PGS has reassessed the offshore acreage of Uruguay which could be hiding reservoirs analogous to the Namibia discoveries.

A recent reservoir characterization study provides insight into the prospectivity of Uruguay’s offshore exploration acreage. The methods used for the analysis include two separate technologies applied independently. Firstly, angle stacks were used in a workflow based on AVO (Amplitude Versus Offset) theory to obtain the relative acoustic properties of the subsurface. The second method applied a machine learning (ML) interpretation algorithm for automatic tracking of seismic reflectors by the assessment of the wavelet properties along each reflector and quantifying wavelet similarity. 

The derivation of attributes from the seismic data enables a more robust identification of possible hydrocarbon indicators than the interpretation of the seismic data in isolation. The images above shows multiple arrows which highlight attribute anomalies in the mid-section within the same geological unit. The anomalies are laterally separated from each other. Two conclusions can be made from this observation: firstly, the anomalies are stratigraphically bound and secondly, the anomaly viewed in section could, in reality, be a single feature coming in and out of the section such as a channel feature. Channel features are evident elsewhere in the section such as the annotated fan system shown on the horizon extraction. It is important to remember that the analogous Namibian discoveries were made in a similar deepwater sand channel setting.

The shallower part of the section shows a prominent water bottom simulating reflector (BSR). A BSR event trends parallel to the water bottom irrespective of lithological units and in this case is labeled with the blue arrow. BSR usually signals the presence of gas hydrates which could be of further exploration potential as the BSR is indicative of a working petroleum system. A BSR layer can provide an effective seal for gas accumulations and some evidence for this is observed (marked with the orange arrow). A strong amplitude relative impedance anomaly, low Vp/Vs and a structural pull-down caused by slow P-wave velocity are indicative of the presence of gas. 

The machine learning outcome provides access to hundreds of tracked surfaces across the entire 15 600 sq. km area. These surfaces were combined along similar geological levels but the differing attributes of interpreted segments were retained. The display above was extracted from a Lower Cretaceous interval and is shown with some features of interest indicated. At this interval, there is evidence of multiple turbidite fan systems along with associated feeder channels present. The identified features are converging towards the outboard side of the data coverage suggesting that their predominant clustering happens within the extent of this seismic survey.

It is important to understand that the colored attributes shown on the extracted surfaces are quite different from the traditional windowed attributes like average, minimum, maximum and root mean squared amplitudes (RMS). The attributes derived from machine learning show wavelet similarity which helps identify subtle changes in the waveform caused by lithological variations. These subtle features are not easily identified on conventional attributes such as RMS amplitude which is inherently averaged over a window.

AVO-based seismic analysis, as well as the application of the ML for detailed tracking of seismic reflectors, has proven to be very efficient for screening seismic data within the study area. Features not previously observed have now been identified such as prospective turbidite channel complexes and fan systems. The stratigraphic levels of these features are likely to be Cretaceous in age indicating an analogous hydrocarbon play to those of the Venus and Graff discoveries as well as the Gazania lead in the Orange Basin.