Challenges of Tunnelling in Volcanic Rock Masses

Volcanic rock masses exhibit temporal and spatial variability, even at the scale and duration of engineering projects. Volcanic processes are dynamic, resulting in rock masses ranging from high-porosity, clay-rich, fractured, and soil-like to low-porosity, high-strength, brittle, and massive. Based on a number of studies in a variety of geological settings, such as active and fossil geothermal systems, on the surface of active volcanoes and up to 3000 m below the surface, the work presented in this article shows the relationship between geological characteristics and mechanical parameters of volcanic rocks. These are then linked to the resultant challenges to tunnelling associated with the mechanical behaviour of volcanic rocks and rock masses, ranging from ductile failure such as squeezing and swelling to dynamic failure such as spalling and rockburst. This article highlights some of the key parameters that should be incorporated in site and laboratory investigations to build representative ground models in volcanic rocks and rock masses. Rock mass characterisation needs to address the highly variable and anisotropic nature of volcanic rocks, ranging from millimetre to decametre scale. Ground models must include not only the mechanical properties, such as strength and stiffness, of typical lab investigations, but also petrophysical properties, such as porosity, and geological conditions, such as alteration. Geomechanical characterisation of these rock masses requires an understanding of geological processes to select appropriate field, lab and design tools. In volcanic rocks, perhaps more than any other rock types, the geology is critical to characterising and understanding the behaviour in response to tunnelling.


Characteristics of Volcanic Rocks and Rock Masses
It is reasonably well understood that composition and texture will affect rock physical and mechanical properties (e.g. [1]). Because of the dynamic systems in which volcanic rock masses are typically generated (Fig. 1), composition and texture are highly spatially and temporally variable, with interplay between primary composition and texture, as well as external processes, leading to a variety of different secondary compositions and textures (Fig. 2). In hydrothermal systems, often associated with volcanic environments, the temperature of the hydrothermal fluids plays a key role in the type of alteration, and hence the porosity and secondary mineralisation, that will occur (Fig. 3). Alteration can affect porosity by infilling pores and fractures as well as dissolving primary minerals [2], leading to high compositional and textural variability. This macro-scale variability is associated with variability in the mechanical properties of volcanic rocks, in partic- Fig. 1: Example2-Dschematics-ageologicalground modelshowing materialpropertiesand volcanicand hydrologicalprocessesin Whakaarivolcano, New Zealand [3]; b Conceptual model of a conventional, hot, liquid dominated geothermal field. The model has been split into the alteration zones typical for a geothermal field, with temperature profiles and surface expressions [4]. Smectitic alteration refers to low temperature alteration dominated by smectite clays, while propylitic alteration refers to high temperature alteration dominated by chlorite, epidote and quartz. Argillic alteration occurs between the two, and is also clay dominated, but in this case with illite clays ular with respect to differences in porosity and secondary mineralogy associated with alteration. Porosity describes the proportion of voids in a rock, and for a similar mineralogy, porosity will have a direct relationship with density. The presence of pore (connected and not connected) negatively impacts uniaxial compressive strength (Fig. 4) and Young's modulus (Fig. 5), two of the key mechanical parameters used for excavation design. These two relationships are well established in the rock physics and rock mechanics literature, as highlighted by [6]. Both relationships show high variability, which demonstrates that, while porosity is important, it is not the only characteristic affecting these parameters. The variability is particularly high at low porosity, suggesting that, at low porosity, the other geological characteristics are more important than porosity. Conversely, Poisson's ratio is not sensitive to porosity [7].
While the elastic moduli and intact compressive strength are important for tunnel design using analytical techniques, numerical modelling of tunnel behaviour requires full fail-ure criteria, such as Mohr-Coulomb or Hoek-Brown. The failure criteria are also highly dependent on the texture, porosity and alteration of volcanic rocks (Fig. 6). To build failure criteria, the Hoek-Brown parameter, mi, and/or the Mohr-Coulomb parameters, cohesion and friction angle, are derived from curve-fitting the failure criterion to triaxial test data [11]. Triaxial testing is not always possible during site investigation for tunnel design and the Hoek-Brown parameter mi, which controls the steepness and curvature of the failure envelope, is often extracted from tables in the literature [12]. demonstrate that the commonly used tabulated mi values (e.g. those based on Hoek and Brown, 1997 [11]) fail to capture the wide variability in these values, especially in volcanic rocks. Read and Richards (2011) [12] also highlight that different mi will result in vastly different failure criteria even with the same intact uniaxial compressive strength (Fig. 7). [13] and [14] provide transfer functions to estimate mi, as well as cohesion and friction angle, in the   [15]). UBLM unaltered brecciated lava margin (high porosity), ABLM altered brecciated lava margin (moderate porosity), UDCL unaltered densecoherentlava(very low porosity), ADCL altered dense coherent lava (moderate porosity), UI unaltered intrusion (very low porosity), AI altered intrusion (low porosity), arrows show the change in the failure criterion from the unaltered to the altered rock Fig. 7: Example of differences in Hoek-Brown failure criteria (in principal stress space) for two ignimbrites with the same uniaxial compressive strength [12] absence of laboratory triaxial data using porosity, an easily and commonly measured value (Fig. 8). [16] demonstrated that geothermal system exploitation can sufficiently disturb the induced stresses through drilling, fluid extraction and injection that, depending on their strength, the rocks near the wellbore can either compact, thereby causing large deformations, or be induced to tensile failure, possibly leading to instability (Fig. 9). Tunnelling will induce similar stress changes by inducing zones of increased and decreased stress at the tunnel face, crown, sides and invert, and further changes to effective stress will occur if the rock masses are dewatered during excavation.

Mechanical Behaviour of Volcanic Rocks and Rock Masses
What is important to consider is that depending on the texture, porosity and alteration, different rock units will respond differently to these induced stresses, depending on the magnitude of the induced stresses compared to the strength. [17] show that altered volcanic and volcano-sedimentary rocks contain clay minerals, such as smectites and zeolites, which can lead to significant swelling potential. They suggest that simple tests, such as the ethylene glycol durability index, can be useful to identify rocks with potential for swelling, which could then be further tested using the oedometer swelling test to estimate swelling pressure.
The rapidly changing lithologies, often resulting in weak, soft lithologies adjacent to strong, stiff lithologies can lead to various design and construction issues, including those associated with mixed-face and mixed-wall conditions, such as heterogeneous deformation, blocky ground, and stress concentration in strong, stiff rock masses. While swelling and squeezing are clear challenges in weak, porous and altered volcanic rocks, spalling and strain burst are also key challenges in strong, massive volcanic rocks at high depth depending on stress magnitude and stress ratios, as observed at the Olmos tunnel in Peru [18]. Particularly relevant to volcanic rock masses is their commonly layered nature (Fig. 2), which can result in layer buckling and rock fall under low stress conditions, or buckling and unpredictable dynamic release under high stress [19].  178. Initial stress state represents the undisturbed reservoir conditions; reinjection and extraction induced stresses represent the perturbed reservoir stresses resulting from changes in pore pressures during reservoir utilisation [16]. Analogies can be made to induced stresses in tunnel excavation

Site Investigation and Design in Volcanic
Rocks and Rock Masses [20] highlighted many lessons learned from the Majes tunnelling project in Peru, and suggest that a qualified geologist should be involved in the site investigation to correctly identify the different lithologies and their associated negative behaviours. [21] also highlight that careful site investigation is needed to determine as closely as possible the different rock and rock mass types that will be encountered by the tunnel. As demonstrated in the previous sections, volcanic rocks and rock masses are highly variable, in particular with respect to macro texture, porosity, and composition. Laboratory experiments must be carefully planned and undertaken to characterise as many of the anticipated lithologies as possible since empirical and published values cannot capture this variability. The work regarding the impact of porosity on mechanical parameters provides a means for estimating these parameters using simple porosity measurements [20]. also show the potential for originally dry units to be wetted during and after excavation, with resulting slaking and swelling. They suggest that all lithologies should be tested for swelling and slaking, regardless of expected water ingress. [21] show that tunnelling in volcanic rock masses requires modification to existing techniques to take into consideration the particular characteristics associated with volcanic processes. For example, rock mass classification, such as the Q system, needs to be adjusted to reduce the impact of columnar jointing in basalt lava, and transitioning away from rock mass classification in highly porous or poorly lithified rock masses, such as scoria and unwelded tuff. [22] also conclude that rock mass classification systems should only be used with caution in these rock masses, and that while many rock units are not sampleable for laboratory testing, they must be taken into account in the design. [23] also say that rock mass classification schemes need to be modified for volcanic rock masses, and that advance borings in the tunnel face are an important tool for reducing the risk of encountering unexpected poor geological conditions.
Adaptable tunnelling methods, whether conventional or mechanised, are best suited to the highly variable geological conditions as highlighted by [24]. These methods must be capable of mitigating behaviour ranging from high plastic deformations to dynamic brittle rupture, often within close succession. Support types also need to consider the wide range of expected behaviour, particular the potential for high pressures on the lining.

Funding. Open access funding provided by Montanuniversität Leoben.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permis-