The team combined two techniques that had never before been used together in a river: smartrocks and seismometers. Smartrocks are artificial pebbles, cast in high-density plastic with a bulk density of 2600 kilograms per cubic meter, that hide an inertial measurement unit inside a triaxial ellipsoid body measuring 6.4 by 7.2 by 13.0 centimeters. Each device records gyroscope velocities at 10 hertz, powered by a battery that allows roughly a month of continuous operation, and carries an RFID tag so researchers can find it again after a flood. Because the sensors ride inside the grain itself, they capture bedload transport from the Lagrangian perspective of the moving particle, distinguishing between rest, vibration, and genuine downstream displacement. Seismometers buried half a meter deep in the channel banks, meanwhile, record the ambient ground vibration generated by countless grain collisions and grain vibrations across the whole reach, providing the Eulerian, channel-scale complement to the tracer’s point of view.

The field sites were two gravel-bed ephemeral channels in the northern Negev Desert of Israel, Nahal Anim and Nahal Yatir, which drain the southern Hebron hills into the Beer Sheva Basin. The choice was deliberate: both experience abrupt stage rises during winter flash floods, but they differ sharply in bed character. Nahal Anim has a catchment of 35 square kilometers, a gentle 0.4 percent slope, a mean channel width of 6 meters, and a median bed grain size of 14 millimeters. Nahal Yatir drains 180 square kilometers, slopes at 1.3 percent, runs 10 meters wide, and carries a much coarser bed with a median grain size of 76 millimeters. Comparing the two allowed the researchers to ask whether the physics of rapid stage rise depends on the morphology of the bed or follows a more universal pattern.

Over two winter seasons the team recorded five flow events, one in Nahal Anim and four in Nahal Yatir, containing 25 distinct rapid stage rises. These were not gentle pulses. Rise durations ranged from 2 to 27.5 minutes, water depth increases spanned 4 to 57 centimeters, and maximum rates of rise reached 61.4 centimeters per minute in Nahal Yatir. Pressure transducers logged the hydrographs, while the seismometers sampled ground motion at 500 hertz across a 10 to 100 hertz analysis band. By computing the power spectral density of the seismic signal and correlating it, frequency band by frequency band, with the gyro velocities of the smartrocks, the researchers identified the specific frequency windows that track bedload activity: 35 to 60 hertz in Nahal Anim and event-dependent bands between 30 and 90 hertz in Nahal Yatir, with Spearman correlation coefficients reaching as high as 0.96.

The central result is that the bed’s response to a rapid stage rise is not a single behavior but a sequence of three distinct stages, and remarkably, the transitions between them occur at the same relative water depths in both channels despite their very different morphologies. When water depth is normalized by the bed roughness length scale d84, the first stage occurs at relative depths below about 0.9, where the water is shallower than the largest roughness elements. Here, rapid stage rise dramatically enhances bed activity: seismic energy ratios between rising and steady conditions reached 2 to 46 in Nahal Yatir and up to 2.2 in Nahal Anim, and gyro velocities during rises exceeded steady-flow values by a mean factor of 3.2 in the shallow bins of Nahal Anim.

What the smartrocks revealed about this shallow stage is perhaps the most surprising finding of the study. The instrumented pebbles were not rolling downstream at all. Their gyro velocities stayed below the displacement threshold of 0.3 radians per second, but they vibrated intensely, and vibration was far more frequent during rapid stage rises than during steady flow. This suggests that the seismic energy radiated during shallow bores comes largely from grains rattling in place under intense near-bed turbulence rather than from actual transport. The implication for seismology is significant: vibrating grains generate seismic noise without contributing to bedload flux, so inversion methods that convert river seismic signals into transport rates may overestimate sediment movement at low stages unless they account for this vibrational source alongside the better-known mechanisms of grain impact and rolling.

In the intermediate stage, at relative depths between roughly 0.9 and 2.5, the enhancement effect largely vanished. Seismic energy ratios converged toward unity, and gyro velocities under rising and steady conditions became nearly indistinguishable. The researchers attribute this buffering to the presence of pre-existing water: many of the intermediate-depth rises occurred when a layer of flow already covered the bed, and previous work has shown that bores propagating over flowing water generate far lower bed shear stresses from turbulent fluctuations than bores over dry beds. The existing water dampens the turbulence burst that makes shallow bores so disruptive. Even so, the smartrocks recorded more frequent genuine displacement during rises than during steady flow in this range, hinting that transport was beginning even while the overall seismic signature suggested parity.

Then, at relative depths greater than about 2.5, the pattern reversed again and rapid stage rise once more enhanced bed activity. Seismic energy ratios climbed back above unity, reaching 1.2 to 2.1 in Nahal Yatir and 2.3 to 3.2 in Nahal Anim, while gyro velocity ratios in Nahal Yatir rose with increasing depth beyond 0.65 meters. Crucially, this deep-flow stage was marked by a shift in motion mode: displacement became more frequent during rapid rises, and at high gyro velocities above 1.2 radians per second the seismic energy ratio spanned 1.4 to 3.3. In other words, when the water is deep enough to fully submerge the roughness elements, a passing bore appears to mobilize the bed across the entire channel, not just the tracer grains, producing genuinely higher transport rates than steady flow at the same depth.

The fact that both channels, one fine-grained and one coarse-grained, showed stage transitions at the same relative depths of roughly 0.9 and 2.5 points to channel roughness as the controlling variable. Turbulence intensity in rivers scales with bed roughness and grows with relative depth, so in shallow water the extra turbulence injected by a passing bore dominates, while in deeper water the roughness-driven turbulence of steady flow already does much of the work. This framework gives modelers a physically grounded way to predict when unsteady flow matters: not by absolute depth, but by depth relative to the bed’s own texture.

The practical stakes extend well beyond desert wadis. Rapid stage rises occur in tidal bores, tsunami run-up, glacial lake outburst floods, and dam or reservoir breaches, and bedload transport sets erosion patterns, reservoir sedimentation, habitat structure, and the design life of bridges and restoration projects. By validating seismic monitoring against in-grain measurements during exactly the conditions that defeat conventional samplers, the study provides both a caution and a tool: seismic ratios can serve as an upper-bound proxy for relative bedload flux, provided the vibrational contribution is recognized. As seismic networks spread across the world’s rivers, knowing that the ground itself remembers how a flash flood begins may prove one of the most useful lessons the Negev’s ephemeral streams have to offer.

Subject of Research: Bedload sediment transport dynamics during rapid stage rises in ephemeral desert streams, measured with instrumented smartrocks and seismic monitoring

Article Title: Integrating smartrock and seismic monitoring to investigate bedload transport dynamics during rapid increase of stages in ephemeral streams

Article References: Integrating smartrock and seismic monitoring to investigate bedload transport dynamics during rapid increase of stages in ephemeral streams. (n.d.). https://doi.org/10.5194/esurf-14-821-2026

Image Credits: AI Generated

DOI: 10.5194/esurf-14-821-2026

Keywords: bedload transport, flash floods, ephemeral streams, smartrocks, seismic monitoring, geomorphology, sediment transport, hydraulic bores, channel roughness, turbulence, Negev Desert, river monitoring

The core insight is that every building acts as a mechanical filter sitting between the earth and its occupants. When seismic waves arrive at a foundation, the structure amplifies, attenuates, and delays them according to its own dynamic properties, so the acceleration recorded on an upper floor is not the ground motion itself but a transformed version of it. For decades, engineers have treated this transformation as a nuisance to be corrected for. The new work inverts that logic: if the filtering behavior of a building can be modeled accurately enough, then the building’s own response becomes a rich source of information about the shaking it experienced at its base. In effect, the city’s building stock becomes a distributed, already-installed seismic sensing array.

Technically, the challenge is an inverse problem: given the acceleration time histories measured by structural response sensors, recover the peak ground acceleration, peak ground velocity, and peak ground displacement at the building’s location. The researchers begin by processing the raw structural acceleration records to extract multiple response parameters, which are then fused together to estimate the three ground-motion intensity measures. Fusing several parameters rather than relying on a single feature gives the learning algorithm a more robust signature of the underlying ground motion, because different parameters carry complementary information about amplitude, frequency content, and duration of the shaking.

To make the framework work at the scale of an entire city, the study area is discretized into grid cells, and within each cell a representative building is modeled as a nonlinear multi-degree-of-freedom shear structure. This is a standard idealization in earthquake engineering in which each story of the building is represented by a lumped mass connected to its neighbors by nonlinear springs, capturing how the structure yields and deforms under strong shaking. By simulating earthquakes acting on these representative buildings, the team generates the training data needed to teach a model how structural responses map back to ground motions for each building type, without waiting decades to accumulate real paired observations of ground truth and building response.

The machine learning architecture at the heart of the method, called ECA-Deep Net, is a hybrid framework that combines convolutional neural networks with long short-term memory networks, and it is trained separately for each representative building. The convolutional layers excel at extracting local features from the sensor waveforms, while the LSTM layers capture the temporal dependencies that matter in a seismic record, where the ordering and persistence of pulses carry physical meaning about the wavefield. A distinctive ingredient is the efficient channel attention mechanism, borrowed from the computer vision architecture ECA-Net. Channel attention allows the network to learn which feature channels are most informative for the task and to weight them accordingly, using a lightweight mechanism that adds almost no parameter overhead. In this application, attention helps the model decide which aspects of the multi-sensor structural response deserve emphasis when inferring ground-motion intensity.

The city-scale validation used a simulated earthquake, comparing the ground-motion maps generated by the framework against reference ground-motion fields computed by conventional means. The results showed that the predicted maps effectively reproduce the spatial variability of the reference fields, which is a critical property: ground shaking in real earthquakes varies sharply over short distances because of soil conditions, basin geometry, and source directivity, and a mapping method that smooths away that variability would mislead emergency responders about which neighborhoods were hit hardest. The framework also proved stable under varying levels of sensor noise at city scale, an essential robustness test for a method intended to run on low-cost hardware in real buildings rather than laboratory-grade instruments.

Perhaps the most compelling evidence comes from the field. The team validated the approach using recorded responses from an instrumented six-story real building, testing the models on earthquake records the network had never seen before, without any architectural modification or hyper-parameter retuning. The models achieved correlation coefficients of up to 0.97 between predicted and actual ground-motion parameters, with error values ranging from 7.33 percent to 12.05 percent. That level of transfer performance, on unseen events and with no recalibration, suggests the learned mapping is capturing genuine physics of the building-ground interaction rather than overfitting to the training set.

The practical implications reach well beyond seismology. Rapid ground-motion maps feed directly into shake maps, damage estimation, and emergency routing decisions, and the cost barrier of conventional strong-motion networks is one of the main reasons many earthquake-prone cities in developing regions lack adequate coverage. Structural health monitoring sensors are increasingly being installed in buildings anyway, for condition assessment and safety management, so a method that doubles them as ground-motion sensors offers a scalable, data-driven, and rapidly updatable alternative. Because each model is tied to a specific representative building, the network can grow incrementally: as more instrumented buildings come online, more grid cells gain coverage, and the city map fills in organically.

The work also fits into a broader movement in earthquake science toward machine learning-based ground motion modeling, from neural network prediction of peak ground acceleration to deep learning approaches using single-station waveforms and volunteer-hosted MEMS accelerometer networks. What distinguishes this study is the explicit use of the building as the sensing element, with the inversion handled by a building-specific model rather than a one-size-fits-all network. The authors acknowledge that the approach relies on representative structural models and simulated training scenarios, and the datasets and code files will be made available on request, which should allow other research groups to test the framework on their own instrumented buildings and urban grids.

If the approach matures from simulation and single-building validation into operational deployment, the aftermath of the next major urban earthquake could look very different. Within minutes of the shaking stopping, emergency managers could have a continuously updated, block-by-block picture of ground-motion intensity assembled from the very structures the earthquake was trying to destroy, informing search-and-rescue priorities, inspection queues, and utility shutdown decisions. The buildings that shelter a city, the study suggests, can also become the instruments that watch over it, provided the deep learning models translating their swaying into seismology are built with the care that this new framework demonstrates.

Subject of Research: City-scale seismic ground motion mapping using building-embedded structural response sensors and hybrid deep learning

Article Title: City-scale ground motion mapping from building-embedded structural response sensors using building-specific hybrid deep learning models

Article References: Zar, A., Li, S., Li, C., Li, C., & Zhao, J. (2026). City-scale ground motion mapping from building-embedded structural response sensors using building-specific hybrid deep learning models. Bulletin of Earthquake Engineering. https://doi.org/10.1007/s10518-026-02659-7

Image Credits: AI Generated

DOI: 10.1007/s10518-026-02659-7

Keywords: ground motion mapping, structural response sensors, deep learning, earthquake engineering, channel attention, LSTM networks, convolutional neural networks, peak ground acceleration, seismic monitoring, structural health monitoring, shear building model, sensor data fusion

The magnitude of the intrusion is modest by volcanic standards. Stamps compares the volume to about 400 Olympic-size swimming pools, an amount that is not in itself a cause for alarm. The magma, according to the team’s numerical modeling, does not appear to be migrating closer to the crater, and the reservoir it is entering already existed prior to the event. Yet the significance of the finding extends well beyond the volume of rock involved. It demonstrates that the monitoring infrastructure installed on the volcano can detect subtle, transient deformation signals and, more importantly, that the team can interpret those signals with sufficient confidence to attribute them to specific subsurface processes. That capability lies at the heart of any credible early-warning effort for a volcano that alternates between gentle and violent behavior.

The detection itself was made possible by a ground-based geodetic network that Stamps’ group began building a decade ago. In 2016, the team installed six Global Navigation Satellite System instruments on the flanks and surroundings of Ol Doinyo Lengai. These continuously operating stations track the horizontal and vertical motions of the ground surface to a precision of about one millimeter, a level of sensitivity that allows researchers to distinguish genuine deformation of the volcanic edifice from noise introduced by atmospheric effects or equipment drift. The GNSS network is complemented by two broadband seismic stations that record earthquake activity across a wide range of frequencies, and the continuous instrumentation is supplemented by episodic benchmark measurements taken during field campaigns.

The combination proved decisive in 2024. When the GNSS stations recorded an uplift signal around the volcano, the team turned to numerical modeling to determine what subsurface source could reproduce the observed pattern of surface movement. By fitting deformation models to the geodetic data, the researchers concluded with a high degree of certainty that the signal was generated by an influx of magma into an already existing magma reservoir roughly three kilometers underground. The result validated both the sensitivity of the network and the interpretive framework the team has developed over more than ten years of observation at the site. The work was published on September 18, 2026, in Frontiers in Earth Science, with the DOI 10.3389/feart.2026.1881885.

Ol Doinyo Lengai occupies a singular position in volcano science. Its carbonatite lavas erupt at temperatures far lower than the silicate lavas produced by virtually every other volcano on the planet, and during quiet periods the volcano maintains an active lava lake that bubbles and flows effusively within its summit crater. This effusive behavior, however, is only half of the volcano’s personality. On average, Ol Doinyo Lengai produces an explosive eruption every 10 to 15 years, and these explosive episodes can pose serious hazards to the communities that live on and around its slopes. Stamps notes that the team expects another explosive eruption within their lifetime, and the central motivation of the research program is to give residents and authorities enough lead time to prepare and respond to evacuation decisions made by the Tanzania Geological Survey.

Understanding what distinguishes a benign magma intrusion from a precursor to violence is the crux of the forecasting problem, and the 2024 event provided a valuable calibration point. According to Stamps, the last time the volcano erupted explosively, the event was preceded by a magnitude 5.9 earthquake. That historical sequence suggests a set of warning indicators that the team now watches for: a significant earthquake followed by observable changes in surface deformation, and, most critically, evidence that magma is moving from a deeper storage zone to a shallower one over time. Such upward migration indicates that magma is ascending the volcanic conduit, and it is precisely the kind of progression that would elevate concern. The 2024 intrusion, by contrast, showed magma entering an existing reservoir without any sign of ascent toward the crater, which is why the team assessed it as low risk.

One of the most stubborn uncertainties in volcanology is timing. Some volcanoes erupt explosively every time they erupt, making their behavior relatively predictable once unrest is detected. Ol Doinyo Lengai is more complicated, alternating between effusive and explosive eruptions in a pattern that is not fully understood. A key reason the team continues to monitor the volcano intensively is to constrain this particular volcano’s time delay, the interval between detectable precursory activity and an actual explosive eruption. Every additional episode of recorded deformation, seismicity, and magma transport adds to a decade-long baseline of observations that gradually reveals how the volcano’s subsurface system transitions from storage to ascent to eruption. The longer and richer the record becomes, the more skillfully the team can match patterns of surface observation to processes occurring underground.

The monitoring effort at Ol Doinyo Lengai is also notable for its ground-based character. Satellite-based observations, including those contributed through NASA, provide valuable context and broad spatial coverage of deformation across the region. But Stamps emphasizes that her team is currently the only group conducting continuous ground-based GNSS monitoring and consistent ground-based seismic monitoring at the volcano. Ground instruments offer continuous temporal sampling and millimeter-level precision that orbital observations alone cannot match, particularly for small, short-lived deformation episodes like the 2024 bulge, which might be missed or ambiguously characterized by intermittent satellite acquisitions. The redundancy and complementarity of the two approaches strengthen the overall early-warning picture.

The practical payoff of this work accrues directly to the people who live in the shadow of the volcano. More than a decade of observations has given the team a substantially better understanding of which underground magma movements correspond to which surface signals, and that empirical link is the foundation of any effort to anticipate when the volcano is edging closer to an explosive eruption. For residents of the surrounding area, the difference between a well-instrumented volcano and a poorly instrumented one can be measured in the time available to act. The Tanzania Geological Survey, which holds formal responsibility for evacuation decisions, now has access to a continuously refreshed, quantitatively interpreted stream of deformation and seismicity data from one of the rift’s most active and most unpredictable volcanoes.

The 2024 bulge may have been modest, equivalent to 400 swimming pools of magma settling into a reservoir it already knew, but as a test of scientific and operational readiness it was passed convincingly. A transient deformation signal was detected in near real time, modeled rigorously, attributed to a specific source at a specific depth, and correctly evaluated as non-threatening. Each such episode sharpens the interpretive tools that will be needed when the volcano’s behavior changes in more consequential ways. Given the volcano’s average recurrence interval of 10 to 15 years between explosive eruptions, the question is not whether the next one will come but whether the warning signs will be recognized in time. On the evidence of this study, the instruments are in place, the baseline record is deepening, and the science of translating millimeters of ground motion into meaningful forecasts is steadily maturing at one of Earth’s most remarkable volcanoes.

Subject of Research: Magma intrusion and surface deformation monitoring at Ol Doinyo Lengai volcano in Tanzania

Article Title: Interpreting a volcano’s ‘bulges’ and predicting the next explosive eruption

Article References: Interpreting a volcano’s ‘bulges’ and predicting the next explosive eruption. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: Ol Doinyo Lengai, volcano monitoring, magma intrusion, GNSS, Tanzania, explosive eruption, East African Rift, surface deformation, carbonatite lava, seismic monitoring, early warning system, Frontiers in Earth Science