MELVIS: An open-source web application for computing and modelling the viscosity of volcanic melts

Danilo Di Genova 

The viscosity of volcanic melts governs magma transport, eruption dynamics, and a
broad range of magmatic processes, so predicting it as a function of temperature,
composition, and dissolved water content is central to physical volcanology. Yet no
freely accessible tool integrates global machine-learning predictions, physically
motivated viscosity equations, and composition-specific models calibrated to avoid
nanocrystallization and water-loss artefacts — experimental biases that can lead
conventional viscometry to overestimate melt viscosity. Nor can existing tools show
how different assumptions about the dependence of melt fragility on dissolved water
propagate into viscosity predictions at eruptive temperatures — a debated question
with direct consequences for conduit-flow modelling and eruption-style
assessment.,Here we present MELVIS (MELt VIScosity), an open-source web
application built with Python and Streamlit. MELVIS comprises three integrated
modules: (1) a viscosity calculator based on the artificial neural network of
Langhammer et al. (2022), developed using a database of 3,482 measurements
spanning basanite to rhyolite including hydrous compositions, with batch processing
and TAS-based compositional classification; (2) a hydrous modelling module that
parameterises the glass transition temperature Tg(H2O) and melt fragility m(H2O)
under three alternative fragility–water relationships, making explicit how model
assumptions propagate into viscosity predictions at eruptive temperatures; and (3) a
library of seven composition-specific, artefact-free MYEGA models for key volcanic
systems. All modules yield closed-form MYEGA parameterisations suitable for direct
implementation in numerical eruption models.