Melt embayments record multi-stage magma decompression histories

Over the last decade, the melt embayment has proven its merit as a robust petrological tool capable of recording magma decompression rates for explosive eruptions. However, the models developed and applied to extract this information from embayments have not accounted for the complexity and nonlinearity of magma flow in the conduit. We present Embayment Decompression in Two Stages (EDiTS): a numerical model for extracting magma decompression rates from measured volatile diffusion profiles preserved in crystal-hosted embayments, approximating magma acceleration using two constant-rate decompression paths. This model solves for three unknown parameters: initial (deeper) and final (shallower) decompression rates, as well as the pressure where a transition occurs. We successfully benchmark EDiTS against existing numerical diffusion models, and use controlled multi-stage decompression experiments on natural quartz-hosted embayments to test the ability of our model to recover known decompression paths. We find that EDiTS is able to closely approximate the known two-stage path in the mixed-volatile (H₂O + CO₂) experiment, while a constant-rate modeling approach is unable to simultaneously fit H₂O and CO₂ gradients. However, in the H₂O-saturated experiment, there is no unique solution to the resulting gradient, with both constant-rate and two-stage models reproducing the measured profile, and EDiTS notably overestimating the known total ascent time by several hours. Using decompression experiments, we show that constant-rate models can provide misleadingly good fits to embayment H₂O gradients produced by more complex decompression histories, and thus the measurement and modeling of multiple diffusing species, when available, can provide crucial constraints. We then apply EDiTS to re-evaluate mixed-volatile embayment datasets from explosive silicic arc and caldera-forming eruptions from five volcanic centers (Yellowstone, WY, USA; Valles, NM, USA; Long Valley, CA, USA; Taupo, NZ; Mount St. Helens, WA, USA). In contrast to the minutes to hours of total ascent time extracted from embayment volatile profiles using constant-rate models, our two-stage model resolves slower initial ascent times that span 3.5–11 h. Final ascent rates are 1–2 orders of magnitude faster than the initial extracted rates, in agreement with theoretical conduit flow model predictions. Reassessment of embayments from the May 18th, 1980 eruption of Mount St. Helens results in an initial stage of ascent consistent with the timing of magma arrival at the surface from the seismically-inferred storage region (7–9 km) ∼3.5 h after the initial blast, and a final stage of ascent (<1–5 min) in close agreement with time-integrated bubble number densities. Our combined numerical and experimental results, and reevaluation of natural datasets, suggest that, with the application of advanced models, the melt embayment can provide a more nuanced picture of magma decompression timescales from the deep conduit to the surface.