Now a team of researchers at the Rocket Force University of Engineering in Xi’an, China, reports a new artificial intelligence system that appears to crack much of that difficulty. Writing in the journal Natural Hazards and Earth System Sciences, Hongru Li, Xihai Li, and colleagues describe Infra-Net, a parallel decision-making network built around what they call multi-view feature learning. On a public benchmark of infrasound recordings, the system achieved a classification accuracy of 100 percent. On far messier field data supplied by the Comprehensive Nuclear-Test-Ban Treaty Organization (CTBTO), it reached 82.07 percent, a clear lead over competing methods tested on the same material. The work arrives at a moment when automated infrasound classification is becoming central both to disaster early warning and to the verification regime that underpins the nuclear test ban treaty.

The core insight behind Infra-Net is deceptively simple: when data are scarce, squeeze every drop of information out of what you have rather than trying to manufacture more. Major natural hazards are, by their nature, rare. Extreme events in particular yield so few recorded examples that conventional deep learning models, which typically hunger for thousands of training samples, struggle to generalize. Many earlier approaches have leaned on data augmentation techniques such as slicing and shifting signals to inflate the sample count, but the authors argue such tricks can distort the underlying physical distribution of the phenomena. Infra-Net takes a different route, extracting richer representations from each individual signal instead of multiplying them artificially.

The mathematical engine of the new system is the wavelet scattering transform, a technique that shares architectural DNA with convolutional neural networks but uses predefined wavelet filters rather than learned ones. Because those filters require no training, the method sidesteps overfitting in small-sample settings, a persistent weakness of deep networks. The scattering transform computes a semi-discrete wavelet decomposition of a signal and then applies a nonlinear modulus operation, producing features that are translation-invariant and stable under small deformations. That stability matters enormously for infrasound, whose signals are non-stationary and routinely warped by atmospheric turbulence on their way from source to sensor.

Li and colleagues add two twists of their own. First, they apply a logarithmic transformation to the scattering coefficients. Because infrasound frequencies sit between a few tenths of a hertz and a few hertz, the raw scattering features tend to have small magnitudes, and the logarithm, being monotonic but disproportionately sensitive to differences among small values, amplifies the subtle distinctions between, say, an earthquake’s extended rupture signal and a chemical explosion’s sharp impulsive burst. A one-way analysis of variance confirmed that the transformation significantly increased the statistical separability of the feature classes. Second, and more radically, the team abandons the standard practice of treating the full scattering matrix as a single input. Instead, each column of scattering coefficients, corresponding to a distinct set of wavelet scales and scattering paths, is fed into the network as an independent feature vector. A single 150-second recording thus becomes six parallel views of the same acoustic event, each carrying complementary physical information.

Those multiple views then flow into a dual-branch architecture. The first branch, dubbed GA-BiGRU, is a bidirectional gated recurrent unit network equipped with a global attention mechanism. Recurrent networks of this kind are built to capture long-range temporal dependencies in sequential data, and the simplified gating of the GRU keeps the parameter count low enough to converge on limited samples. The attention layer addresses a known weakness of recurrent models: the tendency of the final hidden state to forget early portions of a long sequence. By weighting every hidden state across the entire recording, the mechanism preserves information from beginning to end. The second branch, MSCI-Net, is a deliberately shallow convolutional network built around parallel multi-scale kernels. Small kernels catch fine-grained local fluctuations, while larger ones track energy trends over extended time windows, letting the network respond simultaneously to local anomalies and global patterns without the parameter bloat that deep architectures like VGG-16 impose.

The final piece is a confidence-based fusion module that merges the verdicts of the two branches. Rather than simply averaging their probability outputs, the module first averages the confidence scores across the six feature views within each branch, then takes an inner product of the two branches’ probability vectors, so that a final classification requires joint support from both independent evaluators. In visualized test cases where one branch was wrong and the other right, the fusion module consistently corrected the erroneous branch and delivered the correct answer. Ablation experiments showed that the dual-branch design outperformed either branch alone, and that the confidence-based fusion edged out naive averaging, with the improvement over the closest competing configuration holding up in statistical significance testing.

The benchmark results are striking. On the LOTIS library of typical infrasonic signals, recorded by an array at Windless Bight in Antarctica and containing aurora-driven atmospheric gravity waves, microbaroms, mountain-associated waves, and volcanic eruption infrasound, Infra-Net stabilized at perfect classification accuracy. A comparable prototype-network method previously reported needed 10,000 training iterations to reach 99.07 percent; Infra-Net’s convolutional branch alone hit 99.35 percent in just 300 iterations. Against a field of modern rivals including WST-AMResNet-18, WST-Trans, WST-BiLSTM, and MS-SE-ResNet, the new system posted the highest accuracy and the highest Cohen’s Kappa, a statistic that corrects for agreement expected by chance and therefore guards against inflated scores on imbalanced datasets. On the CTBTO field data, which pit natural earthquakes against chemical explosions under real-world noise, the 82.07 percent figure is lower but still clearly ahead of the pack, with an improvement of 3.55 percent over the strongest competitor and roughly 9 percent over traditional methods.

The authors are candid about the remaining gap between laboratory and field. Long-range propagation distorts signals in ways the current model does not fully capture, and cross-scene generalization remains a work in progress. Their roadmap includes domain adaptation through transfer learning, expansion of the training catalog to cover more event types at varied distances and signal-to-noise ratios, and uncertainty-aware fusion schemes that can better arbitrate when different feature views disagree. Even so, the study demonstrates that a mathematically grounded feature extractor, a multi-view reading of the same signal, and a consensus-driven decision mechanism can together deliver robust classification from remarkably little data. For monitoring stations that must run around the clock, distinguishing a volcanic eruption from a quarry blast in near real time, that combination could mean the difference between a false alarm and a life-saving warning.

Beyond hazard response, the implications reach into arms control verification, where the CTBT’s global network depends on reliably separating nuclear-relevant explosions from natural interference. The researchers note that adapting the system to new event categories requires only adjusting the output layer, not redesigning the architecture, which bodes well for deployment across diverse monitoring scenarios. If subsequent work closes the generalization gap, Infra-Net and its descendants may soon become standard equipment in the growing infrastructure that listens to the Earth’s low-frequency voice, turning an inaudible hum into actionable intelligence about the hazards brewing across the planet.

Subject of Research: Machine learning classification of infrasound signals for natural hazard monitoring and discrimination from anthropogenic events

Article Title: Infra-Net: a robust parallel decision-making network for discriminating natural hazards and anthropogenic infrasound events via multi-view feature learning

Article References: Li, H., Li, X., Liu, J., Luo, S., & Zhang, Y. (2026). Infra-Net: a robust parallel decision-making network for discriminating natural hazards and anthropogenic infrasound events via multi-view feature learning. Natural Hazards and Earth System Sciences, 26(9), 4529-4548. https://doi.org/10.5194/nhess-26-4529-2026

Image Credits: AI Generated

DOI: 10.5194/nhess-26-4529-2026

Keywords: infrasound, natural hazards, machine learning, wavelet scattering transform, deep learning, early warning systems, CTBTO, volcanic eruptions, earthquakes, signal classification, neural networks, geophysical monitoring

The size of the particles an eruption produces is not a trivial detail. It is the single most important factor controlling how far pyroclasts travel, how long they remain aloft, and what hazards they pose once they land. Coarse bombs and lapilli fall close to the vent, while fine ash can drift for hundreds or thousands of kilometers, fouling jet engines, coating crops, and settling into water supplies. Yet despite decades of study, volcanologists have lacked a complete picture of exactly how and where these particles form inside and around an erupting volcano. A new study published in Geology, led by Jacopo Taddeucci, a volcanologist at Italy’s Istituto Nazionale di Geofisica e Vulcanologia, now offers the most detailed microtextural evidence yet that the answer is far more complicated than simplified eruption models have assumed.

The stakes are easy to underestimate. As Taddeucci explains, volcanic ash and larger pyroclasts interfere with nearly every aspect of life downwind of an eruption. They disrupt air traffic and road traffic, and once the eruption is over the work has only begun: the material must be removed, and its effects on crops, plants, vegetation in general, and wildlife must be assessed. Understanding how these particles are formed, he argues, is essential precisely because the formation process controls their features—their size, shape, density, and ultimately their behavior in the atmosphere and on the ground.

To interrogate that process, the research team turned to techniques more often associated with materials science than with volcanology. Using scanning electron microscopy and X-ray microtomography, they analyzed pyroclasts collected from three mafic, or basaltic, volcanoes: Etna and Stromboli in Italy, and Cumbre Vieja in Spain. These instruments allowed the researchers to peer inside individual particles at microscopic scales, resolving features far too small to see with the naked eye but large enough to record the physical history of the magma that produced them. Scanning electron microscopy reveals surface textures and fine-scale fractures, while X-ray microtomography builds three-dimensional images of a particle’s interior without cutting it open, preserving the spatial relationships between crystals, voids, and welded fragments.

What the team found inside these particles was a record of violence and repair. By identifying and quantifying characteristic microtextures—including broken crystals, sutures, and incorporated clasts—the researchers demonstrated that fragmentation does not occur at a single depth and a single moment, as many simplified eruption models assume. Instead, the process is protracted, unfolding across several distinct stages. As Taddeucci puts it, this is not a process occurring at a single point in space and time; it covers a span of time and a span of spaces that reaches from inside the volcano to outside it. The particles that rain down on cities like Catania are, in effect, archives of a prolonged sequence of breaking and healing events that began deep within the plumbing system.

Some of this protracted behavior had already been documented in the open air. Volcanologists have long observed bombs and lapilli breaking apart in flight, shattering into smaller fragments as they arc away from the vent. What has remained invisible is everything that happens before the magma exits the volcano, where direct observation is impossible. The new study provides the first microtextural evidence that fragmentation is also protracted inside a volcano, extending the known sequence of breaking events from the observable exterior into the hidden interior of the edifice. The microscopic scars preserved in pyroclasts—fractured crystals, sutured contacts, embedded fragments of earlier generations of magma—now serve as a proxy for processes no camera can reach.

Perhaps the most consequential finding concerns the opposite of breaking. The team showed that crack healing and the welding together of pyroclasts can be just as important as fragmentation in controlling the final particle size distribution. Taddeucci notes that researchers are accumulating more and more evidence that not only the breaking of the magma controls the size of the particles, but also the fact that different particles can stick together after they are broken. In other words, an eruption’s ash output is not simply the residue of how thoroughly the magma shattered; it is the net result of a dynamic competition between fracturing and reassembly. Cracks can close, and fragments that separate can fuse back together, particularly in hot, low-viscosity basaltic systems where freshly exposed surfaces remain plastic enough to weld.

This reframing has direct implications for hazard modeling. Particle size governs how pyroclasts are dispersed through the atmosphere, and computer models of eruptions rely on assumptions about that size distribution to forecast ash cloud trajectories, ground deposition, and aviation impacts. If fragmentation is protracted and healing is significant, then models that treat fragmentation as a single instantaneous event at a fixed depth may systematically misestimate the grain size of the erupted material—and with it, the reach of the hazard. Incorporating the full break-and-heal history into eruption models could therefore sharpen forecasts of how far ash will travel and how long airports, roads, and agricultural areas will remain affected.

The microtextures identified in the study are not unique to the three volcanoes examined. Asked whether the same features would appear in material from Etna’s eruption this summer, Taddeucci did not hesitate: yes, of course, he says, adding that most basaltic eruptions that are explosive in nature will produce these kinds of features. That generality is what makes the finding powerful. Rather than a quirk of one volcano or one eruption, the break-and-heal signature appears to be a common feature of explosive basaltic volcanism worldwide, from the pulsing fountains of Stromboli to the sustained activity of Cumbre Vieja on La Palma.

There is also a retrospective dimension to the work. Because these microtextures can be preserved in rocks from past eruptions, researchers can use them to reconstruct the fragmentation conditions of events that occurred centuries or millennia ago, extending the observational record far beyond the era of modern monitoring. Taddeucci highlights this as a significant advantage volcanology holds over other geophysical hazards: unlike earthquakes, which leave no comparable physical archive of their internal dynamics, eruptions preserve a durable record of the conditions under which their products formed. Reading that record could improve models of future eruptions at well-studied volcanoes and at sites whose behavior is known only from ancient deposits. Taddeucci also emphasizes that the research depended on collaboration, comparing eruptions across different volcanoes and countries—a scope possible only through cooperation with colleagues at observatories and institutions, including researchers beyond the paper’s official list of co-authors. As ash clouds continue to test the resilience of aviation and communities, the microscopic scars inside volcanic particles are proving to be one of the most informative records science has of what really happens when magma comes apart.

Subject of Research: Microtextural evidence for protracted magma fragmentation and healing during basaltic explosive eruptions

Article Title: Scientists find microscopic evidence that magma repeatedly breaks and heals during eruptions

Article References: Scientists find microscopic evidence that magma repeatedly breaks and heals during eruptions. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: volcanology, pyroclasts, magma fragmentation, Mount Etna, Stromboli, Cumbre Vieja, scanning electron microscopy, X-ray microtomography, volcanic ash, basaltic eruptions, Geology journal, eruption modeling