AGU Advances最新文献

Surface meltwater impacts Greenland Ice Sheet mass balance indirectly by reducing albedo and promoting hydrofracture. However, fully understanding both processes requires accurate mapping of small-scale features such as ponds, channels, and moulins that govern meltwater formation and drainage. Here we investigate surface water dynamics at high spatial (∼1 m) and temporal resolution by applying deep learning to high-resolution imagery from the SkySat constellation. We develop a U-Net model that robustly classifies surface meltwater with higher accuracy than a conventional thresholding approach. Our mapping reveals that small water features (<0.015 km²) account for a substantial fraction of surface water area in the western Greenland Ice Sheet ablation zone, especially during May (67%) and August (38%). However, we find that seasonal variability in surface water area is primarily driven by the filling and draining of 12 large supraglacial lakes. The high spatial resolution of the SkySat imagery reveals that much of this variability can be attributed to the development of narrow supraglacial channels that facilitate the drainage of upstream lakes into a single downstream lake. When the downstream lake rapidly drains, we observe synchronous lake drainage across our study site between 13–16 June. This cascading drainage event explains how lakes drain even when they are situated in compressive ice flow regimes and provides an alternative mechanism for synchronous lake drainages typically attributed to transmission of stress perturbations. Our study demonstrates that deep learning applied to high-resolution satellite imagery can provide valuable insights into supraglacial hydrology.

The exponential growth of artificial intelligence (AI) has driven the rapid global expansion of data centers, raising serious concerns about their environmental impact—particularly water use. While national and global water consumption by data centers may seem modest compared to other users, their localized impacts can be significant—especially in regions already facing water stress or drought. This commentary examines the multi-faceted water footprint of data centers, encompassing direct cooling, electricity generation, and supply chain water demands. It highlights major gaps in transparency around how much water data centers use, which undermine effective regulation, innovation, and community planning. To ensure the sustainable growth of digital infrastructure and the preservation of water resources, comprehensive monitoring and public disclosure of water use are essential. Equally important are resilient water infrastructure planning and stronger collaboration between industry and communities.

Nutrient availability is a major driver of ocean biodiversity and productivity, yet long-term global changes remain poorly constrained. Using over 14 million nitrate and phosphate measurements collected between 1925 and 2025, we quantify long-term trends in nutrient concentrations across ocean biomes and depths. We find that surface waters in oligotrophic and mesotrophic regions show significant declines in phosphate, whereas nitrate is also slightly decreasing in oligotrophic but increasing in mesotrophic regions. Many coastal areas exhibit nutrient enrichment. Subsurface waters reveal widespread nitrate accumulation, suggesting an imbalance driven by biological nitrogen fixation and reduced vertical mixing. Comparison with Earth system models indicates that current simulations underestimate the pace of nutrient shifts. Our results highlight a large-scale biogeochemical reorganization of ocean nutrient distributions that may intensify under future climate warming.

Changes in the vertical structure of atmospheric temperature are an important “fingerprint” of human effects on global climate. These changes are mainly driven by human-caused increases in atmospheric levels of $��{\text{CO}}_2�$ and other well-mixed greenhouse gases. Key features of this fingerprint are warming of the troposphere, the lowest layer of the atmosphere, and cooling of the stratosphere, the layer above the troposphere. Cooling in the lower stratosphere (from roughly 15–20 km above Earth's surface) also arises from human-caused depletion of stratospheric ozone. While lower stratospheric cooling diminished in the 21st century, largely due to the emerging “healing” of stratospheric ozone levels after the Montreal Protocol, strong cooling of the mid- to upper stratosphere continued unabated. Satellite observations of this distinctive fingerprint are in accord with current state-of-the-art climate model estimates of human-caused temperature changes. The claim to the contrary made in the recent US Department of Energy review of climate science is factually incorrect.

A more “efficient” biological pump is thought to have played a key role in glacial CO₂ drawdown, with two principal mechanisms invoked to stem the CO₂ “leak” from the modern Southern Ocean. The first sees a strengthened soft tissue pump, associated with a reduction in the ocean's “preformed” nutrient inventory. The second sees an increase in air-sea CO₂ disequilibrium (termed the disequilibrium pump). We use an Earth system model (CM2Mc) to show the tracers radiocarbon (Δ¹⁴C) and oxygen (O₂) exhibit distinct sensitivities to these two pumps within the deep ocean: Δ¹⁴C is more sensitive to disequilibrium pump changes, whereas O₂ is more sensitive to soft tissue pump changes, as expected from the underlying processes. We apply these pump-specific tracer stoichiometries to available deep ocean Δ¹⁴C and O₂ proxy data from the Last Glacial Maximum (LGM). Despite the sparsity of O₂ data, the results show a consistent increase in soft tissue and disequilibrium DIC within the deep ocean of ∼135 μmol/kg, with broadly comparable contributions from the two pumps. Our results imply a reduction in air-sea gas exchange (likely linked to expansion and/or increased isolation of Antarctic Bottom Water), as well as a slowdown of deep ocean overturning and Southern Ocean upwelling at the LGM. Notably, current ocean models struggle to simulate both of these changes simultaneously under glacial forcings. The combined changes in the soft tissue and disequilibrium DIC are of sufficient magnitude to explain the glacial reduction in atmospheric CO₂ once a whole ocean alkalinity increase is accounted for.

Many climate model simulations and limited observations indicate that regions of tropical ascent and precipitation contract in response to surface warming. This response has well-studied implications for the width of the zonal- and annual-mean Intertropical Convergence Zone, but its applicability to zonally asymmetric circulations such as the Pacific Walker circulation remains unknown. Here, we investigate the impact of warming on the area of large-scale ascent in kilometer-scale, mock-Walker simulations with both fixed and interactive surface temperatures. Contrary to the “wet-gets-wetter” and “upped-ante” paradigms of precipitation change, the simulations show a “wet-gets-drier” response to warming in which the ascent region becomes larger and, on average, drier. We attribute these changes to rapid circulation weakening, which limits the transport of moisture into the ascent region. To meet the growing moisture demand for precipitation, local evaporation within the ascent region must increase rapidly, and the ascent region expands to draw moisture from a larger surface area. We link the slowdown of the circulation to increases in gross moist stability driven by a previously unknown mechanism. Central to this mechanism are changes in the vertical structure of the circulation, which features two vertically stacked overturning cells reminiscent of some tropical convergence zones. These results challenge long-held paradigms of tropical precipitation change and show that the vertical structure of tropical circulations can play a critical role in the hydrological response to warming.

Back-propagating earthquakes, characterized by a secondary front reversing into previously ruptured areas, challenge conventional models of rupture dynamics and have been increasingly documented in recent years. The conditions for this phenomenon remain poorly understood: while rupture complexity is often attributed to fault heterogeneity, back-propagating fronts have also been reported on simple faults. Here, we employ earthquake simulations with slip-rate and state-dependent friction to reveal that back-propagating fronts arise spontaneously during unilateral rupture propagation on frictionally homogeneous faults, when the rupture exceeds a critical length about 100 times larger than the nucleation dimension. We propose simple theoretical arguments suggesting that back-propagating fronts are an intrinsic feature of unilateral ruptures under velocity-weakening friction. These conditions preclude self-similar crack-like solutions with constant stress drop; instead, ruptures alternate between pulse and crack-like modes, with back propagation generated at pulse-to-crack transitions. The crack-to-pulse transition is driven by restrengthening and happens when the local slip rate decreases below a specified threshold. An analytical criterion for the extent of back propagation shows that they are enhanced by low rupture velocities and stress drops, and observational examples of back propagation on simple faults are consistent with these requirements. Our study presents a simple, fundamental model for back-propagating earthquakes, suggesting that they may be more prevalent than previously recognized.

The Amazon rainforest is a critical regulator of the global carbon cycle, yet its resilience to compound climate extremes remains uncertain. The severe drought–heat event of 2023, marked by record temperatures and hydrological stress, offers an opportunity to examine the vulnerability of Amazon carbon sink under extreme conditions. In this issue, Botía et al. (2026, https://doi.org/10.1029/2025AV001658) presents a comprehensive, multi-constraint assessment of the impact of 2023 drought-heat extremes on the Amazon carbon cycle. Their synthesis suggests that the Amazon acted as a weak net carbon source in 2023, primarily due to suppressed plant growth during the drought-affected second half of the year. The study also reveals discrepancies between local flux measurements, basin-scale inversions, and dynamical global vegetation model (DGVM) simulations, highlighting persistent challenges related to scale dependence, model structure, and representation of drought processes in DGVMs. Together, these findings underscore both the growing vulnerability of the Amazon carbon sink and the urgent need for sustained, integrated observing and modeling systems to constrain its response to future climate extremes.

The atmospheric lifetime of mercury (Hg) determines its global spread and the delivery of this toxic pollutant to remote ecosystems. Previous studies have generally assumed that the chemical lifetime of elemental mercury (Hg⁰) has remained constant across historical time periods, mirroring present-day (PD, 2010–2019) conditions. However, since preindustrial times (PI, 1850) anthropogenic emissions have altered the concentrations of key oxidants that affect the Hg⁰ lifetime, including bromine radicals (Br), hydroxyl radicals (OH), and ozone (O₃). Here, we use chemistry-climate modeling to analyze the changes in Hg redox chemistry between PI and PD and consequent impacts on Hg transport and deposition. While increasing concentrations of OH and O₃ lead to 16% faster Hg⁰ oxidation in the PD Northern Hemisphere, the increased partitioning of Br to reservoir species slows Hg⁰ oxidation by 20% in the Southern Hemisphere compared to PI. On the global scale, these competing mechanisms lead to an overall buffering of the tropospheric chemical lifetime of Hg⁰. The shift from PI to PD atmospheric composition drives 15% more Hg deposition to tropical and subtropical fisheries, which are the major global source of toxic methylmercury for human exposure. The PI atmosphere was more conducive to the spread of Hg to the remote Southern Hemisphere extratropics, impacting the interpretation of historical records of Hg deposition from natural archives and the supply of Hg to the Southern Ocean marine sediment sink. This study reveals the previously overlooked role of changing atmospheric composition in aggravating human Hg exposure risk via altered deposition patterns.

In 2023, the biogeographic Amazon experienced temperature anomalies of 1.5°C above the 1991–2020 average from September to November. These conditions were driven by high sea surface temperature in the Atlantic and Pacific oceans, together with reduced moisture advection from the Atlantic, causing large vapor pressure and water deficits in the second semester of 2023. Here, we evaluate the response of the Amazon carbon cycle to this extreme event across different spatial scales. We combined atmospheric CO₂ mole fractions and eddy covariance flux data from the Amazon Tall Tower Observatory (ATTO, −2.1441, −58.99), low-latency simulations by Dynamic Global Vegetation Models (DGVMs), an atmospheric inversion, and remote sensing data. We find that in 2023 the Amazon region was, including fires, a net carbon source of 0.01–0.17 PgC. Fire emissions (0.15 [0.13–0.17] PgC) were within typical variability of the 2003–2023 period, thus we attribute the weak carbon source to reduced vegetation uptake during the dry season (August–October). A stronger-than-normal vegetation uptake early in the year (January–April), consistent across data streams and spatial scales, mitigated the total carbon losses by the end of the year. We find a shift from carbon sink to source in May and a peak source in October. Our findings show a reduced vegetation carbon uptake over the Amazon region, leading to a weak carbon source that contributed up to 30% of the net carbon loss in the tropical land in 2023.