Machine learning for graph-based representations of three-dimensional discrete fracture networks

Structural and topological information play a key role in modeling flow and transport through fractured rock in the subsurface. Discrete fracture network (DFN) computational suites such as dfnWorks (Hyman et al. Comput. Geosci. 84, 10–19 2015) are designed to simulate flow and transport in such porous media. Flow and transport calculations reveal that a small backbone of fractures exists, where most flow and transport occurs. Restricting the flowing fracture network to this backbone provides a significant reduction in the network’s effective size. However, the particle-tracking simulations needed to determine this reduction are computationally intensive. Such methods may be impractical for large systems or for robust uncertainty quantification of fracture networks, where thousands of forward simulations are needed to bound system behavior. In this paper, we develop an alternative network reduction approach to characterizing transport in DFNs, by combining graph theoretical and machine learning methods. We consider a graph representation where nodes signify fractures and edges denote their intersections. Using random forest and support vector machines, we rapidly identify a subnetwork that captures the flow patterns of the full DFN, based primarily on node centrality features in the graph. Our supervised learning techniques train on particle-tracking backbone paths found by dfnWorks, but run in negligible time compared to those simulations. We find that our predictions can reduce the network to approximately 20% of its original size, while still generating breakthrough curves consistent with those of the original network.     

Fractionated matrix composition in CV3 Vigarano and alteration processes on the CV parent asteroid

Abstract— Although CV3 Vigarano is one of the most primitive CV chondrites, it has lost most of the S from the matrix; matrix Na is also depleted relative to the concentration in bulk CV chondrites. We used a matrix‐grid technique to study thirteen 50 × 50 μm regions in Vigarano; in each area, we used an electron microprobe to gather data with an electron beam 3 μm in width. We found two end‐member types of matrix textures. One is coarse and porous, has lower Fe contents and lower analytical totals; it appears to be contaminated with comminuted chondrule debris. The other is finer grained and appears smooth; its mean composition has higher Fe, but lower S and Al contents, than the coarse matrix areas. Our tentative interpretation is that the larger grain size of the coarse areas resulted from the admixing of comminuted chondrule materials, and thus that the initial fraction of nebular fines was higher in the fine matrix regions. Aside from volatiles, the overall composition of Vigarano matrix is similar to that observed in matrix‐grid studies of other carbonaceous chondrites: Al, Si, Fe, and Mn have high whole‐chondrite‐normalized abundance ratios; Ca concentrations are low and highly variable. Because asteroidal alteration effects are present in our sample, it is difficult to resolve nebular signatures in the compositions of the grid areas.     

The carbon Cepheid RT Trianguli Australis: additional evidence of triple-α and CNO cycling

We have used echelle spectra of resolving power 35 000 to derive chemical abundances and the ¹²C/¹³C ratio in the 1.9-d carbon Cepheid RT TrA and the Cepheid U TrA, employed as a comparison star. We confirm that RT TrA is very metal-rich with [Fe/H]=+0.4. In addition, C and N are substantially in excess, and a small deficiency in O is present. We interpret these anomalies as resulting from the appearance on the stellar surface of material enriched in ¹²C by the 3- α process, followed by CNO cycling to convert ¹²C to ¹³C and ¹⁴N. In addition, some ¹⁶O has been processed to ¹⁴N. The partial processing of ¹⁶O to ¹⁴N indicates that substantial ¹⁷O may be present. Proton capture seems to have enhanced ²³Na from the Ne isotopes.     

High-resolution simulations of the final assembly of Earth-like planets I. Terrestrial accretion and dynamics

The final stage in the formation of terrestrial planets consists of the accumulation of ∼1000-km “planetary embryos” and a swarm of billions of 1-10 km “planetesimals.” During this process, water-rich material is accreted by the terrestrial planets via impacts of water-rich bodies from beyond roughly 2.5 AU. We present results from five high-resolution dynamical simulations. These start from 1000-2000 embryos and planetesimals, roughly 5-10 times more particles than in previous simulations. Each simulation formed 2-4 terrestrial planets with masses between 0.4 and 2.6 Earth masses. The eccentricities of most planets were ∼0.05, lower than in previous simulations, but still higher than for Venus, Earth and Mars. Each planet accreted at least the Earth's current water budget. We demonstrate several new aspects of the accretion process: (1) The feeding zones of terrestrial planets change in time, widening and moving outward. Even in the presence of Jupiter, water-rich material from beyond 2.5 AU is not accreted for several millions of years. (2) Even in the absence of secular resonances, the asteroid belt is cleared of >99% of its original mass by self-scattering of bodies into resonances with Jupiter. (3) If planetary embryos form relatively slowly, then the formation of embryos in the asteroid belt may have been stunted by the presence of Jupiter. (4) Self-interacting planetesimals feel dynamical friction from other small bodies, which has important effects on the eccentricity evolution and outcome of a simulation.     

Marine and estuarine reservoir effects in central Queensland,Australia: Determination of ΔR values

As a component of archaeological investigations on the central Queensland coast, a series of five marine shell specimens live‐collected between A.D. 1904 and A.D. 1929 and 11 shell/charcoal paired samples from archaeological contexts were radiocarbon dated to determine local ΔR values. The object of the study was to assess the potential influence of localized variation in marine reservoir effect in accurately determining the age of marine and estuarine shell from archaeological deposits in the area. Results indicate that the routinely applied ΔR value of −5 ± 35 for northeast Australia is erroneously calculated. The determined values suggest a minor revision to Reimer and Reimer's (2000) recommended value for northeast Australia from ΔR = +11 ± 5 to +12 ± 7, and specifically for central Queensland to ΔR = +10 ± 7, for near‐shore open marine environments. In contrast, data obtained from estuarine shell/charcoal pairs demonstrate a general lack of consistency, suggesting estuary‐specific patterns of variation in terrestrial carbon input and exchange with the open ocean. Preliminary data indicate that in some estuaries, at some time periods, a ΔR value of more than −155 ± 55 may be appropriate. In estuarine contexts in central Queensland, a localized estuary‐specific correction factor is recommended to account for geographical and temporal variation in ¹⁴C activity. © 2002 Wiley Periodicals, Inc.