36] A. Hördt, P. Andrieux, F. M. Neubauer, H. Rüter, and K. Vozoff, “A first attempt at monitoring underground gas storage by means of time-lapse multichannel transient electromagnetics,” Geophys. Prospect., vol. 48, no. 3, pp. 489–509, 2000.

[37] G. Weir, “Transient electromagnetic fields about an infinitesimally long grounded horizontal electric dipole on the surface of a uniform half-space,” Geophys. J. Astron. Soc, vol. 61, no. 1, pp. 41–56, 1980.

[38] D. Wright and A. Ziolkowski, “Travel time to resistivity mapping using the multi-transient electromagnetic (MTEM) method,” in Proc. Extended Abstracts, 69th EAGE Conf. Exhibition, London, June 11–14, 2007, p. D041.

[39] P. Bannister, “New simplified formulas for ELF subsurface-to-subsurface propa-gation,” IEEE J. Ocean. Eng., vol. OE-9, no. 3, pp. 154–163, 1984.

[40] J. I. Nordskag and L. Amundsen, “Asymptotic airwave modeling for marine controlled-source electromagnetic surveying,” Geophysics, vol. 72, no. 6, pp. F249–F255, 2007.

[41] C. J. Weiss, “The fallacy of the shallow-water problem in marine CSEM explora-tion,” Geophysics, vol. 72, no. 6, pp. A93–A97, 2007.

[42] A. Ziolkowski and D. Wright, “Removal of the airwave in shallow marine transient EM data,” in Proc. SEG Expanded Abstracts, 2007, pp. 534–538.

[43] S. C. Webb and C. S. Cox, “Observations and modeling of seafloor microseisms,” J. Geophy. Res., vol. 91, no. B7, pp. 7343–7358, 1986.

[44] G. H. Hoversten, S. Constable, and H. F. Morrison, “Marine magnetellurics for base salt mapping: Gulf of Mexico field-test at the Gemini structure,” Geophysics, vol. 65, no. 5, pp. 1476–1488, 2000.

[45] S. de la Kethulle de Ryhove and F. Maaø, “On the removal of MT signals from SBL data,” in Proc. EAGE Expanded Abstracts, 2008, p. H047.

[46] A. Ziolkowski, R. Parr, D. Wright, V. Nockles, C. Limond, E. Morris, and J. Lin-foot, “Multi-transient electromagnetic repeatability experiment over the North Sea Harding Field,” Geophys. Prospect., vol. 58, no. 6, pp. 1159–1176, 2010.

[47] M. Commer, and G. Newman, “A parallel finite-difference approach for 3-D tran-sient electromagnetic modeling with galvanic sources,” Geophysics, vol. 69, no. 5, pp. 1192–1202, 2004.

[48] Y. Li and K. Key, “2-D marine controlled-source electromagnetic model-ing, Part 1: An adaptive finite element algorithm,” Geophysics, vol. 72, no. 2, pp. WA51–WA62, 2007.

[49] M. S. Zhdanov, S. K. Lee, and K. Yoshioka, “Integral equation method for 3-D modeling of electromagnetic field in complex structures with inhomoge-neous background conductivity,” Geophysics, vol. 71, no. 6, pp. G333–G345, 2006.

[50] G. Newman, M. Commer, and J. Carazzone, “Imaging CSEM data in the presence of electrical anisotropy,” Geophysics, vol. 75, no. 2, pp. F51–F61, 2010.

[51] S. E. Johansen, H. E. F. Amundsen, T. Røsten, S. Ellingsrud, and T. Eidesmo, “Subsurface hydrocarbons detected by electromagnetic sounding,” First Break, vol. 23, pp. 31–36, 2005.

[SP]

Documents

ieee signal processing