Hydrothermal calderas

Abstract

The standard model of caldera formation is related to the emptying of a magma chamber and ensuing roof collapse during large eruptions or subsurface withdrawal. Although this model works well for numerous volcanoes, it is inappropriate for many basaltic volcanoes (with the notable exception of Hawaii), as these have eruptions that involve volumes of magma that are small compared to the collapse. Many arc volcanoes also have similar oversized depressions, such as Poas (Costa Rica) and Aoba (Vanuatu). In this article, we propose an alternative caldera model based on deep hydrothermal alteration of volcanic rocks in the central part of the edifice. Under certain conditions, the clay-rich altered and pressurized core may flow under its own weight, spread laterally, and trigger very large caldera-like collapse. Several specific mechanisms can generate the formation of such hydrothermal calderas. Among them, we identify two principal modes: mode 1: ripening with summit loading and flank spreading and mode II: unbuttressing with flank subsidence and flank sliding. Processes such as summit loading or flank subsidence may act simultaneously in hybrid mechanisms. Natural examples are shown to illustrate the different modes of formation. For ripening, we give Aoba (Vanuatu) as an example of probable summit loading, while Casita (Nicaragua) is the type example of flank spreading. For unbuttressing, Nuku Hiva Island (Marquesas) is our example for flank subsidence and Piton de la Fournaise (La Réunion) is our example of flank sliding. The whole process is slow and probably needs (a) at least a few tens of thousands of years to deeply alter the edifice and reach conditions suitable for ductile flow and (b) a few hundred years to achieve the caldera collapse. The size and the shape of the caldera strictly mimic that of the underlying weak core. Thus, the size of the caldera is not controlled by the dimensions of the underlying magma reservoir. A collapsing hydrothermal caldera could generate significant phreatic activity and trigger major eruptions from a coexisting magmatic complex. As the buildup to collapse is slow, such caldera-forming events could be detected long before their onset.

Appendix

Details about the scaling procedure and the experimental devices for each experiment may be found in published studies or theses (Merle and Lénat 2003; Cecchi 2003; Cecchi et al. 2005; Merle et al. 2006; Barde-Cabusson and Merle 2007; Barde-Cabusson 2007). Following a well-founded procedure established for volcanic terranes (e.g., Tibaldi 1995; Merle and Borgia 1996), similar conditions are achieved by selecting dimensionless numbers that need to maintain the same value in nature and experiments. According to the Buckingham П theorem, seven dimensionless numbers are required to verify that experiments are geometrically, kinematically, and dynamically well scaled.

From those, it can be shown that volcanic cones may be simulated by a dry sand/plastic mixture, a cohesionless brittle material, which properly reproduces the brittle behavior of nonaltered volcanic rocks. To simulate the weak volcanic core, we used silicone putty, a Newtonian material able to flow under its own weight, as is hypothesized for the clay-rich core of the volcano. Depending on the time during which motion occurs, silicone is appropriate to simulate ductile materials having a rather high viscosity ranging from 10¹⁰ to 10¹⁸ Pa s. Thus, it may be used to simulate either high-viscosity acid lavas (i.e., 10¹⁰–10¹¹ Pa s; see experiments about cryptodome intrusion in Mount St. Helens, Donnadieu and Merle 1998) or natural clays, which display viscosities of about 10¹⁷–10¹⁸ Pa s. This is why using silicone to simulate the clay-rich core of a volcano, as in the experiments under consideration, makes them properly scaled. In contrast, low-viscosity lavas (10²–10³ Pa s) of basaltic shields cannot be simulated by silicone; a better analog material is golden syrup or various types of oils (Galland et al. 2006; Mathieu et al. 2008).