The Challenges of Data Rate and Data Accuracy in the Analysis of Volcanic Systems: An Assessment Using Multi-Parameter Data from the 2012-2013 Eruption Sequence at White Island, New Zealand

Wednesday, 17 December 2014: 1:55 PM
Arthur D Jolly1, Bruce W Christenson2, Jurgen W Neuberg3, Nicolas Fournier4, Agnes Mazot1, Geoff Kilgour4 and Gillian Elizabeth Jolly1, (1)GNS Science-Institute of Geological and Nuclear Sciences Ltd, Lower Hutt, New Zealand, (2)GNS Science-Institute of Geological and Nuclear Sciences Ltd, National Isotope Centre, Lower Hutt, New Zealand, (3)University of Leeds, Leeds, United Kingdom, (4)GNS Science, Taupo, New Zealand
Volcano monitoring is usually undertaken with the collection of both automated and manual data that form a multi-parameter time-series having a wide range of sampling rates and measurement accuracies. Assessments of hazards and risks ultimately rely on incorporating this information into usable form, first for the scientists to interpret, and then for the public and relevant stakeholders. One important challenge is in building appropriate and efficient strategies to compare and interpret data from these exceptionally different datasets.

The White Island volcanic system entered a new eruptive state beginning in mid-2012 and continuing through the present time. Eruptive activity during this period comprised small phreatic and phreato-magmatic events in August 2012, August 2013 and October 2013 and the intrusion of a small dome that was first observed in November 2012. We examine the chemical and geophysical dataset to assess the effects of small magma batches on the shallow hydrothermal system. The analysis incorporates high data rate (100 Hz) seismic, and infrasound data, lower data rate (1 Hz to 5 min sampling interval) GPS, tilt-meter, and gravity data and very low data rate geochemical time series (sampling intervals from days to months). The analysis is further informed by visual observations of lake level changes, geysering activity through crater lake vents, and changes in fumarolic discharges.

We first focus on the problems of incorporating the range of observables into coherent time frame dependant conceptual models. We then show examples where high data rate information may be improved through new processing methods and where low data rate information may be collected more frequently without loss of fidelity. By this approach we hope to improve the accuracy and efficiency of interpretations of volcano unrest and thereby improve hazard assessments.