Chakraborty, Samarjit

Ocean bottom seismic surveys have been a key part of the seismic imaging strategy at several large deepwater fields in the Gulf of Mexico (Howie et al., 2008; Beaudoin et al., 2008; Li et al, 2013). Benefits of these surveys include a wider and more even distribution of azimuths and offsets, a generally lower background receiver noise level, and better receiver coverage in congested offshore development areas. In this paper we present how finite difference modeling improved our understanding of existing seismic data and predicted significant improvement in subsalt image quality with long offset and full azimuth Ocean Bottom Node (OBN) geometry. Once the imaging criteria for acquisition design were met we optimized the survey for full waveform inversion (FWI) with ultra-long offsets. The modeling results were essential for designing an OBN survey that was acquired on the Thunder Horse field in 2015. We observed improved imaging of salt flank with OBN data that helped during velocity model building process.

Presentation Date: Thursday, September 28, 2017

Start Time: 10:10 AM

Location: 371F

Presentation Type: ORAL

Inverse Q compensation is most accurately performed when included in the imaging step. However, this is a computationally expensive process and not always justified, therefore methods which can perform the Q compensation as part of the pre-migration data conditioning are used regularly. In its simplest form, a locally invariant 1D Q-model is applied at either zero offset or offset varying water bottom times. In a deep water environment, the application of Q compensation needs to take into account that no dispersion occurs over the large thickness of the water layer. Therefore, methods which correctly handle the water layer under the straight ray assumption (like in Xia 2005) provide more optimal results. To provide a more accurate, yet still a simple compensation not limited by the straight ray assumption, we propose a new method using a combination of two straight ray approximation methods.

In this study, we present a 3D application of inverse Q filtering prior to imaging based on the inverse Q application method developed by Xia in 2005 and a new straight ray Q-effective method with its corresponding scheme for spatial averaging. We demonstrate that this new approach is more accurate for far offsets, as well as in the presence of a 3D spatially variant depth Q model. The comparison between this method and inverse Q application within Kirchhoff Pre-Stack Depth Migration (Q-KDM) validates the advantage of using the proposed new method for accurate imaging of both supra-salt and subsalt events.

Presentation Date: Wednesday, September 27, 2017

Start Time: 2:40 PM

Location: 371A

Presentation Type: ORAL

Ocean bottom seismic surveys have been a key part of the seismic imaging strategy at several large deepwater fields in the Gulf of Mexico (Howie et al., 2008; Beaudoin et al., 2008; Li et al, 2013). Benefits of these surveys include a wider and more even distribution of azimuths and offsets, a generally lower background receiver noise level, and better receiver coverage in congested offshore development areas. In contrast, the best seismic dataset at BP’s Thunder Horse field complex before 2015 was constructed by merging three narrow azimuth towed streamer (NATS) surveys with two orthogonal speculative wide azimuth towed streamer (WATS) surveys. Maximum offset recorded for this data was 9 km. Large strides have been made since 2002 at Thunder Horse in the quality of the image produced from towed streamer data (Howie and Trout, 2010; Hartman et al., 2015), but progressively more difficult infill wells necessitated a step change in seismic quality to improve reservoir characterization and fault definition. Accordingly, in 2015, an ocean-bottom nodes (OBN) survey was acquired for the first time at Thunder Horse. The fast-track dataset from this survey provided a significant uplift over the merged streamer dataset, and impacted several business decisions within two months of delivery. This paper will show several examples of the impact of the new image, and discuss which aspects of the acquisition method were responsible.

Presentation Date: Thursday, September 28, 2017

Start Time: 9:20 AM

Location: 350D

Presentation Type: ORAL

We report case study results for attenuation of free-surface multiples from deep-water ocean bottom node (OBN) data using a data-driven multiple prediction method that combines OBN and towed-streamer data through multidimensional convolution, similar to the well-known surface-related multiple elimination (SRME) method.

We illustrate the properties of the proposed multiple prediction method using synthetic and field data and note that availability of suitably acquired and processed streamer data is critical to the success of this approach.

In our case study, we have data from five streamer surveys with offsets up to 10 km and broad range of azimuths. Correspondingly, the results of data-driven multiple attenuation are good for the OBN data with offsets up to about the maximum offset of the streamer data. We also compute a model-based prediction of the free-surface multiples using the anisotropic velocity model and prior depth images available for this field.

The data-driven and the model-based approaches of predicting free-surface multiples have complementary properties. We combine models computed with both approaches to attenuate free-surface multiples in the OBN upgoing and downgoing data, as needed for subsalt imaging.

Presentation Date: Monday, October 17, 2016

Start Time: 4:10:00 PM

Location: 142

Presentation Type: ORAL

Deepwater U.S Gulf of Mexico remains one the world’s most prolific producing regions. However the complex geological setting created by the salt tectonics presents multiple technical challenges for drilling. The characterization of highly variable pore pressure, overburden stress and salt-induced stress perturbations requires advanced methods of geomechanical modelling. To address these challenges, we have employed a calibrated 4D finite element (FEM) geomechanical model. The model shows stress variations, in both magnitude and direction, induced by the varying overburden, presence of faults and, most importantly, near salt. Stresses were highly dependent on salt body geometry, with higher compressive values found closer to the salt body. Away from and below the salt, the stress regime returns to a more normal ‘relaxed’ state. The field wide stress calculation is substantiated by two independent field observations. Firstly, simulated stress values were corroborated with observed losses during casing and cementing. Secondly, observed time-lapse seismic shifts associated with reservoir compaction and overburden elongation agreed with compaction features and magnitude simulated by the model. This 3D/4D stress characterization overcomes many of the limitations of the traditional 1D geomechanical modeling assumptions and provides a richer input for drilling design and decisions.

Presentation Date: Wednesday, October 19, 2016

Start Time: 10:20:00 AM

Location: 146

Presentation Type: ORAL

3D VSP data provides a unique opportunity to improve image resolution and fault definition in the vicinity of a well. However, the processing and imaging of VSP data requires special accommodations for its distinctive acquisition geometry. In this abstract, we demonstrate two key VSP pre-processing steps that greatly impacted the final image from the Mad Dog 3D VSP data, including XYZ vector field reorientation based on 3D elastic finite-difference modelling, and shot-to-shot directional de-signature using near field hydrophone data. We also discuss how utilizing the multiple energy - in addition to primary - extends our capability to image the shallow overburden.

Presentation Date: Thursday, October 20, 2016

Start Time: 8:30:00 AM

Location: 146

Presentation Type: ORAL

In 2014, BP contracted FairfieldNodal to acquire and process two ocean bottom node (OBN) surveys in deep water in the Gulf of Mexico at the Atlantis and Thunder Horse fields, on and near the Sigsbee Escarpment. At 1912 and 2031 nodes, respectively, these were the largest deep water node surveys ever acquired. Concurrently, FairfieldNodal was contracted to reprocess the 2005-2006 and the 2009 vintages of the Atlantis OBN acquisition in an effort to match the processing of the 2014 OBN data. Because the Thunder Horse survey is expected to be the first of a number of OBN surveys on the field over time, the same processing strategy was employed to enhance its applicability as a 4D baseline survey as well. This document and presentation will describe some of the methods and procedures used to mitigate known and measurable obstacles to the repeatability of the different vintages of OBN data in an effort to simplify 4D analyses. As the 2005/6 Atlantis survey was the first deep water node survey of its kind, there has been a natural evolution in the ancillary measurements and data acquired in support of the seismic data as well as significant advances in the technology of the nodes. An example of the ancillary data would be the types and quantity of measurements related to characterizing the acoustic qualities of the water column.

Presentation Date: Tuesday, October 18, 2016

Start Time: 9:15:00 AM

Location: 150

Presentation Type: ORAL

Summary

Accurately predicting the effective image area of a 3D VSP can be a difficult task to accomplish. The creation of a field wide Finite difference earth model at the Mad Dog Field provided the opportunity to perform forward modeling of an already existing 3D VSP in order to calibrate expectations of image area on possible future VSP acquisition. The acquired 3D VSP, which had recently been reprocessed, was modeled by conducting acoustic wave equation forward modelling and migrated with reverse time pre-stack depth migrations (RTM) and wave equation migrations (WEM) of the synthetic data. These data were then stacked and compared to the images from the acquired VSP.

At first look, the finite difference model VSP image area was much larger than the image area of the acquired VSP. One problem was that the RTM produced a much more extensive image both above and below the receivers, a function of the algorithm being able to handle both upgoing and down-going energy. The WEM, however, could handle only up-going energy, and produced an image closer in appearance to the VSP processed image. Although the acquired VSP was processed with a reverse time migration, the down-going energy contribution was much less robust than that of the modeled data; this could be due to the velocity model errors in the real world case. Differences between the acquired VSP and the modeled WEM may also be due to velocity model errors in the actual VSP processing. With both migration types, it became apparent that the modeled image areas would likely be interpreted as being significantly more extensive than the actual realized image. However, through detailed inspection, it was possible to determine which areas of the modeled image needed to be discounted. This ultimately resulted in a predicted image that was very similar to the actual VSP image. These learnings were subsequently applied to the modeling of a planned VSP acquisition in a different well, and impacted both the source pattern design and the expected image.

Introduction

The Mad Dog Field is a giant subsalt field in the Gulf of Mexico, discovered by BP in 1998. The field started producing in 2005. The field lies beneath the edge of the Sigsbee Escarpment in 4000 to 7000 feet of water approximately 190 miles south of New Orleans

Summary

Subsalt imaging at the Thunder Horse Field in the Gulf of Mexico is challenging primarily because the salt canopy, overlying roughly 75% of the structure, greatly distorts subsalt illumination and causes imaging and resolution problems. Since the Thunder Horse discovery, advancements in seismic acquisition techniques and imaging technologies have significantly improved subsalt images. The latest successful application is from a tilted transverse isotropy (TTI) reverse time migration (RTM) project combining two wide azimuth towed streamer (WATS) data sets and three narrow azimuth towed streamer (NATS) data sets. The addition of an extra WATS data set and the application of the recent imaging technologies are key contributors to the dramatic structural image improvements with better defined three-way events and a higher signal-to-noise ratio (S/N).

Introduction

The Thunder Horse Field has been producing since 2008 and is located in the south-central part of the Mississippi Canyon protraction area in the Gulf of Mexico. A large overlying allochthonous salt body causes rapid spatial and temporal changes in illumination and image quality, making interpretation difficult, especially near the steeply dipping three-way closure against the salt stock. During the course of discovery and development, BP has made continuous efforts to better understand and improve Thunder Horse’s subsalt image with new seismic data sets and more advanced imaging technologies (Pfau et al., 2002; Ray et al., 2002, 2005; Gherasim et al., 2012). The latest successful TTI RTM project with two WATS data sets and three NATS data sets is the continuation of this effort to improve Thunder Horse subsalt images.

This project aimed to improve the structural image in poorly illuminated areas and to maximize the usable vertical and horizontal resolution for well targeting and planning. The latest image shows a dramatic improvement over the previous TTI RTM image produced in the 2012 project for three reasons. First, the additional WATS data in the NE-SW direction illuminated some key areas that the NW-SE WATS and three NATS surveys did not. Second, the majority of the NATS traces were migrated rather than just used to infill missing traces in the NW-SE WATS shot gathers, as was done in 2012. Finally, more advanced imaging workflows and technologies were used to address specific problem areas in the data. Shot patch-based angle gather illumination weighting (AGILW) and input data selection technologies, which were applied in this project, effectively attenuate noise while preserving signal. Specular imaging using RTM dip gathers also helped enhance the S/N. We also discovered one of the reasons for frequency loss underneath the salt.

Summary

In this case study we present the results from a time-lapse (4D) feasibility study that includes analysis of core data and reservoir simulator to seismic modeling to predict time-lapse seismic response from dynamic reservoir model from the Mad Dog field.

Ultrasonic velocities (UV’s) for both P and S waves were obtained from nine axially oriented sandstone core plug samples. These experiments were conducted under uniaxial strain conditions, in which radial strain was kept constant. The samples were cleaned, dried and then subjected to multiple loading and unloading cycles designed to simulate reservoir stress conditions of pressure depletion and build-up due to production and water injection. Travel times were obtained by careful picking of first arrival events on P and S full waveforms. These picks were used to compute P and S velocities, which were then used together with density data to compute dynamic bulk and shear moduli. These core data were then used to characterize the sensitivity of the dry rock frame to changes in effective stress. Sensitivity of porosity to changes in effective stress was also characterized.

Dry frame moduli were obtained from sonic log measurements using Gassmann fluid substitution for comparison with core data. Computed dry frame moduli from logs show close agreement with the core measurements. This comparison provides more confidence in the laboratory core measurements with less likelihood of core damage. Moreover, we could estimate the effective stress coefficient by matching the dry frame bulk moduli from cores to logs.

Next the core dry frame parameters were used to model rock properties changes in the reservoir for a single interface. Finally, full-field time-lapse feasibility study with the 3D reservoir simulation model was done to model the expected time-lapse response of the Mad Dog field. Results from the feasibility study show that the model-based time-lapse signal could be observed in extra-salt reservoirs using a time-lapse noise threshold derived from time-lapse ocean bottom nodes data from analog fields.

The core analysis and time-lapse feasibility studies demonstrate the need to integrate field-based core measurements to predict time-lapse signal in deepwater reservoirs under complex overburden.