Imaging Soufrière Hills shallow magma chamber: Integrating tomography and thermal models M. Paulatto (michelep@earth.ox.ac.uk), C. Annen, T. J. Henstock, E. Kiddle, T. A. Minshull, R.S.J. Sparks, B. Voight Fig. 4. (A-D) Model A: Under-accretion of 250-m-thick sills with 1.5 km radius at 182-year intervals starting at 4.5 km depth and lasting 4,000 years. (A) Temperature. (B) Melt fraction. (C) Velocity anomaly. (D) Filtered velocity anomaly. (E-H) same as (A-D), for Model B: two events of under-accretion of 300-m-thick sills with 2 km radius at 400-year intervals, each starting at 5 km depth and lasting 4000 years, with a 15,000-year repose period. Fig. 5. Integrated model of the magmatic system of Montserrat. Layer 1, recent sediments and volcanics. Layer 2: Mesozoic sediments and volcanics; Layer 3: intermediate plutons and active magma chamber. Fig. 3. Inverse temperature and melt estimates. (A) Seismic velocity anomaly beneath SHV. (B) Estimated temperature assuming no melt is present. (C,D) Predicted melt content assuming melt resides in interconnected flat pockets (C) or in isolated spherical pockets (D). 1. Introduction The ongoing eruption at the Soufrière Hills volcano (SHV), Montserrat, is thought to be fed by a shallow magma chamber, but its size and geometry are debated [1]. We use travel-time seismic tomography and numerical magma chamber growth models to constrain the shallow magma chamber with unprecedented detail and infer possible chamber growth scenarios. 3. Melt estimate The LVV may be caused by differences in lithology, by elevated temperatures or by the presence of partial melt. Lithology is not expected to contribute significantly as the petrology of the three volcanoes is the same. Temperature was estimated with the method of Karato [5]. At the centre of the anomaly temperature is well above the solidus of SHV andesites (~700°C) (Fig. 3B). Melt content was estimated with a self-consistent approach for multi-phase composites [6]. Our estimate is 3-9% depending on melt geometry (Fig. 3C,D), much lower than melt estimates from petrology (30-35%). 4. Magma chamber accretion A finite difference numerical algorithm [7] was used to calculate predicted temperature and melt distributions for a magma chamber formed by incremental sill accretion (Fig. 4A,B,E,F). The seismic velocity anomaly expected from such a magma chamber is much sharper and stronger than the observed anomaly (Fig. 4C,G). Predicted velocity anomalies were smoothed with a filter with width equal to seismic resolution (Fig. 4D,H) and compared to the tomographic results to select a best fitting magma chamber model. References [1] Voight et al. (2010), Geophys. Res. Lett,37, L00E05. [2] Le Friant, A. et al. (2004), J. Geol. Soc., 161,147-160. [3] Hobro, J. W. D. et al. (2003), Geophys. J. Int. 152,79-93. [4] Zelt, A. C. (1998), Geophys. J. Int., 135, 1101-1112. [5] Karato, S. (1993), Geophys. Res. Lett., 20, 1623-1626. [6] Berryman, J. G. (1980), J. Acoust. Soc. Am. 68, 1809-1831. [7] Annen, C. et al. (2008), J. Geophys. Res., 113, B07209. 5. Conclusions • The seismic tomography is consistent with a 13 km 3 magma chamber formed between 5.5 and at least 7.5 km depth (Model B, Fig 4E-H). • The volume could be smaller (~5 km 3 ), with a different melt geometry, or larger (~18 km 3 ), if it extended beyond 7.5 km depth. • Shallow magma chambers are transient and can form in a few thousand years. • Integration of tomography with numerical models and other constraints reveals the details of the magmatic system with unprecedented detail (Fig. 5). Fig. 2. Tomographic model. (A,B,C) W-E vertical sections through the three volcanic centres of (A) SHV, (B) Centre Hills (CH) and (C) Silver Hills (SH). (D) S-N section. (E,F) Horizontal sections at 2 and 7 km depth respectively. The three volcanoes have a shallow structure consisting of high- velocity cores (marked with white dashed contours) and low-velocity aprons (dark grey dashed ). The LVV (marked in red) is well resolved at up to 7.5 km depth. Fig. 1. Recording array and shooting geometry. The data collected include the recordings of over 600 km of active source seismic data on an array of land and ocean bottom seismometers. DEM from [2]. Shooting track in red. The black dashed lines mark the sections shown in Fig. 2. 2. Seismic tomography Our analysis is based on active-source wide-angle seismic data collected on and around Montserrat in December 2007 as part of the SEA-CALIPSO project (Fig. 1). The 3D tomographic model (Fig. 2) was obtained by regularised inversion of first arrival travel-times [3] and covers an area of 45x50 km. The resolution length was assessed by calculating the resolvability of checkerboard patterns [4]. A low-velocity volume (LVV), at 4-8 km depth, corresponds to the shallow magma chamber. LVV LVV