Deep neural network application for 4D seismic inversion to changes in pressure and saturation: Optimizing the use of synthetic training datasets

In this work, we tackle the challenge of quantitative estimation of reservoir dynamic property variations during a period of production, directly from four-dimensional seismic data in the amplitude domain. We employ a deep neural network to invert four-dimensional seismic amplitude maps to the simultaneous changes in pressure, water and gas saturations. The method is applied to a real field data case, where, as is common in such applications, the data measured at the wells are insufficient for properly training deep neural networks, thus, the network is trained on synthetic data. Training on synthetic data offers much freedom in designing a training dataset, therefore, it is important to understand the impact of the data distribution on the inversion results. To define the best way to construct a synthetic training dataset, we perform a study on four different approaches to populating the training set making remarks on data sizes, network generality and the impact of physics-based constraints. Using the results of a reservoir simulation model to populate our training datasets, we demonstrate the benefits of restricting training samples to fluid flow consistent combinations in the dynamic reservoir property domain. With this the network learns the physical correlations present in the training set, incorporating this information into the inference process, which allows it to make inferences on properties to which the seismic data are most uncertain. Additionally, we demonstrate the importance of applying regularization techniques such as adding noise to the synthetic data for training and show a possibility of estimating uncertainties in the inversion results by training multiple networks.     

Fluvial dynamics and 14C-10Be disequilibrium on the Bolivian Altiplano

Determining sediment transfer times is key to understanding source-to-sink dynamics and the transmission of environmental signals through the fluvial system. Previous work on the Bolivian Altiplano applied the in situ cosmogenic ¹⁴C-¹⁰Be-chronometer to river sands and proposed sediment storage times of ~10–20 kyr in four catchments southeast of Lake Titicaca. However, the fidelity of those results hinges upon isotopic steady-state within sediment supplied from the source area. With the aim of independently quantifying sediment storage times and testing the ¹⁴C-¹⁰Be steady-state assumption, we dated sediment storage units within one of the previously investigated catchments using radiocarbon dating, cosmogenic ¹⁰Be-²⁶Al isochron burial dating, and ¹⁰Be-²⁶Al depth-profile dating. Palaeosurfaces appear to preserve remnants of a former fluvial system, which has undergone drainage reversal, reduction in catchment area, and local isostatic uplift since ~2.8 Ma. From alluvium mantling the palaeosurfaces we gained a deposition age of ~580 ka, while lower down fluvial terraces yielded ≤34 ka, and floodplains ~3–1 ka. Owing to restricted channel connectivity with the terraces and palaeosurfaces, the main source of channel sediment is via reworking of the late Holocene floodplain. Yet modelling a set of feasible scenarios reveals that floodplain storage and burial depth are incompatible with the ¹⁴C-¹⁰Be disequilibrium measured in the channel. Instead we propose that the ¹⁴C-¹⁰Be offset results from: (i) non-uniform erosion whereby deep gullies supply hillslope-derived debris; and/or (ii) holocene landscape transience associated with climate or human impact. The reliability of the ¹⁴C-¹⁰Be chronometer vitally depends upon careful evaluation of sources of isotopic disequilibrium in a wide range of depositional and erosional landforms in the landscape. © 2018 John Wiley & Sons, Ltd.     

Intra-survey reservoir fluctuations – implications for quantitative 4D seismic analysis

During the time taken for seismic data to be acquired, reservoir pressure may fluctuate as a consequence of field production and operational procedures and fluid fronts may move significantly. These variations prevent accurate quantitative measurement of the reservoir change using 4D seismic data. Modelling studies on the Norne field simulation model using acquisition data from ocean-bottom seismometer and towed streamer systems indicate that the pre-stack intra-survey reservoir fluctuations are important and cannot be neglected. Similarly, the time-lapse seismic image in the post-stack domain does not represent a difference between two states of the reservoir at a unique base and monitor time, but is a mixed version of reality that depends on the sequence and timing of seismic shooting. The outcome is a lack of accuracy in the measurement of reservoir changes using the resulting processed and stacked 4D seismic data. Even for perfect spatial repeatability between surveys, a spatially variant noise floor is still anticipated to remain. For our particular North Sea acquisition data, we find that towed streamer data are more affected than the ocean-bottom seismometer data. We think that this may be typical for towed streamers due to their restricted aperture compared to ocean-bottom seismometer acquisitions, even for a favourable time sequence of shooting and spatial repeatability. Importantly, the pressure signals on the near and far offset stacks commonly used in quantitative 4D seismic inversion are found to be inconsistent due to the acquisition timestamp. Saturation changes at the boundaries of fluid fronts appear to show a similar inconsistency across sub-stacks. We recommend that 4D data are shot in a consistent manner to optimize aerial time coverage, and that additionally, the timestamp of the acquisition should be used to optimize pre-stack quantitative reservoir analysis.     

Sensitivity of the elastic anisotropy and seismic reflection amplitude of the Eagle Ford Shale to the presence of kerogen

The Eagle Ford Shale of Central and South Texas is currently of great interest for oil and gas exploration and production. Laboratory studies show that the Eagle Ford Shale is anisotropic, with a correlation between anisotropy and total organic carbon. Organic materials are usually more compliant than other minerals present in organic‐rich shales, and their shapes and distribution are usually anisotropic. This makes organic materials an important source of anisotropy in organic‐rich shales. Neglecting shale anisotropy may lead to incorrect estimates of rock and fluid properties derived from inversion of amplitude versus offset seismic data. Organic materials have a significant effect on the PP and PS reflection amplitudes from the Austin Chalk/Upper Eagle Ford interface, the Upper Eagle Ford/Lower Eagle Ford interface, and the Lower Eagle Ford/Buda Limestone interface. The higher kerogen content of the Lower Eagle Ford compared with that of the Upper Eagle Ford leads to a negative PP reflection amplitude that dims with offset, whereas the PS reflection coefficient increases in magnitude with increasing offset. The PP and PS reflection coefficients at the Austin Chalk/Upper Eagle Ford interface, the Upper Eagle Ford/Lower Eagle Ford interface, and the Lower Eagle Ford/Buda Limestone interface all increase in magnitude with increasing volume fraction of kerogen.     

Quantifying potential propagule pressure of aquatic invasive species from the commercial shipping industry in Canada

We quantify and compare different measures of potential propagule pressure (PPP) of aquatic invasive species (AIS) from commercial vessels in Canada. We used ship arrivals and ballast water discharge volumes as proxies for PPP from ballast water organisms, and wetted surface area (WSA) as a proxy for hull fouling PPP, to determine their relative contributions to total PPP. For three regions studied, PPP proxies correlated significantly across ports and some vessel categories. Relative contributions of ship arrivals, ballast discharge, and WSAs to PPP, evidenced by non-significant correlations across these measures, varied across regions, ports, vessel types, and seasons. Flow-through (dominant on east and west coasts) and empty-refill (in Great Lakes-St. Lawrence region) were the major ballast water exchange methods employed by the vessels surveyed. These methods have different biological efficacy for AIS removal, influencing PPP. Our study illustrates benefits and limitations of using different PPP proxies to estimate invasion risk.     

Setting new thermal standards for transitional and coastal (TraC) waters

The construction of a new generation of coastal power stations in the UK and other western European countries has highlighted the absence of robust standards for thermal discharges to transitional and coastal waters. The proposed discharge volumes are greater than hitherto, yet there has been little independent critical examination of their potential impact, whilst much of the existing guidance has been adapted from freshwater practice. This review considers the available knowledge on the tolerance and behaviour of fish and other marine biota to heated effluents. Appropriate ways are suggested of grouping fish species to reflect their sensitivity to thermal effects. The plethora of existing standards are considered and their validity assessed in a framework of predicted seawater temperature rise. Those species or groups of organisms most likely to be affected are identified and finally specific recommendations for thermal standards consistent with long term sustainability are proposed.     

Modelling converted seismic waveforms in isotropic and anisotropic 1-D gradients: discontinuous versus continuous gradient representations

Over the past decade, there have been numerous receiver function studies directed at imaging the lithosphere-asthenosphere boundary (LAB). Although it is generally accepted that receiver function phases observed in these studies are derived from physical mode conversions at depth within the lithosphere-asthenosphere transition, it is still debatable as to whether these phases are directly indicative of the LAB. This is because interpretation of receiver function LAB signals relies on understanding the elastic characteristics of the Earth??s outer thermal boundary layer. The main issues for receiver function imaging are the sharpness of the elastic material property transition and, more importantly, what specifically are the material gradients. To test the various transition models, a forward modelling approach is required that allows accurate waveform synthetics for a range of discontinuous and continuous gradients in anisotropic, elastic media. We present a derivation of the reflection and transmission response for continuous one-dimensional (1-D) gradients in generally anisotropic elastic media. We evaluate the influence of 1-D isotropic and anisotropic elastic gradients on the seismic waveform by comparing numerical results of models for discontinuous and continuous transitions. The results indicate that discontinuous representations using layers each with uniform parameters and with thicknesses on the order of approximately 1/3 to 1/8 of the dominant seismic wavelength can be used to accurately model P-to-S and S-to-P mode conversions due to continuous transitions of both isotropic and anisotropic elastic properties. From a practical point of view, when comparing synthetic modelling with observation, this constraint can be relaxed further. The presence of signal noise and/or the result of receiver function stacking techniques will likely obscure these subtle waveform e ff ects. Hence this study suggests that accurate synthetic waveforms for LAB transitions can be modelled with discontinuous gradient representations using a reasonable number of discrete transition layers with layer thicknesses no greater than 1/2 to 1/3 the dominant seismic wavelength.     

Aminostratigraphy and thermoluminescence dating of coastal aeolianites and the later Quaternary history of a failed delta: The River Murray mouth region,South Australia

A geochronological framework for the sequential development of coastal barrier aeolianite complexes in the mouth region of the River Murray, Australia's largest river system is presented based on amino acid racemization and thermoluminescence dating. The sedimentary successions represent a foreshortened and condensed sequence of coastal barriers compared with those of the Coorong Coastal Plain in southern South Australia where the barrier complexes are more widely separated in response to tectonic uplift. The barriers have formed during interglacial sea-level highstands and are correlatives of genetically equivalent landforms of the Coorong Coastal Plain. Thermoluminescence dating and the extent of amino acid racemization in aeolianite ‘whole-rock’ sediment samples, reveal a general increase in age of the barriers landwards from the modern coastline. In detail, however, the individual barriers represent composite structures having formed in more than one interglaciation, due to the reoccupation of Pleistocene shoreline positions during sea-level highstands of similar amplitude, in a zone of gradual basin subsidence. The most seaward Pleistocene aeolianite at Surfer Beach is of interstadial age (Marine Isotope Stage 5c, 105 ± 5 ka; MIS 5c), and correlates with the Robe Range of the Coorong Coastal Plain. The last interglacial shoreline (130 ± 15 ka; MIS 5e) is particularly well-defined in the River Murray mouth region. It is represented by a complex association of coastal parabolic dunes superimposed on a transverse dune system, which runs parallel with the former coastline, and also includes associated estuarine, lagoonal and open ocean beach facies. Landward of the last interglacial succession are distinct barriers relating to the penultimate interglaciation (215 ± 35 ka; MIS 7), as well as earlier interglaciations (350 ± 65 ka; MIS 9 or 11 and 470 ± 70 ka; MIS 11 or 13). The coastal barriers have been successively breached by the ancestral River Murray at times of lower sea level during glacial cycles. Former mouths of the River Murray during interglacial sea-level highstands are likely to have existed near Tauwitchere Island during MIS 7, and between Goolwa and Hindmarsh Island and near the southern-most part of Lake Albert during the last interglacial (MIS 5e). The River Murray mouth region represents a failed delta as the limited sediment brought to this area since late middle Pleistocene time has been either rapidly incorporated within aeolian deposits during sea-level highstands, or transported to the edge of the Lacepede Shelf during glacial maxima. The Holocene and modern River Murray has not established a marine delta, but deposits its load in the settling basins of the terminal lakes. Only a small digitate delta has formed where the river enters Lake Alexandrina.     

Shaking‐table tests of flat‐bottom circular silos containing grain‐like material

According to Eurocode 8, the seismic design of flat‐bottom circular silos containing grain‐like material is based on a rough estimate of the inertial force imposed on the structure by the ensiled content during an earthquake: 80% of the mass of the content multiplied by the peak ground acceleration. A recent analytical consideration of the horizontal shear force mobilised within the ensiled material during an earthquake proposed by some of the authors has resulted in a radically reduced estimate of this load suggesting that, in practice, the effective mass of the content is significantly less than that specified. This paper describes a series of laboratory tests that featured shaking table and a silo model, which were conducted in order to obtain some experimental data to verify the proposed theoretical formulations and to compare with the established code provisions. Several tests have been performed with different heights of ensiled material – about 0.5 mm diameter Ballotini glass – and different magnitudes of grain–wall friction. The results indicate that in all cases, the effective mass is indeed lower than the Eurocode specification, suggesting that the specification is overly conservative, and that the wall–grain friction coefficient strongly affects the overturning moment at the silo base. At peak ground accelerations up to around 0.35 g, the proposed analytical formulation provides an improved estimate of the inertial force imposed on such structures by their contents. Copyright © 2015 John Wiley & Sons, Ltd.