Page 277 - Caldera Volcanism Analysis, Modelling and Response

252 J. Martı ´ et al.

column’s weight (i.e. a positive pressure gradient with depth; otherwise magma in
the conduit would not flow out), yet at the same time there is a negative pressure
gradient with respect to the host rock, i.e. the lithostatic load. The presence of
volatiles and their effect on magma density can thus, in part, justify the application
of such models. However, what remains unexplained is the fact that during the vast
majority of volatile-rich eruptions conduit closure occurs before the underpressure
condition is reached and consequently caldera formation is avoided. This
phenomenon indicates that geometric and/or dynamic conditions required to
keep a conduit open are only achieved occasionally. The question as to why and
how this is achieved remains unanswered. The timescales of magma ascent and host
rock response to pressure variations are, in our opinion, key aspects that should
draw further attention.

3.2. Host rock models

This group of models (Komuro et al., 1984; Chery et al., 1991; Gudmundsson
et al., 1997; Gudmundsson, 1998; Burov and Guillou-Frottier, 1999; Guillou-Frottier
et al., 2000; Roche and Druitt, 2001; Folch and Martı ´, 2004; Gray and Monaghan,
2004; Gudmundsson, 2008) focus on the conditions necessary for the initiation
of ring faults. A ring fault is essentially a subvertical normal fault that requires
specific stress conditions to form or slip (e.g. Gudmundsson et al., 1997, 2008). A
ring fault may initiate at any depth between the margins of the chamber and the
surface. If failure initiates at the margin of the chamber, the resulting dykes would
relax the stress difference and may hinder the development of a ring fault.
Furthermore, ring faults may extend to considerable depth, but calderas are surface
features and do not form unless the stress field at surface favours ring-fault
formation. The stress field in the vicinity of the surface is thus a controlling factor
for caldera formation. Existing numerical results suggest that faults controlling
caldera collapse commonly develop from tension fractures at the surface of the
associated volcano and propagate to greater depths, towards the boundary of the
associated magma chamber (Gudmundsson, 1988a, 1988b). At a certain depth,
these tensional ring fractures change into normal-fault ring fractures. In general,
models assume that for this to occur the maximum shear stress must concentrate at
the lateral margins of the magma chamber and, simultaneously, the maximum
tensile stress must peak both at the surface and at a radial distance close to the
surface projection of the chamber walls. The latter condition is supported by field
evidence which suggests that ring faults are nearly vertical, and by analogue
experiments which show a correlation between the area of collapse and the cross-
sectional area of the reservoir (see previous sections).
In addition to the definition of the appropriate stress field, models also need to
specify a fracture criterion. Rocks behave as brittle materials at rapid loads, low
confining pressures, and low temperatures, whereas they tend to be ductile at high
confining pressures and temperatures (Rutter, 1974). Consequently, it seems
reasonable to assume both brittle and ductile fracture criteria near the magma
chamber walls. One possible choice (Gudmundsson, 1988b, 1998; Folch and Martı ´,