Feed aggregator

MCNet: A Multi-source Remote Sensing Image Cascade Network for Accurate and Efficient Tailings Pond Segmentation

Publication date: Available online 25 August 2025

Source: Advances in Space Research

Author(s): Pan Wang, Hengqian Zhao, Zhiguo Liu, Fei Xu, Hancong Fu, Jihua Mao

Application of object-based image analysis and some machine learning models in gully erosion susceptibility mapping

Publication date: Available online 25 August 2025

Source: Advances in Space Research

Author(s): Soraya Yaghobi, Mohsen Hosseinalizadeh, Chooghi Bairam Komaki, Mauro Rossi, Alessandro Cesare Mondini, Ali Najafinejad, Hamid Reza Pourghasemi

A real-time high accurate positioning method for vehicles in long-mileage transportation

Publication date: Available online 23 August 2025

Source: Advances in Space Research

Author(s): Chenxi Deng, Wei Jiang, Baigen Cai, Jian Wang, Chris Rizos, Xiaohui Ba, Jiang Liu

A broad climatology of very high latitude (<math xmlns:mml="http://www.w3.org/1998/Math/MathML" altimg="si5.svg" class="math"><mrow><mo>∼</mo><mn>75</mn><mi>°</mi></mrow></math>) substorms: An update

Publication date: Available online 22 August 2025

Source: Advances in Space Research

Author(s): Anand K. Singh, Navin Parihar, Shailendra Saini, Sunil Kumar, Rashmi Rawat, Vladimir Belakhovsky, Wojciech J. Miloch, Gopi Seemala, A.K. Sinha

Retrieving soil heavy metal concentrations using hyperspectral remote sensing images based on three-band indices at a coal mine in China

Publication date: Available online 22 August 2025

Source: Advances in Space Research

Author(s): Haorui Bai, Bin Guo, Lin Pei, Bo Zhang, Pingping Luo, Tengyue Guo, Min Wu, Xiaohan Ma

Change detection in optical and SAR images by integrating fine-grained feature extraction and contextual transformation

Publication date: Available online 22 August 2025

Source: Advances in Space Research

Author(s): Man Zhang, Liang Huang, Bohui Tang, Tonggen Yang, Siming Pu

Researchers discover massive geo-hydrogen source to the west of the Mussau Trench

Phys.org: Earth science - Fri, 09/05/2025 - 18:00
Hydrogen is the most abundant element in the solar system. As a source of clean energy, hydrogen is well-suited for sustainable development, and Earth is a natural hydrogen factory. However, most hydrogen vents reported to date are small, and the geological processes responsible for hydrogen formation—as well as the quantities that can be preserved in geological settings—remain unclear.

Discovery of North America's role in Asia's monsoons offers new insights into climate change

Phys.org: Earth science - Fri, 09/05/2025 - 18:00
A study published in the journal Science Advances, indicates how the heating in North America can trigger remote effects in Asia—this could be further exacerbated by anthropogenic global warming and human modification of the North American land surface.

Pulsed biogenic methane identified as key driver of oceanic anoxia during the Mesozoic Era

Phys.org: Earth science - Fri, 09/05/2025 - 17:37
The Toarcian Oceanic Anoxic Event (T-OAE), a major environmental upheaval occurring approximately 183 million years ago during the Mesozoic Era, stands as one of the most severe perturbations to Earth's carbon cycle in geological history.

Mirror image molecules reveal drought stress in the Amazon rainforest

Phys.org: Earth science - Fri, 09/05/2025 - 15:50
In 2023, the Amazon rainforest experienced its worst recorded drought since records began. River levels dropped dramatically and vegetation at all levels deteriorated due to intense heat and water shortages. In such conditions, plants release increased amounts of monoterpenes—small, volatile organic compounds that act as a defense mechanism and help communication with their environment. Some molecules, such as α-pinene, which smells like pine, occur as mirror-image pairs, known as enantiomers.

Physics-based indicator predicts tipping point for collapse of Atlantic current system in next 50 years

Phys.org: Earth science - Fri, 09/05/2025 - 15:45
The Atlantic Meridional Overturning Circulation (AMOC) is an enormous loop of ocean current in the Atlantic Ocean that carries warmer waters north and colder waters south, helping to regulate the climate in many regions. The collapse of this critical circulation system has the potential to cause drastic global and regional climate impacts, like droughts and colder winters, especially in Northwestern Europe.

Quantifying Predictability of the Middle Atmosphere

EOS - Fri, 09/05/2025 - 13:43
Editors’ Highlights are summaries of recent papers by AGU’s journal editors. Source: Journal of Geophysical Research: Atmospheres

Atmospheric circulations are chaotic and unpredictable beyond a certain time limit. Quantifying predictability helps determine what forecast problems are potentially tractable. However, while predictability of weather close to the surface is a much-studied problem, showing a prediction limit of approximately 10 days, less is known about how predictable the atmosphere is at higher layers.

Garny [2025] applies a high-resolution global model to study atmospheric predictability from the surface to the mesosphere/lower thermosphere (MLT; 50-120 kilometers altitude), providing new understanding of coupling between atmospheric levels and fundamental behavior of the upper atmosphere. The author shows that the MLT is somewhat less predictable than lower atmospheric layers due to rapid growth of ubiquitous small-scale waves, with predictability horizons of about 5 days. However, atmospheric flows in the MLT on larger horizontal scales of a few thousand kilometers can remain predictable for up to 3 weeks.

This research highlights the importance of high-resolution, ‘whole atmosphere’ models to understand and predict circulations in the middle atmosphere and coupling from the surface to the edge of space.

Citation: Garny, H. (2025). Intrinsic predictability from the troposphere to the mesosphere/lower thermosphere (MLT). Journal of Geophysical Research: Atmospheres, 130, e2025JD043363. https://doi.org/10.1029/2025JD043363

—William Randel, Editor, JGR: Atmospheres

Text © 2025. 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.

Dust Is the Sky’s Ice Maker

EOS - Fri, 09/05/2025 - 13:10

Dust plays a major role in the formation of ice in the atmosphere. A new analysis of satellite data, published in Science, shows that dust can cause a cloud’s water droplets to freeze at warmer temperatures than they otherwise would. The finding brings what researchers had observed in the laboratory to the scale of the atmosphere and may help climate scientists better model future climate changes.

In 1804, French scientist Joseph Louis Gay-Lussac ascended to about 23,000 feet (7,000 meters) in a hydrogen balloon from Paris, without supplemental oxygen, to collect air samples. He noted that clouds with more dust particles tended to have more frozen droplets.

In the 20th century, scientists found that pure water can remain liquid even when cooled to −34.5°C. But once even tiny amounts of material, such as dust, are introduced, it freezes at much warmer temperatures.

“It’s like Schrödinger’s cat. Either there’s an ice crystal, or there’s a liquid droplet.”

In 2012, researchers in Germany were finally able to test this directly in a cloud chamber experiment. They re-created cloud conditions in the lab, introduced different types of desert dust, and gradually cooled the chamber to observe the temperatures at which droplets froze.

For Diego Villanueva, an atmospheric scientist at ETH Zürich in Switzerland and lead author of the new study, it was striking that scientists had uncovered these processes in the lab, yet no one had examined them in such detail in nature.

The challenges were obvious. To watch an ice crystal nucleate, researchers would need instruments on an aircraft or balloon to catch a micrometer-sized droplet in a cloud at just the right moment. “It’s like Schrödinger’s cat,” said Daniel Knopf, an atmospheric scientist at Stony Brook University who was not involved in the work.. “Either there’s an ice crystal, or there’s a liquid droplet.”

In the new study, Villanueva and his colleagues analyzed 35 years of satellite data on cloud tops across the Northern Hemisphere’s extratropics—a region spanning the U.S. Midwest, southern Canada, western Europe, and northern Asia. The researchers wanted to see whether dust influenced whether cloud tops were liquid or ice. They focused on cloud tops, rather than entire clouds, simply because the tops are visible in satellite imagery.

Desert Dust and Cold Clouds

Villanueva and his colleagues examined two satellite datasets covering 1982–2016, trying to infer microscopic details of cloud tops such as the number of ice crystals or droplet sizes. One dataset tracked whether cloud tops were liquid or ice, and the other measured how much dust was in the air at the same time. Although the team examined global patterns, they focused on the northern extratropical belt, where mixed-phase clouds are common and large amounts of dust from deserts like the Sahara and Gobi circulate.

But the “dataset quality was just so poor that everything that came out was basically just noise,” Villanueva added. In the end, the researchers focused on a simpler detail: the fraction of clouds with ice at their tops. “This took me nearly 3 years,” Villanueva said.

The analysis revealed that regions with more dust had more ice-topped clouds. The effect was strongest in summer, when desert winds lift the most dust.

A distinctive pattern emerged: A tenfold increase in dust roughly doubled the likelihood of cloud tops freezing. “You’d need 100 times more dust to see freezing become 4 times as frequent,” Villanueva explained.

“I think the study is quite elegant.”

The new work showed that the same processes researchers have observed at the microscale in laboratories occur at much larger scales in Earth’s atmosphere. Even after accounting for humidity and air movement, dust remained the key factor for ice nucleation in most instances, though there are exceptions. In some places, such as above the Sahara, few clouds form despite the presence of dust, perhaps, the authors suggest, because the movement of large swaths of hot air prevents freezing.

“I think the study is quite elegant,” Knopf said. He explained that taking 35 years of satellite data, finding a relationship between dust levels and frozen cloud top rates, and then showing that it lines up perfectly with lab experiments is basically “the nail in the coffin” for proving dust’s role in ice nucleation. Scientists now have robust satellite evidence of dust aerosols directly affecting cloud freezing, matching what laboratory experiments had predicted.

The finding has implications for climate modeling. To predict the effects of climate change more accurately, models must account for dust and the ways it affects cloud freezing and helps shape precipitation. Liquid-topped clouds reflect more sunlight and cool the planet, whereas ice-topped clouds let in more sunlight and trap heat.

However, Knopf noted that there is more work to be done to understand exactly what the new observations mean for scientists’ understanding of climate. “If you want to really know the precipitation or climate impacts [of dust], you really need to know the number of liquid droplets or the number of ice crystals,” he said.

Villanueva is motivated to keep looking at clouds and aerosols. In the next 10–20 years, the Earth may have drier surfaces because of climate change, which will likely produce more dust aerosols in the atmosphere. He added, “I want to know how clouds will respond in the scenario.”

—Saugat Bolakhe (@saugat_optimist), Science Writer

Citation: Bolakhe, S. (2025), Dust is the sky’s ice maker, Eos, 106, https://doi.org/10.1029/2025EO250328. Published on 5 September 2025. Text © 2025. 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.

Cruise to Measure Gulf Dead Zone Faces Stormy Funding Future

EOS - Fri, 09/05/2025 - 13:05

This story was originally published in the Louisiana Illuminator.

Despite being called a “cruise,” the people on board The Pelican described the experience on the hypoxia monitoring expedition as very different from the elaborate dinners on a towering vacation ship or booze- and buffet-filled Caribbean itinerary.

Passengers describe waves up to 5 feet high in the Gulf of Mexico, swinging the 116-foot research vessel like a pendulum, plaguing anyone who didn’t have sturdy sea legs with bouts of seasickness. Daytime temperatures in late July soared ever higher as sweat dripped down the backs of hard-hat covered heads.

The guests on board The Pelican weren’t seeking pleasure or status. They were unpaid students and researchers who say they weathered the conditions in the name of science itself.

“It’s not glamorous, but it is very important.”

“It’s not glamorous, but it is very important,” said Cassandra Glaspie, assistant professor at Louisiana State University and the chief scientist for the National Oceanic and Atmospheric Administration’s annual hypoxia cruise.

The 11-day voyage provides vital information on the sea life and environmental conditions within the seasonal low-oxygen zone that develops off the coast of Louisiana. The data the cruise collects informs state and federal efforts to reduce the size of the “dead zone” and sheds light on impacts to those who rely on the water for their livelihoods, like shrimpers and fishermen.

Now, after its 40th year and 38th hypoxia cruise, The Pelican’s annually planned journey faces challenges to stay afloat, potentially undermining decades of research and future plans to get the dead zone under control.

A Decades Long Struggle

Biologists, undergraduate student researchers and crew alike celebrated the cruise’s 40th anniversary aboard The Pelican with a party that had an “old bird” theme, chosen to honor the boat, which has also been sailing for 40 years.

The Pelican and the hypoxia cruise’s 40th anniversaries party on the water. Credit: Yuanheng Xiong

More than just an excuse to eat cake (with rainbow sprinkles), the purpose of the cruise is to capture information snapshots of just how bad conditions get in the dead zone.

“We bring water up to the surface. We have a little chemistry lab…to figure out what the oxygen level is chemically, and then we can validate that against what our sensors are telling us,” Glaspie said.

The low-oxygen area appears annually as nutrients, primarily from agricultural fertilizers from the massive Mississippi River Basin, drain downriver and spur algae overgrowth.

Algae eat, defecate and die, using up the oxygen in the water when they decompose and sink to the bottom. Fish, shrimp and other marine life leave the low oxygen area when they can and suffocate when they can’t, putting pressure on the vital commercial Gulf fishery and the people who rely on it. Exposure to low-oxygen waters can also alter reproduction, growth rates and diet in fish species.

Glaspie took over the work of coastal scientist Nancy Rabalais, who launched the maiden cruise in 1985 and led it for decades after. Every summer begins with a forecast of the zone’s predicted size, estimated by various scientific models and measurements of nitrogen and phosphorus throughout the river basin taken throughout the year.

“A lot of times with pollution, you hear anecdotal evidence of how it might be increasing cancer rates or it might be causing fisheries to fail,” Glaspie said. “Here, we have an actual, measurable impact of nutrient pollution in the Mississippi River watershed.”

The Mississippi River/Gulf of America Hypoxia Task Force, an interagency federal, state and tribal effort to reduce the size of the dead zone, uses data from the cruise to determine whether it is meeting its goals.

In the past five years, the dead zone has been as large as 6,700 square miles, and even larger in previous years, reaching nearly the size of New Jersey.

While still more than two times the size that the Task Force wants, the Gulf dead zone was slightly smaller than forecasted this year, about the size of Connecticut at around 4,400 square miles.

Federal and state officials lauded the limited success of the zone’s smaller size in a July 31 press conference held to discuss the results of the hypoxia cruise’s 2025 findings. They also called for continued work.

“It requires strong collaboration between states, tribes, federal partners and stakeholders,” said Brian Frazer, the EPA’s Office of Wetlands, Oceans and Watersheds director.

Mike Naig, Iowa’s agriculture secretary, said states should be “scaling up” initiatives to reduce nutrient pollution.

Whether or not this will actually happen is uncertain.

Funding Cuts

Since the Trump administration took office, funding for nutrient reduction efforts upriver as well as money to operate the cruise itself have been scaled back or cut entirely.

The Environmental Protection Agency’s 319 and 106 funding programs under the Clean Water Act are the main funding mechanisms for states to reduce nutrient pollution throughout the Basin. Those grants aren’t funded in President Trump’s proposed FY 2026 budget, said Frazer.

The 106 programs have historically doled out $18.5 million annually, according to the EPA, with additional money sometimes allocated from Congress. The 319 program provided $174.3 million in FY 2025.

The cuts to these programs are not yet final. Congress can decide to add in additional funding, and has in past years.

States rely on both funds to reduce and monitor nutrient runoff in their waters, said Matt Rota, senior policy director for Healthy Gulf, a nonprofit research group. Rota has monitored policy changes surrounding the Gulf dead zone for more than 20 years, and he questions whether current reduction strategies can be maintained, let alone efforts redoubled.

“It’s always good to see a dead zone that’s smaller than what was predicted,” Rota said. “I am not confident that this trend will continue.”

“It’s a relatively inexpensive program. … This is baseline stuff that our government should be doing.”

Aside from cuts to reduction efforts, money for The Pelican’s annual cruise is also slipping away. Glaspie said that, ideally, the cruise has 11 days of funding. It costs about $13,000 a day to operate the vessel, she said.

“It’s a relatively inexpensive program” with big payoffs for seafood industry workers who rely on the water for their livelihoods, Rota said. “This is baseline stuff that our government should be doing.”

Funding for the hypoxia cruise has been part of the National Oceanic and Atmospheric Administration’s annual operational budget, making it a more reliable source than grant funding. But with the Trump administration taking a hatchet to government-backed research, there is increasing uncertainty over whether The Pelican and its crew will embark upon future missions.

This year, Glaspie said, NOAA defunded a day of the cruise. The Gulf of America Alliance, a partnership group to support the Gulf’s economic and environmental health amongst the five bordering states, stepped in to make up the difference. Glaspie said having that additional day was a saving grace for the research.

“This is a fine-tuned machine, and the consequences for cutting it short are really predictable and well-known,” she said. “If I’m asked to create an estimate of the surface area of hypoxia, and we’re not able to cap off the end in Texas waters, I’m not really going to be able to give a reliable estimate.”

Even without additional cuts, Glaspie said she already conducts the hypoxia cruise “on a shoestring budget.” Researchers on board don’t get paid, and every person who supports its mission—besides the crew that runs the boat—are volunteers.

“It’s tough for me to not pay people. I mean, they’re working solid 12-hour shifts. It is not easy, and they are seasick for a lot of this, and they can’t call home,” Glaspie said. “It doesn’t sit well with me to not pay people for all this work, but this is what we’ve had to do because we don’t have the money to pay them.”

Students Jorddy Gonzalez and Lily Tubbs retrieve the CTD sensor package after measuring dissolved oxygen at a regular stop on the annual hypoxia cruise while students watch. Credit: Cassandra Glaspie A Rapidly Changing Gulf

Defunding research as climate change intensifies—creating extreme heat in the Gulf—could further undermine hypoxia containment efforts and the consistency of decades worth of data collection.

“I think the rising temperatures is a big question,” Rota said.

“We have 40 years of data, which is almost a gold standard,” Glaspie said. “We’ve just reached that threshold where we can really start to ask some more detailed questions about the impacts of hypoxia, and maybe the future of hypoxia.”

Despite this year’s smaller zone surface area, low oxygen levels went deeper into the water than Glaspie had ever seen before.

“The temperature drops [as the water gets deeper], the salinity increases, and the oxygen just goes basically to zero,” she said.

In some areas, Glaspie’s measurements showed negative oxygen levels.

“Oxygen doesn’t go in the negative. It was just so low that the sensor was having trouble with it,” she said. “It’s the first time I’ve seen it like this.”

She also noticed unusually large amounts of algae on the surface of the water “like ectoplasm in Ghostbusters.”

The smaller-than-forecasted size of the dead zone surprised researchers on The Pelican who saw just how deep the low oxygen levels went.

“None of us really thought until the estimate came out that it was below average size because we’re able to see the three-dimensionality of it. That’s not really incorporated into that estimate,” Glaspie said.

She also noticed unusually large amounts of algae on the surface of the water “like ectoplasm in Ghostbusters.” Toxic algae blooms can kill fish and other sea life as well as poison humans.

“If I had to say what would be important for us to monitor in the future, it would be these algal blooms, and making sure that we’ve got a good handle on which ones have harmful species,” she said.

This is why Glaspie, donned in her sun-protective clothes and work boots, braves the waves, the heat and the journey across the Gulf every year.

“This is our finger on the pulse of our nutrient pollution problem that Louisiana is inheriting from the entire country,” Glaspie said. “We cannot take our finger off that pulse. It is unfair to Louisiana. We have this pollution problem. We need to understand it.”

—Elise Plunk (@elise_plunk), Louisiana Illuminator

This story is a product of the Mississippi River Basin Ag & Water Desk, an independent reporting network based at the University of Missouri in partnership with Report for America, with major funding from the Walton Family Foundation.

Minimizing phase-space energies

Physical Review E (Plasma physics) - Fri, 09/05/2025 - 10:00

Author(s): Michael Updike, Nicholas Bohlsen, Hong Qin, and Nathaniel J. Fisch

A primary technical challenge for harnessing fusion energy is to control and extract energy from a nonthermal distribution of charged particles. The fact that phase space evolves by symplectomorphisms fundamentally limits how a distribution may be manipulated. While the constraint of phase-space vol…


[Phys. Rev. E 112, 035202] Published Fri Sep 05, 2025

Convective nature of the stimulated Raman side scattering in inertial confinement fusion

Physical Review E (Plasma physics) - Fri, 09/05/2025 - 10:00

Author(s): F.-X. Zhou, C.-W. Lian, R. Yan, Y. Ji, J. Li, Q. Jia, and J. Zheng

Absolute growth of stimulated Raman side scattering (SRSS) in inertial confinement fusion appears to be absent in experiments. Based on simulations the authors find that absolute growth of SRSS occurs only in the limit of an infinite laser beam width. This finding may have implications for the design of experiments.

#AdvancingField #OpenDebate


[Phys. Rev. E 112, L033201] Published Fri Sep 05, 2025

High-resolution multi-parameter characterization of the subsurface using full waveform inversion on broadband data: application to the oceanic crust in the North Sea using a dense ocean bottom cable dataset

Geophysical Journal International - Fri, 09/05/2025 - 00:00
SummaryThis study focuses on the hydrophone component of a dense ocean bottom cable dataset from the North Sea. This data had already been used in the past to illustrate the high resolution power of full waveform inversion based strategies. We have developed a highly scalable implementation of a visco-acoustic full waveform inversion engine making it possible to double the frequency content of the inverted data compared to previous studies, using simultaneously up to almost 50,000 CPU. This results in a multi-parameter reconstruction of the subsurface, where the P-wave velocity, the density and the quality factor are reliably reconstructed down to 2 km depth, with a resolution of about 10 meters.

Combining interferometry and wave equation tomography for near surface characterization: 3D imaging of the Harmaliére alpine landslide

Geophysical Journal International - Fri, 09/05/2025 - 00:00
SummaryHarmaliére is an active landslide located in a mountainous region of southern France where the presence of a thick layer of clays provides favorable conditions for the development of slowly moving landslides. However, at Harmaliére, the alternation of sudden reactivation and quiet episodes suggests that specific structural and geomechanical properties control its kinematics. In order to shed light on its subsurface properties, we deployed for one month a dense network of seismic nodes within the landslide and recorded active-source and ambient noise seismic data. These datasets have been first independently processed with dedicated interferometry-based processing and inversion workflows to reconstruct P-wave (active sources) and S-wave (seismic noise) velocity models. Full wave-equation tomography is then performed to improve the reliability and resolution of the obtained elastic model by iteratively fitting the virtual gathers obtained by cross-correlation of the ambient-noise recordings. As opposed to conventional ambient-noise tomography, the approach fully accounts for topography and 3D elastic heterogeneities. The obtained high-resolution 3D models are then qualitatively interpreted in terms of landslide properties and geological lithologies, that can influence landslide kinematics.

Characterizing and Clustering Debris Flow and Environmental Noise Seismic Signals Using Unsupervised Deep Learning

Geophysical Journal International - Fri, 09/05/2025 - 00:00
SummaryDebris flows pose a significant threat to the sustainable development of mountainous regions. As an effective real-time sensing technique, microseismic monitoring plays a critical role in the detection and analysis of debris flow activity. However, current microseismic monitoring technologies face challenges in distinguishing mixed signals originating from different sources, limiting our understanding of the full dynamic evolution of debris flow events. To address this issue, we propose an unsupervised deep clustering-based signal classification framework, which focuses on analyzing the signal characteristics at various stages of debris flow events. A two-dimensional spectrogram dataset was constructed, encompassing signals from debris flows, rockfalls, earthquakes, and environmental noise. A deep autoencoder was employed to compress spectral features into a 16-dimensional latent space, followed by clustering using deep embedded clustering and Gaussian Mixture Models. Experimental results demonstrate that, after optimizing the feature space and data partitioning strategy, the proposed method achieves an average classification accuracy of 96.81 per cent across the four signal types. Power spectral density distribution analysis further confirms that this method not only accurately identifies debris flow signals but also effectively captures their energy distribution and dynamic evolution at different stages. Interpretability analysis reveals strong correlations between the extracted latent features and conventional seismological parameters, particularly the peak count of the time-domain autocorrelation function and the first quartile of the central frequency. Based on this method, a complete segmentation of debris flow events was successfully achieved, revealing the typical signal characteristics and temporal evolution of each stage. Cross-station validation indicates that the proposed framework demonstrates strong robustness and generalization across different monitoring locations. In addition, preliminary exploration of its integration with supervised learning suggests its potential applicability in real-time monitoring scenarios, offering a novel approach for debris flow early warning. This study presents an efficient and intelligent method for debris flow signal recognition and dynamic monitoring.

Bayesian full waveform inversion with sequential surrogate model refinement

Geophysical Journal International - Fri, 09/05/2025 - 00:00
SummaryBayesian formulations of inverse problems are attractive due to their ability to incorporate prior knowledge, account for various sources of uncertainties, and update probabilistic models as new information becomes available. Markov chain Monte Carlo (MCMC) methods sample posterior probability density functions (pdfs) provided accurate representations of prior information and many evaluations of likelihood functions. Dimensionality-reduction techniques such as principal component analysis (PCA) can assist in defining the prior pdf and the input bases can be used to train surrogate models. Surrogate models offer efficient approximations of likelihood functions that can replace traditional and costly forward solvers in MCMC inversions. Many problem classes in geophysics involve intricate input/output relationships that conventional surrogate models, constructed using samples drawn from the prior pdf fail to capture, leading to biased inversion results and poor uncertainty quantification. Incorporating samples from regions of high posterior probability in the training may increase accuracy, but identifying these regions is challenging. In the context of full waveform inversion, we identify and explore high-probability posterior regions using a series of successively-trained surrogate models covering progressively expanding wave bandwidths. The initial surrogate model is used to invert low-frequency data only as the input/output relationship of high-frequency data are too complex to be described across the full prior pdf with a single surrogate model. After a first MCMC inversion, we retrain the surrogate model on samples from the resulting posterior pdf and repeat the process. By focusing on progressively narrower input domain regions, it is possible to progressively increase the frequency bandwidth of the data to be modeled while also decreasing model errors. Through this iterative scheme, we eventually obtain a surrogate model that is of high accuracy for model realizations exhibiting significant posterior probabilities across the full bandwidth of interest. This surrogate model is then used to perform an MCMC inversion yielding the final estimation of the posterior pdf. Numerical results from 2D synthetic crosshole Ground Penetrating Radar (GPR) examples demonstrate that our method outperforms ray-based approaches, as well as results obtained when only training the surrogate model using samples from the prior pdf. Our methodology reduces the overall computational cost by approximately two orders of magnitude compared to using a classical finite-difference time-domain forward scheme.

Theme by Danetsoft and Danang Probo Sayekti inspired by Maksimer