Feed aggregator

Solar activity affecting Earth and its planetary neighbors encompasses a wide range of phenomena, from the steady solar wind and the interplanetary magnetic field to extreme events like solar flares, coronal mass ejections (CMEs), and solar energetic particle (SEP) events. These space weather phenomena interact in complex ways with planetary magnetospheres and atmospheres. On Earth, we see the results in the dancing lights of stunning auroras and in less frequent but sometimes severe disruptions to telecommunications, navigation, and energy infrastructure.

Forecasting conditions throughout the heliosphere (the region influenced by the solar wind), understanding the variety of Sun-Earth interactions, and predicting arrivals of space weather events—both benign and potentially hazardous—are a grand challenge.

The Sun-Earth challenge requires tracking and predicting conditions—from routine and quiet to rare and extreme—across tens of millions of kilometers of interplanetary space.

Solar flares emit electromagnetic radiation that spreads in all directions. In contrast, the propagation of CMEs and SEP events depends on their source location on the Sun and on the heliospheric magnetic field, which is carried outward by the solar wind. The impacts these events have on magnetosphere systems further vary depending on particle energies and intensities in SEPs and on particle speeds and the magnetic field orientation in CMEs. The Sun-Earth challenge thus requires tracking and predicting conditions—from routine and quiet to rare and extreme—across tens of millions of kilometers of interplanetary space.

This tracking and prediction is powered by petabyte-scale datasets from solar observatories and spacecraft measurements that provide rich observational archives. Researchers use these data to deduce physically meaningful quantities describing the heliosphere and to identify patterns to distinguish quiet from active conditions. The resulting insights not only answer fundamental science questions but also provide critical prediction time frames needed by space weather forecasters.

Even with all these data, the enormity of space between the Sun and Earth presents a major obstacle to our predictive capabilities. Another obstacle is that the data are obtained by different instruments operating at different locations and times. These factors combine to create a unique data sparsity challenge that complicates large-scale analysis.

These fundamental issues—the massive yet still insufficient supply of data available, the extreme differences in the scales of the processes we must illuminate, and the need for actionable predictions—suggest opportunities for artificial intelligence (AI) and machine learning (ML) to complement traditional physics-based analytical approaches [Camporeale, 2019]. In a series of workshops—insights from which inform the discussion below—scientists explored such opportunities and how they can advance heliophysics research and operational space weather forecasting.

The Need for Space Weather Forecasting

Space weather events can have significant impacts on infrastructure and humans. They can disrupt satellite operations (e.g., by enhancing atmospheric drag on satellites), damage electronics in space, interfere with radio communications and GPS, and even affect power grids (e.g., through geomagnetically induced currents) during the most severe events. They can also pose risks to people, especially astronauts beyond the protection of Earth’s atmosphere and airline crews and passengers on long-distance polar flights, during which exposure to energetic particles is elevated. Forecasting offers a first line of defense in preparing for or preventing damaging and hazardous effects of space weather.

In assessing major CMEs, forecasters consider whether and when events will reach Earth and whether they will trigger geomagnetic storms and substorms. For SEP events, predictions must include arrival times, peak intensities, durations, and energy characteristics.

Predicting extreme space weather phenomena is vital, but equally important is forecasting periods when no significant activity is expected, which is critical information for satellite operators and other stakeholders. Making such predictions requires understanding physics spanning 8 orders of magnitude in space and time, from subsecond processes in Earth’s magnetic environment to multiday solar eruptions propagating across the 150 million kilometers between the Sun and Earth (Figure 1) and long-term interactions at scales associated with the 11-year solar cycle.

Fig 1. Length scales and Sun-to-Earth transit times vary greatly for different types of space weather (SW), including solar flares, solar energetic particle (SEP) events, coronal mass ejections (CMEs), and interplanetary coronal mass ejections (ICMEs). High-speed particles are the first to arrive, usually within minutes of a flare, whereas CMEs arrive in 2–4 days. Credit: Georgoulis et al. [2026], CC BY-NC-ND 4.0

In addition to operational forecasting, these challenges are fundamental in heliophysics research. Such research includes work to reveal how the Sun generates its magnetic field, how solar wind accelerates and evolves, how planetary magnetospheres respond to external forcing, how particles are accelerated, and how energy transfers across multiple scales and regimes.

Unique Challenges in Heliophysics

Modern AI and ML algorithms excel at analyzing well-curated, extensive datasets that include millions of training examples. For example, AI-aided terrestrial weather forecasting relying on continuous, high-resolution coverage from thousands of ground stations, weather balloons, and satellites has advanced dramatically in recent years.

Fewer than a dozen spacecraft monitor Earth’s magnetosphere, a region spanning tens of Earth radii. Solar wind observations are even sparser.

Heliophysics, however, presents a unique and somewhat opposite scenario. Fewer than a dozen spacecraft monitor Earth’s magnetosphere, a region spanning tens of Earth radii (about 6,371 kilometers). Solar wind observations are even sparser, with just a handful of monitors scattered across the space between the Sun and Earth. This fundamental scarcity poses a challenge for data-driven approaches, which typically depend on abundant observations that are well distributed in space and time to produce trustworthy (i.e., generalizable and reproducible) models.

Data sparsity is further compounded by the relative rarity of intense space weather phenomena such as CMEs, major geomagnetic storms, and extreme substorms, which occur only a few times per solar cycle. Most heliophysical observations capture quiet, low-activity conditions when the solar wind is steady and magnetospheres are calm. Standard ML approaches trained on such imbalanced datasets may achieve high statistical accuracy by simply predicting a “nothing-will-happen” outcome but completely fail when extreme events occur.

Although solar eruptions and geomagnetic storms are relatively rare, they exhibit recurring patterns and consistency in their physical drivers. This regularity suggests that historical observations, when properly clustered and analyzed, can be used to enhance prediction capabilities. The challenge therefore lies in extracting meaningful patterns from sparse measurements of rare events while avoiding models that work well for average conditions but fail when they matter most [Chu et al., 2025].

AI Solutions for Data Sparsity

Heliophysics research employs clever approaches to extract maximum information from the limited available observations. One strategy is to mine multidecade observational records from various satellites and to match and group together measurements collected at times with similar solar wind and geomagnetic activity conditions.

This process clusters tens of thousands of data points from similar magnetospheric states. Such clustering enables reconstruction of dynamic features like nightside magnetic field changes during substorms [Stephens et al., 2019] and the presence of near-Earth magnetotail reconnections [Angelopoulos et al., 2020].

Another, more universal approach is to embed fundamental physical laws directly into ML models through physics-informed neural networks [Raissi et al., 2019], ensuring that predictions respect physical reality even when training data are limited. Data assimilation techniques used in weather forecasting similarly blend sparse observations with physics-based simulations and update models as new measurements arrive.

This animated model shows Earth’s magnetosphere during a powerful May 2024 geomagnetic storm that involved strong solar flares and multiple CMEs. The visualization uses the Multiscale Atmosphere-Geospace Environment (MAGE) model from the Johns Hopkins Applied Physics Laboratory to depict wind rushing toward Earth and disturbing its magnetic field (orange and purple lines). The green cloud represents electric field current intensity; the blue squiggles are tracers of solar wind velocities. Credit: NASA Scientific Visualization Studio and NASA DRIVE Science Center for Geospace Storms

These methods converge on a common theme: building gray box models (so named because they’re less opaque than black box models) that are data driven but grounded in physically real constraints. For data-starved applications, hybrid approaches can outperform purely data-driven or purely physics-based methods [Liu et al., 2025].

Satellite instruments are generating increasingly large solar wind datasets. However, the variables obtained (e.g., solar wind speed and pressure) are highly intercorrelated [Borovsky, 2018], making it difficult to identify which ones truly drive magnetospheric responses. New algorithms are helping to distill datasets without losing critical scientific information [e.g., Camporeale, 2025]. Meanwhile, advanced statistical and ML methods can cut through dataset complexity by reducing dimensionality, identifying causal relationships among variables, and providing clues about dominant drivers.

For instance, information theory provides tools to detect dependencies in complex systems, establish causality, and rank variables that most effectively predict space weather outcomes [Wing et al., 2022]. Such techniques can be paired with other “explainable” tools, such as SHAP (SHapley Additive exPlanations) values, a method inspired by game theory, to pinpoint physical variables (e.g., solar wind speed or magnetic orientation) that drive a prediction [Ma et al., 2023].

Distilling datasets and improving model interpretability help make ML more practical and more scientifically trustworthy and its predictions more robust. But fully trusting ML models in operational environments requires rigorous validation and uncertainty quantification. These models must not only make predictions but also indicate their confidence levels for operational decisionmaking.

When a model forecasts a major geomagnetic storm, operators need to know whether that prediction carries 60% or 95% confidence, for example.

When a model forecasts a major geomagnetic storm, operators need to know whether that prediction carries 60% or 95% confidence, for example. Ensemble approaches, in which multiple models provide a range of outcomes, help quantify this uncertainty, while using standardized, well-documented datasets enables fair model intercomparisons.

The research community is developing ML-ready benchmark datasets with consistent formatting and clear metadata to establish such validation procedures [e.g., Angryk et al., 2020]. These resources allow researchers to test new algorithms against common baselines, accelerating progress while ensuring that advances are robust and reproducible rather than artifacts of specific data processing choices.

Notably, one domain in heliophysics that is not affected by severe data sparsity is solar imaging. Decades of continuous, high-resolution observations from the Solar Dynamics Observatory (SDO), which delivers 1.5 terabytes of data every day, have created enormous data archives. Because the Sun drives space weather throughout the heliosphere, these datasets offer an ideal opportunity for use in foundation models, large-scale ML systems trained to learn comprehensive internal representations that can then be easily adapted to specific scientific tasks with minimal additional training.

Surya, a foundation model designed to construct a digital representation of the Sun, represents one such effort. It is still in early development and has yet to be validated, but this approach illustrates how data-rich domains can be leveraged with modern AI techniques to create tools that broadly benefit heliophysics research and space weather forecasting.

Advancing Research and Operational Forecasting Together

In addition to the needs for data and model development and validation, applying AI to address the challenges of heliophysics requires sustained, multidisciplinary collaborations. Fostering those collaborations has been the focus of a series of workshops, with the most recent being 2025’s Machine Learning, Data Mining and Data Assimilation in Geospace (LMAG25) meeting at the Johns Hopkins University Applied Physics Laboratory. The workshops have brought together heliophysicists, machine learning experts, data scientists, and specialists from weather forecasting and applied mathematics to exchange knowledge and establish community standards.

Space weather forecasters need models that are accurate and interpretable and that provide not just statistical metrics but also actionable predictions.

The LMAG forums also serve as gathering spaces for scientists to validate models against diverse datasets, compare physics-based and data-driven approaches, develop performance benchmarks, and discuss how to bridge research and operational requirements. Space weather forecasters need models that are accurate and interpretable and that provide not just statistical metrics but also actionable predictions with known limitations and reliability. Of course, researchers also benefit. These conversations allow them to gain insight into operational constraints that shape how modeling approaches become practical in real-world settings.

LMAG and similar initiatives facilitate direct exchanges among adjacent communities, including by making meeting presentations openly available. These efforts are helping translate cutting-edge AI and ML techniques into practical tools that help protect critical infrastructure and human well-being. They are also deepening our understanding of how the Sun shapes space weather throughout the solar system and its effects—both mundane and major—on Earth.

References

Angelopoulos, V., et al. (2020), Near-Earth magnetotail reconnection powers space storms, Nat. Phys., 16(3), 317–321, https://doi.org/10.1038/s41567-019-0749-4.

Angryk, R. A., et al. (2020), Multivariate time series dataset for space weather data analytics, Sci. Data, 7(1), 227, https://doi.org/10.1038/s41597-020-0548-x.

Borovsky, J. E. (2018), The spatial structure of the oncoming solar wind at Earth and the shortcomings of a solar-wind monitor at L1, J. Atmos. Sol. Terr. Phys., 177, 2–11, https://doi.org/10.1016/j.jastp.2017.03.014.

Camporeale, E. (2019), The challenge of machine learning in space weather: Nowcasting and forecasting, Space Weather, 17(8), 1,166–1,207, https://doi.org/10.1029/2018SW002061.

Camporeale, E. (2025), PARIS: Pruning Algorithm via the Representer theorem for Imbalanced Scenarios, arXiv:2512.06950, https://doi.org/10.48550/arXiv.2512.06950.

Chu, X., et al. (2025), Imbalanced Regression Artificial Neural Network Model for Auroral Electrojet Indices (IRANNA): Can we predict strong events?, Space Weather, 23(5), e2024SW004236, https://doi.org/10.1029/2024SW004236.

Georgoulis, M. K., et al. (2026), Prediction of solar energetic events impacting space weather conditions, Adv. Space Res., in press, https://doi.org/10.1016/j.asr.2024.02.030.

Liu, Y., et al. (2025), Data-driven modeling of electrostatic turbulence by physics-informed Fourier neural operator, Mach. Learn. Sci. Technol., 6(4), 045050, https://doi.org/10.1088/2632-2153/ae19cd.

Ma, D., et al. (2023), Opening the black box of the radiation belt machine learning model, Space Weather, 21(4), e2022SW003339, https://doi.org/10.1029/2022SW003339.

Raissi, M., P. Perdikaris, and G. E. Karniadakis (2019), Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations, J. Comput. Phys., 378, 686–707, https://doi.org/10.1016/j.jcp.2018.10.045.

Stephens, G. K., et al. (2019), Global empirical picture of magnetospheric substorms inferred from multimission magnetometer data, J. Geophys. Res. Space Phys., 124(2), 1,085–1,110, https://doi.org/10.1029/2018JA025843.

Wing, S., et al. (2022), Modeling radiation belt electrons with information theory informed neural networks, Space Weather, 20(8), e2022SW003090, https://doi.org/10.1029/2022SW003090.

Author Information

Savvas Raptis (savvas.raptis@jhuapl.edu), Manolis K. Georgoulis, Mikhail Sitnov, Anthony Sciola, and Simon Wing, Johns Hopkins University Applied Physics Laboratory, Laurel, Md.

Citation: Raptis, S., M. K. Georgoulis, M. Sitnov, A. Sciola, and S. Wing (2026), Vast space, sparse data: An AI answer to twin space weather challenges, Eos, 107, https://doi.org/10.1029/2026EO260188. Published on 11 June 2026. Text © 2026. The authors. CC BY-NC-ND 3.0
Except where otherwise noted, images are subject to copyright. Any reuse without express permission from the copyright owner is prohibited.

Author(s): B. Disson, N. Bonifaci, O. Lesaint, C. Poulain, R. Dussart, and S. Iseni

This experimental work reinvestigates the breakdown voltage in air for electrode gaps ranging from 0.10 to 6.00µm. Rarely addressed in the literature of the field, a special focus is placed on varying the gas humidity and its direct impact on the breakdown voltage. For short distances (<2µm), the…

[Phys. Rev. E 113, 065207] Published Thu Jun 11, 2026

In November 2025, Cyclone Senyar generated extreme rainfall in parts of Sumatra, Indonesia, triggering thousands of landslides. Our new paper in the journal Current Biology demonstrates that these landslides might have a devastating impact on a critically endangered population of Tapanuli orangutan.

In November 2025, Cyclone Senyar brought extreme rainfall to large parts of Sumatra in Indonesia. I have written about this on previous occasions – the rainfall triggered vast numbers of landslides.

In my line of work, we often focus on the landslide impacts on the landscape, on human lives and on infrastructure. We rarely consider the impacts on th eanimal population. This is certainly a weakness that the Cyclone Senyar event brings to focus.

Part of the area devastated by the landslides is that slopes around the Batang Toru rover, an area of forest that is home to a rare species of orangutang. These great apes, Pongo tapanuliensis, live in a habitat known as the West Block of Tapanuli. There are only 800 individuals left in the wild, a situation that is highly precarious. The loss of even a small number of adults could tip the species towards extinction.

I was a part of a consortium of scientists that considered the landslide impacts of Cyclone Senyar on the habitat of these orangutangs. The results have just been published in the journal Current Biology (Meijaard et al. 2026) – the paper is open access and published under a creative commons license.

This image, from the paper, shows the landslide impacts of Cyclone Senyar:-

Before and after satellite imagery of the impacts of Cyclone Senyar. From: Meijaard et al. (2026).

In the study area of 71,161 hectares, the mapping indicates that there were 50, 185 individual landslides, covering a surface area of 8,303 hectares. This is about 11% of the forested area. We then estimate the likely loss of the orangutang population, which is likely to be in the range of 18-120 individuals, with a central estimate of 58 individuals. This is likely to have been a devastating loss for this highly endangered population.

This level of habitat loss might also be placing a severe pressure on the remaining population, so further fatalities are very possible through, for example, reduced food availability.

The intensity of the rainfall was almost certainly supercharged by climate change. The impacts of Cyclone Senyar are being replicated widely – and of course we are now in the northern hemisphere tropical cyclone season again.

Our paper makes some policy recommendations for this population of orangutans. First, the government of Indonesia needs to permanently protect this area of forest against mining , palm oil and hydropower developments. Ideally, the protected area should be expanded. Second, Indonesia needs support for biodiversity-recovery, hazard forecasting and ecological restoration planning.

Reference

Meijaard, E. … Petley. D. … et al. 2026, Extreme rainfall further endangers the world’s rarest great ape. Current Biology. https://doi.org/10.1016/j.cub.2026.05.029

Return to The Landslide Blog homepage Text © 2026. The authors. CC BY-NC-ND 3.0
Except where otherwise noted, images are subject to copyright. Any reuse without express permission from the copyright owner is prohibited.

AbstractSeismicity rates in west Texas and southeast New Mexico have increased over the last nine years and are in large part driven by subsurface wastewater disposal associated with oil-gas operations. Injection-induced seismicity is often explained as the result of fault weakening from fluid pressurization. However, fluid injection also induces poroelastic stresses from fluid-rock coupling, which in some cases are larger than the perturbations induced from pore pressure alone. In this work, we model in-zone changes in pore pressure and poroelastic stressing along four major fault zones that have hosted moderate-large ML 4 – 5 earthquakes in the Delaware Basin (DB) of west Texas. We leverage high-quality downhole pore-pressure measurements to constrain the in-situ hydraulic diffusivity and storativity. The data show that the deep injection interval has a storativity of ∼ 5×10−5 and a diffusivity between 23-65 m2/s, suggesting that this interval can pressurize rapidly and transmit fluid pressure efficiently. We view these as local hydraulic properties and use an ensemble modeling approach that accounts for a large range in diffusivities and storativities to model changes in Coulomb Failure Stresses (CFS). Our 2D fully coupled poroelastic models show that deep subsurface fluid injection can induce between ∼1-1000 kPa in CFS along basement rooted faults that penetrate the injection interval, with the largest values occurring in models that use the hydraulic properties inferred from our in-situ measurements. However, the induced changes in CFS are much smaller (∼ 20-30 kPa) when averaging over a large range in hydraulic properties. Irrespective of the model parameterization, the in-zone perturbations in CFS are dominated by changes in pore pressure, even at distances as far as 20-30 km from the nearest injection source. Our results highlight the importance of obtaining in-situ poromechanical measurements and indicate that such high-resolution measurements are critical to understanding subsurface stressing associated with fluid injection.

Climate change could make historically rare tropical storms in Southern California produce significantly more precipitation in the next few decades, and when they strike, landslides are likely to become a bigger risk across the region, according to new research in Nature Climate Change.

Strong and consistent evidence shows that the entire climate system is continuing to heat, driving rapid global warming. Human activities pushed global warming to 1.37°C in 2025, and its level is projected to surpass 1.5°C in about four years. Crucially, the rate at which heat is accumulating in Earth's system suggests high levels of future warming. These are some of the key findings from the latest Indicators of Global Climate Change (IGCC) report, published in Earth System Science Data.

For almost a century, researchers have known that vertical land motion—the lifting and sinking of the ground—affects sea level locally. As the ground sinks, the sea level rises relative to the land. Scientists also assumed this process generally occurred at a steady rate over time. But a research team that includes Thomas Wahl, a UCF researcher and associate professor in the Department of Civil, Environmental and Construction Engineering, has found that ground subsidence has undergone phases of variable change, creating significant implications for coastal communities.

A brown mass—a mixture of water, boulders and fine matter—plows through the landscape. The mountains wash more than a thousand lorryloads of material into the valley on a fairly regular basis, causing damage in excess of CHF 100 million per year in Switzerland alone. A better understanding of this natural hazard requires data from the debris flow channels, of which there is very little because of the complex surveying process.

Ice-age sea-level declines may have turned seafloor volcanoes into natural iron fertilizer for plankton, potentially enhancing ocean carbon storage, Boston College researchers report in the journal Nature Geoscience.

Publication date: Available online 9 June 2026

Source: Advances in Space Research

Author(s): Shizhan Gao, Jixin Ding, Lin Chen, Yifan Wang, Ming Xu

Publication date: Available online 4 June 2026

Source: Advances in Space Research

Author(s): Mario F. Palos, Anna Ivanova, Giovanni Mengali, Alessandro A. Quarta, Marco Bassetto, Janis Dalbins, Mihkel Pajusalu, Antti Tamm, Mihails Ščepanskis, Perttu Yli-Opas, Iaroslav Iakubivskyi, Pekka Janhunen, Martin Tajmar, Julijanas Zeludevicius, Andris Slavinskis

The quest to identify a new way to potentially counter one of the world's most widely discussed solar geoengineering proposals has taken a new, exciting turn—raising questions about how future climate interventions could be governed.

Human-caused sea-level rise has significantly increased the frequency of extreme coastal flooding worldwide, according to a new study led by a Tulane University researcher. The research, published in the journal Nature Climate Change, found that coastal flooding events expected only once every 100 years are now, on average, about 12 times more likely to occur.

The number of icebergs in the Arctic has increased sharply since the 2000s. This is due to the destabilization of large glaciers in northeast Greenland and parts of the Russian Arctic, as well as the increasing mobility of sea ice. The result: Stones rain down from the melting icebergs, forming new hard-substrate habitats for marine life on the soft seafloor.

Source: Earth’s Future

Hotter, drier conditions in the western United States have led to a rise in wildfire activity that has damaged or destroyed infrastructure, natural ecosystems, and entire towns across the region. As fires grow larger and more destructive, the cost of managing them rises as well.

Fire management agencies in the United States have been feeling the pressure. Between 2014 and 2023, fire management agencies across all levels of government experienced a 131% increase in total area burned and a 268% increase in total fire spending adjusted for inflation compared to the period between 1985 and 1994.

Today, federal agencies like the Department of the Interior (DOI) and the U.S. Department of Agriculture Forest Service (USFS) continue to invest in aiding states and managing hazardous fuel growth on public land, as well as suppressing active fires. Policymakers and federal agencies alike must decide how to manage limited budgets while protecting people, property, and natural resources.

Prestemon et al. built statistical models based on historical data to examine the potential increase in spending by the DOI and the USFS between now and 2100. Their models link wildfire activity to climate variables such as temperature and water vapor deficit and then connect fire activity to suppression costs. To capture a range of possible future conditions on federal lands, the study predicts 10 fire and suppression spending scenarios by applying five different climate models to two different warming pathways (the moderate Representative Concentration Pathways (RCP) 4.5 scenario and the high-emissions RCP 8.5 scenario).

The results varied by region and scenario, but each of the 10 scenarios suggested a rise in area burned as well as inflation-adjusted fire suppression spending, with higher fire activity translating to higher costs. Projected changes in DOI and USFS land burned increased 80% by mid-century and 208% by late century.

By the middle of the century, both agencies are projected to see spending increases: about 0.65% per year for DOI spending and about 0.87% per year for USFS spending from 2020 to 2100. Although uncertainty increased with time and outcomes varied across climate models and warming pathways, the largest increases in both cost and wildfire activity were consistently projected for the northwestern United States. (Earth’s Future, https://doi.org/10.1029/2025EF007985, 2026).

—Rebecca Owen (@beccapox.bsky.social), Science Writer

Citation: Owen, R. (2026), As wildfires increase in the West, so does suppression spending, Eos, 107, https://doi.org/10.1029/2026EO260187. Published on 10 June 2026. Text © 2026. AGU. CC BY-NC-ND 3.0
Except where otherwise noted, images are subject to copyright. Any reuse without express permission from the copyright owner is prohibited.

Human activities have significantly altered the freshwater cycle, threatening its ability to support vital climatic and ecological Earth system processes. A new study led by researchers at the University of Eastern Finland shows that the freshwater cycle has increasingly moved away from a stable state due to climate change and large-scale water and land use. The newly published study updates the status of the freshwater change planetary boundary, which is one of the core components in the annual Planetary Health Check reports.

Researchers at the Institute of Industrial Science, the University of Tokyo (UTokyo-IIS), have developed a real-time data assimilation system that substantially improves streamflow and flood forecasting accuracy across Japan. This new data assimilation system outperforms Japan's current early warning system. Compared with the previous early warning system, the new data assimilation system yields improved forecasts for 80% of the major river reaches in Japan and enables reliable flood prediction across more than 60% of the observation stations in Japan.

The first few billion years of Earth's history saw the rise of life, the atmosphere and the oceans. Still, that time is shrouded in mystery: Not many rocks remain that preserve a record of those early iterations of our modern world. Dynamic geologic processes like erosion, subduction and burial mean the surface is constantly being reshaped, and older rocks aren't very common. But that time period is critical to understanding our own origins and how catastrophic events like asteroid impacts might have affected early life on the planet.

Researchers at Geosciences Barcelona (GEO3BCN-CSIC) have developed a novel way to monitor the silent accumulation of volcanic gases beneath Earth's surface using seismic ambient noise. The results could significantly improve early warning systems for volcanic eruptions.

Following the devastating urban fire that broke out in Wajima City after Japan's 2024 Noto Peninsula earthquake, investigators struggled to identify a clear ignition source, despite widespread destruction and unusual reports of flames emerging from areas with no visible combustible material. While previous studies have linked earthquakes to secondary disasters such as tsunamis and infrastructure failures, the possible role of underground methane release has remained poorly understood.