The challenge of the Marlim Complex deepwater 4D project was to acquire seismic data over 1,520km² in a heavily obstructed oilfield producing more than 550,000bpd from ten fixed production platforms. Seismic quality control was so relevant for this project that, for the first time in Brazil, six floating production, storage and offloading vessels (FPSOs) rotated to align with the seismic vessel, which towed ten cables, each 6km in length.
In the interpretation phase, 4D imaging enabled us to detect the movement of oil, gas and water through a reservoir. It improved static and dynamic modelling, and provided confidence to predict reservoir behaviour, thus reducing risk in new drilling projects.
It also helped locate bypassed reserves and optimise the placement of infill wells. In 1994, we started the first phase of 4D seismic with a pilot onshore steam injection-monitoring project.
After 15 years of experience with 4D we developed a 4D seismic workflow. These are some of the most important steps in the process.
Deepwater Marlim Complex 4D seismic project
The Marlim Complex 4D project has been the largest seismic acquisition survey oriented to reservoir characterisation and production monitoring in Brazilian fields. The giant Marlim turbidite oilfield, discovered in 1985, 130km off the southeastern Brazilian coast, in water depths between 600 and 1,100m, is the largest producing field in Brazil.
It is 100% operated by Petrobras and is exploited through 85 producing wells and 51 injectors connected to seven floating production units. The initial development phase has recently concluded, with about 200 wells (41 horizontal) successfully drilled.
A new development phase is now in its early stages.
Technical and economic feasibility study
A risk analysis related to reservoir thickness, rock quality (facies and net-to-gross distribution), structural context, basic sediments and water (BSW), and gas-oil-ratio (GOR) was carried out for each well. For the economic analysis, we used VOI methodology with a two-path decision tree – one with new seismic and another with no seismic.
For the seismic path, we estimated the "expected monetary value" by considering success and failure paths and did the same for the path without seismic. This simple approach enabled us to estimate the gross value of new seismic information.
The next step was to reduce the cost for acquisition, processing, and manpower for seismic interpretation so as to quantify the net VOI. The Marlim asset team identified six types of potential gain after 4D interpretation: new infill wells, standby drilling wells due to geological uncertainties, reduced drilling investment, increased well productivity, identification of new exploratory opportunities, and increased workover efficiency.
To see if 4D seismic could map the spatial distribution of reservoir fluids, the first step was to build a 3D petro-elastic model. We downscaled the information from the simulations and built a model based on Gassmann's equation. Multiple synthetic seismic and acoustic impedance images were calculated from the flow model to guide future seismic interpretation.
Seismic evaluation and acquisition design
During this phase, we detected a strong compressional velocity and density contrast in the Marlim reservoirs, although there was little contrast in shear velocity. The 30Hz dominant frequency in previous data was not enough for the desired vertical resolution; at least 60Hz was needed.
VSP implied that 60–70Hz was achievable. Analysis of a 3D swath suggested that the reflection signal in the reservoir was still strong at the recorded maximum offset of about 3,800m.
The South Marlim reservoirs required better multiple attenuation, which led to recording a maximum offset of 6,000m. Analysis of a 2D long-offset line showed that the signal at the reservoir level remained up to about 5,000m.
The signal, from reflections 800m deeper, extended beyond 6km. Anisotropy was evident, starting not far from the seabed.
The long-offset signal is lower in frequency and undergoes phase change. Thus, despite not enhancing the stack, it contained useful information on the nature of the reservoir.
A feasibility study demonstrated the difficulty of obtaining full coverage in a region with many obstructions. It was difficult to obtain good coverage simply by shooting at the same direction and undershooting.
A better job could be done by steering the vessel carefully around the obstructions and by more elaborate undershooting, but it would still be very difficult.
There were two major processing concerns: multiple attenuation and prestack migration. The wide-azimuth undershooting of platforms made it difficult for 2D surface multiple attenuation techniques, so 3D SRME techniques were recommended.
We were also aware of the need to obtain something better than one-third fold cover in each of the near/mid/far offset ranges to try to minimise the gaps in the base offset ranges, at least in a 3D sense.
The 2005 data set covered 1,520 km² and took 201 days. When acquiring data over a producing field, it is common to encounter obstructions.
We have fixed installations and non-permanent obstructions that challenge efficient perations: production platforms, floating production, storage and offloading vessels, drilling rigs, tankers, pipe laying barges, marker buoys and general marine traffic. A vessel with ten cables, each 6km long and separated by 50m, would be possible for about two-thirds of the survey.
The final phase would be shot in a two-boat mode, with a source vessel and a recording vessel that reduced cable length to 3km in undershooting areas. The currents were occasionally very high and produced feathering of 15-30°. This caused severe damage in the repeatability of the vintages.
The results of a 2005 pre-stack time migration-migrated section, when compared with the same line from 1997, were outstanding in terms of resolution to improve reservoir characterisation of Marlim and South Marlim. In order to generate a 4D response from the 1997 and 2005 demultipled data sets, we needed to minimise their acquisition differences.
As well as the different acquisition parameters, other differences were the result of the significant increase in surface production installations from 1997 to 2005.
The data were compared to find the most closely colocated pair of field traces on an offset and common mid-point by common mid-point basis. After identifying matching pairs of traces, the data sets were edited to preserve these (the most repeatable) traces before imaging.
After each processing stage, 4D quality control attributes were generated. These are measures of the differences in amplitude, phase, and timing between the data sets.
The 3D stack volumes were also used to create a difference volume and compute the nrms attribute. Anomalies in these quality control attributes highlight areas with possible residual acquisition-related differences.
The most important contribution of 4D interpretation to the geological modelling was to characterise absolute horizontal permeability trends for the Marlim Turbidite System. The 4D amplitude difference (2005-1997) map revealed important anisotropy in water displacement around the reservoir base.
This anisotropy was introduced into the absolute-horizontal-permeability maps of the lower zones in the turbidite system by geostatistics, and the history match of the field was improved by more realistic geological permeability maps. Before updating the geological model with the 4D results, a good history match existed but in order to attain it the engineer spent a year fitting the data and had to introduce big changes in the original absolute distribution of horizontal permeability.
Using the absolute distribution of horizontal permeability from the 4D data improved history matching without many changes in the original permeability map, yielding a more reliable model. Also, the time spent in history matching was reduced by three months.
In the interpretation phase of the Marlim 4D project, we took advantage of the quality of our reservoirs, rock and fluid properties, the thick unconsolidated turbidites sandstones reservoir, the excellent Campos Basin seismic response, good signal-to-noise ratio, and some clear fluid contacts and seismic anomalies in addition to the extensive water-injection effects in the reservoirs to calibrate our 4D signals and integrate 4D interpretation into the 3D geological and flow modelling.
After 4D interpretation, five new wells confirmed the 4D indications. A secondary gas cap, a confined region with overpressure, and heterogeneities were identified by 4D data and confirmed by production and well data.
Many wells were repositioned because of water indications and sealing faults. In addition, permeability maps were improved because of the pattern of water displacement shown by 4D imaging.
Comparing simulation results before and after 4D interpretation, shows that the net value of the project doubles, total oil production increases 4.76% (the number of wells to be drilled decreased from 13 to 10), and production per well increases 24%. Most of this improvement can be attributed to the 4D interpretation that cancelled two wells and repositioned many others.
However, the information provided by 4D seismic must be coupled with additional information such as saturation logs, tracers, formation tests and pressure. Additional work is particularly needed in recognising 4D signatures for tuning particular areas.