|
Titel |
Space-time patterns in ignimbrite compositions revealed by GIS and R based statistical analysis |
VerfasserIn |
Melanie Brandmeier, Gerhard Wörner |
Konferenz |
EGU General Assembly 2017
|
Medientyp |
Artikel
|
Sprache |
en
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 19 (2017) |
Datensatznummer |
250138642
|
Publikation (Nr.) |
EGU/EGU2017-1729.pdf |
|
|
|
Zusammenfassung |
GIS-based multivariate statistical and geospatial analysis of a compilation of 890
geochemical and ca. 1,200 geochronological data for 194 mapped ignimbrites from Central
Andes documents the compositional and temporal pattern of large volume ignimbrites
(so-called “ignimbrite flare-ups”) during Neogene times. Rapid advances in computational
sciences during the past decade lead to a growing pool of algorithms for multivariate statistics
on big datasets with many predictor variables. This study uses the potential of R and ArcGIS
and applies cluster (CA) and linear discriminant analysis (LDA) on log-ratio transformed
spatial data. CA on major and trace element data allows to group ignimbrites according to
their geochemical characteristics into rhyolitic and a dacitic “end-members” and
differentiates characteristic trace element signatures with respect to Eu anomaly, depletion of
MREEs and variable enrichment in LREE. To highlight these distinct compositional
signatures, we applied LDA to selected ignimbrites for which comprehensive data sets
were available. The most important predictors for discriminating ignimbrites are La
(LREE), Yb (HREE), Eu, Al2O3, K2O, P2O5, MgO, FeOt and TiO2. However,
other REEs such as Gd, Pr, Tm, Sm and Er also contribute to the discrimination
functions.
Significant compositional differences were found between the older (>14 Ma)
large-volume plateau-forming ignimbrites in northernmost Chile and southern Peru and the
younger (< 10 Ma) Altiplano-Puna-Volcanic-Complex ignimbrites that are of similar
volumes. Older ignimbrites are less depleted in HREEs and less radiogenic in Sr isotopes,
indicating smaller crustal contributions during evolution in thinner and thermally less evolved
crust. These compositional variations indicate a relation to crustal thickening with a
“transition” from plagioclase to amphibole and garnet residual mineralogy between 13 to 9
Ma. We correlate compositional and volumetric variations to the N-S passage of
the Juan-Fernandéz ridge and crustal shortening and thickening during the past
26 Ma. The value of GIS and multivariate statistics in comparison to traditional
geochemical parameters are highlighted working with large datasets with many
predictors in a spatial and temporal context. Algorithms implemented in R allow taking
advantage of an n-dimensional space and, thus, of subtle compositional differences
contained in the data, while space-time patterns can be analyzed easily in GIS. |
|
|
|
|
|