摘要

A new algorithm for modelling one or two Mogi magma sources in actively deforming volcanoes is proposed using Santorini (Thera, Greece) volcano as a test site. This algorithm is based on a quasi-deterministic grid-scan search followed by a topological inversion in R-n space. It avoids point solutions and local minima and is primarily designed for GPS observations. It can be extended to other types of magma sources defined by closed functions and to other types of data. A validation of the method used an accuracy-oriented approach: a comparison between %26apos;true%26apos; or %26apos;reference%26apos; and inverse-modelled magma sources, which revealed bias-free, precise magma source solutions. A validation of the algorithm for false alarms, i.e. for identification of non-existent sources produced by computation artefacts, was also made. Computed solutions correspond to point sources affecting idealized media (homogeneous elastic half spaces) and hence ignore errors induced by magma source and lithology complexities, which should be superimposed to obtain the total error budget. Still, the solutions allow modelling of the propagation of measurement errors because of the mathematical model adopted; this problem has previously been ignored. A sensitivity analysis of results obtained using this method permit us to predict and quantify expected bias and uncertainties in modelled magma sources as a function of their depth and of the observation networks. Increased bias and noise in solutions for specific observations, networks and types of data are inferred, and this may explain reported discrepancies in various magma-source models. Some implications for the Santorini volcano, which had an unrest episode in 2011-2012, are that the algorithm can lead to less noisy solutions of the Mogi sources, that the surface deformation may be recognized as resulting from inflation of two spherical sources and that differences in magma source models using different types of data may only indicate predictable uncertainties in the solutions.

  • 出版日期2013-10