Complex geology and deepwater presalt targets offshore Brazil lend themselves to seismic imaging challenges. The postsalt sections of the Santos and Campos Basins also featured thin-bed, high-impedance layers of volcanic material and other shallow velocity anomalies. These features were often near the top of salt (TOS) and produced areas of poor presalt focusing. Because accurately modeling the postsalt was imperative for imaging the presalt, we developed a new workflow for velocity model updating. Where previous top-down methods have failed, our method embraced the challenges of imaging beneath the volcanics by incorporating additional presalt information into the updates. Our method is interpretation-dependent but provides a robust and reliable workflow for updating the postsalt velocity model in the presence of complex geology. Using a narrow azimuth data set acquired with variable-depth streamers, we evaluated the effectiveness of our iterative workflow. The result was improved presalt imaging.


Numerous seismic imaging challenges exist in the basins offshore Brazil, mostly due to the region’s complex geology and deepwater presalt targets. The postsalt sections of the Santos and Campos Basins also feature thin-bed, high-impedance layers of volcanic material and other shallow velocity anomalies. These features can produce areas of poor focusing beneath them due to the complex lateral velocity variation related to the movement of the volcanic layers. Because accurately modeling the postsalt is imperative for imaging the presalt, which is today’s focus of exploration in the region, we present a new workflow for velocity model updating. Our method builds on the work of others and incorporates different methods into an iterative and holistic approach to velocity model building.

Geologic motivation

The Santos and Campos Basins (Figure 1) are located in a deepwater region approximately 300 km off the coast of Rio de Janeiro, Brazil. Significant discoveries in these basins have caused an increase in attention toward improving the seismic imaging of the area to aid in understanding the regional geologic setting.

To date, only a few wells have been drilled into the presalt, primarily in areas well-imaged by exploratory seismic surveys. The presalt sequence can be categorized into three phases: lower synrift, upper synrift, and sag sequences (Cainelli, 1999; Carminatti et al., 2009). Synrift sequences display numerous extensional faults (Mann, 2013). The upper synrift section in the northern Santos Basin is significant because it provides a good reservoir rock for the Franco and Libra fields. The postrift sag sequence is also suspected to be of reservoir quality. Figure 2 shows a clear picture of this presalt, postrift sag sequence. New imaging techniques in the area aim to properly image the postrift sag and synrift sequences.

Because the velocity above the salt has large implications for imaging in the presalt sections, we devoted our attention to refining the postsalt sediment velocity model. Some of the biggest imaging difficulties in the region are the result of igneous intrusions above the salt and other high-impedance shallow anomalies. The layers of volcanic material exhibit very high-amplitude seismic response yet are very thin. Well logs in the area indicate that these volcanic events tend to have a thickness of less than 50 m, which can be difficult to accurately update with conventional ray-based tomography. Furthermore, the instances of volcanics are not consistent throughout the basin; in some areas, the volcanics occur in small, isolated areas and appear with more winglike characteristics (Figure 3). Other locations are more challenging because the volcanic layer can be mapped over larger areas (Figure 4). Localized and subregional volcanic layers often can be identified in close proximity to the top of salt (TOS). When the subregional volcanic layers are located near the TOS, differentiating between salt boundaries and volcanic intrusions can be very challenging (Figure 5). Related to the impedance contrast, the sediment layer between the volcanics and TOS is often void of recognizable seismic events. However, even when we can see the events, particularly in the Santos Basin, we face highly faulted sections beneath the volcanic intrusions that can cause additional lateral velocity variations in already poorly imaged areas. Indeed, these thin layers of volcanic material test the limits of current imaging and velocity-model-building techniques and warrant more reliable workflows. Additionally, only a limited number of wells exist in the region, and we rely heavily on the evaluation of seismic data for validation.

Current methods for updating suprasalt sediment velocities

Evaluating the postsalt velocity in the Santos or Campos Basins can be quite challenging because of the complexity in the geology; when volcanic intrusions are present, the problem intensifies. Figure 6 shows a traverse line through the Santos and Campos Basins. The Cabo Frio high, indicated by the blue vertical line, marks the transition among the two basins. Each basin exhibits complex geologic features, and the notable volcanic intrusions are seen as high-amplitude reflections above the salt (indicated by the yellow arrows). Typical solutions for updating the seismic velocity in the area include a combination of ray-tracing tomography and layer-based updates constrained by geologic interpretations in the region.

One top-down strategy to model building in the Santos Basin described by Huang et al. (2010) incorporates at least one iteration of Albian-layer tomography. Because of the abundance of carbonates at the Albian level in the Santos Basin, the velocity in the layer does not follow typical sediment compaction trends. Optimized parameterization for common image gather (CIG) picking and tomography are necessary to update velocities within the Albian layer. Huang et al. (2010) also highlight the importance of incorporating anisotropic parameters into the velocity model.

Hu and Zhou (2011) show the benefit of high-resolution tomographic updates in the Santos Basin using examples around the Lula discovery. The higher density picks in the x-⁠, y-, and z-directions improved the accuracy of the velocity model. When this high-resolution approach is combined with the nonlinear slope tomography method designed by Guillaume et al. (2012), incremental details are added to the sediment overburden. High-resolution, nonlinear slope tomography incorporates an additional term in the cost function involving the dip of kinematically migrated, locally coherent events. Nagir and Malave (2013) use a tomography method that is constrained by a structurally conformant 3D dip field in the Santos Basin, which results in a velocity model that accurately follows the geologic structures and improves the interpretability of the presalt.

Chen and Shen (2012) present an example of weighted high-resolution tomography in the Para-Maranhao Basin in the northern Brazil. Shallow velocity anomalies were updated using weighted constraints, preventing anomalous velocity values from distorting the deeper velocity updates. Strong lateral velocity changes were more accurately handled using this weighted approach to tomography. However, this approach relies on deeper sediment events to update the shallow anomalies and cannot be applied to the Santos and Campos Basins.

Finally, sediment velocity scan is another method that uses deeper events for shallow updates. Velocity scans, however, require sufficiently thick layers to operate. In the Santos and Campos Basins, the sediment layer between the volcanics and the TOS is not very thick, and the salt is too complex to be reliable for update validation.

Each of these methods has its own merits and provides some improvement in the velocity model, but they all struggle in the presence of thin-bed volcanics existing in close proximity to the TOS. All of the methods rely on events that are either inside or below the region of the intended velocity update. However, in this geologic instance, we have neither. Unassisted, they do not solve all of the challenges seen beneath very strong velocity anomalies.

New iterative method

One important facet of our new workflow is that while previous top-down methods wait for a finalized suprasalt sediment model before defining the salt boundaries, our method does not. It moves forward with a salt interpretation, knowing that the postsalt sediment velocity is not finalized in all regions. Then, we use the salt itself and the presalt as criteria to improve the quality of our sediment velocities and to evaluate the updates. This philosophy has implications in each piece of our iterative workflow as outlined below.

Agnihotri (2013) introduces an iterative approach focused on salt-overhang interpretations in the Gulf of Mexico. In his method, approximations of the salt geometries are defined early, and subsalt images are used to first determine if a change in interpretation is moving in the right direction (i.e., honing in on the true geometry). But the subsalt is also used to help identify areas that may require salt modification. We built on this concept and made a few modifications:

  1. Although Agnihotri’s method is primarily used for updating the salt boundaries, our goal is to update sediment above the salt as well as refine the salt geometries.

  2. In addition to using subsalt focusing as a way to evaluate the impact of the update, we attempt to use the image beneath salt to help determine the update. This change provides additional controls for the updates and allows for convergence of the velocity updates.

The first phase of the workflow is defined as the classification phase (Figure 7). Even though the salt geometry may be modified later with further sediment updates, we use a complete salt interpretation for the entire region. From our initial model, we can quickly separate the well-imaged presalt areas from the poorly imaged ones. Areas that have reasonable imaging in the presalt continue with a traditional, top-down approach for velocity model refinement. Areas that are poorly imaged continue onto the next phase of our workflow.

The second phase of our workflow is the subvolcanic velocity scans. Although the subvolcanic scans may have been ineffective as part of a traditional top-down flow, they can provide meaningful updates when coupled with the salt body updates. This coupling enables us to use changes in the presalt to gauge the accuracy of the update in the postsalt. Because the gather and stack focusing improves, additional insight into the model can be derived from the data. Previously unseen events can be seen and often times lead to modifications of the previous salt body and possibly even volcanic interpretations.

With a more accurate postsalt model, previously ambiguous areas beneath the volcanics are now able to be evaluated with tomography. This initiates the third phase of our workflow, the iterative tomographic update. For this phase, we can deploy any or all of the methods we described previously — methods that had once been limited by the poor illumination are now better equipped. Local geology helps us determine which of the tomographic methods we should use.

Structurally guided tomography incorporates the regional geologic interpretation of the BOS and the presalt structure to anticipate velocity errors in the postsalt velocity model. This method further refines the model from the previous velocity scan updating phase and provides additional insight into the presalt structures. Because the velocity model converges, gather curvature begins to become more consistent, allowing for additional high-resolution tomography as well as nonlinear slope tomography in some regions. Updates are performed in these local regions, incorporating gather curvature from the presalt and modifying the salt geometry with each update. Salt geometries are modified through a combination of ray tracing and manual picking as needed. By implementing an iterative process, including reevaluating the TOS, BOS, and postsalt velocities, we are able to converge to an accurate velocity model.

Once the velocity model is of sufficient quality throughout the entire survey, previously segregated areas can be combined into one for the final interpretation phase. In many areas, multiple passes of each stage are applied to further refine the model. Any remaining degrees of detail are added to generate the final, full-volume velocity model.

The role of salt interpretation in the iterative flow

This iterative workflow is heavily dependent upon interpretation. From the onset, we need an accurate interpretation of the top of volcanic intrusions as well as a reasonably well-defined TOS and BOS. Because each iteration requires modifications of the salt geometries, the entire region can potentially be interpreted multiple times over. Initial horizon interpretations may lack some of the accuracy needed to fully image the presalt structures. Because confidence is gained in the sediment model, additional details are gradually incorporated into the interpretations.

To preserve the interpretation accuracy in a cost-effective way, two types of interpretation modifications are used in the flow: 3D ray tracing and manual interpretations. The 3D ray-tracing method performs a demigration of the surface using the previous velocity model and subsequently remigrates the surface with the new model. This method is fast and works well for overall regional velocity changes. Most iterations require subtle changes to a local salt boundary, which the 3D ray-tracing algorithm handles well.

If the updated model has a significant velocity modification or the updated sediment velocity provides a new understanding of the geologic structure, significant change from the previous interpretations can be applied. In these cases, manual salt interpretation is required. It is slower than 3D ray tracing, but it provides much needed improvement and detail to the updated model. Also, the significant change of the salt interpretation itself can reflect the improvement of the updated velocity model. With the modifications to existing salt geometries, a migration is required to verify the update. If the modifications improve the presalt image, then they are kept. If the image does not improve, the interpretations can be adjusted. Although the iterative manual adjustments and validations sound like a time-prohibitive procedure, in practice with our study area, we found that the method worked expeditiously and was worth the additional time to arrive with superior results.

Case study

Our case study uses data from a recently acquired narrow azimuth variable-depth streamer survey (Figure 8). The survey was acquired using 8-km-long cables spaced 100-m apart with a shot interval of 50 m. After processing the data, we used a top-down approach to build an initial tilted transversely isotropic sediment model, found to be accurate in most regions with the exception of areas with significant volcanic layers residing near salt boundaries. Included in the approach was high-resolution, nonlinear slope tomography. Using this flow, we accurately imaged the postsalt in most areas of the survey. However, upon close examination, we saw that some improvement was still needed in areas featuring volcanic intrusions. Postsalt velocity scans beneath the volcanic intrusions were ineffective at producing updates because of limited gather reflectivity between the top of volcanic and the TOS (Figure 9).

To launch our iterative workflow, we interpreted the regional salt body, expecting that some areas may change after the postsalt velocities were updated. We ran a salt body migration to facilitate an initial presalt evaluation. In this classification stage, we separated the well-imaged areas from poorly imaged areas and noticed a trend. Many of the poorly focused presalt sections were related to the volcanic intrusions (Figure 10).

Using the image beneath the autochthonous salt bodies as our reference, we ran velocity scans to estimate the velocity error between the volcanic intrusion and the TOS. Upon determining a reasonable velocity estimate, we ran a subsequent salt-body migration to evaluate the result. After only one iteration of velocity scans and updating, we noticed significant improvements in the gathers and the stacks (Figure 11). By incorporating the salt geometry into the velocity scans, we successfully evaluated the impact of any changes in the sediment velocity. Furthermore, because the accuracy of the velocity model improved, more seismic events became visible.

Although many improvements were made with the velocity scan update, we continued working on improving the model. In several areas, we elected to run a combination of structurally guided tomography and nonlinear slope tomography in the presalt region to refine the local velocity (Figure 12). The structurally guided tomography incorporated the regional geologic interpretation of the BOS and the presalt structure to anticipate velocity errors in the postsalt. This method further refined the model created from the velocity scan updates. After each velocity update iteration, the salt geometry was modified to reflect the new sediment overburden. Throughout the entire iterative process, we reevaluated the postsalt velocities and the interpretations of the TOS and BOS. Doing this enabled us to converge on a more geologically consistent and accurate velocity model than what would have been obtained using a top-down approach. Figure 13 compares an area before and after our entire iterative workflow.

Discussion and limitations

This workflow is interpretation intensive. Horizon modification and migration are critical components of our new workflow. With numerous iterations, some portions of a survey can potentially be interpreted multiple times over, and each iteration requires a human element to evaluate it. Incorporating more advanced velocity model estimation techniques, such as full-waveform inversion, could add additional details into the workflow with testing currently underway. Although we focused this case study on updating the velocity model in areas of thin-bed volcanics near the TOS, this workflow can be applied to other imaging difficulties seen in other regions. In many areas throughout the world, accurately estimating velocities above salt, such as carapaces or shale bodies, can be quite challenging. These structures have many of the same constraints seen in this case study, such as an uncertain suprasalt velocity and a lack of gather reflectivity above salt, and an increasing focus on improving the image below. By implementing an iterative approach to suprasalt updating, we can improve the focusing subsalt and add confidence to the overburden velocity.


Thin-bed volcanics near the TOS and other shallow velocity anomalies are challenging for traditional top-down velocity updating workflows. Because traditional top-down workflows rely heavily on CIG gather focusing either within or below the areas of the update, they lack critical information needed to make reliable updates in these complex areas. By incorporating the salt-body velocity model and presalt reflections into the postsalt velocity updates, it is possible to more accurately estimate the velocity beneath the anomalies. Our iterative workflow is effective for enabling these velocity updating methods that were previously deemed ineffective. By relying on presalt information to simultaneously update the velocities of the suprasalt, the velocity anomaly, and the interpretation of the salt body, our approach makes headway in understanding more pieces of the Brazil presalt geology.

Biographies and photographs of the authors are not available.

Freely available online through the SEG open-access option.