Swiss Polar Institute

The unique conduction properties of condensed matter systems with topological order have recently inspired a quest for the similar effects in classical wave phenomena. Acoustic topological insulators, in particular, hold the promise to revolutionize our ability to control sound, allowing for large isolation in the bulk and broadband one-way transport along their edges, with topological immunity against structural defects and disorder. So far, these fascinating properties have been obtained relying on moving media, which may introduce noise and absorption losses, hindering the practical potential of topological acoustics. Here we overcome these limitations by modulating in time the acoustic properties of a lattice of resonators, introducing the concept of acoustic Floquet topological insulators. We show that acoustic waves provide a fertile ground to apply the anomalous physics of Floquet topological insulators, and demonstrate their relevance for a wide range of acoustic applications, including broadband acoustic isolation and topologically protected, nonreciprocal acoustic emitters.

A novel variable stiffness actuator composed of a dielectric elastomer actuator (DEA) and a low-melting-point-alloy (LMPA) embedded silicone substrate is demonstrated. The device which we call variable stiffness dielectric elastomer actuator (VSDEA) enables functional soft robots with a simplified structure, where the DEA generates a bending actuation and the LMPA provides controllable stiffness between soft and rigid states by Joule heating. The entire structure of VSDEA is made of soft silicones with an elastic modulus of less than 1 MPa providing a high compliance when the LMPA is active. The device has the dimension of 40 mm length × 10 mm width × 1 mm thickness, with mass of ~1 g. We characterize VSDEA in terms of the actuation stroke angle, the blocked force, and the reaction force against a forced displacement. The results show the controllable actuation angle and the blocked force up to 23.7 ° and 2.4 mN in the soft state, and 0.6 ° and 2.1 mN in the rigid state. Compared to an actuator without the LMPA, VSDEA exhibits ~90× higher rigidity. We develop a VSDEA gripper where the mass of active parts is ~2 g, which is able to successfully hold an object mass of 11 g, exhibiting the high performance of the actuator.

We review recent progress in miniaturized dielectric elastomer actuators, sensors and energy harvesters. We focus primarily on configurations where the large strain, high compliance, stretchability and high level of integration offered by dielectric elastomer transducers provide significant advantages over other mm or m-scale transduction technologies. We first present the most active application areas, including: tunable optics, soft robotics, haptics, micro fluidics, biomedical devices and stretchable sensors. We then discuss the fabrication challenges related to miniaturization, such as thin membrane fabrication, precise patterning of compliant electrodes, and reliable batch fabrication of multilayer devices. We finally address the impact of miniaturization on strain, force and driving voltage, as well as the important effect of boundary conditions on the performance of mm-scale DEAs.

Abstract Dense crossbar arrays of non-volatile memory (NVM) can potentially enable massively parallel and highly energy-efficient neuromorphic computing systems. The key requirements for the NVM elements are continuous (analog-like) conductance tuning capability and switching symmetry with acceptable noise levels. However, most NVM devices show non-linear and asymmetric switching behaviors. Such non-linear behaviors render separation of signal and noise extremely difficult with conventional characterization techniques. In this study, we establish a practical methodology based on Gaussian process regression to address this issue. The methodology is agnostic to switching mechanisms and applicable to various NVM devices. We show tradeoff between switching symmetry and signal-to-noise ratio for HfO 2 -based resistive random access memory. Then, we characterize 1000 phase-change memory devices based on Ge 2 Sb 2 Te 5 and separate total variability into device-to-device variability and inherent randomness from individual devices. These results highlight the usefulness of our methodology to realize ideal NVM devices for neuromorphic computing.

Temperature and nutrient supply interactively control phytoplankton growth and productivity, yet the role of these drivers together still has not been determined experimentally over large spatial scales in the oligotrophic ocean. We conducted four microcosm experiments in the tropical and subtropical Atlantic (29°N-27°S) in which surface plankton assemblages were exposed to all combinations of three temperatures (in situ, 3 °C warming and 3 °C cooling) and two nutrient treatments (unamended and enrichment with nitrogen and phosphorus). We found that chlorophyll a concentration and the biomass of picophytoplankton consistently increase in response to nutrient addition, whereas changes in temperature have a smaller and more variable effect. Nutrient enrichment leads to increased picoeukaryote abundance, depressed Prochlorococcus abundance, and increased contribution of small nanophytoplankton to total biomass. Warming and nutrient addition synergistically stimulate light-harvesting capacity, and accordingly the largest biomass response is observed in the warmed, nutrient-enriched treatment at the warmest and least oligotrophic location (12.7°N). While moderate nutrient increases have a much larger impact than varying temperature upon the growth and community structure of tropical phytoplankton, ocean warming may increase their ability to exploit events of enhanced nutrient availability.

During mitochondrial fission, key molecular and cellular factors assemble on the outer mitochondrial membrane, where they coordinate to generate constriction. Constriction sites can eventually divide or reverse upon disassembly of the machinery. However, a role for membrane tension in mitochondrial fission, although speculated, has remained undefined. We capture the dynamics of constricting mitochondria in mammalian cells using live-cell structured illumination microscopy (SIM). By analyzing the diameters of tubules that emerge from mitochondria and implementing a fluorescence lifetime-based mitochondrial membrane tension sensor, we discover that mitochondria are indeed under tension. Under perturbations that reduce mitochondrial tension, constrictions initiate at the same rate, but are less likely to divide. We propose a model based on our estimates of mitochondrial membrane tension and bending energy in living cells which accounts for the observed probability distribution for mitochondrial constrictions to divide.

Abstract Immersive virtual reality (IVR) offers possibilities of creating a learner‐centric environment that can provide more presence and engagement for students leading to an enhanced learning experience compared to conventional classroom practices. However, the potential of IVR in vocational education and training (VET) has not yet been explored in‐depth, and it is an open question of whether it can effectively support learner creation in a designing task. In this paper, we present an IVR application developed to support gardener apprentices in designing gardens. Using this application, we conducted an experimental study with gardener apprentices to investigate the effect of the IVR interface compared to paper sketching and learner behavior on the proportion, composition, and creativity of the design outcome. Additionally, we investigated how it can be combined with a paper sketching activity to improve its effectiveness. Our analysis shows that the IVR interface can be more effective for the proportion aspect, but this may be limited to students that are able to use it after working with paper. In terms of the combination order, the effectiveness of IVR on the design quality was improved when it was carried out after the paper sketching and this ordering produced a more effective outcome for the proportion and composition aspects. Finally, our results show that IVR design quality is related to learner behaviors such as the time spent on designing and the number of simulations used. This study demonstrates the effectiveness of IVR applications in supporting designing skills and how effectiveness can be improved by combining it with a conventional method of practice. Practitioner Notes What is already known about this topic The effectiveness of IVR has been reported in a variety of educational settings. IVR enhances learning through situated experience, multiple perspectives, and simulations that transfer to the real world. Most IVR applications for VET have focused on passive learning or basic procedural skills. What this paper adds Design and implementation of an IVR application in VET to support design‐related learning. The use of IVR may have a significant impact on the quality of design outcomes. The use of IVR after paper sketching may increase the quality of designs during IVR use. Implications for practice and/or policy IVR can enhance the design outcome in terms of the proportion aspect, but this may be limited to students that are able to use it after working with paper. The design quality can be affected by the combination order between IVR and paper sketching where the effectiveness of IVR can be improved when it is carried out after paper sketching. The behavior of learners while using IVR can be used as indicators of the design quality.

While it has been recently demonstrated that both iron (Fe) and manganese (Mn) control Southern Ocean (SO) plankton biomass, how in particular Mn governs phytoplankton species composition remains yet unclear. This study, for the first time, highlights the importance of Mn next to Fe for growth of two key SO phytoplankton groups at two locations in the Drake Passage (West and East). Even though the bulk parameter chlorophyll a indicated Fe availability as main driver of both phytoplankton assemblages, the flow cytometric and microscopic analysis revealed FeMn co-limitation of a key phytoplankton group at each location: at West the dominant diatom Fragilariopsis and one subgroup of picoeukaryotes, which numerically dominated the East community. Hence, the limitation by both Fe and Mn and their divergent requirements among phytoplankton species and groups can be a key factor for shaping SO phytoplankton community structure.

Iron (Fe) is an essential micronutrient that limits primary productivity throughout the surface of the Southern Ocean. Here, we present the first high-resolution depth profiles for dissolved Fe and Fe isotope ratios (δ56Fe) from all major zones of the Southern Ocean, collected during the Antarctic Circumnavigation Expedition in austral summer 2017. Open-ocean surface waters are characterized by remarkably high δ56Fe values (up to +1.6‰) and very low Fe concentrations (<0.05 nmol kg−1). We attribute the elevated δ56Fe values above the ferricline to the effect of continuous shallow cycling processes (uptake, recycling, and binding of Fe to organic ligands), with only a very limited resupply of Fe from below. Below the ferricline, δ56Fe values approach ∼0‰ and remain constant down to our deepest samples at 1000 m, with no obvious isotope signal from regeneration. This overall pattern in δ56Fe is modified near islands, continental shelves and hydrothermal vents, where distinct δ56Fe signatures are associated with different Fe sources. Near the volcanic Balleny Islands, elevated surface Fe concentrations associated with low δ56Fe are indicative of reductive release of isotopically light Fe from sediments. Elevated δ56Fe values at depth near the Balleny seamount chain and near the East Scotia Arc may reflect distal hydrothermal influences, caused by fractionation associated with precipitation or the loss of specific phases of Fe during long-range transport. Sedimentary sources of isotopically light Fe on the Antarctic Peninsula are important for shelf waters. Long-distance transport of this sediment-derived Fe and its influence on surface waters are strongly dependent on the regional circulation, and may ultimately be the source of light Fe previously observed within Antarctic Intermediate Water in the Atlantic sector of the Southern Ocean.

Phytoplankton photosynthetic physiology can be investigated through single-turnover variable chlorophyll fluorescence (ST-ChlF) approaches, which carry unique potential to autonomously collect data at high spatial and temporal resolution. Over the past decades, significant progress has been made in the development and application of ST-ChlF methods in aquatic ecosystems, and in the interpretation of the resulting observations. At the same time, however, an increasing number of sensor types, sampling protocols, and data processing algorithms have created confusion and uncertainty among potential users, with a growing divergence of practice among different research groups. In this review, we assist the existing and upcoming user community by providing an overview of current approaches and consensus recommendations for the use of ST-ChlF measurements to examine in-situ phytoplankton productivity and photo-physiology. We argue that a consistency of practice and adherence to basic operational and quality control standards is critical to ensuring data inter-comparability. Large datasets of inter-comparable and globally coherent ST-ChlF observations hold the potential to reveal large-scale patterns and trends in phytoplankton photo-physiology, photosynthetic rates and bottom-up controls on primary productivity. As such, they hold great potential to provide invaluable physiological observations on the scales relevant for the development and validation of ecosystem models and remote sensing algorithms.

Abstract. Air quality measures that were implemented in Europe in the 1990s resulted in reductions of ozone precursor concentrations. In this study, the effect of these reductions on ozone is investigated by analyzing surface measurements of this pollutant for the time period between 2000 and 2015. Using a nonparametric timescale decomposition methodology, the long-term, seasonal and short-term variation in ozone observations were extracted. A clustering algorithm was applied to the different timescale variations, leading to a classification of sites across Europe based on the temporal characteristics of ozone. The clustering based on the long-term variation resulted in a site-type classification, while a regional classification was obtained based on the seasonal and short-term variations. Long-term trends of deseasonalized mean and meteo-adjusted peak ozone concentrations were calculated across large parts of Europe for the time period 2000–2015. A multidimensional scheme was used for a detailed trend analysis, based on the identified clusters, which reflect precursor emissions and meteorological influence either on the inter-annual or the short-term timescale. Decreasing mean ozone concentrations at rural sites and increasing or stabilizing at urban sites were observed. At the same time, downward trends for peak ozone concentrations were detected for all site types. In addition, a reduction of the amplitude in the seasonal cycle of ozone and a shift in the occurrence of the seasonal maximum towards earlier time of the year were observed. Finally, a reduced sensitivity of ozone to temperature was identified. It was concluded that long-term trends of mean and peak ozone concentrations are mostly controlled by precursor emissions changes, while seasonal cycle trends and changes in the sensitivity of ozone to temperature are among other factors driven by regional climatic conditions.

The goal of these review papers is to summarize the recent advances in the metallurgy of aluminium alloys and from this analysis, to try to outline future developments in this field. Part I deals with the transformation of aluminium alloys from the liquid to the solid state, while Part II will focus on solid-state transformations. These papers are by no means exhaustive since the literature is very abundant, but the authors wish to give a personal view of what they think are the most relevant scientific contributions that can impact future technological developments.

Over the last decades, it has been reported that the habitat of the Southern Ocean (SO) key species Antarctic krill (Euphausia superba) has contracted to high latitudes, putatively due to reduced winter sea ice coverage, while salps as Salpa thompsoni have extended their dispersal to the former krill habitats. To date, the potential implications of this population shift on the biogeochemical cycling of the limiting micronutrient iron (Fe) and its bioavailability to SO phytoplankton has never been tested. Based on uptake of fecal pellet (FP)-released Fe by SO phytoplankton, this study highlights how efficiently krill and salps recycle Fe. To test this, we collected FPs of natural populations of salps and krill, added them to the same SO phytoplankton community, and measured the community's Fe uptake rates. Our results reveal that both FP additions yielded similar dissolved iron concentrations in the seawater. Per FP carbon added to the seawater, 4.8 ± 1.5 times more Fe was taken up by the same phytoplankton community from salp FP than from krill FP, suggesting that salp FP increased the Fe bioavailability, possibly through the release of ligands. With respect to the ongoing shift from krill to salps, the potential for carbon fixation of the Fe-limited SO could be strengthened in the future, representing a negative feedback to climate change.

The international Antarctic Circumnavigation Expedition (ACE), organised by the Swiss Polar Institute (SPI), took place in the austral summer of 2016 / 17. Departing from Cape Town in December 2016, the R/V Akademik Tryoshnikov sailed around the Southern Ocean, navigating the following route: Cape Town (20 December 2016) - Hobart (18 January 2017) via: Marion Island Iles de Crozet et Kerguelen Heard Island Hobart (22 January 2017) - Punta Arenas (22 February 2017) via: Mertz glacier Balleny Islands Mount Siple Scott Island Peter I Island Diego Ramírez Punta Arenas (26 February 2017) - Cape Town (19 March 2017) via: South Georgia South Sandwich Islands Bouvetøya Scientists from 22 distinct projects studied a wide range of disciplines, collecting data and samples from the Southern Ocean and a number of terrestrial sites on islands around Antarctica, as well as the continent itself. In this cruise report, the full details of the purpose of the cruise, on-board data and sample collection, and project methods and preliminary results are described.

Abstract In recent years, the conjecture on the instability of Anti-de Sitter spacetime, put forward by Dafermos–Holzegel (Dynamic instability of solitons in 4 + 1 dimesnional gravity with negative cosmological constant, 2006. https://www.dpmms.cam.ac.uk/~md384/ADSinstability.pdf ) and Dafermos (The Black Hole Stability problem, Talk at the Newton Institute, Cambridge, 2006. http://www-old.newton.ac.uk/webseminars/pg+ws/2006/gmx/1010/dafermos/ ) in 2006, has attracted a substantial amount of numerical and heuristic studies. Following the pioneering work (Phys Rev Lett 107(3):031102, 2011) of Bizon–Rostworowski, research efforts have been mainly focused on the study of the spherically symmetric Einstein-scalar field system. The first rigorous proof of the instability of AdS in the simplest spherically symmetric setting, namely for the Einstein-null dust system , was obtained in Moschidis (A proof of the instability of AdS for the Einstein-null dust system with an inner mirror, 2017. arXiv:1704.08681 ). In order to circumvent problems associated with the trivial break down of the Einstein-null dust system occuring at the center $$r=0$$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mrow> <mml:mi>r</mml:mi> <mml:mo>=</mml:mo> <mml:mn>0</mml:mn> </mml:mrow> </mml:math> , Moschidis (2017) studied the evolution of the system in the exterior of an inner mirror placed at $$r=r_{0}$$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mrow> <mml:mi>r</mml:mi> <mml:mo>=</mml:mo> <mml:msub> <mml:mi>r</mml:mi> <mml:mn>0</mml:mn> </mml:msub> </mml:mrow> </mml:math> , $$r_{0}&gt;0$$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mrow> <mml:msub> <mml:mi>r</mml:mi> <mml:mn>0</mml:mn> </mml:msub> <mml:mo>&gt;</mml:mo> <mml:mn>0</mml:mn> </mml:mrow> </mml:math> . However, in view of additional considerations on the nature of the instability, it was necessary for Moschidis (2017) to allow the mirror radius $$r_{0}$$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:msub> <mml:mi>r</mml:mi> <mml:mn>0</mml:mn> </mml:msub> </mml:math> to shrink to 0 with the size of the initial perturbation; well-posedness in the resulting complicated setup (involving low-regularity estimates of uniform modulus with respect to $$r_{0}$$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:msub> <mml:mi>r</mml:mi> <mml:mn>0</mml:mn> </mml:msub> </mml:math> ) was obtained in Moschidis (The Einstein-null dust system in spherical symmetry with an inner mirror: structure of the maximal development and Cauchy stability, 2017. arXiv:1704.08685 ). In this paper, we establish the instability of AdS for the Einstein-massless Vlasov system in spherical symmetry; this will be the first proof of the AdS instability conjecture for an Einstein-matter system which is well-posed for regular initial data in the standard sense, without the addition of an inner mirror . The necessary well-posedness results for this system are obtained in our companion paper (Moschidis in The characteristic initial-boundary value problem for the Einstein-massless Vlasov system in spherical symmetry, 2018. arXiv:1812.04274 ). Our proof utilises an instability mechanism based on beam interactions which is superficially similar to the one appearing in Moschidis (A proof of the instability of AdS for the Einstein-null dust system with an inner mirror, 2017. arXiv:1704.08681 ). However, new difficulties associated with the Einstein-massless Vlasov system (such as the need for control on the paths of non-radial geodesics in a large curvature regime) will force us to develop a different strategy of proof involving a novel configuration of beam interactions. One of the main novelties of our construction is the introduction of a multi-scale hierarchy of domains in phase space, on which the initial support of the Vlasov field f is localised. The propagation of this hierarchical structure of the support of f along the evolution will be crucial both for controlling the geodesic flow under minimal regularity assumptions and for guaranteeing the existence of the solution until the time of trapped surface formation.

Isoprene is a biogenic trace gas produced by terrestrial vegetation and marine phytoplankton. In the remote oceans, where secondary aerosols are mostly biogenic, marine isoprene emissions affect atmospheric chemistry and influence cloud formation and brightness. Here, we present the first compilation of new and published measurements of isoprene concentrations in the Southern Ocean and explore their distribution patterns. Surface ocean isoprene concentrations in November through April span 1 to 94 pM. A band of higher concentrations is observed around a latitude of ≈40 ∘ S and a surface sea temperature of 15 ∘ C. High isoprene also occurs in high productivity waters near islands and continental coasts. We use concurrent measurements of physical, chemical, and biological variables to explore the main potential drivers of isoprene concentration by means of paired regressions and multivariate analysis. Isoprene is best explained by phytoplankton-related variables like the concentrations of chlorophyll-a, photoprotective pigments and particulate organic matter, photosynthetic efficiency (influenced by iron availability), and the chlorophyll-a shares of most phytoplankton groups, and not by macronutrients or bacterial abundance. A simple statistical model based on chlorophyll-a concentration and a sea surface temperature discontinuity accounts for half of the variance of isoprene concentrations in surface waters of the Southern Ocean.

Abstract. Understanding the dynamics of marine phytoplankton productivity requires mechanistic insight into the non-linear coupling of light absorption, photosynthetic electron transport and carbon fixation in response to environmental variability. In the present study, we examined the variability of phytoplankton light absorption characteristics, light-dependent electron transport and 14C-uptake rates over a 48 h period in the coastal subarctic north-east (NE) Pacific. We observed an intricately coordinated response of the different components of the photosynthetic process to diurnal irradiance cycles, which acted to maximize carbon fixation, while simultaneously preventing damage by excess absorbed light energy. In particular, we found diurnal adjustments in pigment ratios, excitation energy transfer to reaction centre II (RCII), the capacity for non-photochemical quenching (NPQ), and the light efficiency (α) and maximum rates (Pmax) of RCII electron transport (ETRRCII) and 14C uptake. Comparison of these results from coastal waters to previous observations in offshore waters of the subarctic NE Pacific provides insight into the effects of iron limitation on the optimization of photosynthesis. Under iron-limited, low-biomass conditions, there was a significant reduction of iron-rich photosynthetic units per chlorophyll a, which was partly offset by higher light absorption and electron transport per photosystem II (PSII). Iron deficiency limited the capacity of phytoplankton to utilize peak midday irradiance for carbon fixation and caused an upregulation of photoprotective mechanisms, including NPQ, and the decoupling of light absorption, electron transport and carbon fixation. Such decoupling resulted in an increased electron requirement (Φe,C) and decreased quantum efficiency (ΦC) of carbon fixation at the iron-limited station. In both coastal and offshore waters, Φe,C and ΦC correlated strongly to NPQ, albeit with a significantly different slope. We discuss the implications of our results for the interpretation of bio-optical data and the parameterization of numerical productivity models, both of which are vital tools in monitoring marine photosynthesis over large temporal and spatial scales.

Abstract. Evaluation of photosynthetic competency in time and space is critical for better estimates and models of oceanic primary productivity. This is especially true for areas where the lack of iron (Fe) limits phytoplankton productivity, such as the Southern Ocean. Assessment of photosynthetic competency on large scales remains challenging, but phytoplankton chlorophyll a fluorescence (ChlF) is a signal that holds promise in this respect as it is affected by, and consequently provides information about, the photosynthetic efficiency of the organism. A second process affecting the ChlF signal is heat dissipation of absorbed light energy, referred to as non-photochemical quenching (NPQ). NPQ is triggered when excess energy is absorbed, i.e. when more light is absorbed than can be used directly for photosynthetic carbon fixation. The effect of NPQ on the ChlF signal complicates its interpretation in terms of photosynthetic efficiency, and therefore most approaches relating ChlF parameters to photosynthetic efficiency seek to minimize the influence of NPQ by working under conditions of sub-saturating irradiance. Here, we propose that NPQ itself holds potential as an easily acquired optical signal indicative of phytoplankton physiological state with respect to Fe limitation. We present data from a research voyage to the Subantarctic Zone south of Australia. Incubation experiments confirmed that resident phytoplankton were Fe-limited, as the maximum quantum yield of primary photochemistry, Fv∕Fm, measured with a fast repetition rate fluorometer (FRRf), increased significantly with Fe addition. The NPQ “capacity” of the phytoplankton also showed sensitivity to Fe addition, decreasing with increased Fe availability, confirming previous work. The fortuitous presence of a remnant warm-core eddy in the vicinity of the study area allowed comparison of fluorescence behaviour between two distinct water masses, with the colder water showing significantly lower Fv∕Fm than the warmer eddy waters, suggesting a difference in Fe limitation status between the two water masses. Again, NPQ capacity measured with the FRRf mirrored the behaviour observed in Fv∕Fm, decreasing as Fv∕Fm increased in the warmer water mass. We also analysed the diel quenching of underway fluorescence measured with a standard fluorometer, such as is frequently used to monitor ambient chlorophyll a concentrations, and found a significant difference in behaviour between the two water masses. This difference was quantified by defining an NPQ parameter akin to the Stern–Volmer parameterization of NPQ, exploiting the fluorescence quenching induced by diel fluctuations in incident irradiance. We propose that monitoring of this novel NPQ parameter may enable assessment of phytoplankton physiological status (related to Fe availability) based on measurements made with standard fluorometers, as ubiquitously used on moorings, ships, floats and gliders.

The massive amounts of digitized historical documents acquired over the last decades naturally lend themselves to automatic processing and exploration. Research work seeking to automatically process facsimiles and extract information thereby are multiplying with, as a first essential step, document layout analysis. If the identification and categorization of segments of interest in document images have seen significant progress over the last years thanks to deep learning techniques, many challenges remain with, among others, the use of finer-grained segmentation typologies and the consideration of complex, heterogeneous documents such as historical newspapers. Besides, most approaches consider visual features only, ignoring textual signal. In this context, we introduce a multimodal approach for the semantic segmentation of historical newspapers that combines visual and textual features. Based on a series of experiments on diachronic Swiss and Luxembourgish newspapers, we investigate, among others, the predictive power of visual and textual features and their capacity to generalize across time and sources. Results show consistent improvement of multimodal models in comparison to a strong visual baseline, as well as better robustness to high material variance.

Abstract Fluid injection in deep geological formations usually induces microseismicity. In particular, industrial‐scale injection of CO 2 may induce a large number of microseismic events. Since CO 2 is likely to reach the storage formation at a lower temperature than that corresponding to the geothermal gradient, both overpressure and cooling decrease the effective stresses and may induce microseismicity. Here, we investigate the effect of the stress regime on the effective stress evolution and fracture stability when injecting cold CO 2 through a horizontal well in a deep saline formation. Simulation results show that when only overpressure occurs, the vertical total stress remains practically constant, but the horizontal total stresses increase proportionally to overpressure. These hydro‐mechanical stress changes result in a slight improvement in fracture stability in normal faulting stress regimes because the decrease in deviatoric stress offsets the decrease in effective stresses produced by overpressure. However, fracture stability significantly decreases in reverse faulting stress regimes because the size of the Mohr circle increases in addition to being displaced towards failure conditions. Fracture stability also decreases in strike slip stress regimes because the Mohr circle maintains its size and is shifted towards the yield surface a magnitude equal to overpressure minus the increase in the horizontal total stresses. Additionally, cooling induces a thermal stress reduction in all directions, but larger in the out‐of‐plane direction. This stress anisotropy causes, apart from a displacement of the Mohr circle towards the yield surface, an increase in the size of the Mohr circle. These two effects decrease fracture stability, resulting in the strike slip being the least stable stress regime when cooling occurs, followed by the reverse faulting and the normal faulting stress regimes. Thus, characterizing the stress state is crucial to determine the maximum sustainable injection pressure and maximum temperature drop to safely inject CO 2 .