A new algorithm for complex lava flow modeling driven by satellite-derived data

Lava flows represent the main hazardous phenomena that may threaten human buildings in basaltic volcanoes. Physical and chemical characteristics of lava, as well as the rate at which lava is emitted at the vents, strongly control the motion and the extensions of lava flows. Estimation of the likely paths that lava flows may follow during an ongoing eruption can be evaluated through numerical modeling driven by satellite-derived time averaged discharge rate (TADR). However, complex eruptive dynamics, such as the sequential opening of new vents, formation of lava tubes and fluctuations in effusion rates, may lead to the development of composite lava fields, which are challenging to reproduce with a simple 2D modeling. In this study, we present a new optimization algorithm based on the Metropolis-Hastings approach to model complex emplacements of lava flow fields using satellite imagery. In particular, the algorithm performs sequential time-step refinements by assimilating TADR estimations, derived from low spatial resolution satellite imagery (e.g. MODIS, SLSTR, SEVIRI), and the locations of both main and ephemeral vents, obtained from higher spatial resolution imagery (e.g. MSI Sentinel 2, OLI & TIRS Landsat 8/9 or Planetscope). The algorithm automatically chooses the optimal input parameters and the associated uncertainty, minimizing the mismatch between the simulated and the observed lava flow features extracted from multi-source satellite imagery. The testing and validation have been performed using two recent effusive events that occurred at Mt. Etna (Italy): the 27 February - 1 March 2017 and the 27 November 2022 - 6 February 2023 eruptions.