The geometry and development of a lava tube network as deduced from multispectral imaging and InSAR

Detecting lava tubes is challenging in the field due to their hidden nature and inaccessibility, but it can be important for understanding lava flow dynamics and mitigating hazards. Here we show how analysis of multispectral imagery (Sentinel-2 and Landsat) and InSAR (Sentinel-1) can enable the delineation of a ∼14-km long lava-tube network entirely by remote sensing. The lava tube network formed in 2024 during the 68-day long eruption of Fernandina volcano, a highly active, yet remote and uninhabited island in the Galapagos Islands. The arterial tube(s) and main branches of the network were mapped based on: (1) spatio-temporally stable, point-like thermal anomalies (“skylights”) from syn-eruption shortwave and thermal infrared imagery; and (2) a dendritic pattern of horizontal displacements defined by post-eruption InSAR timeseries analysis. Furthermore, elongated perpendicular baselines of Sentinel-1 interferograms enabled us to estimate lava-flow thicknesses of up to ∼17 m locally and a lava-field bulk volume of ∼84 ± 40 × 10⁶ m³. Lastly, we traced the growth of the lava field from a time series of InSAR coherence images. Combined with the lava thickness mapping, the coherence mapping gives initial magma eruption rates of 87 m³s⁻¹, which over two weeks declined rapidly and non-linearly to below 6 m³s⁻¹. This sharp reduction in eruption rate coincides with a transition - observed in multispectral imagery - from initial open channel flow to enclosed tube flow. Although the tube flow phase accounted for only 18% of the total erupted volume, it spanned 75% of the eruption duration and facilitated 35% (5 km) of the total lava run-out. These entirely remotely generated results are consistent with field‐based observations of lava tube development on Hawaii. A multi-sensor approach to remote sensing of lava tubes may therefore contribute in future to modelling of lava flow advance and to assessment of tube-collapse hazard.