Linking hydrothermal alteration and volcanic rock mechanics through VNIR-SWIR spectroscopy

Volcanoes are dynamic and complex natural systems, constantly changing through eruptions, alteration, and erosion. Hydrothermal systems are ubiquitous on volcanoes, causing physical and mechanical change to rock properties via hydrothermal alteration. The most commonly measured rock physical properties are porosity and uniaxial compressive strength (UCS) as they provide insight as to their history and potential mechanical behavior. Porosity and UCS are affected by the primary properties and emplacement history (e.g., volatile content, crystallinity, cooling rate, composition, fragmentation type) and the post-emplacement conditions (e.g., surface weathering and hydrothermal alteration). This creates highly heterogenous rock masses, with variation occurring across mm to m scales. Currently, destructive testing is required to measure UCS and porosity (destructive of the wider sample) accurately and needs a large volume of samples to capture the heterogeneity of volcanic rock masses. This testing is cost and time prohibitive, requiring large sample volume, in terms of sample size and number, which often require shipping to specialist labs. Schmidt hammers can be used to estimate UCS non-destructively, however, they produce inaccurate results on soft rocks such as hydrothermally altered rocks. Here, we present a new non-destructive method for predicting porosity and UCS across a range of volcanic rocks, from non-altered to highly altered.

This study uses visible-near infrared (VNIR) to shortwave infrared (SWIR) wavelengths (350-2500 nm) reflectance spectroscopy to predict porosity and UCS via Partial Least Squares Regression (PLSR). Reflectance spectroscopy is a non-destructive method that is sensitive to both physical (surface roughness and crystal/particle size) and chemical (mineral species and abundance) properties of volcanic rocks. Because these rock attributes also influence the physical and mechanical properties of rock, reflectance spectroscopy could be used to quantitatively predict porosity and UCS. This study used experimentally deformed volcanic rocks from Ruapehu, Ohakuri, Whakaari, and Banks Peninsula (New Zealand), Merapi (Indonesia), Chaos Crags (USA), Styrian Basin (Austria), La Soufrière de Guadeloupe (Eastern Caribbean), Volvic (France), and Cracked Mountain (Canada) to evaluate the accuracy of PLSR-based predictions for porosity and UCS. The training samples encompass a wide range of volcanoes, alteration degree (non-altered, silicic, argillic, and phyllic alteration), mineralogical differences (initial composition from basalt to rhyolite and alteration products), and textural differences (original textures such as lava and pyroclastic, and alteration textures including veins). Model sensitivity is evaluated by adding randomly individual samples to the training database or performing leave one group out cross validation based on characteristics (e.g., alteration mineral types, textural features, or volcano location). From this analysis, specific alteration mineralogy can be evaluated for its effect on porosity and UCS predictions such as the role of phyllosilicate formation causing a reduction in UCS. The proposed non-destructive method via VNIR-SWIR spectroscopy can complement existing rock mechanical testing methods to better quantify highly heterogenous volcanic and hydrothermal systems and their rock successions.