Imaging the geology subsalt and at the transition between extra-salt and subsalt has been a challenge at Mad Dog even with extensive seismic data coverage, including two WATS surveys and multiple NATS surveys. WATS acquisition and TTI velocity model processing generated major improvements in the image at Mad Dog. One of the observations of a previous TTI project is the presence of a strong orthorhombic anisotropic effect in a salt mini basin above the field. This finding led to the decision to reprocess the Mad Dog data with a tilted orthorhombic (TOR) velocity model. The main objective of this project is to build an orthorhombic velocity model with nine parameters compared to five with the TTI processing. The TOR anisotropic parameters are generated with the latest FWI and tomography techniques and take guidance from the stress field from a geomechanical model. The outcome of the project is very encouraging with results including better constructive imaging in crucial areas of the field, an incremental increase in signal-to-ratio everywhere and increased fault resolution. The TOR velocity model will be used to migrate a future ocean bottom nodes survey to address some of the remaining imaging challenges.
Presentation Date: Wednesday, October 17, 2018
Start Time: 8:30:00 AM
Location: 208A (Anaheim Convention Center)
Presentation Type: Oral
The motivation for generating upscaled velocity models includes practical goals such as reducing the computational cost of modeling or data processing. Existing 1-D effective medium solutions such as Backus averaging pose challenges as they can smooth sharp contacts like unconformities without careful application and therefore cannot reproduce the seismic response of the fine scale medium. In this paper, we apply Bayesian inversion, with reversible jump Markov Chain Monte Carlo (rjMCMC) sampling, to implement upscaling velocity logs as an inverse problem. The inversion minimizes the misfit between reference seismograms computed from the original well log and the seismic response of the coarse, upscaled velocity model. The results include an ensemble of upscaled models from which information such as the optimal coarse model and the upscaling uncertainty can be estimated. With this method, we are able to compare upscaling uncertainty when using single-offset and multi-offset seismograms as reference signals. The uncertainty of upscaled models generated using multi-offset seismograms as reference signals is lower than the value obtained using only a zero-offset seismogram as a reference.
Well log upscaling is a traditional approach to compare elastic properties measured at higher frequencies in well logs to those obtained from lower frequencies measurements in surface seismic or vertical seismic profile data. Upscaled velocity log values can provide a good estimate for an initial velocity model during full waveform inversion (FWI). Simple statistical methods (arithmetic, harmonic or geometric averages) or analytical averaging techniques like Backus averaging are used for well log upscaling. Backus averaging can be performed as layer based upscaling (Folstad and Schoenberg, 1992; Pr¨ussmann, 1996; Gibson and Hwang, 2009) and smooth window based upscaling (Rio et al., 1996; Liner and Fei, 2006; Lindsay and Van Koughnet, 2001; Sayers, 1998; Tiwary et al., 2007).
Smooth window based upscaling can preserve gradational contacts (Lindsay and Van Koughnet, 2001) but sharp contacts like unconformities are smoothed over. The ambiguities in layerbased Backus averaging for a well log include the number of layers to be considered and whether the thickness of the layers should be uniform or non uniform. Grechka (2003) showed the average elastic properties, obtained from numerical modeling, deviate from Backus average values in medium with fractures. Gibson and Hwang (2009) applied a stochastic approach of layer-based Backus averaging on a test well with a fixed number of layers but allowed perturbation of layer boundaries in the Markov Chain Monte Carlo (MCMC) implementation. An advantage of this stochastic approach is that it produces probability distributions for depths of layer boundaries, giving quantitative insights into uncertainty in the upscaled models.
This study shows the results from finite difference modeling for optimizing the acquisition design of the Atlantis 2014-2015 Ocean Bottom Nodes (OBN) survey. During the planning of this Atlantis time lapse acquisition, it was realized that additional nodes would be available, which provided an opportunity to use those nodes to improve the 3D static seismic image. A modeling study was initiated with the objective to identify optimum placement of the additional nodes. The model was generated using realistic detailed stratigraphy modeling and several cases were studied by adding node patches in different directions. Observations show that the addition of the nodes to the South, has the largest impact on the imaging of Atlantis Field. To measure the imaging impact of each added node location a more detailed tool was generated; the node areal contribution map, which shows how much each individual node contributes to the image of the reservoir. Improved imaging was also shown by modeling of reduced node spacing and increased node density. Altogether, the insights from this modeling work enabled optimized acquisition design for the survey that was acquired on the Atlantis Field in 2014-2015.
The Atlantis Field sits about 300 km south of the Louisiana coast in the Gulf of Mexico in around 7000 feet of water (Figure 1). It began production in October 2007 from Middle Miocene turbidite reservoirs that lie about 17,000 feet below sea level. As the field is sitting under the Sigsbee escarpment and the edge of a very complicated salt body with multiple salt fingers, seismic imaging has been challenging (Roberts et al., 2011). Only the southern end that sits outside of the salt can be imaged with confidence.
The primary objective of the 2005-2006 OBN survey was to obtain a consistent, high-quality image of the subsalt portion of the reservoir. This survey was the world’s first large-scale deepwater survey to employ autonomous nodes (Beaudoin and Ross, 2007). OBN technology also allows for operationally highly repeatable time-lapse seismic and overcomes the challenges presented by surface and subsea installations. Therefore, in 2009 a monitor survey was acquired (Reasnor et al., 2010). This time lapse data had excellent geometric repeatability and low 4D noise levels. The survey showed the depletion signature of the field in the time shift and the amplitude response (van Gestel et al., 2013).