Feed aggregator

The impacts were severe: Within a very short time, tropical storm Doksuri intensified into a super typhoon in July 2023. Exceptionally strong winds tore roofs from houses along the coasts of China and the Philippines, trees were uprooted, and torrential rain flooded streets and residential areas. In many places, everyday life came to a temporary halt.

California is no stranger to the hot, dry summer weather that makes wildfires more likely. But wildfire season in the state is now stretching into the heart of winter, when it has historically been protected by cool, wet weather. In January 2025, Southern California experienced some of the deadliest and costliest wildfires in the state’s history.

Now, a new study published in Nature Communications shows that the climatic changes that increase the risk of these winter wildfires could be driven by low autumn snow levels thousands of miles away, in western Eurasia. The authors said that tracking snowfall in Eurasia could help forecast winters in California that will have higher chances of wildfires.

The researchers were motivated by the catastrophic 2025 wildfires to search for climate drivers of winter wildfire conditions in California. First, they looked for correlations between winter wildfires and ocean temperatures, especially La Niña events that are associated with drier-than-average conditions in California. They also examined variability in sea ice, which can affect global weather patterns. But they saw only weak connections.

Compared to oceans and sea ice, the influence of snow cover on global weather patterns is less studied, said Shineng Hu, a climate scientist at Duke University and lead author of the paper. But another climate researcher in Hu’s lab, Shizuo Liu (first author of the study), had previously studied the connection between snow cover and weather patterns and suggested the team look for connections between snow and fires. That’s when they found significant correlations between the winter wildfires in California and low snow cover in western Eurasia.

“When I saw the result, I was suspicious,” Hu said, “because we all know that correlation doesn’t mean causality.” But they ran hundreds of climate model simulations reducing snow cover in Eurasia and saw an increased probability of winter fires in California. “At that point, we were pretty much convinced that there could be something interesting happening over there,” Hu said.

Propagating Pressure

“I’m glad to see this group saying snow can do something similar to what ocean temperature anomalies can do.”

The scientists determined that this intercontinental link starts because the land absorbs more energy when snow cover is low, disturbing the atmosphere above it. This disturbance, like a stone thrown into water, generates large waves of air called Rossby waves that travel eastward along the jet stream across the Pacific Ocean. The Rossby waves drive the formation of a high-pressure zone that creates the hot, dry, windy conditions conducive to wildfires.

“I’m glad to see this group saying snow can do something similar to what ocean temperature anomalies can do,” said Judah Cohen, a climatologist at the Massachusetts Institute of Technology who was not involved in the study but has also studied the links between snow in North America and Eurasia. “I’ve been surprised by how important this mechanism is for U.S. weather in the winter and how little there is about it in the literature.”

“This is just one missing gap that people didn’t even realize. We want to add that to the table.”

But Cohen suggested the study tells only part of the story. In North America, dry winters in the west are paired with wet, cold winters in the east. The same is true in Eurasia, and according to Cohen’s past research, when snow levels are low in western Eurasia but high in eastern Eurasia, a temperature and pressure gradient is created across the continent. The energy released as the atmosphere works to equalize that pressure drives the Rossby waves. Cohen said the disparity between snow levels in eastern and western Eurasia would likely strengthen the Rossby waves and then the warming in California. “If all of Eurasia [had] below normal [snow levels], I don’t think you could easily excite this wave energy that propagates across the hemisphere.” He also stressed that Rossby waves don’t just travel eastward. They also travel upward into the stratosphere, where they bounce back down over North America and intensify the high pressure over the western United States.

Both Cohen and the study authors insisted that many other factors influence whether wildfires ignite in winter. “This is just one missing gap that people didn’t even realize. We want to add that to the table,” said Hu. But monitoring snow levels in Eurasia could offer signs of bad wildfire winters to come. The January 2025 Southern California fires were preceded by low snow levels in November and December in Eurasia, Hu said. “So there’s a 1‑month lag, which gives us some hope that we can use that for prediction.”

—Andrew Chapman (@andrewchapman.bsky.social), Science Writer

Citation: Chapman, A. (2026), Low snow in Eurasia linked to wildfires in California, Eos, 107, https://doi.org/10.1029/2026EO260138. Published on 13 May 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.

For decades, regulators built their ocean monitoring programs mainly around pesticides and pharmaceuticals, treating them as the primary chemical threat to ecological and human health.

That assumption left a much larger category of compounds largely unexamined: the industrial chemicals embedded in packaging, furniture, and everyday personal care products. Those chemicals, it turns out, have been spreading widely. And they’re now showing up even in the places some might consider pristine, such as coral reefs in the Caribbean.

These compounds are biologically active, some interfere with microbial metabolism, and according to a sweeping meta-analysis published in Nature Geoscience, they may be altering how the ocean cycles carbon, one of our planet’s most critical biogeochemical processes.

“Beyond the usual [pesticides and pharmaceuticals], what really surprised us was that everyday industrial chemicals are showing up at even higher levels and not just in coastal or polluted areas, but pretty much everywhere,” said Daniel Petras, a biochemist at the University of California, Riverside.

Led by Petras and Jarmo-Charles Kalinski, a postdoctoral fellow at the Rhodes University Biotechnology Innovation Centre, the study reanalyzed 21 publicly available datasets comprising seawater samples collected over more than a decade across the Pacific, Indian, and North Atlantic Oceans, including the Baltic and Caribbean Seas.

All groups the researchers examined—industrial pollutants, pharmaceuticals, and pesticides—belong to a class called xenobiotics: human-made organic compounds that are foreign to natural systems. Pesticides and pharmaceuticals were prevalent in coastal samples, as expected, given their well-documented entry through agricultural runoff and wastewater outfalls.

But industrial compounds behaved differently. Polyalkylene glycols used in hydraulic fluids, phthalates from polyvinal chloride (PVC) packaging, organophosphate flame retardants from furniture and electronics, and surfactants from personal care products proved far more widespread across all ecosystem types than either pesticides or pharmaceuticals. “These are chemicals we use all the time,” Petras said, “so they end up spreading widely.”

Glimpsing What Was Always There

To map the ocean’s full chemical landscape, the researchers analyzed more than 2,300 samples from temperate coastal zones, coral reefs, and the open ocean, searching for the presence of xenobiotics and examining the share of dissolved organic matter (DOM), a pool of carbon-containing molecules dissolved in seawater. In total, the team identified 248 known xenobiotic molecules. Their work offers the most comprehensive chemical map of anthropogenic organic pollution in the ocean to date.

Researchers used nontargeted mass spectrometry paired with scalable computational tools. Unlike conventional targeted analysis, which tests only for a predefined list of known hazardous molecules, this open-ended approach can detect thousands of chemicals simultaneously, even at low concentrations. The team then applied molecular networking, a computational technique that enables the identification of not only known substances but also their “families” or derivatives.

Coral Reefs as Far-Flung Hot Spots

“Our traditional idea of ‘pristine’ needs a serious rethink, as anthropogenic potential sources are now present nearly everywhere.”

For Petras, it was surprising to find these compounds in coral reefs like those in French Polynesia, which are typically viewed as perfect, “postcard-style” paradises. Yet closer examination reveals that these areas are, indeed, rarely isolated. Agriculture, urban runoff, hotel infrastructure, and cruise ship traffic all contribute pollutants. Remnants of human activity, such as sunscreen, wastewater, and boat fluids, are concentrated near reefs.

“We specifically detected plasticizers and flame retardants even in these remote areas,” Petras said. “This suggests that our traditional idea of ‘pristine’ needs a serious rethink, as anthropogenic potential sources are now present nearly everywhere.”

Anastazia T. Banaszak, a researcher at the Reef Systems Unit of the Universidad Nacional Autónoma de México who was not involved in the study, stressed the broader implications for reef conservation: “Inadequately treated urban wastewater discharges pose a risk to coral reefs and the success of restoration projects,” she said. Such discharges raise nutrient levels, fueling macroalgal blooms that grow faster than corals and compete with them for space. This pressure on ecosystems is intensifying as climate change shifts the baseline against which restoration outcomes are measured, Banaszak noted.

Carbon…and Microbes?

Beyond reefs, these synthetic compounds could be affecting the ocean’s carbon cycle. DOM is one of Earth’s largest carbon reservoirs, comparable in size to all the carbon dioxide (CO2) in the atmosphere. Marine microbes transform it from readily degradable forms into biologically resistant ones; refractory DOM that escapes microbial consumption accumulates in the ocean and acts as an important climate regulator.

But with industrial compounds representing up to 63% of DOM in some estuarine samples (with a global estimate of 10%), the microbial loop is, perhaps, facing chemical conditions it did not evolve to handle. This shift means the efficiency of the ocean’s carbon pump, the mechanism that pulls CO2 from the atmosphere, could be compromised in ways that are not yet understood.

“The data suggest they are present at substantial levels,” Petras said. “Enough that they should be considered in models of carbon cycling.”

Handling the Invisible

Finding xenobiotics is only the first step, the authors say. They laid out several suggestions for next steps. For instance, governments should mandate open-ended approaches as a standard monitoring tool, not just targeted testing of preselected chemicals. Oceanographic data also should be publicly available and standardized, following FAIR (findable, accessible, interoperable, reusable) principles.

“There’s already a strong track record of building long-term datasets for things like trace metals and nutrients. I hope that nontargeted analysis could become part of such long-term efforts,” Petras concluded. “We’ve been quite active in establishing these tools for the community.”

—Mariana Mastache-Maldonado (@deerenoir.bsky.social), Science Writer

Citation: Mastache-Maldonado, M. (2026), Have we been focusing on the wrong ocean pollutants? This study maps what we’ve been missing, Eos, 107, https://doi.org/10.1029/2026EO260151. Published on 13 May 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.

Geological faults hold many secrets that may help us answer important questions about the nature of our planet and what really happens deep underground. One of the biggest mysteries lies within the Atotsugawa Fault System in Japan. What makes the area unusual is that, despite being in a tectonically active region where Earth's plates are constantly shifting, it does not produce as many large earthquakes as other major faults.

Editors’ Vox is a blog from AGU’s Publications Department.

Ensuring the sustainability of water resources and ecosystems in a changing world requires a thorough understanding of how water moves through Earth’s Critical Zone, a dynamic interface where air, water, soil, plants, and rocks interact. Researchers can track and model this movement of water using naturally occurring markers or “tracers.”

A recent article in Reviews of Geophysics explores the latest advancements in tracer-aided mixing models and how they can help us to better understand the Critical Zone. Here, we asked the authors to give an overview of the Critical Zone, how tracer-aided mixing modeling works, and future directions for research.

What is the Critical Zone (CZ)?

The Critical Zone is Earth’s “living skin”—the dynamic layer where the atmosphere, hydrosphere, biosphere, and lithosphere interact. It stretches from the top of the vegetation canopy and, in cold regions, from the surface of snowpacks and glaciers, down through soils and into the deeper aquifers. It encompasses lakes, streams, and wetlands at the surface, and extends beyond the soil layer to underlying groundwater aquifers. It is where rainfall, snowmelt and glacier melt become soil moisture, where plants take up water and return it to the atmosphere, where aquifers get recharged, and where streamflow is generated. In short, the Critical Zone is where most processes that sustain terrestrial life and freshwater resources unfold.

Why is it important to understand how water moves through the Critical Zone?

Virtually every freshwater resource we rely on (e.g., drinking water, irrigation) passes through the Critical Zone.

Virtually every freshwater resource we rely on (e.g., drinking water, irrigation) passes through the Critical Zone at some point. Global warming, land-use changes, and intensifying water demand emerging from rapid urbanization and changes in agriculture are reshaping how water is stored and released within the Critical Zone, often in ways we cannot yet predict. Understanding how much water is stored within the Critical Zone, how this water is both recharged from rainfall and snowmelt and eventually discharged into streams, and the timescale of these dynamic processes is essential for protecting ecosystems, safeguarding water supplies, and adapting to a changing climate.

How would you explain a tracer-aided mixing model to a non-specialist?

Imagine mixing a glass of orange juice with a glass of apple juice, and trying afterwards to work out how much of each went into the glass. If the juices had distinctive “fingerprints” (imagine its color, sugar content, or a specific chemical) and these fingerprints primarily changed because of the mixing of these two juices, you can then measure the fingerprint in the final mixture and back-calculate the proportion of its distinct sources.

Tracer-aided mixing models work in a similar way but can track the entire water cycle. Different water sources (e.g., rainfall, snowmelt, glacier melt, soil water, groundwater) can have distinct “fingerprints” in a naturally occurring tracer, such as stable isotopes of water or specific dissolved elements. By measuring these fingerprints in the streamwater or groundwater and in its potential sources for example, hydrologists can estimate how much each source contributed to the streamwater or groundwater.

Conceptual model of the different components of the Critical Zone. “Gw” stands for groundwater. Credit: Popp et al. [2025], Figure 2

What are some of the most significant and exciting recent advances in tracer-aided mixing models?

Classical mixing models relied on demanding assumptions: that all water sources can be identified and sampled, and that their signatures were distinct and constant in time. Much of the recent progress has been about relaxing these assumptions.

Bayesian approaches now estimate full probability distributions and provide a more realistic picture of uncertainty. Methods like Convex Hull End-Member Mixing Analysis (CHEMMA) use machine learning to infer the distinct sources directly from data, while ensemble hydrograph separation exploits tracer fluctuations over time, thereby making un-mixing feasible even when multiple sources have overlapping signatures. Perhaps the most conceptually novel advance is end-member splitting, which flips the question from “where does streamflow come from?” to “where does precipitation go?”

Alongside these modeling advances, there have been immense advances in how tracers are measured. Portable laser and mass spectrometers now enable high-frequency, in-situ tracer measurements which allows us to capture critical hydrological events such as storms and snowmelt in near-real time.

What are stable water isotope tracers and what are their advantages?

Stable water isotopes are naturally occurring non-radioactive atoms of hydrogen and oxygen that make up a water molecule but have slightly different molecular masses. The two stable isotopes widely used in hydrology are 2H (deuterium) and 18O (oxygen-18). Because these isotopes are part of the water molecule itself, they directly travel with the water molecule. Their key advantages are: (1) they are conservative, meaning they do not react chemically as water moves through soils and aquifers, and (2) they carry distinct signatures resulting from climatic variables such as air temperature.

These properties make stable water isotopes the most versatile and widely used tracer in Critical Zone hydrology.

Consequently, in the European Alps, winter precipitation has a different isotopic signature than summer precipitation because winters are cooler than summers. Other hydrological processes such as evaporation and sublimation leave a recognizable fingerprint on the remaining water, thereby allowing us to estimate how much evaporation or sublimation occurred. Stable water isotopes can be measured in essentially every water compartment, from atmospheric vapor and precipitation to snowpack, plant xylem, soil water, streams, and groundwater. Together, these properties make stable water isotopes the most versatile and widely used tracer in Critical Zone hydrology.

What are the current limitations of tracer-aided mixing models?

Despite their power, mixing models still face many constraints. End-member signatures vary in space and time, are sometimes too similar to distinguish, and some sources may be overlooked entirely. Non-conservative tracers such as nitrate or sulfate can react with their environment along their journey, thereby biasing results if these reactions are not explicitly accounted for.

Sampling is another major bottleneck. Capturing the spatial heterogeneity of soils, snowpacks, and groundwater requires a lot of measurements that are often logistically or financially prohibitive, especially in remote regions. Many of the newer, more powerful tracers such as noble gases or stable isotopes of trace elements, can only be analyzed by a handful of specialized laboratories. As a result, global coverage remains highly uneven, with key regions such as the Arctic and the global South still under-sampled.

What are some of the major unsolved questions and where is more research needed?

There are several fronts where more research is needed. Source signatures are not static, and methods that explicitly capture their variability in time are still underdeveloped. Embedding tracers within global Earth System Models would, in theory, enable more accurate assessment of hydrological partitioning e.g., how rainfall, snowmelt, and glacier melt are split between sublimation, evapotranspiration, groundwater, and streamflow. These will directly inform more robust climate projections, but this remains technically demanding.

Expanding data coverage in under-sampled regions is critical, and citizen science and low-cost sensors may help. Machine learning is a promising approach for uncovering non-linear relationships and gap-filling sparse datasets, but requires training data that often do not yet exist. Greater interdisciplinary integration, e.g., combining tracers with remote sensing, ecological indicators, and biogeochemical data, could yield a more holistic view of the Critical Zone. Finally, the field would benefit from shared protocols and open data practices to enhance progress.

—Andrea L. Popp (andrea.popp@smhi.se; 0000-0003-3911-8105), Swedish Meteorological and Hydrological Institute, Sweden; Harsh Beria (hberia@ethz.ch; 0000-0003-2597-9449), ETH Zurich, Switzerland

Editor’s Note: It is the policy of AGU Publications to invite the authors of articles published in Reviews of Geophysics to write a summary for Eos Editors’ Vox.

Citation: Popp, A. L., and H. Beria (2026), Tracing water’s hidden journey through the Earth’s living skin, Eos, 107, https://doi.org/10.1029/2026EO265019. Published on 13 May 2026. This article does not represent the opinion of AGU, Eos, or any of its affiliates. It is solely the opinion of the author(s). 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.

Tropical forests store more than 60% of the world's vegetation biomass and are among the most important ecosystems for regulating the global carbon cycle and climate. However, their regulatory role is greatly influenced by the forests' carbon residence time—how long carbon remains in the vegetation biomass pool before it is released again into the atmosphere. This figure is tied to the rate of biomass turnover—how quickly vegetation is replaced through growth and mortality.

It's autumn in the Southern Hemisphere, which means it's fog season in the Victorian Alps. NASA's Terra satellite captured this view of morning fog filling valleys in several national parks across the mountains of eastern Victoria in May.

Abstract

To meet the high-precision and high-integrity positioning demands of safety–critical applications, monitoring the quality of precise satellite products in global navigation satellite system (GNSS) precise point positioning (PPP) is crucial. This work employs ionosphere-free (IF) PPP with ambiguity resolution (PPP-AR) phase residuals to construct test statistics for monitoring the quality of precise satellite corrections. By utilizing precise satellite orbit and clock products from CODE, WUM, and GRG, the PPP-AR phase residuals were first analyzed with sample moments, Allan variance and power spectral density (PSD). The key findings are as follows: (1) The skewness and kurtosis results indicate that ambiguity-fixed phase residuals deviate from an ideal zero-mean Gaussian distribution and exhibit a super-Gaussian distribution. (2) Allan variance and PSD analysis reveal that flicker noise dominates the phase residuals. (3) The noise amplitudes are similar for all satellites, but certain differences are observed among different GNSS systems and satellite types. (4) The noise level of phase residuals is influenced by the receiver types, antenna types, and precise products from different analysis centers. Leveraging the error characteristics, the two-step Gaussian overbounding (OB) method was employed to estimate the corresponding OB parameters of the phase residuals. The overbounding results demonstrate that, under similar conditions, phase residuals can be bounded by the calculated bound within the acceptable integrity risk after removing the detected outliers. Anomaly monitoring experiments further show that phase residuals can effectively capture anomalies in precise satellite corrections, with the set threshold successfully detecting such anomalies.

The measurement of virtual height of the sporadic E layer (h'Es) is very sensitive to the type of ionosonde used and the calibration processes. The ionosondes used by the national institute of communication an...

Abstract

Solar photospheric abundances and CI-chondrite compositions are reviewed and updated to obtain representative solar system abundances of the elements and their isotopes. The new photospheric abundances obtained here lead to higher solar metallicity. Full 3D NLTE photospheric analyses are only available for 11 elements. A quality index for analyses is introduced. For several elements, uncertainties remain large. Protosolar mass fractions are H (X = 0.7060), He (Y = 0.2753), and for metals Li to U (Z = 0.0187). The protosolar (C+N)/H agrees within 13% with the ratio for the solar core from the Borexino experiment. Elemental abundances in CI-chondrites were screened by analytical methods, sample sizes, and evaluated using concentration frequency distributions. Aqueously mobile elements (e.g., alkalis, alkaline earths, etc.) often deviate from normal distributions indicating mobilization and/or sequestration into carbonates, phosphates, and sulfates. Revised CI-chondrite abundances of non-volatile elements are similar to earlier estimates. The moderately volatile elements F and Sb are higher than before, as are C, Br and I, whereas the CI-abundances of Hg and N are now significantly lower. The solar system nuclide distribution curves of s-process elements agree within 4% with s-process predictions of Galactic chemical evolution models. P-process nuclide distributions are assessed. No obvious correlation of CI-chondritic to solar elemental abundance ratios with condensation temperatures is observed, nor is there one for ratios of CI-chondrites/solar wind abundances.

The January 2022 Tonga volcanic eruption generated atmospheric pressure waves that propagated over the ocean’s surface and triggered a meteotsunami. This meteotsunami caused significant amplitudes exceeding 10...

Abstract

Accurately modeling and prediction the nonlinear motion of GNSS (Global Navigation Satellite System) coordinate time series holds significant theoretical and practical value for the study of geodynamics. A novel integrated network, named Ensemble Learning method based on Signal Source Driver (ELSSD), is proposed, which leverages the strengths of Long Short-Term Memory (LSTM) and Deep Self-Attention Neural Network (DSANN), while integrating GNSS loading data as an additional data source. Additionally, a multi-track synchronous sliding window data processing strategy is designed to address the challenge of multi-source data fusion input. The effectiveness of this algorithm is validated using GNSS coordinate time series from 186 global stations over a period of 10 years. Experimental results initially illustrate that, when accounting for displacement caused by environmental loading effects, there is a marked improvement in the modeling and prediction accuracy compared with GNSS input-only. Furthermore, the application of three ensemble network strategies-Bagging, Boosting, and Stacking-have further been demonstrated to enhance modeling and prediction accuracy. Compared with LSTM and DSANN networks, the proposed ELSSD algorithm achieves an average RMSE (Root Mean Square Error) of 3.6 mm for both modeling and prediction, with modeling accuracy improvements of 4.8% and 6.2%, while prediction accuracy improvements of 5.4% and 5.9%, respectively. With respect to the traditional Least Square method, there is an improvement of 22.1% and 27.9% in modeling and prediction accuracy, respectively. Regarding noise characteristics, there is a significant reduction in colored noise amplitude, with decreases of 36.7% and 36.0% observed in modeling and prediction, respectively. Simultaneously, the velocity uncertainty experiences an average reduction of 27.1% and 27.5%. The average velocity differences are measured at 0.06 mm/year and 0.24 mm/year, respectively. Hence, our findings suggest that the ELSSD algorithm emerges as an effective methodology for handling multi-source data input in GNSS coordinate time series, presenting promising practical applications in the field.

The Mw 7.5 Noto Peninsula earthquake, which occurred on January 1, 2024, was considerably hazardous to the peninsula and surrounding regions owing to a strong motion, large-scale crustal deformation, and subse...

Nature Geoscience, Published online: 21 February 2025; doi:10.1038/s41561-025-01653-z

Quartz-rich clasts in Martian meteorite NWA 7533 indicate the presence of granitic rocks on early Mars that formed via hydrothermal activity and impact melting, according to petrologic and in situ geochemical analyses.

Abstract

The predictive learning of total electron content (TEC) spatiotemporal sequences aims to generate future TEC maps by learning from historical data, where both the spatial appearances and temporal variations are crucial for accurate predictions. However, the state-of-the-art TEC map prediction models typically employ sequential stacking of ConvLSTM, ConvGRU, and their variants. These models focus more on modeling temporal variations, and the spatial features extracted from the historical sequence are highly abstracted, resulting in the fine-grained spatial appearances not being adequately memorized or transmitted, leading to fuzzy prediction results during storm time. In this paper, we used PredRNN to propose a storm-time ionospheric TEC spatiotemporal prediction model with multichannel features, named Multichannel PredRNN, which can simultaneously remember the temporal patterns and spatial appearances in input sequence. The temporal memory as well as the spatial memory are updated repeatedly over time, ensuring that both temporal memory and spatiotemporal memory are fully utilized in prediction. According to Dst index, 60 magnetic storm events from 2011 to 2019 were selected as the dataset. We first discussed the impact of feature combinations on predictive performance. The results show that using multichannel feature (TEC + Dst&F10.7), the Multichannel PredRNN and the comparison models ConvGRU and ConvLSTM have the best prediction performance. Then we used the optimal feature combination for prediction. We compared Multichannel PredRNN with IRI-2016, COPG, ConvLSTM and ConvGRU under various conditions, including the entire test magnetic events, periods of quiet and storm, different phases of geomagnetic storms, and the most severe geomagnetic storms. Finally, we compared the performance of different output steps. The experimental results indicate that in all cases, Multichannel PredRNN with dual memory state and zigzag flow is superior to four compared models.

Abstract

The gravimetry measurements from the Gravity Recovery and Climate Experiment (GRACE) and its follow-on (GRACE-FO) mission provide an essential way to monitor changes in ocean bottom pressure ( \(p_b\) ), which is a critical variable in understanding ocean circulation. However, the coarse spatial resolution of the GRACE(-FO) fields blurs important spatial details, such as \(p_b\) gradients. In this study, we employ a self-supervised deep learning algorithm to downscale global monthly \(p_b\) anomalies derived from GRACE(-FO) observations to an equal-angle 0.25  \( ^{\circ }\) grid in the absence of high-resolution ground truth. The optimization process is realized by constraining the outputs to follow the large-scale mass conservation contained in the gravity field estimates while learning the spatial details from two ocean reanalysis products. The downscaled product agrees with GRACE(-FO) solutions over large ocean basins at the millimeter level in terms of equivalent water height and shows signs of outperforming them when evaluating short spatial scale variability. In particular, the downscaled \(p_b\) product has more realistic signal content near the coast and exhibits better agreement with tide gauge measurements at around 80% of 465 globally distributed stations. Our method presents a novel way of combining the advantages of satellite measurements and ocean models at the product level, with potential downstream applications for studies of the large-scale ocean circulation, coastal sea level variability, and changes in global geodetic parameters.

Abstract

Data-driven technologies have shown promising potential for improving GNSS positioning, which can analyze observation data to learn the complex hidden characteristics of system models, without rigorous prior assumptions. However, in complex urban areas, the input observation data contain task-irrelevant noisy GNSS measurements arising from stochastic noise, such as signal reflections from tall buildings. Moreover, the problem of data distribution shift between the training and testing phases exists for dynamically changing environments. These problems limit the robustness and generalizability of the data-driven GNSS positioning methods in urban areas. In this paper, a novel deep reinforcement learning (DRL) method is proposed to improve the robustness and generalizability of the data-driven GNSS positioning. Specifically, to address the data distribution shift in dynamically changing environments, the robust Bellman operator (RBO) is employed into the DRL optimization to model the deviations in the data distribution and to enhance generalizability. To improve robustness against task-irrelevant noisy GNSS measurements, the long-term reward sequence prediction (LRSP) is adopted to learn robust representations by extracting task-relevant information from GNSS observations. Therefore, we develop a DRL method with robust augmented reward sequence prediction to correct the rough position solved by model-based methods. Moreover, a novel real-world GNSS positioning dataset is built, containing different scenes in urban areas. Our experiments were conducted on the public dataset Google smartphone decimeter challenge 2022 (GSDC2022) and the built dataset Guangzhou GNSS version 2 (GZGNSS-V2), which demonstrated that the proposed method can outperform model-based and state-of-the-art data-driven methods in terms of generalizability across different environments.

Nature Geoscience, Published online: 18 February 2025; doi:10.1038/s41561-025-01648-w

Global sea-level rise during Meltwater Pulse 1A followed sequential ice loss from the Laurentide, Eurasian and then West Antarctic ice sheets, according to a fingerprinting approach.