Atmospheric and Oceanic Sciences

The El Niño–Southern Oscillation (ENSO) is a dominant mode of interannual climate variability with profound global socioeconomic and ecological impacts 1–3 . Skillful long-lead prediction of ENSO is critical but remains hindered by nonlinear ocean–atmosphere coupling and the formidable spring predictability barrier (SPB) 4–6 . While recent deep learning models have shown promise in extending ENSO forecasts beyond a lead time of one year, they largely operate as opaque “black boxes” that mask underlying physical mechanisms. Here we leverage dynamical system deep learning (DSDL) 7,8 to construct a fully transparent, multivariate “glass-box” ENSO prediction model. Our model achieves skillful predictions up to 19 months ahead, robustly circumventing the SPB and significantly outperforming state-of-the-art models. Notably, it successfully hindcasts the onset, intensity, and decay of the 2015–2016 super El Niño event more than a year in advance, even when initialized during the boreal spring. Physical interpretation reveals that the “glass-box” DSDL model functions as a multi-basin synergistic prediction framework, which further predicts a potential super El Niño development in 2026–2027. By rendering all model terms explicit, this approach transforms the “black box” into a physically accountable dynamical system, representing a fundamental advance in interpretable climate forecasting.

Abstract Propagating waves on the ocean surface can be represented as a stochastic process whose statistics are characterized by a spectrum. This paper reviews methods for measuring the wave spectrum and related quantities. Observations begin by sensing fluid dynamical properties of the sea surface over space and/or time. Visual observations, collected routinely since the mid‐18th century, comprise the longest‐running wave record. Nearshore measurement methods continue to advance, including traditional pressure and acoustic sensing as well as newer technologies like distributed acoustic sensing and LiDAR. Detailed small‐scale wave physics can now be explored with measurement techniques using light, including stereo‐imaging and polarimetry. Reductions in the size, cost, and power consumption of microelectronics have propagated through ocean wave instrumentation, most notably in wave buoys. Global networks of freely drifting miniature wave buoys offer novel observational capabilities. Remote sensing techniques based on radar and LiDAR continue to evolve and are widely deployed from land, ships, aircraft, autonomous vehicles, and satellites. Spaceborne altimeters form one of the most important records of wave height, and new spaceborne sensors now observe directional spectra globally with sampling akin to traditional altimetry. Aircraft and autonomous systems provide strategic sampling capabilities for detailed process studies and access to extreme storm environments. The quality and quantity of ocean wave measurements have never been greater. This review aims to help make sense of it all.

Abstract The Atlantic Meridional Overturning Circulation (AMOC), a tipping element of the climate system, currently has an estimated global warming threshold for collapse of +4.0 °C (uncertainty range 1.4–8 °C). However, such a threshold may not be meaningful because AMOC stability depends on the rate of radiative forcing change, not a set temperature. Here we identify an AMOC stabilizing mechanism that operates on timescales slower than present-day warming rates. Slow forcing permits coherent adjustment of surface and interior ocean properties, supported by enhanced evaporation and reduced sea-ice extent, counteracting destabilizing feedbacks. Using a slow CO 2 ramp (+0.5 ppm yr −1 ) climate model simulation, we explicitly demonstrate the AMOC remains stable up to +5.5 °C of global warming. By contrast, under faster CO 2 ramps, the AMOC collapses at substantially lower warming levels (+2 °C). Our findings demonstrate rate-induced AMOC tipping and imply that limiting the rate of emissions is critical for reducing the risk of an AMOC collapse.

Abstract Europe has experienced extreme floods in recent decades. However, even larger floods are possible and must be considered in flood risk management 1,2 . Their characteristics can be clarified by analysing the largest documented historical floods. In Central Europe, the Magdalena Flood of July 1342 is usually considered the largest of the last millennium; however, knowledge of its characteristics is incomplete 3–7 . Here we show that 16 major flood events occurred across much of Europe between late 1341 and 1343. Four of these events had return periods of 500–1,000 years (the Magdalena, Bartholomew, Candlemas and Jacob Floods). Although Magdalena was thought previously to be the only extreme European flood in 1342 3,7 , our new documentary dataset suggests that it formed part of a broader sequence. The year with the greatest number of extreme floods during the past 700 years was 1342, and 1343 ranks among the top ten. This highly unusual sequence of floods had substantial socio-economic impacts, including a paradigm shift in flood mitigation measures in Europe. A series of volcanic eruptions along with multi-annual Arctic sea ice retreat is a plausible cause of this flood sequence. Clusters of extreme floods occurring within a few months are rarely considered in risk management 8 . Quick and proactive risk strategies are needed that account for this eventuality.

Abstract Extreme vapor pressure deficit (EVPD), representing the 5% hottest and driest surface air conditions, has risen rapidly over the Southern Amazon during the dry season, especially over forested areas. Its frequency has increased by 7 1.5% (or 11 more dry season days) per decade, and its intensity has increased at rates faster than that of mean vapor pressure deficit (VPD), primarily due to rising extreme surface temperatures. Our observation‐based attribution analysis shows that anthropogenic forcings are primarily responsible for 93% and 64% of the increases in EVPD frequency and intensity, respectively. Under a scenario close to current greenhouse gas emission trends (SSP 2–4.5), climate models project that dry season EVPD intensities will reach 7–15 standard deviations above climatological levels by the end of the century, substantially increasing the risks of extreme weather conditions conducive to wildfires in the Southern Amazon.

Abstract This paper investigates the performance of a unique proof-of-concept hybrid model in a cycling data assimilation scheme. This previously published model combines the Simplified Parameterization, primitive-Equation Dynamics model (SPEEDY) with an ML-based component that itself is capable of modeling the global atmospheric dynamics. Analysis and forecast experiments are carried out assuming that ERA5 reanalyses, interpolated to the model grid, represent the “true” spatiotemporal evolution of the atmosphere. Six-hourly simulated observations are generated for a 30-year training period and a one-year testing period by randomly perturbing the “true” states. To investigate the effect of the training data on the model performance, the model is trained on different data sets in the different experiments: the training data are either ERA5 reanalyses, analyses prepared using SPEEDY for cycling, or analyses prepared using the hybrid model for cycling. The simulated observations are assimilated with a Local Ensemble Transform Kalman Filter (LETKF) and the length of the ensuing forecasts is 10 days in all experiments. The cycled LETKF remains stable for the entire testing period in all experiments. When the hybrid model is trained on ERA5 reanalyses, the biases of the analyses are negligible and the variance of the analysis error is greatly reduced compared to the experiment in which SPEEDY rather than the hybrid model is used for cycling. The gains in analysis accuracy are more modest when the hybrid model is trained on analyses obtained with SPEEDY or a prior trained version of the model. All forecasts with the hybrid model are more accurate than with SPEEDY.

Abstract Ozone-temperature sensitivity measures the increase in ozone concentration per Kelvin rise in temperature, serving as a critical metric for assessing ozone responses to climate warming. Here, from extensive observations across the Northern Hemisphere, we show that China exhibits a remarkably higher present-day (2017–2024) summertime surface ozone-temperature sensitivity (mΔO3-ΔTmax; 3.7 ppbv K–1) compared to the US (1.6 ppbv K–1) and Europe (1.8 ppbv K–1), with extremely high values reaching 8–10 ppbv K–1 at the site level. GEOS-Chem model simulations with improved accuracy in reproducing observed mΔO3-ΔTmax attribute this elevated sensitivity predominantly to temperature-indirect effects (66%), comprising comparable contributions from temperature-associated changes in radiation (44%) and transport patterns (51%). These indirect effects dominate the spatial and interannual variability of mΔO3-ΔTmax in China. Direct temperature effects on emissions, chemistry, and deposition account for the remaining 34%. Global models project that stringent emission reductions will decrease China’s future mΔO3-ΔTmax to levels comparable to those of the US and Europe, highlighting emission levels as the primary determinant of the magnitude of ozone-temperature sensitivity. These findings emphasize that sustained emission controls are crucial for mitigating the risks associated with compound heat and ozone pollution.

Abstract. Soil radiocarbon (14C) measurements are crucial for understanding soil carbon cycling over timescales ranging from years to millennia. However, the global synthesis and comparison of radiocarbon data have been limited due to the variety of measurement methodologies and data formats. The International Soil Radiocarbon Database (ISRaD) is an open-access, community-driven archive designed to compile soil radiocarbon data and facilitate large-scale research on soil carbon dynamics. Here, we present ISRaD version 2 (v2), which has grown significantly since its initial release in 2020 (https://doi.org/10.5281/zenodo.17860507, Beem-Miller et al., 2025). It now contains data from 515 unique studies spanning 1669 sites globally, with over 20 000 radiocarbon observations across multiple hierarchical levels, including bulk soil layers, soil fractions, laboratory incubations, interstitial carbon in soil pores, and in situ fluxes of CO2 and CH4. Major updates include expanded metadata structures to capture emerging measurement techniques and an improved soil fractionation template to better capture diverse methods. There has also been a substantial increase in data from underrepresented ecosystems, including cultivated soils and wetlands. Despite this growth, significant geographic and data-type gaps persist. Tropical and arid regions, soils deeper than 100 cm, and certain types of measurements, including incubation, interstitial, and flux, are severely undersampled. We discuss the scientific advances enabled by ISRaD v1 and the major updates to the database and data representation. We also explore future opportunities for ISRaD and the soil radiocarbon community. ISRaD v2 continues to serve as a living archive and dynamic platform for the soil radiocarbon research community. It supports synthesis efforts that are critical for predicting how soil carbon will respond to environmental and climatic changes.

Abstract Transboundary river basins often operate under information asymmetry, where upstream dam operations remain obscure due to limited monitoring and/or restricted data sharing. Under increasing climate variability and intensifying hydrological extremes, such operational opacity can amplify downstream flood risk, undermine preparedness, and heighten transboundary tensions. This study demonstrates that upstream dam operations can be inferred using satellite observations alone in such data-restricted settings. Using a politically constrained transboundary basin for illustration, we reconstruct key hydrological signals—reservoir water level, inflow, and release behavior—by integrating multi-source satellite observations and hydrological modeling. Reservoir water levels are reconstructed by combining satellite altimetry, optical imagery, and synthetic aperture radar (SAR), substantially reducing temporal gaps in observations. Reservoir inflow is estimated using a rainfall–runoff framework forced by satellite-based precipitation and evaporative demand, with surrogate discharge derived from L-band microwave signals. Independently, downstream SAR-derived surface water dynamics are used to construct a surrogate discharge signal that provides an external constraint on inferred releases and shows strong agreement with available gauge observations (r > 0.8). The inferred dam operation patterns are physically coherent, reproduce observed water-level dynamics with low error, and capture extreme flood conditions, including near pass-through behavior during peak inflow events. Comparison of alternative operation modules shows that incorporating even simple hydraulic constraints substantially improves physical plausibility and consistency with downstream observations, highlighting the importance of operational realism in satellite-informed inference. By transforming spaceborne observations into operational insight without reliance on in situ measurements or data sharing, this framework demonstrates how satellite- informed hydroinformatics can partially reduce information asymmetry in politically sensitive basins. The approach is transferable to data-scarce regions worldwide and provides a pathway toward improved transparency, downstream risk awareness, and integrated water resources management under climate stress.

ABSTRACT Climate change is driving shifts in the spatial distribution of climate zones worldwide, with particularly strong impacts in mid‐latitude regions. The Argentine Pampas, one of the world's most important agro‐productive areas, provides a key case for assessing these dynamics. This study analyses Köppen–Geiger climate transitions across four consecutive 30‐year periods (1901–1930, 1931–1960, 1961–1990 and 1991–2020) using high‐resolution maps and a pixel‐by‐pixel approach. Transitions were classified into nine categories, distinguishing single‐driver processes—changes in aridity, summer temperature or precipitation seasonality alone—from compound transitions in which precipitation seasonality and summer temperature shifted simultaneously within the same pixel. Results reveal that climate transitions affected between 8.6% and 14.3% of the regional area across consecutive periods, consistently exceeding global averages. The temporal trajectory is non‐linear, characterised by successive reversals in the dominant process: early 20th‐century aridity increase, mid‐century aridity reduction with the retreat of semi‐arid conditions and a recent dominance of winter drought development. Compound transitions remained spatially marginal in every period (0.5% or less), indicating that hydroclimatic and thermal shifts generally operated as spatially distinct processes across the region. In the most recent period (1961–1990 to 1991–2020), transitions affected 11.6% of the Pampas, with winter drought development accounting for over two‐thirds of all changes, primarily through shifts from Cfa to Cwa climates—a pure single‐driver hydroclimatic transition. Crop exposure analysis indicates that more than 3.1 million hectares of cropland fall within transition zones, with 91% concentrated in the winter drought development belt, encompassing the core soybean–maize–wheat production system. These results demonstrate that climate change in the Pampas is expressed primarily as a reorganisation of hydroclimatic regimes and precipitation seasonality rather than as uniform trends in annual conditions. The identified transition rates and spatial patterns position the region as a mid‐latitude climate change hotspot and highlight the need for territorially differentiated adaptation strategies in highly productive agricultural systems.