The role of laboratory experiments in volcanology

Experiment is central to scienti¢c methodology. Within science in general, experiments are used for four primary purposes, i.e. as a tool to explore novel phenomena and to provide systematic observations of processes, to determine the values of key parameters, to test hypotheses and theoretical models, and to validate computational models, which themselves can be considered as numerical experiments. The study of volcanic eruptions and magmatic processes is becoming increasingly systematic, quantitative and rigorous. The drive is towards providing detailed quantitative descriptions and models of the £ow phenomena in terms of fundamental physico^chemical processes. This is motivated in no small way by an increasing need for reliable and quantitative predictions of volcanic eruptions on which to base rational hazard and risk management decisions during volcanic crises. A natural consequence of this development of the scienti¢c study of volcanic £ow phenomena is that rigorous laboratory experimentation is becoming an increasingly important feature within volcanic research. The contribution of experimental research to our understanding of volcanic processes is di⁄cult to overstate. The determination of the values of key parameters has traditionally taken place in the laboratory. However, more recently experiments have become much more widely applied to the problem of understanding complex dynamical processes and their underlying mechanisms. Without models (experimental, theoretical, numerical) we are limited to qualitative interpretations of ¢eld observations and remote sensing. Field data are of course essential and provide the measure against which ultimately the applicability of observations from other sources must be assessed. However, there are several reasons why the scienti¢c study of volcanic phenomena cannot rely solely on ¢eld data. Direct observations of eruption processes in the ¢eld are limited to those parts of an eruption that are accessible. This restricts detailed observations to a limited range of above-surface processes. Moreover, logistical constraints mean that the systematic collection of data will usually only be possible for volcanic eruptions near an established observatory and for activity for which there has been some forewarning. As a result of these constraints, direct ¢eld-based observations of natural volcanic eruption processes are generally incomplete and uncertain. Another source of natural data derives from observations made on volcanic deposits. In principle, processes can be inferred from the features