Shale reservoir characterisation in the Haynesville

Paul Boughton

Jo Firth looks at an integrated geophysical approach to characterising shale gas reservoirs

CGGVeritas has developed an integrated geophysical approach to characterising shale gas reservoirs, based on pre-stack azimuthal seismic data analysis calibrated with well log and core measurements to identify sweet spots.
Relative production estimates across the reservoir may be derived by combining seismic estimates of lithological, geomechanical and stress properties, correlated to existing well measurements to predict porosity, volume of shale, carbonate content and water saturation. This technique has been applied to the CGGVeritas Tri-Parish multi-client survey in the Haynesville Shale with positive results (Fig. 1). However, all shale plays are different, so the selection of geomechanical and lithological parameters that provide the best correlation with production will vary from one play to another and must be derived for each survey.
A quantitative understanding of a host of rock properties such as acoustic impedance, Poisson’s ratio, and Young’s modulus can be obtained from prestack seismic. These properties are in turn related to quantitative reservoir properties such as porosity, Total Organic Content (TOC) and mineral content.
CGGVeritas combines the elastic rock properties, derived from seismic inversion, with azimuthal velocity and AVO analysis of conventional 3D seismic data to estimate principal stresses. The Differential Horizontal Stress Ratio (DHSR) is an important parameter for prediction of hydraulic fractures and can be estimated from seismic data alone. These estimated stresses are then calibrated to the stress state of the reservoir, derived from drilling and completion data, microseismic analysis and regional information. Zones with relatively high brittleness (derived from isotropic Young’s Modulus, Lambda-Mu-Rho, etc) and low DHSR (no preferential stress orientation) are more prone to fracturing and tend to produce fracture swarms when completed, potentially increasing production.

[Page Break]

CGGVeritas has applied this workflow to its Tri-Parish data. This is a high-pressure, high-temperature (HPHT) field with no large structural features and so has a relatively homogeneous stress field. Although there is an east-west regional orientation of the maximum horizontal stress field over the Haynesville area, there are lateral variations in the local stresses. The understanding of this variability is crucial for optimal completions. Potentially brittle zones have been identified and their associated DHSR, fracture initiation pressure and closure stress have been estimated. Calibration and validation are critical. These seismically derived predictions were calibrated with existing production and well core and test measurements to determine optimal zones for drilling and completion.
The azimuthal signature needs to be adequately recorded, processed and measured In order to estimate the stress field distribution. After gather conditioning a dataset should have maximized signal-to-noise ratios, flattened reflector events and preserved amplitude variations.
The key steps in compliant amplitude-versus-offset and amplitude-versus-azimuth seismic data conditioning are random noise attenuation, angle muting to remove spurious energy beyond a threshold angle of incidence, high-resolution de-aliased radon transform de-multiple and high-density velocity analyses for both anisotropic and azimuthal velocity derivation. Ray bending for offset angle gather conversion and reservoir-oriented gather conditioning for prestack inversion are also critical.
Multi-attribute analysis of the Tri-Parish data suggests that better development locations are found here in areas that have a combination of certain key rock properties, for example, better porosity development, high siliceous mineralogical content, and high values of TOC. Detailed rock property analyses showed that properties such as Poisson’s ratio and Lambda-Rho (incompressibility) are useful for identifying these areas in this field. Low Poisson’s ratio areas indicate the more siliceous, low-carbonate content, normally associated with better porosity development. Bulk volume of gas can be estimated by combining these properties via multi-attribute analysis.
In general, we found that no single attribute provided enough information to correlate seismic data to production, but multiple elastic- and stress-related attributes can be correlated to average production and horizontal well length at well locations with 95per cent correlation. The predicted production shows several undrilled areas with potentially high productivity.
For validation, we compared the predicted local stress fields to tri-axial measurements from core samples at two locations. The full strain tensor and the principal stress directions were measured from these core samples, which then served as baseline values for correlating seismic predictions. We found that the direction of maximum horizontal stress, predicted from seismic observations, matched the corresponding core stress measurements to within 5 per cent.
Although a statistical method of exploitation, where several wells are drilled near a productive well and areas around failed wells are abandoned, has proved successful so far in the relatively homogeneous Haynesville Shale, there are opportunities to increase production by targeting predicted sweet spots. This is what CGGVeritas can provide through the analysis and calibration of the stress and related attribute volumes derived from this seismic shale gas workflow. This analysis and workflow have the potential to predict fracture behavior and reservoir drainage geometry, enabling optimal well placement and completion designs including stage zoning in hydraulic fracture stimulation. Significantly, this will also mitigate the risks associated with drilling hazards and hydraulic stimulation.

[Page Break]
Meanwhile, since launching BroadSeis at Barcelona last year, CGGVeritas has recorded BroadSeis data in many locations around the world, in different water depths and over different geologies. In all cases the increases in bandwidth at both low and high frequencies (six octaves have been recorded in the Gulf of Mexico) have provided significant improvements in imaging and data quality.
Since the beginning of the year, CGGVeritas has four commercial BroadSeis projects awarded and is currently acquiring and processing a multi-client 3D Survey in Quad 29 of the Central North Sea. Initial results from this survey are causing considerable excitement.
Off North West Australia the extra low-frequency energy recorded allows far better differentiation of layers than conventional seismic, due to the lack of side lobes to the wavelet and the excellent phase control of the low frequencies. The broad bandwidth provides high resolution as well as stunning textural and stratigraphic detail and, perhaps for the first time, allows for direct discrimination of rock or fluid properties. These characteristics have also been particularly noticeable offshore Guyana.
Where penetration sub-salt or sub-basalt is an issue BroadSeis performs well.  
In these areas resolution is limited by a lack of high frequencies, which are mainly attenuated due to scattering and transmission losses, lack of illumination, multiple contamination and velocity model errors. Three octaves of signal are generally considered necessary for adequate seismic resolution. Therefore, in order to achieve sufficient bandwidth in these areas, it is necessary to extend the low frequencies. Conventional marine streamer acquisition lacks sufficient signal-to-noise ratio in the 2-7 Hz bandwidth due to streamer depth, tow noise and source array configuration. BroadSeis acquisition achieves usable frequencies down to 2.5 Hz, providing three octaves of data below 20Hz, so allowing seismic resolution sub-salt and sub-basalt.
West of Shetlands considerable increases in signal-to-noise ratio and continuity have been observed sub-basalt on BroadSeis data. Clear definition of intra and base basalt layers has also been seen, including probable multiple basalt flows pinching out.  In the Gulf of Mexico considerable improvement to the base salt illumination was achieved. On this survey six octaves of usable data was recorded, providing impressive resolution and texture in the near surface, as well as continuity at depth.
In a pilot 3D survey in the Central North Sea BroadSeis provided increased lateral resolution, with the timeslices in the near surface (see images) showing clear definition of channels and yielding detail similar to a topographic map. Despite being recorded with short streamers (4km) and having a limited aperture, being only 7.5km wide, the images of the Base Cretaceous Unconformity (BCU) provide considerable improvements in structural resolution and imaging over the reference high-quality conventional long-offset (6km) data in the area.
The BroadSeis 3D data is less noisy and the additional low-frequency energy helps delineate deep structures, showing clear layer differentiation and local impedance contrasts and heterogeneities, without the confusion caused by wavelet side-lobes. The broader frequency content, especially at the low end, improves sedimentary package differentiation and delineation within the sub-BCU Upper Jurassic strata as well as in the Cretaceous above.
The pilot 3D was deliberately acquired over a known AVO anomaly. The BroadSeis deghosting algorithm is true-amplitude preserving and angle stacks over this anomaly showed the AVO effect more clearly than the conventional data. The improved low frequencies achieved using BroadSeis give better stability and accuracy to the inversion process and the prestack inversion of this dataset provided better correlation with the well data than inversion of the conventional dataset.
Interpreters find it much easier and faster to work with the BroadSeis 3D data. The subtlety, texture and continuity within the volume mean that auto-picking of horizons is far quicker and more reliable. The frequency bandwidth allows the auto-tracking tool to work optimally with greatly reduced intervention, allowing interpreters more time to focus on interpretation subtleties.
The many examples of BroadSeis data that CGGVeritas has now acquired from around the world all show considerable improvement in data quality, increasing both lateral and temporal resolution. BroadSeis decreases confusion caused by wavelet side lobes and so clarifies impedance contrasts. The extra low-frequency energy allows layers to be easily differentiated. The increased bandwidth and resolution of BroadSeis provides a step-change in seismic stratigaphy and allows us to directly infer lithology and fluid effects in the data. BroadSeis 3D data has more than lived up to the expectations of the earlier 2D data and we look forward to further BroadSeis success later this year.

Jo Firth is with CGGVeritas, Crawley, UK.
Fig.1. Mapping of the predicted first six months of production (calibrated to horizontal well length), based on correlation of calibrated geomechanical and lithological attributes with actual production rates. High-productivity areas are shown in red.
Fig. 2. Shallow timeslice (262ms) from UK CNS Quad 20 showing exceptional resolution, character and detail resembling a topographic map.

Recent Issues