Feed aggregator

SummaryThe Sanriku-Oki subduction region in northeastern Japan is a tectonically active zone where the Pacific Plate subducts beneath the Okhotsk Plate, generating frequent earthquakes. In this study, we present a three-dimensional unstructured S-wave traveltime tomography model of the off-Sanriku forearc using local earthquake data recorded by ocean-bottom Distributed Acoustic Sensing (DAS). The dense spatial sampling provided by DAS enables cost-effective, high-resolution imaging of the forearc region that is difficult to achieve with conventional ocean-bottom seismometer deployments. Eight local earthquakes recorded near the DAS cable provided 209,193 high-quality S-wave arrival times after applying data-selection criteria. These earthquakes were recorded using two DAS interrogator units: the AP Sensing N5200A (70 km coverage, 5 m channel spacing) and the OptaSense QuantX (100 km coverage, 2 m channel spacing). The initial reference models were constructed on a 3D tetrahedral mesh with target cell sizes of 1.5 and 3.0 km, accurately incorporating the DAS cable geometry and earthquake hypocentres. Synthetic traveltimes were efficiently computed using the Fast Sweeping Method (FSM) and the traveltime models were subsequently refined through traveltime inversion. The resulting S-wave tomography models reveal significant spatial heterogeneity within the off-Sanriku forearc region. Low S-wave velocities (Vs ∼ 0.78–0.85km s−1) indicate shallow, unconsolidated Neogene sediments, underlain by higher velocities in the Cretaceous basement (Vs ∼ 1.2–2.4km s−1). The lower crust and uppermost mantle wedge exhibit clear along-strike segmentation of the forearc. In the landward domain, high S-wave velocities in the lower crust (4.15–4.4km s−1) and uppermost mantle (4.6–4.65km s−1) indicate a mechanically strong overriding plate and a thick, cold mantle wedge, with no evidence of significant partial melting or serpentinisation associated with subduction processes. In contrast, the central offshore region exhibits moderately reduced S-wave velocities (3.9–4.1km s−1), suggesting a mechanically weaker overriding plate. Toward the trench, S-wave velocities decrease markedly within the forearc crust, defining a low velocity zone associated with crust–crust interaction above the subducting Pacific Plate and likely reflecting fluid-rich, deformed forearc material in the shallow subduction environment.

Sea ice is not just solid frozen water. It's riddled with tiny pockets and channels of liquid brine. Whether those pockets connect to form pathways determines whether seawater, nutrients and gases can move through the ice, according to decades of research by University of Utah mathematician Ken Golden.

Dramatic droughts linked to the decline of the Classic Maya civilization approximately 800 to 1000 CE may not have required any external trigger, according to a new climate modeling study. Instead, they could have emerged from Earth's own natural climate variability—shifts within the climate system that, when aligned, are capable of producing prolonged dry periods on their own.

Because methane has around 80 times the warming potential of CO2 over a 20-year period, it has been a major focus for climate action groups. The Global Methane Pledge, launched at COP26 in November 2021, aims to cut human-caused methane emissions by 30% by 2030.

Organic soils cover less than 9% of Norway's land area, and about 65,000 hectares are currently used as agricultural land. Emissions from these areas are presently estimated at more than 2 million tons of CO₂ equivalents per year. This is equivalent to the emissions from 400,000 fossil-fueled cars.

Lead-acid batteries are omnipresent. An integral part of most electric vehicles and all conventional vehicles globally, they also serve as backup energy storage systems in developing countries. But if lead-acid batteries are recycled in smelting units without adequate pollution control measures, they can cause elevated lead pollution that persists in local soils for thousands of years. However, because recycling sites with pollution control measures cost millions of dollars, most efforts are informal and unregulated.

In a recent study, researchers reported that scraping lead-contaminated soil in the vicinity of an abandoned recycling site for used lead-acid batteries and treating it with phosphate was linked to a 22% reduction in the blood lead levels (BLLs) of children who were living close to that site in a Bangladeshi town. The research was published in the International Journal of Hygiene and Environmental Health.

“Informal battery recycling is rampant in Bangladesh.”

“Informal battery recycling is rampant in Bangladesh,” said study coauthor Mahbubur Rahman, an environmental health scientist at the International Centre for Diarrhoeal Disease Research, Bangladesh. “Used lead-acid batteries are broken up and smelted in close proximity to residential and agricultural areas, which exposes those communities to lead emissions that contaminate their soil and water sources.”

Rahman and colleagues analyzed the BLLs of 130 children living close to two recycling sites for used lead-acid batteries (ULAB) in the Tangail District of Bangladesh that were abandoned in early 2019. They also assessed the BLLs of 37 children who did not live anywhere near ULAB recycling sites. The researchers then carried out soil remediation efforts at one of the ULAB sites but not the other. Prior to the work, the team members held informational sessions for the community about the dangers of lead pollution so locals could provide informed consent to participate.

The team observed that following remediation efforts, the lead content of the soil in and around the former battery recycling site decreased from more than 20,000 parts per million to less than 400 parts per million, which was considered acceptable by the U.S. EPA when the study was conducted, from 2022 to 2023. (The EPA reduced the limit to 200 parts per million in 2024.)

The researchers collected and cleaned up soil from children’s play areas, roadsides, and courtyards of 68 households that belonged to the intervention group. A year after the lead-contaminated soil was cleaned up, the 89 children from those households had the most significant decreases in their BLLs: from 90.1 to 70.4 micrograms per liter, a decrease of more than 21%.

“We know for sure that the areas close to abandoned ULAB recycling sites are as contaminated as areas around abandoned lead mines.”

The children in the group who lived close to the second abandoned ULAB recycling site, where soil remediation was not conducted, experienced only about an 8.4% decrease in their BLLs, from 88.5 to 81.1 micrograms per liter. The reduction in the control group’s BLLs could be attributed to a government initiative focused on reducing lead levels in turmeric, which was happening over the same time period as the study, Rahman said.

Anne Riederer, an environmental health scientist at the University of Washington who was not involved in the new study, said the dangers of lead exposure from ULAB recycling sites are well documented.

“We know for sure that the areas close to abandoned ULAB recycling sites are as contaminated as areas around abandoned lead mines. This study fits with the bigger picture of what we have learned to date about cleaning up contaminated sites and how that could improve children’s health,” she said.

A Widespread Issue

Similar studies conducted in Brazil and Bangladesh reported 46% and 35% reductions, respectively, in children’s BLLs following soil remediation initiatives around ULAB recycling sites.

Despite those drastic improvements, the children’s BLLs were still far above the World Health Organization’s threshold of 50 micrograms per liter. “This could mean there are other sources of lead exposure, like paints and cookware items,” said Rahman. “Or the persistently high BLLs could be because of chronic and long-term lead exposure, due to which lead gets deposited deep into the bones for several decades, even if [people] move away from toxic sites.”

Rahman explained that while soil remediation is an effective mitigation measure for lowering childhood lead exposure, it is also labor-intensive and expensive. Though the team identified hundreds of toxic sites borne from informal ULAB recycling, it wasn’t possible for them to remediate the soil at every site.

“The reason why this issue is so widespread is [that] informal recycling is cheap,” he said. “That makes the formal sector reluctant to invest in costly pollution control measures.”

—Anuradha Varanasi, Science Writer

Citation: Varanasi, A. (2026), Cleanup of battery recycling sites may lower childhood lead exposure, Eos, 107, https://doi.org/10.1029/2026EO260120. Published on 15 April 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.

Source: Earth’s Future

Coastal landscapes are constantly being reshaped by natural forces, and as climate change causes more frequent storms and sea level rise, that change will only intensify. Because these areas are densely populated with homes, tourist destinations, and industries, understanding how and where the coast will change is a pressing issue. However, reliable predictions that lead to actionable knowledge are rare.

Lentz et al. describe the state of knowledge regarding coastal evolution, highlight gaps in scientists’ understanding, and describe opportunities for integrating information from various models, data sources, and end users.

Current coastal evolution predictions are often focused on too specific a location and are therefore hard to generalize or analyze too large a region and therefore lack detail, the authors say. In addition, it’s challenging for researchers to link the effects of acute events, such as storms, with long-term trends like sea level rise.

Improving these simulations will likely require combining many different types of models, including physics-based numerical models, models based on empirical measurements, and statistical models that include machine learning. To fully understand potential changes, the authors note that it is also essential to consider both coastal processes and human actions.

The researchers recommend several ways to improve consistency and collaboration in the field of coastal change forecasting. First, standardizing approaches and outcomes would make it easier to produce national-scale predictions. Right now, the variety of tools used across different locations makes it difficult for scientists to compare results and communicate effectively. They also emphasize the need for using coordinated research approaches. Stronger transdisciplinary collaboration, accompanied by essential training and support, would also enable scientists to make better predictions, the researchers say.

Comparing predictions to real-world observations of coastal landscape change could also help untangle this multifaceted challenge. By studying how coastlines have already changed, researchers can validate models and choose those that are performing best. Such comparisons require datasets that adequately capture coastal landscape change across both time and space. Remote sensing data and the use of artificial intelligence (AI) for data processing may help provide these improved datasets, the researchers suggest.

Engaging end users during the project planning process is also helpful because only end users truly know what kind of information they need to adapt to landscape change. Knowing how to engage end users can be difficult for physical scientists, but various tools and specialized personnel exist who can help coordinate these interactions, the authors say. (Earth’s Future, https://doi.org/10.1029/2024EF005833, 2026)

—Saima May Sidik (@saimamay.bsky.social), Science Writer

Citation: Sidik, S. M. (2026), How to study coastal evolution, Eos, 107, https://doi.org/10.1029/2026EO260115. Published on 15 April 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.

Editors’ Highlights are summaries of recent papers by AGU’s journal editors. Source: AGU Advances

Solar eruptions can trigger geomagnetic storms that disrupt satellites, GPS, and power grids, affecting daily activities and technology. Therefore, it is extremely important to understand these storms in order to mitigate their impact. Previous studies mainly focused on interplanetary conditions.

Ghag et al. [2026] investigate the interaction between solar ultraviolet light (EUV) during storms and the Earth magnetic field, taking into account its misalignment and offset with respect to the Earth’s rotational axis, which depend on time. Such misalignment and offset induce variations in EUV exposure in turn influencing the ionosphere and its interaction with the magnetosphere.

The study applies the Multiscale Atmosphere-Geospace Environment (MAGE), a physics based fully coupled whole geospace model. The causal relationship between storm timing and storm effect is explored revealing insights on our capability to predict storm impact based on the time dependent Earth system state.

The rotation of the magnetic pole around the rotational pole in the NH and SH. The location of the rotational pole is denoted in blue and the magnetic pole in red. Credit: Ghag et al. [2026], Figure 6c

Citation: Ghag, K., Lotko, W., Pham, K., Lin, D., Merkin, V., Raghav, A., & Wiltberger, M. (2026). Universal time influence on stormtime magnetosphere ionosphere coupling. AGU Advances, 7, e2025AV002071. https://doi.org/10.1029/2025AV002071

—Alberto Montanari, Editor-in-Chief, AGU Advances

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.

Although the fundamental constitutive laws for steel and concrete were established more than a century ago, weak layers in snow remain a mystery. There are currently two theories about how they fail. A study published in Geophysical Research Letters has now identified a clear favorite.

New research reveals a powerful yet overlooked driver of climate change: intensifying ocean eddies. These swirling currents—that break off from major currents—are redistributing heat and nutrients in the ocean and amplifying climate extremes in key coastal ecosystems.

The Landslide Blog is written by Dave Petley, who is widely recognized as a world leader in the study and management of landslides.

I’ve not posted about Landslides in Art much in recent years – the most recent edition was almost two years ago – but loyal readers will know that this is a long running series of posts.

Anyway, I came across a page recently about the major landslide that struck the village of Runswick Bay in North Yorkshire. It includes a painting of the village with the above name, by an artist who signed themselves as “Jotter”. The painting is now in the collection of the Kirkleatham Museum:-

The Great Landslide of 1664 of Runswick Bay on The day Village That Slipped into the Sea by “Jotter”. The painting is held by the Kirkleatham Museum.

Now, there is a twist in that the landslide actually occurred in 1662, not 1664!

Runswick Bay is a picture postcard village in North Yorkshire of the UK, located at [54.53356, -0.75015]. The coastal part of the village is built on landslide debris, and there has been some movement in recent decades. In the late 1990s a very large scheme was put in place to mitigate the ongoing movement.

This is a Google Earth view of the village:-

Google Earth view of Runswick Bay.

The Tees Valley Museums site describes the landslide of 1662, noting that there were two major failure events. It is very fortunate that no-one was killed. The village was essentially destroyed and then rebuilt to the south of the original site.

It is probably true to say that the painting by Jotter is not a classic, but it does capture some interesting aspects of the site. First, it appears that the morphology is that of an existing landslide mass – this was probably a reactivation rather than a first time failure. Second, the toe was actively eroding, so maybe the two phase failure involved a collapse at the toe, which then destabilised the mass upslope? This would fit the eyewitness reports. Finally, note the mass in the background, which is also the result of a series of failure events.

There are many other major landslides along this section of coast – it is a classic area of UK mass movement geology. And it is truly beautiful too – visit if you can.

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.

A common mineral hiding in plain sight could hold the key to making copper production cleaner, faster and more efficient, just as global demand for the metal surges to power the energy transition. In an article published in Nature Geoscience, researchers from Monash University's School of Earth, Atmosphere and Environment describe why chalcopyrite, the source of around 70% of the world's copper, has remained so difficult to process, and how its hidden chemistry could be harnessed to unlock more sustainable extraction.

SummarySeismic wave speed monitoring is important for the non-destructive evaluation of material properties in response to external forcing. Coda wave interferometry (CWI) uses travel time perturbations in multiply-scattered seismic wave trains – the seismic coda – to detect subtle perturbations in bulk wave speed. However, conventional body-wave CWI cannot separate the coupled contributions of P and S waves, which are sensitive to different material properties. We introduce energy partitioning inversion which decouples these modes by combining a scattering model with CWI measurements within non-equipartitioned coda windows. We applied this methodology to repeated ultrasonic pulse surveys during two laboratory loading experiments on Clashach sandstone: a dynamic experiment (constant strain rate until brittle failure) and a quasi-static experiment (modulating stress to maintain constant acoustic emission rate and slow down the failure process). Relative travel time perturbations and their full covariance between all pairs of surveys were measured across multiple coda windows and inverted for a single perturbation profile using a least-squares method to minimise the variance of the profile. Using an isotropic point scatterer model to predict mode partitioning with respect to the coda lapse time, we invert travel time perturbations for the scattering mean free path travel time and relative P and S wave speed perturbations via Markov-chain Monte Carlo inversion to quantify uncertainty. P and S wave speed perturbations were resolved with 95 % credible intervals of 0.025 % and 0.008 %, respectively. During the quasi-static experiment the temporal resolution was sufficient to capture a quasi-linear decrease in P and S wave speeds by ~ 50 % and ~ 14%, respectively, from peak to failure. The peak P and S wave speed perturbations were ~ 33% lower and ~ 75% higher, respectively, compared to those found in the dynamic experiment. These results demonstrate that CWI and energy partitioning inversion enables the robust, uncertainty-quantified evaluation of separate relative bulk P and S wave speed perturbations in strongly-scattering media.

Models of glacial flow and retreat rely on estimates of glacial ice viscosity, the measure of the ice's resistance to flow. Ice viscosity is dependent on the stress applied to the glacier. Most ice sheet models use a standard equation to model ice flow that includes the variable n, called the stress exponent. A larger value of n means ice viscosity is more sensitive to changes in stress. For decades, glaciologists have, almost exclusively, used an assumed n value of 3 in the models they use to predict ice flow.

Coastal groundwater is a key source of drinking water in many regions of the world. However, it is threatened by overabstraction and the potential for salinization. Rising sea levels are further exacerbating the situation. This is demonstrated by a recent study published in Nature Water by a research team led by Professor Robert Reinecke from the Institute of Geography at Johannes Gutenberg University Mainz (JGU) and Annika Nolte from the Climate Service Center Germany (GERICS) in Hamburg.

Mark Clementz, a professor in the University of Wyoming Department of Geology and Geophysics, and colleagues have produced a compelling study that shows that an increase in volcanic activity in the Andes in the Late Miocene Epoch likely resulted in a cooling of Earth between 5.4 million and 7 million years ago. Their findings are in a new article, titled "Andean volcanism, ocean fertilization, marine ecosystem turnover, and global cooling in the Late Miocene," published in Communications Earth & Environment.

How can magma buried 5, 10, or even 15 km underground be detected without any surface indicators? The answer lies in ambient noise tomography, a technique that analyzes natural ground vibrations with high precision. A team from the University of Geneva (UNIGE), the Institute of Geosciences and Earth Resources (CNR-IGG), and the National Institute of Geophysics and Volcanology (INGV) has identified a vast reservoir containing approximately 6,000 km3 of magma beneath Tuscany.

Source: AGU Advances

Models of glacial flow and retreat rely on estimates of glacial ice viscosity, the measure of the ice’s resistance to flow.

Ice viscosity is dependent on the stress applied to the glacier. Most ice sheet models use a standard equation to model ice flow that includes the variable n, called the stress exponent. A larger value of n means ice viscosity is more sensitive to changes in stress. For decades, glaciologists have, almost exclusively, used an assumed n value of 3 in the models they use to predict ice flow.

However, through recent experiments and observations, researchers have found that an n value of 4 may actually better represent the conditions of Earth’s ice sheets and glaciers.

Martin et al. created a model representation of the fast-retreating Pine Island Glacier in West Antarctica. The ice sheet in their model had a true n value of 4, but they ran model projections using both n = 4 and n = 3. That allowed them to observe how their model would incorrectly predict glacial flow and resulting sea level change, given an incorrect n value.

The researchers modeled glacial retreat for 100 years under both equations with two different glacial melting scenarios. They then modeled glacial recovery for another 300 years. Under a moderate scenario, the n = 3 model underestimated glacial retreat by 18% and sea level change contributions by 21%. Under an extreme melting scenario, the model underestimated sea level contributions by 35%.

Notably, those disparities in glacial retreat and sea level change contribution predictions increased more than would be expected between the two scenarios, potentially increasing the level of uncertainty in current projections of sea level change. The researchers also suggest that incorrect n values may be mistakenly attributed to other physical processes in current ice sheet models.

The results could have far-reaching implications for predictions of future glacial melt and may prompt investigations into its effects on sea level, the authors say. (AGU Advances, https://doi.org/10.1029/2025AV001946, 2026)

—Madeline Reinsel, Science Writer

Citation: Reinsel, M. (2026), Glaciers may flow into the ocean more quickly than we think, Eos, 107, https://doi.org/10.1029/2026EO260107. Published on 14 April 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.

In the winter of 923, a magnitude 7.5 earthquake struck the heart of Puget Sound. Shorelines slid into the water, the seafloor rose up, and a tsunami swept through the region.

The Seattle fault zone, actually a mesh of faults that runs right under its eponymous city, was responsible for this quake. The fault continues to pose one of the deadliest threats to the Pacific Northwest; if a similar quake were to hit today, it would threaten millions of lives and cause billions of dollars in damage.

Two new papers dig into recurrence intervals, or the quiescent periods between earthquakes, for the Seattle fault zone. They offer good news and bad news: One study, published in Geology, found that in the past 11,000 years, the massive 923 event was the only quake of magnitude 7.5 or greater. The other study, published in GSA Bulletin, found that smaller, but still damaging, quakes occur more frequently than previously thought.

The Seattle fault zone runs east-west under the city and the surrounding Puget Sound. Credit: Washington Geological Survey (Washington Department of Natural Resources)

The new research indicates the worst-case scenario of frequent 923-style events is less likely than some scientists thought, said Harold Tobin, a geophysicist at the University of Washington and head of the Pacific Northwest Seismic Network, who was not involved in either study. But researchers also found that “the less worse, but still bad scenarios” are more likely than previously thought.

Meet the Seattle Fault

“For a fault that has had so much attention, there’s so much we still don’t know.”

The Seattle fault zone is a thrust fault system that stretches about 75 kilometers (46 miles) from the foothills of the Cascades east of Seattle to the Hood Canal, which runs along the shores of the Olympic Peninsula to the city’s west, passing under Seattle along the way.

Geologists began rigorously exploring the fault system in the early 1990s, intrigued by gravitational anomalies, uplifted marine terraces (stair-step geological formations along coastlines), and evidence of a roughly 1,000-year-old tsunami. All these features hinted at a major, shallow earthquake on a local fault zone—likely the 923 event.

But “for a fault that has had so much attention, there’s so much we still don’t know,” said Elizabeth Davis, an earthquake geologist at the University of Washington who led the Geology study.

The most pressing questions are how big quakes on the fault get, how often they hit, and, ultimately, what risks the fault poses to people who live in the Puget Sound area.

“It takes some real geologic sleuthing to get at those tough questions,” Tobin said.

Biggest Seattle Fault Quakes Are Rare

Davis focused on the activity of the main fault, which can generate the biggest quakes in the Seattle fault zone complex. It was responsible for the 923 quake. But the existing record went back only about 5,000 years.

“We just don’t know what the recurrence interval for these big quakes is,” Davis said. “We wanted to lengthen the record.”

To do so, Davis and her collaborators turned to marine terraces, the oldest of which date back to the end of the last ice age about 11,000 years ago. The quake in 923 raised terraces by about 8 meters (26 feet), and scientists wanted to look for similar-scale uplift in terraces all around the sound.

The researchers mapped more than 150 terraces around Puget Sound and measured their depths. After accounting for regional slopes, they estimated uplift over time that could have been caused by quakes.

They found that in that 11,000-year period, only the 923 event generated significant uplift. Thick sediment mantles could mask smaller events but not 923-scale quakes, Davis said.

Estimating true recurrence intervals requires knowing the timing of multiple events. But the finding is “not bad news,” she said. It provides some evidence that the recurrence interval is likely not shorter than about 5,000 years.

“That could give us more of a buffer between now and when the next big one like that will happen,” said Stephen Angster, a U.S. Geological Survey geologist who led the GSA Bulletin study.

Smaller, Damaging Quakes Are More Frequent

Angster’s work focused on Seattle’s secondary faults, which are smaller, mostly blind faults (those not visible at the surface) capable of generating damaging earthquakes. Previous work had shown that one of these secondary faults generated a magnitude 6.7 earthquake, highlighting the risk they pose. Angster wanted to explore rupture histories of these secondary faults, particularly whether they could rupture independently from the main fault.

The researchers used a suite of paleoseismic tools, including magnetic data, field and lidar mapping, trenches dug across faults, and geochronology. They studied two newly identified secondary faults that have orientations similar to the main fault.

They found three new earthquakes to add to the region’s seismic history, including the oldest and youngest events in the known record, which were around 11,000 years ago and in the early 1800s, respectively. The earthquakes appear to be evidence of ruptures that occurred independently of the main fault, suggesting that the smaller—but still dangerous—secondary faults should be considered in hazard modeling.

With that lengthened record and the addition of three quakes, the recurrence interval the researchers found was about every 350 years over the past 2,500 years. This timing refined the previous estimate of every several hundred years.

There also appears to be an increase in activity over the past 2,000 years.

“Maybe we should be paying attention to that,” Angster said.

What Happens Next

“There are other earthquakes that aren’t as big but that occur more frequently. Those might not be as catastrophic, but it would be a very bad scenario for Seattle” if such events occurred.

“These are both carefully done studies,” Tobin said. “We now have evidence that the 923 event was the biggest in 11,000 years. But there are other earthquakes that aren’t as big but that occur more frequently. Those might not be as catastrophic, but it would be a very bad scenario for Seattle” if such events occurred.

It’s still to be determined whether the risk from secondary faults will be incorporated into the National Seismic Hazard Model, which includes the 923 quake but not smaller ones along the Seattle fault zone. The secondary faults were left out in previous efforts because they are shorter than the minimum length required to be included and because of uncertainties in their potential rupture magnitude.

—Rebecca Dzombak (@rdzombak.bsky.social), Science Writer

Citation: Dzombak, R. (2026), On the Seattle Fault, the biggest quakes aren’t the most likely, Eos, 107, https://doi.org/10.1029/2026EO260114. Published on 14 April 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.

Editors’ Highlights are summaries of recent papers by AGU’s journal editors. Source: Journal of Advances in Modeling Earth Systems

The purpose of atmospheric data assimilation is to obtain a 3-dimensional gridded representation of the fields of the atmospheric state variables (temperature, wind, pressure, etc.) for a specific time based on atmospheric observations. The product of data assimilation, called analysis, can be used to prepare weather maps and to start model-based weather forecasts. Analyses collected over a long period of time can also be used for research and to monitor variability and changes in the climate.

The main challenges of data assimilation are that observations are not collocated with the grid-points of the analysis, and most observations do not observe the variables of interest directly and have errors. For example, satellite-based observations, which form the bulk of the operationally assimilated observations, measure the intensity of electro-magnetic waves at the top of the atmosphere; a physical quantity that depends on the atmospheric state in highly complicated ways. The background-error covariance matrix is a key component of a data assimilation system, responsible for spreading information from observations to the unobserved locations and state variables. A good estimate of this matrix is essential to produce analyses in which the fields of the state variables are realistic and consistent with each other. Obtaining such an estimate is particularly challenging for tropical locations, where physics-based knowledge does not lead to a straightforward practical formulation.

In a new study, Melinc et al. [2026] propose a novel machine learning-based (ML-based) approach to define a background-error matrix that is equally effective in the midlatitudes and tropics. This approach takes advantage of the power of ML to learn quantitative relationships between different state variables at different locations-relationships that are either not known, or cannot be easily used for the formulation of a background-error matrix based on physics-based knowledge.

Citation: Melinc, B., Perkan, U., & Zaplotnik, Ž. (2026). A unified neural background-error covariance model for midlatitude and tropical atmospheric data assimilation. Journal of Advances in Modeling Earth Systems, 18, e2025MS005360. https://doi.org/10.1029/2025MS005360

—Istvan Szunyogh, Associate Editor, JAMES

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.