towards improved time-lapse seismic repetition...
Post on 25-Jul-2018
Embed Size (px)
2013 EAGE www.firstbreak.org 73
special topicfirst break volume 31, November 2013
1 Schlumberger.2 Apache North Sea Limited.3 Apache Corporation.* Corresponding author, E-mail: firstname.lastname@example.org
Towards improved time-lapse seismicrepetition accuracy by use of multi-measurement streamer reconstruction
Patrick Smith1*, Jason Thekkekara1 , Julie Branston1, Grant Byerley2, David Monk3 and Jeffrey Towart2 describe, with an example from the North Sea Forties field, how multi-measurement streamer technology enables the reconstruction of a monitor survey to the streamer locations and datum of a previous survey, thus improving time-lapse repeatability.
M onitoring producing hydrocarbon reservoirs by time-lapse (4D) seismic measurements has become routine, and it is widely accepted that accurate repetition of source and receiver locations between
surveys is key to time-lapse seismic data quality. Most marine time-lapse seismic surveys are acquired with towed streamer techniques. While the source locations of these surveys can be accurately repeated (Paulsen and Brown 2008), exact rep-etition of streamer locations can still be challenging.
Eiken et al. (2003) introduced the concept of interpolat-ing multi-streamer shot records from their acquired positions to the desired receiver locations. Duplicating shotpoints and interpolating to the same receiver locations for each survey resulted in well repeated time-lapse seismic data. Unfortunately, accurate interpolation required the use of such small crossline streamer separations that the technique was uneconomic for large-scale reservoir monitoring. Multi-domain interpolation techniques (e.g., Sharma et al., 2011) show promise, but to date have not eliminated the need to accurately repeat source and receiver locations.
The introduction of multi-measurement streamer technol-ogy (Robertsson et al., 2008) enabled wavefield reconstruction from data acquired with economically feasible streamer separations. Two or more multi-measurement surveys may be reconstructed at common locations, or a multi-measurement monitor survey may be reconstructed at the receiver locations of a previous conventional marine streamer dataset. This arti-cle describes the latter application using multi-measurement test lines acquired by Apache North Sea Limited over the Forties field in 2012 with WesternGecos IsoMetrix marine isometric seismic technology.
Reconstruction of multi-measurement marine streamer seismic dataThe multi-measurement streamer used by Apache contained
densely spaced hydrophones and accelerometers that record the pressure wavefield and the vertical and crossline com-ponents of the particle motion vector. Commercial surveys acquired in 2012 used eight multi-measurement streamers, typically with 75m crossline streamer spacing. The streamers were towed at depths of between 15 and 20 m to minimise streamer noise.
Vassallo et al. (2012) showed how the generalized matching pursuit (GMP) algorithm can be used to recon-struct from each multistreamer record, shot-by-shot upgoing and downgoing pressure wavefields sampled on a dense 6.25m inline by 6.25m crossline grid. The streamer array records signal energy over an apparent velocity range of roughly -1480 to +1480 m/s, implying that the reconstructed wavefield supports spatially unaliased signal for temporal frequencies up to about 120Hz. This is more than sufficient for most hydrocarbon reservoir imaging applications. The GMP algorithm outputs the reconstructed wavefields at a user-defined datum.
Repeating a conventional marine streamersurvey with multi-measurement streamer dataThe multi-measurement monitor survey should accurately repeat the source locations of the base survey. The streamer spread width must be sufficient to enable reconstruction of the base survey receiver locations, taking into account any feather mismatch that occurs during acquisition. GMP gener-ates densely gridded shot gathers for the up and downgoing pressure wavefields at the datum of the base survey. These are combined to create the total pressure wavefield, which is then interpolated to the X,Y co-ordinates of the receiver locations of the equivalent base survey shots. The process is shown schematically in Figure1. The total wavefield gener-ated by this process should closely match that recorded by the baseline survey. Performing the reconstruction at the ini-
www.firstbreak.org 2013 EAGE74
first break volume 31, November 2013
source locations of the 2000 survey, and the acquisition parameters of the two surveys are summarized in Table 1 and Figure3. The 50 m streamer spacing of the 2010 survey provides substantial oversampling of receiver locations relative to the 2000 survey and maximizes the chances of accurate receiver repetition. During processing, time-lapse binning was used to select the best repeated traces from the 2000 and 2010 surveys, and to discard the redundant ones.
Having used the Q-Marine point-receiver marine seismic system to acquire the 2010 monitor survey, Apache noted improved signal-to-noise ratio in the 6 to 10Hz frequency range relative to previously acquired datasets. This was attributed to better spatial and temporal sampling, along with the lower noise floor delivered by digital group form-ing technology. The additional usable low frequency energy helped to better delineate the thicker turbidite channel sand complexes that define the Forties reservoir (Figure4). Apache therefore decided, in 2012, to test whether acquisi-
tial stages of the processing flow ensures identical operation, between the baseline and monitor survey, of subsequent steps such as wavelet processing and demultiple.
Time-lapse survey positional repetition accuracy is typi-cally assessed by evaluating, for each time-lapse survey trace pair, the sum of the absolute distance between the source locations and the absolute distance between the receiver locations. This sum is commonly referred to as S+R. A multi-measurement monitor survey reduces R to zero, and so should provide seismic data that closely matches the baseline survey. However the multi-measurement reconstruc-tion process is not error free (Eggenberger et al. in review) and becomes progressively less accurate as D (the distance between the reconstructed output location and the nearest multi-measurement streamer) increases. This may, to some degree, offset the benefits of reducing R to zero.
Time-lapse seismic monitoring of the Forties fieldThe Forties field was discovered by BP in 1970 and is located in the North Sea, about 175km east of Aberdeen. Apache North Sea Limited bought the field in 2003, by which time production had declined from a peak of about 500,000 barrels of oil per day (bopd) in 1979 to about 35,000 bopd. Apache implemented an intensive re-evalu-ation and drilling programme that increased production to around 60,000 bopd by targeting further bypassed pay opportunities with infill wells. The black dots on Figure2 represent the approximately 120 infill wells drilled since Apache began operating the field in 2003. Time-lapse seis-mic technology played a key role in de-risking those infill wells by clearly imaging areas where water has replaced oil in the reservoir (Rose et al., 2011). The red dots represent the 90 remaining infill targets in the current portfolio which are the focus of continued time-lapse monitoring efforts at Forties.
Towed streamer seismic surveys were acquired over Forties in 1988, 1995, 2000 and 2010, with all four surveys being co-processed in 2010. The 2010 dataset repeated the
Figure 1 Schematic representation of the recon-struction of a multi-measurement shot record. The pale blue lines represent multi-measurement streamers at 75 m spacing. The dotted lines represent the dense 6.25 x 6.25 m grid of traces generated by the GMP reconstruction process. The red lines represent the streamer locations of the previous baseline survey. The dense grid of traces is interpolated to the XY co-ordinates of the receiv-ers within the baseline survey streamers for this shot. The reconstruction accuracy is a function of D (the distance between the reconstructed out-put location and the nearest multi-measurement streamer).
Figure2 Outline of the Forties field, and locations of two multi-measurement test lines.
2013 EAGE www.firstbreak.org 75
first break volume 31, November 2013
ation of how the accuracy of the reconstructed data would vary with D. The source array was identical to that used in 2010.
Each multi-measurement multi-streamer shot gather was reconstructed at the receiver locations and receiver datum of the equivalent multi-streamer shotpoint from the 2000 survey. Summation of the reconstructed up and down-going wavefields introduced a streamer ghost consistent with that of the 2000 data.
The reconstructed multi-measurement test line and the equivalent sail lines from the 2000 and 2010 surveys were processed through the flow described in Table 2. This flow is identical to that used during the 2010 time-lapse
tion with a multi-measurement streamer system could accurately repeat the previous surveys while simultaneously enabling the creation of data with broad spatial and tempo-ral bandwidths.
Forties 2012 multi-measurement marine streamer acquisition testTwo of the previous sail lines were reacquired in 2012, as shown in Figure 2. This article describes the results from line 1294 only. The streamer spread configuration, shown in Figure3, was designed so that some of the previous streamer positions would be accurately repeated by multi-measure-ment streamers and others would not. This facilitated evalu-
Figure3 Overview of the basic acquisition configurations for th