We have developed a method based on an encoder-decoder convolutional neural network for automatic channel detection in 3D seismic volumes. We use two architectures borrowed from computer vision: SegNet for image segmentation together with Bayesian SegNet for uncertainty measurement. We train the network on 3D synthetic volumes and then apply it to field data. We test the proposed approach on a 3D field data set from the Browse Basin, offshore Australia, and a 3D Parihaka seismic data in New Zealand. Applying the weights estimated from training on 3D synthetic volumes to a 3D field data set accurately identifies channel geobodies without the need for any human interpretation on seismic attributes. Our proposed method also produces uncertainty volumes to quantify the trustworthiness of the detection model.
The Miocene and Pliocene interval of the Veracruz Basin, southeast Mexico, experienced an evolving array of shortening, strike-slip, and volcanic forces in response to plate-scale interactions. The basin is divided into six structural domains that define regions of comparable timing and type of structural deformation, and the basin fill is separated into two long-term depositional phases, each of which can be tied to a waning and then waxing of major basin-bounding tectonic events. The first phase of deposition took place from the early to late Miocene and is tied to the waning effects of the Laramide orogeny. The Miocene basin inherited a tectonically steepened basin margin, across which deep canyons were carved and variably filled with mudstone and thin remnants of coarse sandstone and conglomerate. This zone of erosion and bypass grades into thick, sandstone-rich basin-floor fans. Later in phase I, subaqueous volcanoes, tied to distant plate subduction, developed offshore and formed a bathymetric barrier that prevented turbidity currents from entering the ancestral Gulf of Mexico. The volcanoes also served as immovable buttresses, around which intrabasinal thrust belts developed in response to regional shortening. The second depositional phase is tied to the onset of internal basin shortening and uplift of the north basin margin known as the trans-Mexican volcanic belt. This uplift caused a dramatic reconfiguration of the sediment-dispersal system, whereby large shelf clinoforms prograded from north to south across the basin. In contrast to the onlapping stacking pattern of phase I units, phase II units stack in a strongly offlapping pattern. Proven and postulated reservoir-trap combinations, ranging from four-way to three-way combination (stratigraphic), to pure stratigraphic traps are common. Four-way closures mapped from the two-dimensional and three-dimensional seismic data are large (P50: 5000 km2) and are covered with thick, lower Miocene fan sandstones. Traps that depend on a stratigraphic component are thinner and smaller in size (P50: 1000 km2), but more numerous than the four-way closures. Because many structures have experienced prolonged pulses of compression, top seal is considered an important geologic risk to the retention of substantial gas-column heights.
The lower continental slope of Morocco's west coast consists of Triassic-age salt manifested in the form of diapirs, tongues, sheets, canopies, and toe thrusts. Active salt diapirism and regional tectonics greatly influence the morphology of the modern sea floor, forming a severely rugose expression with ongoing minibasin development and episodic submarine failure. Detailed mapping of a 1064-km2 (411-mi2) seismic survey acquired in the Safi Haute Mer area revealed that Jurassic to Holocene salt mobilization continually affected distribution of sediment, causing a range of depositional flow styles, from slumps to sheet slides and mass-transport complexes (MTCs). Large sediment waves (20 km [12 mi] long, 1.5-km [0.9-mi] wavelength) were also documented at the end of the Aptian. An east-west–trending structural anticline downdip of the salt activated during initiation of the Atlas uplift in the latest Cretaceous to earliest Tertiary and shaped much of the lower slope into the Tertiary with a persistent canyon system and slope channels. The largest of the debris flows is a Cretaceous-age MTC, a 500-m (1640-ft)-thick flow that spans an area of up to 20,000 km2 (7722 mi2). Composing the MTC are (1) chaotic, mounded seismic facies; (2) internal syndepositional thrusts; and (3) transported megablocks (3.3 km2 [1.3 mi2]) with preserved internal stratigraphy. The MTC originated from an upslope collapse of a narrow shelf during the earliest phases of the Alpine orogeny.
A 15,000-km2 (5792-mi2) three-dimensional seismic data survey that covers the shelf-slope transition of the eastern offshore Trinidad continental margin reveals the geometry and depositional history of the last maximum glacial lowstand shelf-margin succession. Despite the lack of well information at these shallow depths, the quality and continuity of the seismic data allow us to pursue a detailed seismic stratigraphic interpretation of the last lowstand margin succession. The basin-fill stratal architecture of the studied stratigraphic interval shows a great deal of lateral and vertical variability along the continental margin during the Pleistocene to Holocene. Three geomorphological elements controlled the character of the accommodation within the basin and were crucial in transporting, delivering, and storing sediment supply from shelf to slope areas: (1) the Columbus sedimentary pathway on the shelf, (2) bypass zones in the shelf-break region, and (3) deep-water depocenters. The location and geometry of these geomorphological elements within the basin are clearly controlled by underlying structures, transpressional to the north and gravity driven to the south. Migration of the paleo-Orinoco delta system across the shelf was also a key factor in defining the stratigraphic geometries that are observed within the shelf break. Development of shelf-edge versus outer-shelf deltaic systems on the continental margin was controlled by the nature of sediment supply at specific times, as well as by the availability of accommodation, although, to a lesser extent, to relative sea level fluctuations. The interpretation also showed that, for time-equivalent units, parts of the shelf-edge region could develop as an erosional margin (sediment bypass zones), whereas other parts of the shelf edge could behave as an accretionary margin (sediment accumulation). The sequence-stratigraphic interpretation that was attempted in this work also demonstrated that the characteristics of systems tracts can abruptly change along strike in the shelf-edge region for time-equivalent units. These changes can be misleading if a genetic interpretation is pursued only on the basis of the definition of system tracts in the shelf-edge region without the consideration of a complete regional framework.
This integrated study provides a geological and geochemical framework for upper Miocene and Pliocene siliciclastic gas plays in the Macuspana Basin. Structural controls for the plays are deep-seated faults that tap Mesozoic thermogenic gas sources, areas of intense shale diapirism and folding, and areas with structural inversion that could enhance trapping and reservoir productivity. Early Neogene thrusting south of the basin triggered evacuation of Oligocene shale along northwest-dipping listric faults in the eastern and southeastern basin margin. These faults are associated with large-scale rollover structures and thick (>500 m) upper Miocene shoreface and wave-dominated deltaic complexes. A second phase of extension in the early Pliocene formed a set of broad, southeast-dipping listric faults in the western basin, controlling thick accumulations of stacked Pliocene shoreface deposits. Trap formation and enhancement in the southern basin margin are linked to late Miocene to Pliocene inversion. The primary stratigraphic controls on play occurrence in the upper Miocene in the onshore part of the basin are the regional facies distribution of northwest-prograding shoreface and wave-dominated deltaic systems. There was a shift in Pliocene sedimentation from the southeast to the west and northwest parts of the basin, where thick successions of aggradational shoreface and wave-dominated deltaic deposits accumulated in depocenters defined by shale evacuation along growth faults. Valley-fill deposits in both the upper Miocene and Pliocene resulted from shortlived periods of base-level change induced by either uplift on the southern basin margin or eustasy. The offshore part of the basin is inferred to consist of deep-water turbidite deposits that formed downdip (westward) of a hypothesized mixed clastic-carbonate prograding complex from the Yucatan platform. Three petroleum systems (Mesozoic, Paleogene–lower Neogene, and upper Miocene–Pliocene) contributed to the hydrocarbon accumulations and hydrocarbon generation and migration in the basin. Principal Upper Jurassic/Lower Cretaceous source rocks generated wet thermogenic gases and oil. Secondary lower Tertiary source rocks generated dominantly dry biogenic gases. Mixtures of the two gas types are common. Numerous deep-seated growth faults and faults serve as pathways for Mesozoic-sourced hydrocarbons. Surface seeps and abundant gas shows suggest that hydrocarbons are being generated today.
Detection and interpretation of fault systems and stratigraphic features and the relationship between them are crucial for seismic interpretation and reservoir characterization. To provide better interpretation insight and to be able to extract overlooked features out of seismic data volumes, we have developed a new attribute that detects faults and other discontinuities while handling local nonstationary variations across them. First, we used predictive painting to form a structural prediction of seismic events from neighboring traces (left and right neighboring traces in 2D and neighboring traces in all directions around a reference trace in 3D) according to the local structural slopes. Then, we computed prediction residuals by subtracting each prediction from the original data, and we found the smallest prediction-error interval for each point that best represented discontinuity information at that point. The extracted fault information changed with location (spatially and temporally), and it was nonstationary. Conventional coherence measures operate on a spatial window of neighboring traces and a temporal (vertical) analysis window of samples above and below the analysis point, and they can hardly cope with nonstationarity in fault information. In contrast, in our method, neither temporal nor spatial windows were involved in coherence computation, which allowed us to honor nonstationary changes of fault information and to achieve high resolution in the vertical and lateral directions. To assess the performance of the proposed attribute, we compared it with the conventional coherence attribute over the same data set. The comparison demonstrated the effectiveness of discontinuity detection using predictive coherence and showed its value in extracting additional information from seismic data.