Machine learning classifiers for attributing tephra to source volcanoes: an evaluation of methods for Alaska tephras

Glass composition-based correlations of volcanic ash (tephra) traditionally rely on extensive manual plotting. Many previous statistical methods for testing correlations are limited by using geochemical means, masking diagnostic variability. We suggest that machine learning classifiers can expedite correlation, quickly narrowing the list of likely candidates using well-trained models. Eruptives from Alaska's Aleutian Arc-Alaska Peninsula and Wrangell volcanic field were used as a test environment for 11 supervised classification algorithms, trained on nearly 2000 electron probe microanalysis measurements of glass major oxides, representing 10 volcanic sources. Artificial neural networks and random forests were consistently among the top-performing learners (accuracy and kappa > 0.96). Their combination as an average ensemble effectively improves their performance. Using this combined model on tephras from Eklutna Lake, south-central Alaska, showed that predictions match traditional methods and can speed correlation. Although classifiers are useful tools, they should aid expert analysis, not replace it. The Eklutna Lake tephras are mostly from Redoubt Volcano. Besides tephras from known Holocene-active sources, Holocene tephra geochemically consistent with Pleistocene Emmons Lake Volcanic Center (Dawson tephra), but from a yet unknown source, is evident. These tephras are mostly anchored by a highly resolved varved chronology and represent new important regional stratigraphic markers.     

COCORP: new perspectives on the deep crust

Summary. Relict sutures from colliding continents, regions characterized by a "young" Mono, layering and faulting throughout the crust, mid-crustal magma traps, and seismic "bright spots" which suggest deep crustal fluids are among recent COCORP findings. In addition, new studies of signal penetration, noise mitigation, recording geometry, and coherency filtering have yielded better understanding of, and substantial improvements in, data quality. Amplitude anomalies, or "bright spots", in the Basin and Range may be due to magma at mid-crustal levels; in one case, a normal fault appears to link the deep magma with young surface volcanics. Another bright spot. 15 km deep in southeastern Georgia, has a flat geometry that suggests a gas/liquid interface, perhaps within fluids underthrust along an Appalachian suture. The Mono continues to appear relatively undisturbed in many regions of past deformation, suggesting that its formation post-dates these major tectonic episodes. The diversity of reflection patterns from the U.S. Cordillera casts further doubt on the generality of the common model of a reflective, layered lower crust underlying a transparent upper crust.     

INTEROBJECT STRUCTURE OF UNGROUPED IRON METEORITES AS REVEALED BY ADVANCED CLUSTERING: METHODS AND CHEMICAL FEATURES

Using the MASLOC clustering strategy, 61 ungrouped iron meteorites are probed for physico-chemical clusters. The existing classification of 13 resolved chemical groups is used as an internal coherence reference. Results include tentative establishment of a new group: ID (7 members), with chemical features similar to those of the non-magnetic groups IAB and IIICD. Six additional pairings, all with magmatic chemical signatures, are also recognized.     

Formation and photochemistry of Methylamine in Jupiter's atmosphere

The formation of methylamine (CH₃NH₂) in the upper troposphere and lower stratosphere of Jupiter is investigated. Translationally hot hydrogen atoms are produced in the photolysis of ammonia, phosphine, and acetylene which react with methane to produce methyl (CH₃) radicals; the latter recombine with NH₂ to form CH₃NH₂. Also, methane is catalytically dissociated to CH₃ + H by the species C₂ and C₂H produced in the photolysis of acetylene. It is shown that the combined production of CH₃NH₂ and subsequent photolysis to HCN is unlikely to account for the HCN observed near Jupiter's tropopause. Recombination of NH₂ and C₂H₅N followed by photolysis to HCN is the preferred path. Production of C₂H₆ by these two processes is negligible in comparison to the downward flux of C₂H₆ from the Lyman α photolysis region of CH₄. An upper limit column density on CH₃PH₂ is estimated to be ～10¹³ cm^?2 as compared to 10¹⁵ cm^?2 for CH₃NH₂. Hot H atoms account for a negligible fraction of the total ortho-para conversion by the reaction H + H₂     

Ice-free conditions in Novaya Zemlya 35 000–30 000 cal years B.P., as indicated by radiocarbon ages and amino acid racemization evidence from marine molluscs

Novaya Zemlya was covered by the eastern part of the Barents–Kara ice sheet during the glacial maximum of marine isotope stage 2 (MIS 2). We obtained ¹⁴C ages on 37 samples of mollusc shells from various sites on the islands. Most samples yielded ages in the range of 48–26 ¹⁴C Ky. Such old samples are sensitive to contamination by young ¹⁴C, and therefore their reliability was assessed using replicate analyses and amino acid geochronology. The extent of aspartic acid racemization (Asp D/L) indicates that many of the ¹⁴C ages are correct, whereas some are minimum ages only. The results indicate that a substantial part of Novaya Zemlya was ice-free about 35–27 ¹⁴C Kya, and probably even earlier. Corresponding shorelines up to >140 m a.s.l. indicate a large Barents–Kara ice sheet during early MIS 3. These results are consistent with findings from Svalbard and northern Russia: in both places a large MIS 4/3 Barents–Kara ice sheet is postulated to have retreated about 50 Kya, followed by an ice-free interstadial that lasted until up to ca. 25 Kya. The duration of the MIS 2 glaciation in Novaya Zemlya was calculated by applying the D/L values to a kinetic equation for Asp racemization. This indicates that the islands were ice covered for less than 3000 years if the basal temperature was 0^oC, and for less than 10 000 years if it was −5^oC.     

Late Holocene storm-trajectory changes inferred from the oxygen isotope composition of lake diatoms,south Alaska

The oxygen isotope ratios of diatoms (δ¹⁸O_diatom), and the oxygen and hydrogen isotope ratios of lake water (δW) of lakes in south Alaska provide insight into past changes in atmospheric circulation. Lake water was collected from 31 lakes along an elevation transect and diatoms were isolated from lake sediment from one lake (Mica Lake) in south Alaska. In general, δW values from coastal lakes overlap the global meteoric water line (GMWL). δW values from interior lakes do not lie on the GMWL; they fall on a local evaporation line trajectory suggesting source isotopes are depleted with respect to maritime lakes. Sediment cores were recovered from 58 m depth in Mica Lake (60.96° N, 148.15° W; 100 m asl), an evaporation-insensitive lake in the western Prince William Sound. Thirteen calibrated ¹⁴C ages on terrestrial macrofossil samples were used to construct an age-depth model for core MC-2, which spans 9910 cal years. Diatoms from 46, 0.5-cm-thick samples were isolated and analyzed for their oxygen isotope ratios. The analyses employed a newly designed, stepwise fluorination technique, which uses a CO₂ laser-ablation system, coupled to a mass spectrometer, and has an external reproducibility of ±0.2‰. δ¹⁸O_diatom values from Mica Lake sediment range between 25.2 and 29.8‰. δ¹⁸O_diatom values are relatively uniform between 9.6 and 2.6 ka, but exhibit a four-fold increase in variability since 2.6 ka. High-resolution sampling and analyses of the top 100 cm of our lake cores suggest large climate variability during the last 2000 years. The 20th century shows a +4.0‰ increase of δ¹⁸O_diatom values. Shifts of δ¹⁸O_diatom values are likely not related to changes in diatom taxa or dissolution effects. Late Holocene excursions to lower δ¹⁸O_diatom values suggest a reduction of south-to-north storm trajectories delivered by meridional flow, which likely corresponds to prolonged intervals when the Aleutian Low pressure system weakened. Comparisons with isotope records of precipitation (δP) from the region support the storm-track hypothesis, and add to evidence for variability in North Pacific atmospheric circulation during the Holocene.

Using soil and contaminant properties to assess the potential for groundwater contamination to the lower Great Lakes,USA

Contaminant risk factors in surface soil were evaluated within the urbanized Rouge River watershed in southeastern Michigan, USA, which includes metropolitan Detroit. An analytical risk factor model and Geographic Information Systems overlays were used to quantify and characterize the potential impacts of five categories of contaminants including DNAPLs (dense nonaqueous phase liquids), LNAPLs (light nonaqueous phase liquids), PAHs (polynuclear aromatic hydrocarbons), PCBs (polychlorinated biphenyls), and lead. The results indicate that DNAPL compounds released into sand, moraine, and sandy and silty clay soil types have the greatest potential to affect groundwater, and impact the Great Lakes ecosystem and the public health.