Enhanced Mineral Quantification and Uncertainty Analysis from Downhole Spectroscopy Logs Using Variational Autoencoders

This paper describes an innovative machine-learning application, based on variational autoencoder frameworks, to quantify the concentrations and associated uncertainties of common minerals in sedimentary formations using the measurement of atomic element concentrations from geochemical spectroscopy logs as inputs. The algorithm comprises an input(s), encoder, decoder, output(s), and a novel cost function to optimize the model coefficients during training. The input to the algorithm is a set of dry-weight concentrations of atomic elements with their associated uncertainty. The first output is a set of dry-weight fractions of 14 minerals, and the second output is a set of reconstructed dry-weight concentrations of the original elements. Both sets of outputs include estimates of uncertainty on their predictions. The encoder and decoder are multilayer feed-forward artificial neural networks (ANN), with their coefficients (weights) optimized during calibration (training). The cost function simultaneously minimizes error (accuracy metric) and variance (precision or robustness metric) on the mineral and reconstructed elemental outputs. Training of the weights is done using a set of several-thousand core samples with independent, high-fidelity elemental and mineral (quartz, potassium-feldspar, plagioclase-feldspar, illite, smectite, kaolinite, chlorite, mica, calcite, dolomite, ankerite, siderite, pyrite, and anhydrite) data. The algorithm provides notable advantages over existing methods to estimate formation lithology or mineralogy relying on simple linear, empirical, or nearest-neighbor functions. The ANN numerically capture the multidimensional and nonlinear geochemical relationship (mapping) between elements and minerals that is insufficiently described by prior methods. Training is iterative via backpropagation and samples from Gaussian distributions on each of the elemental inputs, rather than single values, for every sample at each iteration (epoch). These Gaussian distributions are chosen to specifically represent the unique statistical uncertainty of the dry-weight elements in the logging measurements. Sampling from Gaussian distributions during training reduces the potential for overfitting, provides robustness for log interpretations, and further enables a calibrated estimate of uncertainty on the mineral and reconstructed elemental outputs, all of which are lacking in prior methods. The framework of the algorithm is purposefully generalizable so that it can be adapted across geochemical spectroscopy tools. The algorithm reasonably approximates a “global-average” model that requires neither different calibrations nor expert parameterization or intervention for interpreting common oilfield sedimentary formations, although the framework is again purposefully generalizable so it can be optimized for local environments where desirable. The paper showcases a field application of the method for estimating mineral type and abundance in oilfield formations from wellbore-logging measurements.

Characterizing a weathering profile over the Kuantan Basalt

Three broad morphological zones can be differentiated; the top pedological soil (Zone I) being 3.60 m thick and comprising brown, soft to stiff, clays. The intermediate saprock (Zone II) is 1.12 m thick and consists of brown, very stiff, sandy silt with many lateritic concretions, whilst the bottom bedrock (Zone III) is an outcrop of vesicular olivine basalt with weathering along joints. Constant volume samples show the saprolite (sub-zone IC) to have dry unit weights of 11.78 to 12.80 kN/m3, whilst the solum (sub-zones IA and IB), and saprock, have values ranging from 10.65 to 11.09, and from 11.35 to 11.50, kN/m3, respectively. Porosities are variable; the saprolite with the lowest values of 52 to 56% and the solum and saprock with values of 57 to 60%. Clay and silt contents increase up the profile with a corresponding decrease in sand and gravel contents. Colloid (<1 μm size) contents especially increase up the profile from 10 to 15% in saprock through 30 to 40% in saprolite and exceeding 57% in the solum, These increasing colloid contents point to the increasing effects of pedological processes. Thin-sections of weathered rims (1-2 cm thick) show alteration of basalt to start with formation of micro-cracks (Stage 1) that become stained by secondary iron oxides and hydroxides. Decomposition of the essential minerals then occurs in the order: olivine (Stage 2), augite (Stage 3), and plagioclase feldspar (Stage 4). An increase in apparent porosity, but a decrease in unit weights and specific gravity, reflect these stages of weathering; the boundary between ‘rock’ and ‘soil’ material occurring when all olivine and augite crystals have decomposed. It is concluded that the weathering profile results from in situ alteration of basalt due to lowering of an unconfined groundwater table; pedological processes giving rise to further alteration.

Study of Radioactivity in Bajaur Norite Exposed in the Himalayan Tectonic Zone of Northern Pakistan

Radioactivity in Granites of Pakistan systematically increases from south to north. The Ambella Granite found at the northern edge of Pakistan is highly radioactive. Radioactivity measurements made on, so called, Bajaur Granite, located in northern Pakistan, have been found to be lowest among all the granitic rock of the area. In order to find out the exact nature of Bajaur rocks, mineralogical studies were carried on rock chips and powdered samples. The Bajaur Norite contains plagioclase feldspar more than 45% as the chief constituent. Orthopyroxene and clinopyroxene are 27% and 18%. Quartz, biotite, and some opaque minerals are also found in accessory amounts. Bajaur Granite is in fact not a granite but Norite, which is rich in Na-Ca plagioclase series of feldspars. The plagioclase feldspar rich in Na-Ca are low in radioactivity. Moreover, the average gamma activities of 226Ra, 232Th, and 40K (4.98 ± 0.13 Bqkg−1, 4.03 ± 0.31 Bqkg−1, 204.40 ± 4.72 Bqkg−1 and a total of all three radionuclides are 214.00 ± 5.39 Bqkg−1) for Bajaur Norites are found too be much less than the average of the world’s Granites. Indoor and outdoor hazard indices of Bajaur Norite are much below building materials used throughout the world and largely beneath their criterion restrictions. As per radiations’ hazards are concerned, the Bajaur Norite as a building stone may be considered as the safest material available in the area that does not pose any radiological hazard.

Carbon Mineralization with North American PGM Mine Tailings—Characterization and Reactivity Analysis

Global efforts to combat climate change call for methods to capture and store CO2. Meanwhile, the global transition away from fossil energy will result in increased production of tailings (i.e., wastes) from the mining of nickel and platinum group metals (PGMs). Through carbon mineralization, CO2 can be permanently stored in calcium- and magnesium-bearing mine tailings. The Stillwater mine in Nye, Montana produces copper, nickel, and PGMs, along with 1 Mt of tailings each year. Stillwater tailings samples have been characterized, revealing that they contain a variety of mineral phases, most notably Ca-bearing plagioclase feldspar. Increases in inorganic carbon in the tailings and ion concentration in the tailings storage facilities suggest carbonation has taken place at ambient conditions over time within the tailings storage facilities. Two experiments were performed to simulate carbon mineralization at ambient temperature and pressure with elevated CO2 concentration (10% with N2), revealing that less than 1% of the silicate-bound calcium within the tailings is labile, or easily released from silicate structures at low-cost ambient conditions. The Stillwater tailings could be useful for developing strategies of waste management as production of nickel and PGM minerals increases during the global transition away from fossil energy, but further work is needed to develop a process that can realize their full carbon storage potential.

Nonclimatic and extrabasinal processes controlled pedogenesis in paleosols of the Paleocene Nacimiento Formation, New Mexico, USA

Paleosols can represent intervals of nondeposition in sedimentary packages and are used with increasing frequency as proxies for paleoenvironmental conditions during basin filling. However, the complexities of factors both internal and external to pedogenesis require consideration of paleosols in any basin or sedimentary package in a broader context than is often assumed. With this in mind, we measured and analyzed pedogenic features, stratigraphic position, geochemical composition, and petrography of paleosols in the Nacimiento Formation to gain insight into the paleoenvironmental conditions of the early Paleocene in the San Juan Basin. During this interval, the San Juan Basin was located in middle northern latitudes (∼40°N) and saw rapid terrestrial siliciclastic sedimentation related to Laramide tectonism. Evidence from earlier researchers suggests that prevailing climate conditions during Nacimiento Formation deposition were warm and humid. We used morphological properties of paleosols to categorize paleosols into pedotypes indicative of distinct pedogenic conditions. The general stratigraphic distribution of these pedotypes shows an up-section increase in soil drainage conditions through the Nacimiento Formation that cannot be correlated with known climate changes. We suggest that fluvial system evolution was the major control on pedogenic conditions. We investigated Nacimiento Formation paleosols with widely used paleosol geochemical climate analyses, which provided paleoclimate estimates that are in disaccord with independently derived estimates. We show that in alluvial depositional systems with source areas in weathered sedimentary rocks, these analyses can be difficult to interpret and likely lead to estimates that do not reflect true pedogenic conditions during the postdepositional near-surface alteration of sediments. Petrographic analysis of Nacimiento Formation paleosols showed that some likely formed under semiarid to subhumid conditions that allowed pedogenic accumulation and illuviation of smectite clays yet did not substantially chemically alter primary detrital plagioclase feldspar grains in paleosol B horizons. The paleosols of the Nacimiento Formation, when analyzed at the basin scale, show that sedimentary aggradational processes can overpower climate processes in creation of a sequence of paleosols in a stratigraphic section. In addition, the incorporation of clays from sedimentary rocks in a basin’s source area can drastically skew the geochemical signatures and therefore interpretations of paleosols in that basin.

Petrology and Geochemistry of Nakora Ring Complex with Emphasis on Tectonics and Magmatism, Neoproterozoic Malani Igneous Suite, Western Rajasthan, India

The present contribution reports on the field, petrographical and geochemical observations of the volcano-plutonic rocks of the Nakora Ring Complex (NRC) from the Neoproterozoic, Malani Igneous Suite (MIS) (Northwestern Peninsular India) and confers about their magmatic evolution and tectonic implications. Three magmatic phases are notable in the NRC which is Extrusive, Intrusive and Dyke phase where with small quantities of basaltic flows was initiated and accompanied by extensive/voluminous acidic flows. Petrographically, rhyolite shows flow bands, porphyritic, spherulitic, aphyritic and perlitic textures whereas basalt flows are distinguished by the presence of labradorite in lath-shaped crystals (plagioclase feldspar) and clinopyroxene (augite). The presence of high silica and total alkalis in NRC rocks, as well as high field strength elements (HFSE), enrichment of trace elements and negative anomalies of Sr., Eu, P, and Ti indicates that the emplacement of the lava flows was controlled by complex magmatic processes such as fractional crystallization, crustal contamination and partial melting. The association of basalt-trachyte-rhyolite means that the magma chamber was supplied a significant amount of heat to the crust before the eruption. Moreover, a volcanic vent was also reported at NRC where rhyolite was associated with agglomerate, volcanic breccia, perlite and tuff. The current research proposed that the Neoproterozoic magmatism at NRC was controlled by rift-related mechanism and produced from crustal source where the heat was supplied by mantle plume.

Revisiting the Bøggild Intergrowth in Iridescent Labradorite Feldspars: Ordering, Kinetics, and Phase Equilibria

The enigmatic Bøggild intergrowth in iridescent labradorite crystals was revisited in light of recent work on the incommensurately modulated structures in the intermediated plagioclase. Five igneous samples and one metamorphic labradorite sample with various compositions and lamellar thicknesses were studied in this paper. The lamellar textures were characterized with conventional transmission electron microscopy (TEM) and scanning transmission electron microscopy (STEM). The compositions of individual lamellae were analyzed with high-resolution energy-dispersive X-ray spectroscopy (EDS) mapping and atom probe tomography (APT). The average structure states of the studied samples were also compared with single-crystal X-ray diffraction data (SC-XRD). The Na-rich lamellae have a composition of An44–48, and the Ca-rich lamellae range from An56 to An63. Significant differences between the lamellar compositions of different samples were observed. The compositions of the Bøggild intergrowth do not only depend on the bulk compositions, but also on the thermal history of the host rock. The implications on the subsolidus phase relationships of the plagioclase feldspar solid solution are discussed. The results cannot be explained by a regular symmetrical solvus such as the Bøggild gap, but they support an inclined two-phase region that closes at low temperature.

Tectonic Setting and Provenance of Eocene Sandstones of Disang Group, Tirap District, Arunachal Pradesh

The petrographic study of sandstone belonging to Disang Group of Eocene age shows that quartz and rock fragments are the main constituents among the framework grains. Feldspar percentage is low whereas plagioclase feldspar is dominant with the rare occurrence of K-feldspar. XRF-analysis reveals that SiO2, AI2O3, Fe2O3, MnO, TiO2, Na2O, K2O, CaO, MgO and P2O5 are the major and minor elements. SiO2 constitutes the major proportion of the oxides in the sediments. The study reveals that the provenance of these sandstones is mainly the igneous and metamorphic rocks, and the tectonic setting was an active continental margin. The presence of higher content of chert and mafic rock fragments points towards the ophiolite zone as a provenance. Detritus were possibly derived from the uplifted fold thrust belt of the Myanmar’s landmass with subordinate contribution from the Mishmi Hills region lying to the northeast of the study area.

Petrophysical and petrographic characteristics of Barail Sandstone of the Surma Basin, Bangladesh

The Barail sandstone in the Surma Basin is a medium- to coarse-grained pinkish-colored rock exposed near the northeastern margin of Bangladesh. In this study, we evaluated the reservoir quality of the Barail sandstone based on its petrophysical and petrographic characteristics. Petrophysical analyses of outcropped samples showed that sandstones are made up of 16.48% porosity and 132.48 mD permeability. Sandstone density ranges from 1.94 g/cm3 to 2.37 g/cm3, with a mean value of 2.12 g/cm3, shown as moderately compacted sandstone. Integrated data such as bulk density, porosity, permeability, Rock Quality Index (RQI), Normalized Porosity Index (NPI), Flow Zone Indicator (FZI), compressive strength, etc. with their relationships indicate that Barail sandstone owing characters to become a good petroleum reservoir. The rock samples consisted mainly of quartz with an insignificant amount of rock fragments and plagioclase feldspar and are categorized as sub-arkose to sub-litharenite. The rock samples also contains lithic (andesine, microcline, muscovite, biotite, etc.) of granitic and gneissic fabric and some volcanic product like aguite, albite, andesine, garnet, spinel and ulvo-spinel indicating the source of nearby orogeny. The euhedral to subhedral shape of the quartz grain in a porphyritic texture, moderately sorted with a smaller amount of clay minerals indicating the moderately mature rock type. The iron oxide border around the quartz grain also indicates that the Barail sandstone was deposited under dry climatic condition.

ENHANCED MINERAL QUANTIFICATION AND UNCERTAINTY ANALYSIS FROM DOWNHOLE SPECTROSCOPY LOGS USING VARIATIONAL AUTOENCODERS

This paper describes an innovative machine learning application, based on variational autoencoder frameworks, to quantify the concentrations and associated uncertainties of common minerals in sedimentary formations using the measurement of atomic element concentrations from geochemical spectroscopy logs as inputs. The algorithm comprises an input(s), encoder, decoder, output(s), and a novel cost function to optimize the model coefficients during training. The input to the algorithm is a set of dry-weight concentrations of atomic elements with their associated uncertainty. The first output is a set of dry-weight fractions of fourteen minerals, and the second output is a set of reconstructed dry-weight concentrations of the original elements. Both sets of outputs include estimates of uncertainty on their predictions. The encoder and decoder are multilayer feed-forward artificial neural networks (ANN), with their coefficients (weights) optimized during calibration (training). The cost function simultaneously minimizes error (the accuracy metric) and variance (the precision or robustness metric) on the mineral and reconstructed elemental outputs. Training of the weights is done using a set of several-thousand core samples with independent, high-fidelity elemental and mineral (quartz, potassium-feldspar, plagioclase-feldspar, illite, smectite, kaolinite, chlorite, mica, calcite, dolomite, ankerite, siderite, pyrite, and anhydrite) data. The algorithm provides notable advantages over existing methods to estimate formation lithology or mineralogy relying on simple linear, empirical, or nearest-neighbor functions. The ANN numerically capture the multi-dimensional and nonlinear geochemical relationship (mapping) between elements and minerals that is insufficiently described by prior methods. Training is iterative via backpropagation and samples from Gaussian distributions on each of the elemental inputs, rather than single values, for every sample at each iteration (epoch). These Gaussian distributions are chosen to specifically represent the unique statistical uncertainty of the dry-weight elements in the logging measurements. Sampling from Gaussian distributions during training reduces the potential for overfitting, provides robustness for log interpretations, and further enables a calibrated estimate of uncertainty on the mineral and reconstructed elemental outputs, all of which are lacking in prior methods. The framework of the algorithm is purposefully generalizable that it can be adapted across geochemical spectroscopy tools. The algorithm reasonably approximates a ‘global-average’ model that requires neither different calibrations nor expert parameterization or intervention for interpreting common oilfield sedimentary formations, although the framework is again purposefully generalizable so it can be optimized for local environments where desirable. The paper showcases field application of the method for estimating mineral type and abundance in oilfield formations from wellbore logging measurements.