The seismic amplitude and its variation with offset – or angle of incidence, as it is also known – is important for characterization of conventional reservoirs, but the transformation of seismic data into precise reservoir properties is fraught with challenges. A notable complication is the interference caused by converted waves.
Converted seismic waves are the downward-traveling compressional, or P-waves, that transform into upward-traveling shear, or S-waves, upon reflection at an interface. These waves can contaminate the primary P-wave reflections utilized in AVO analysis, leading to false positive anomalies or masking genuine AVO anomalies. Such interference can have serious consequences, such as misclassifying prospective targets as non-reservoir zones, or vice versa. While it is possible to get rid of converted waves during data processing, geoscientists might not always have the time, resources or capability to reprocess the data. When faced with these limitations, they must rely on the data at hand. Detailed log analysis and incorporation of geological insights can aid in detecting potential converted wave interference, enabling geophysicists to refine their interpretations and mitigate the effects of converted waves on AVO analysis.
This case study illustrates the impact of converted wave interference on seismic amplitudes and AVO responses, which can result in incorrect interpretations of reservoir properties. Also discussed are strategies to prevent such misinterpretations and ensure a more accurate reservoir characterization.
Geologic Setting
The Lower Magdalena Valley basin, a prolific region located onshore in northwest Colombia, is a forearc basin that was formed through a series of subduction and accretion events. This gas-rich basin primarily features structural traps, such as anticlines and three-way fault-dependent closures. The principal source rock is derived from the early Cieneraga de Oro formation dating from the late Oligocene to the early Miocene. The main reservoir sands are situated in the upper segment of the CdO formation, which dates back to the late Miocene. The CdO formation’s depositional environment transitions from fluvial to deltaic and into shallow marine settings. The CdO formation’s thickness spans from 1,500 to 3,000 feet, with porosity ranging from 18 to 28 percent. The CdO reservoir units vary from massive to laminated sands, are medium to coarse in grain size, and sometimes interbedded with limestone, coal and shale. The deeper segments of the CdO exhibit more fluvial characteristics and a greater abundance of coal. The overlying cap rock consists of the Porquero formation from the Mid Miocene, characterized by a deep marine depositional environment. This formation is truncated by the regional Mid-Miocene Unconformity and is topped by the Tubara formation’s shallow marine deposits from the Upper Miocene.
Available Seismic Data
The acquisition and processing of the 3-D seismic survey were completed in early 2024. The seismic data, acquired using dynamite as the source, featured a maximum offset of 4,590 meters, a two-millisecond sample interval, a six-second record length and a bin size of 20-by-30 meters. The data processing sequence included surface-consistent deconvolution, velocity analysis, trim statics, surface-consistent scaling, 5-D trace interpolation, prestack time migration, residual and VTI velocity analysis, as well as high-resolution radon transform for multiple suppression and noise attenuation, all of which are now available in a standard workflow for reservoir characterization and quantitative interpretation.
In addition to the pre-stack seismic data, well logs for some wells were also available. These logs could be utilized to identify the anomalous zones through AVO analysis and impedance inversion. Such analyses typically commence with the preconditioning of seismic gathers to improve the signal-to-noise ratio. Figure 1 illustrates a comparison of several input and conditioned gathers near one of the wells. After conditioning, as shown in figure 1b, there is a noticeable reduction in high-frequency noise and enhanced alignment of reflection events. The well depicted, superimposed on the section in figure 1, revealed a substantially thick gas-bearing layer at the depth marked by the lower yellow block arrow, with a thinner gas sand layer above it. Consequently, conducting AVO analysis is of interest not only to verify the presence or absence of gas in the available wells but also to identify other areas within the 3D seismic volume where gas presence might be detected.
The next step in the AVO workflow involves the offset-to-angle transformation, as AVO formulation is carried out in the angle domain. This means that offset gathers are transformed into angle gathers. Typically, this transformation is performed using a seismic velocity or a velocity field derived from wells. More information on this comparison can be found in the Geophysical Corner column of March 2019. On examining the gathers, the maximum angle utilizable for AVO analysis was found to be 36 degrees.
Once the range of angle of incidence (0-36 degrees) is determined, AVO attribute volumes can be calculated using the corresponding AVO equation, tailored to the specific AVO attributes needed. For instance, to obtain the intercept and gradient volumes, one can apply Shuey’s approximation to the Zoeppritz equations. Likewise, to create P- and S-reflectivity volumes, Fatti’s approximation is suitable, which is also necessary for constructing the fluid stack. The type of AVO anomaly anticipated dictates the suitable interpretation of these attributes, whether individually or in combination.
Figure 2a displays a crossline section traversing the fluid stack volume, overlaid with the well depicted in figure 1. The absence of an anomaly at the location indicated by the cyan arrows is perplexing. It suggests a meticulous review of the data at hand and consideration of possible pitfalls.
Factors Affecting Seismic Amplitudes
In AVO analysis, the focus is usually on primary events observed on amplitude-preserved prestack seismic data. However, P-SV mode-converted waves, which are generated at oblique incidence angles or when there’s a significant contrast in subsurface rock layers, appear at larger offsets. These waves intersect primary events, leading to notable amplitude variations. While these converted waves are sometimes suppressed during processing, if they persist and are within the target layers, it’s crucial to identify them through forward modeling and reduce their impact in AVO analysis by applying offset or angle mutes. To assess the effect of converted waves on seismic data, we undertook a series of steps involving a detailed review of well logs, utilizing them for forward modeling, comparing the modeled seismic response with the actual data, and adjusting angle gathers to produce the fluid factor AVO attribute.
Examination of Well Data
In reservoir characterization, the accuracy of well log data is crucial. Compressional (DT) and shear (DTS) sonic logs are essential for precise porosity estimation, pore pressure prediction and elastic impedance calculations, but they are also highly sensitive to borehole conditions and environmental factors.
For the study area, we used data from adjacent wells to improve the accuracy of compressional and shear sonic logs. This method relies on the principle of spatial continuity, which assumes that formations in a geologically consistent environment should exhibit similar log responses. By incorporating trends and constrained values from nearby well logs, we can correct potential discrepancies and inaccuracies in the sonic data for the well in question.
The procedure begins by carefully selecting nearby wells that ideally intersect the same target formations and exhibit minimal lateral variations in depositional patterns. Once the reference wells are chosen, the correction process leverages two key aspects of their sonic log responses: trend analysis and constrained values within the region. It involves identifying characteristic response patterns in the sonic logs of neighboring wells for both DT and DTS across the targeted formations. By establishing these trends, a baseline for the expected sonic behavior within the target reservoir is set. In geological strata with low expected variability, such as clean sandstones or shales, sonic values from adjacent wells provide constraints on the target well data. This approach promotes consistency and reduces potential borehole effects on the sonic logs of the target well.
The integration of trend analysis and constrained values enables a data-driven calibration of the target well sonic logs through linear regression for DT and DTS.
Figure 3a presents a crossplot of P-velocity (DT) against S-velocity (DTS) using data from several nearby wells with available shear logs. Notably, the cluster of points in the lower left, marked by a red dashed ellipse, seems unusual. This cluster originates from a shallower section (12.5-inch section) of the well.
Following the trend analysis of the well data and constraining the values within the specified strata, a new crossplot was created, depicted in figure 3b. The previously noted cluster of points has now been eliminated.
Modeling AVO Response
After correcting the well log data for sonic and shear measurements, an elastic gather (forward model) was created to assess amplitude variations with offset at the target levels. Figure 4a illustrates the modeled elastic gather with only primary reflections, while figure 4b displays the amplitude variation as a function of offset for two reflective events (marked in red and green on the elastic gather), which represent the top and base of the gas sand. A Class III AVO response is evident at the interface between shale and gas-charged sand (red event), with a contrasting response for the blue event. The crossplot of intercept and gradient for the data, depicted in figure 4c, shows a red square in the third quadrant and a blue square in the first quadrant, indicating a Class III anomaly. However, this anomaly was not present in the actual seismic data, as previously indicated in figure 2a.
To determine the reason for this discrepancy, two elastic gathers were produced: one featuring only primary events and the other including both primary events and converted waves. Figure 5 illustrates the comparison between the modeled gathers and their anticipated AVO trends from the elastic gathers with solely primary events (figure 5a) and those with both primary events and converted waves (figure 5b).
The AVO trends for two seismic events, indicated by the trough and peak with pale yellow and green arrows, are shown in figures 5c and 5d. The red and blue lines represent the best fit lines for synthetic data generated with only primary events and with both primary events and converted waves, respectively. These displays suggest that converted waves can introduce a sinusoidal or wobbly pattern in amplitude, which may alter the AVO gradient and, consequently, the interpretation of derived DHI volumes. Upon analyzing the AVO trend of real seismic data across various seismic events and locations, a sinusoidal amplitude pattern was observed, as depicted in figure 6. This pattern, also seen on the elastic gathers, suggests potential contamination of real seismic data with converted waves, necessitating extra caution during AVO anomaly interpretation.
Mitigating the Influence of Converted Waves
The recognition of converted waves in actual seismic data presents a subsequent challenge: their mitigation. One strategy is to return to the processing phase to carefully eliminate these converted waves, although this could lead to extra expenses that the management might not approve. Another option is to diminish the impact of converted waves by restricting the maximum angle in AVO analysis, since converted waves are typically more noticeable at greater offsets, a fact corroborated by the forward modeling shown in figure 5.
Figure 7 demonstrates the impact of selecting the maximum angle for AVO analysis. Figures 7b and 7c display the AVO analysis corresponding to a pronounced trough when maximum angles of 36 and 30 degrees are applied, respectively. An interesting observation arises: the AVO trend seems to coincide with a non-reservoir zone when angles up to 36 degrees are used, but it transitions to a Class III anomaly trend, which is anticipated at this level, when the angle is limited to 30 degrees. Figure 2b presents the fluid-factor section along a crossline that intersects the well, with the P-velocity curve in red superimposed on both images. The fluid factor has been calculated to reduce the effects of converted waves by limiting the maximum angle in the analysis. The anticipated anomalies for the two sands are now visible at the location indicated by the two cyan block arrows, which were absent in figure 2a.
The spatial distribution of AVO anomalies can significantly differ based on the chosen maximum angle for the analysis. This is demonstrated in figure 8, which shows horizon slices from the fluid factor volume across a 10-millisecond window at the gas sands level. Figure 8a reveals the spatial distribution of fluid factor anomalies without reducing the impact of converted waves, whereas figure 8b displays the distribution with a reduced effect of the converted waves.
Conclusions
The occurrence of converted waves within P-wave seismic data can obscure the expected AVO response for identified target gas sands, which is usually evaluated to determine a technique’s effectiveness. Our research demonstrates that converted waves can be pinpointed through elastic gather modeling using well data, considering scenarios with only primary events as well as those with both primary and converted wave events. By comparing these models with actual seismic gathers, we can accurately gauge the level of contamination. In our study, neglecting converted waves led to the fluid stack attribute missing the signature of the target gas sands at a well location. Nevertheless, restricting the AVO analysis to an angular range of 30 degrees diminishes the impact of converted waves on P-wave data, thereby highlighting the fluid stack anomalies linked to the target sands. This method streamlines the lateral interpretation of fluid anomalies, enhancing their potential for exploitation.