Physical and mechanical rock properties of a heterogeneous volcano: the case of Mount Unzen, Japan

Abstract

Volcanoes represent one of the most critical geological settings for hazard modelling due to their propensity to both unpredictably erupt and collapse, even in times of quiescence. Volcanoes are heterogeneous at multiple scales, from porosity, which is variably distributed and frequently anisotropic, to strata, which are laterally discontinuous and commonly pierced by fractures and faults. Due to variable and, at times, intense stress and strain conditions during and following emplacement, volcanic rocks span an exceptionally wide range of physical and mechanical properties. Understanding the constituent materials' attributes is key to improving the interpretation of the hazards posed by the diverse array of volcanic complexes. Here, we examine the spectrum of physical and mechanical properties presented by a single dome-forming eruption at a dacitic volcano, Mount Unzen (Japan), by testing a number of isotropic and anisotropic lavas in tension and compression with acoustic emission (AE) monitoring. The lava dome erupted as a series of 13 lobes between 1991 and 1995, and its ongoing instability means that much of the volcano and its surroundings remain within an exclusion zone today. During a field campaign in 2015, we selected four representative blocks as the focus of this study. The core samples from each block span a range in total porosity from 9.14 % to 42.81 % and a range in permeability from 1.65×10-15 to 1.88×10-9 m2 (from 1065 measurements). For a given porosity, sample permeability varies by >2 orders of magnitude and is typically lower for macroscopically anisotropic samples than for isotropic samples of similar porosity. An additional 379 permeability measurements on planar surfaces of both an isotropic and anisotropic sample block showed consistent minimum, maximum, and average permeabilities, and comparable standard deviations to measurements on core and disc samples; this indicated a negligible impact of sample size on recorded permeability across the range of sample sizes and absolute permeabilities tested. Permeability measured under confined conditions showed that the lowest permeability samples, whose porosity largely comprises microfractures, are most sensitive to effective pressure and that anisotropy of permeability is enhanced by confinement. The permeability measurements highlight the importance of the measurement approach, scale, and confinement conditions in the description of permeability. The uniaxial compressive strength (UCS) ranges from 13.48 to 47.80 MPa, and tensile strength (UTS) using the Brazilian disc method ranges from 1.30 to 3.70 MPa, with crack-dominated lavas being weaker than vesicle-dominated materials of equivalent porosity. UCS is lower in saturated conditions, whereas the impact of saturation on UTS is variable. UCS is between 6.8 and 17.3 times higher than UTS, with anisotropic samples forming each endmember. The Young's modulus of dry samples ranges from 4.49 to 21.59 GPa and is systematically reduced in water-saturated tests. The interrelation of porosity, UCS, UTS, and Young's modulus was modelled with good replication of the data, and empirical relationships are provided. Acceleration of monitored acoustic emission (AE) rates during deformation was assessed by fitting Poisson point process models in a Bayesian framework. An exponential acceleration model closely replicated the tensile strength tests, whilst compressive tests tended to have relatively high early rates of AEs, suggesting failure forecast may be more accurate in tensile regimes, though with shorter warning times. The Gutenberg–Richter b value has a negative correlation with connected porosity for both UCS and UTS tests which we attribute to different stress intensities caused by differing pore networks. The b value is higher for UTS than UCS, and it typically decreases (positive Δb) during tests, with the exception of cataclastic samples in compression. Δb correlates positively with connected porosity in compression and correlates negatively in tension. Δb using a fixed sampling length may be a more useful metric for monitoring changes in activity at volcanoes than the b value with an arbitrary starting point. Using coda wave interferometry (CWI), we identify velocity reductions during mechanical testing in compression and tension, the magnitude of which is greater in more porous samples in UTS but independent of porosity in UCS and which scales to both b value and Δb. Yet, saturation obscures velocity changes caused by evolving material properties, which could mask damage accrual or source migration in water-rich seismogenic environments such as volcanoes. The results of this study highlight that heterogeneity and anisotropy within a single system not only add variability but also have a defining role in the channelling of fluid flow and localisation of strain that dictate a volcano's hazards and the geophysical indicators we use to interpret them.