Tracking time-dependent eruption source unsteadiness and local entrainment in ground-based thermal imagery using spectral-clustering

Abstract:

Volcanic eruptions have unsteady source conditions where variations in source mass flux are comparable in magnitude to the mean mass flux or occur over timescales equal to or less than plume-rise times. As a quantitative measure of unsteadiness, we consider a ratio of two timescales: the Pulsation Number, Pu₀=τ_p/τ_e (see Gilchrist et al., Abstract ID 743400), where τ_p is the time interval between pulses of erupting material, and τ_e is the overturn time of the largest eddies (at the source) associated with each pulse. When eruption pulses are well separated in time, Pu₀>>1 and we expect pulses to behave as individual transient thermals, while very closely spaced pulses (Pu0<<1) approximate a classic steady plume. Critically, where Pu₀≈1, we expect transitional behavior between that of a pure thermal and a pure plume. For unsteady plumes, the entrainment rate, α_e, of ambient atmosphere is a local function of time and height, and constant mass flux assumptions common in numerical and experimental simulations may not accurately model plume entrainment and rise height. To investigate the effects of source unsteadiness on entrainment, we use high resolution, ground-based thermal infrared imagery of a range of unsteady explosive events at Sabancaya Volcano, Peru. We compare three events: (1) a large, transient explosion dominated by a single initial pulse followed by several smaller pulses, with Pu₀≈3, (2) a long-lived (∼4 hours duration) sustained plume with quasi-periodic emissions (1<Pu₀<2), (3) and an “emergent explosion” ( 2-3 minutes duration) with multiple distinct pulses (0.6<Pu₀<2). We use a novel algorithm based on spectral clustering to track the time-dependent evolution of individual coherent turbulent structures rising from the vent. We use two independent methods, evolution with height of (1) thermal statistics and (2) cluster dimensions, to estimate local entrainment for each tracked feature. Results show local entrainment rates in a continuum between that of a pure jet (α_e≈0.1) and that of a pure transient thermal (α_e≈0.25). Ongoing efforts will compare time-dependent entrainment as a function of source unsteadiness against predictions from experimental results. This study further demonstrates the potential for quantifying entrainment and mixing properties from detailed analysis of thermal imagery.