Page 283 - Caldera Volcanism Analysis, Modelling and Response

258 J. Martı ´ et al.

Guillou-Frottier, 1999). Their conclusion is that the presence of far-field
stresses can significantly modify the distribution and geometry of faulting.
Regional extensional stresses shift the locations of the zones of minimal and
maximal bending stresses, leading to both the formation of multiple faults and
favouring the occurrence of deep faulting centred over the magma chamber
roof. As long as the extension continues, fractures develop at a more vertical
angle and propagate from the centre to the borders to create a wider dispersed
fault zone.
In conclusion, results from the available models agree that magma chamber
geometry and load conditions are the two main controlling factors for ring-fault
formation and, hence, the generation of collapse calderas. With few exceptions,
models concur in predicting that sill-like chamber geometries are a necessary
requirement. In addition, only the satisfaction of specific conditions seems to
generate a favourable stress field for collapse: tension, regional doming, chamber
overpressure combined with load increase by erupted products, and chamber
underpressure.

3.3. Restrictions and limitations of theoretical models
Theoretical models contribute important semi-quantitative information comple-
menting experimental models and field studies. However, as with all models, they
also have restrictions and limitations that we discuss in this section.
Similar to the analogue models, numerical models assume, with few exceptions,
a homogenous crust. It is certainly a gross approximation to nature because country
rocks are normally heterogeneous in composition and properties. Lithological
heterogeneities can influence the stress field and the rock strength profile and,
consequently, the development and propagation of fractures (e.g. Gudmundsson
and Brenner, 2005).
One of the main problems of existing models on caldera-collapse-formation
processes is that fluid dynamics and rock mechanics are, in all cases, uncoupled.
As mentioned above, in an ideal case, all physical processes should be simulated
simultaneously, as processes occurring inside the chamber may affect those taking
place in the country rock and vice versa.
None of the discussed models succeeds in simulating dyke injection neither
during collapse nor during tumescence. However, as mentioned before, the
possibility of dyke injection is important during pre-caldera episodes as it can
regulate (dis)equilibrium conditions inside the magma chamber. The impossibility
of simulating dyke injection has further consequences: (i) dyke intrusions may
significantly modify the physical properties of country rock, for example, its tensile
or shear strengths, evidently affecting fault nucleation and propagation; and
(ii) some interpretation of results obtained may be misinterpreted. Whereas some
authors claim that ring fault or other collapse-controlling structures develop from
the top of the magma chamber to the surface (e.g. Gray and Monaghan, 2004),
others state that ring-fault nucleation at depth is not possible as any magma chamber
rupture would lead to a dyke intrusion (e.g. Gudmundsson, 1998). Of course, these