Micro to Macroscale: the three-dimensional network characteristics of serpentinite dehydration veins

On Earth, subduction zones facilitate the cycling of volatiles between the Earth’s surface and interior. Volatile cycling has significant effects on the long-term state of the Earth’s climate and tectono-magmatic events, including volcanism and earthquakes. A key stage in the volatile cycle is the devolatilization of the subducting oceanic lithosphere, in which volatiles can escape the previously hydrated rocks. However, it is not well known how efficiently volatiles are transported at this stage. To better understand how volatiles escape at these conditions, we have analyzed the dehydration-related vein networks of the Erro-Tobbio meta-serpentinites (ET-MS), Italy. The ET-MS display well preserved networks of metamorphic olivine veins. These veins are the result of the dehydration reaction of antigorite and brucite to produce H2O and olivine. However, due to the low permeability of serpentinite at depth, the dehydration reaction requires the formation of self-organizing vein networks to allow the produced fluid to escape [1]. Thus, the metamorphic olivine veins in ET-MS may be used as a proxy for fluid flow pathways. We took a multiscale approach to analyzing the network architectures. For microscale (~16 µm voxel size) and mesoscale (~200 µm voxel size) resolutions, X-ray tomography methods are sufficient to visualize the three-dimensional structure of the networks. However, for large scale observations these methods are inapplicable. To solve this, we apply a novel workflow to analyze outcrop scale (~10 m) network systems in three dimensions using only two-dimensional data. By training a generative adversarial network (GAN) with two-dimensional data conditioned by spatial orientation, we can generate statistically representative three-dimensional networks that mimic those of the ET-MS. These representations also display similar characteristics in their respective pore-network-models. With this method, it is possible to produce reasonable three-dimensional approximations of the ET-MS vein networks using only photogrammetry data of the outcrops. In turn, this allows us to extract metrics, such as permeability, that describe the volatile transport efficiency of the ET-MS, and further, how these characteristics change at a broad range of scales.  

[1] Plümper et al. (2017) Nature Geoscience 10(2), 150-156.