Atmospheric and Oceanic Sciences

Abstract Accurate radar echo extrapolation is critical for short‐term weather forecasting, yet existing deep learning methods often suffer from echo ambiguity, intensity decay, and insufficient global context utilization. To address these limitations, this paper proposes Global‐Frequency Spatiotemporal Long Short‐Term Memory ( GFST‐LSTM), a novel model that integrates a global attention mechanism and Fourier convolutional modules into the Spatiotemporal LSTM ( ST‐LSTM) architecture. The attention module dynamically weights multi‐scale spatiotemporal features by enhancing channel and spatial correlations, while the Fourier convolution module captures global periodic patterns via frequency‐domain transformations. Evaluated on the Moving Modified National Institute of Standards and Technology database (Moving MNIST) benchmark and Jiangsu Province radar data sets (2019–2021), GFST‐LSTM achieves a 22.9% improvement in Critical Success Index and 13.1% in Heidke Skill Score over Predictive Recurrent Neural Network at the 40 dBZ threshold. Notably, it excels in preserving strong echo regions during 60–120 min predictions, reducing positional bias by 6.6% compared to the Motion Gated Recurrent Unit (MotionGRU). Ablation studies confirm the synergistic effect of both modules, with the full model outperforming variants that lack either component.

Abstract The wide‐swath altimeter Surface Water and Ocean Topography (SWOT) provides unprecedented two‐dimensional sea‐level observations, whose ability to capture upper‐ocean dynamics requires assessment. The dynamically balanced signal and noise contributions in SWOT‐KaRIn Level‐3 (L3) sea level products are here originally quantified and contrasted with those from Level‐4 (L4) gridded nadir‐only products, combining sea level data, 137 trajectories from drifters deployed in the Western Mediterranean, ERA5 winds, and the framework of Demol et al. (2025, https://doi.org/10.1029/2024JC021637 ). The filtered L3‐2km product, or interestingly the unfiltered L3‐2km product with a 25 km Gaussian filter, offers the best compromise for fine‐scale studies, though residual noise still accounts for about one‐third of total variance. L4 products contain less balanced signal but are noise‐free and better suited for large‐scale analyses. SWOT KaRIn adds value mainly at scales smaller than ∼100 km and shorter than ∼10 days. This studies provides a benchmark for global sea‐level assessments.

Abstract In 2023, Yunnan in southwestern China experienced the most severe compound drought–heatwave (CDHW) event on record. To effectively address these extremes, it is crucial to understand the atmospheric physical processes that initiate and sustain them. This study provides a comprehensive diagnosis of the formation and amplification mechanisms of this event. The results indicate that the CDHW in MAM 2023 was exceptionally intense. An anomalously westward‐extending subtropical high placed Yunnan under persistent high‐pressure anticyclonic conditions that suppressed cloud development. As a result, there were marked reductions in cloud cover and cloud thickness, which weakened shortwave reflection and enhanced surface solar absorption. This led to positive anomalies in surface shortwave cloud radiative forcing (SWCF; +15.12 W m −2 ) and surface net cloud radiative forcing (NCF; +11.86 W m −2 ), while surface longwave cloud radiative forcing (LWCF) exhibited a negative anomaly (−3.26 W m −2 ). In MAM 2023, the SWCF value (−55.98 W m −2 ) accounted for 27% of the net surface shortwave radiation (NSW; 209.08 W m −2 ), whereas the LWCF value (25.31 W m −2 ) contributed 32% of the net longwave radiation (NLW; −77.80 W m −2 ). Meanwhile, strengthened sensible heat flux (SHF) and reduced latent heat flux (LHF) indicate a transition from a “moist evaporative” to a “dry sensible‐heating” surface regime. These processes together established a positive cloud–radiation–land–atmosphere feedback in which reduced cloudiness enhanced radiative warming, thereby suppressing evaporation and strengthening sensible heating, and further favored cloud reduction, thus sustaining and intensifying the CDHW.

Abstract Tropical stratospheric aerosol injections are known to strengthen the wintertime Stratospheric Polar Vortex (SPV). Here, we revisit the circulation response to aerosol perturbations during the first month following injection using chemistry‐climate model simulations of regional nuclear war scenarios. We diagnose the atmospheric heat and momentum budgets to assess the thermal and dynamical responses to tropical soot injection. The results reveal that, during the initial adjustment period of 30 days, radiative heating from aerosols is confined to the tropics. In contrast, temperature and circulation changes in the mid‐to‐high latitudes are governed primarily by dynamical processes, with changes in eddy momentum fluxes driving the intensification of the SPV. Together, these findings demonstrate that circulation responses to stratospheric aerosol perturbations—through the redistribution of heat and momentum to remote regions—play a key role in the strengthening of the winter polar jet.

Abstract Flood risk in semi-arid regions is intensifying due to climate change; however, reliable prediction remains a fundamental challenge in catchments with severe hydrometric data scarcity (station density < 0.2 per 1000 km²). To address this gap, we developed and validated a novel hybrid modeling framework that integrates a physically based hydrological model (Hydrologic Engineering Center - Hydrologic Modeling System, HEC-HMS) with interpretable, Bayesian-optimised machine learning techniques (Random Forest, eXtreme Gradient Boosting , XGBoost ). The SHapley Additive exPlanations (SHAP) method was employed to interpret and identify the key drivers of flooding. SHAP analysis revealed that antecedent precipitation and soil permeability are the primary controlling factors in flood occurrence. The validated framework demonstrates robust performance, achieving excellent accuracy in flood event classification (Precision-Recall- Area Under the Curve (PR-AUC) = 0.98) and reducing false alarm rates by 87.5% compared to the standalone physical model. This study provides a practical, validated tool that enhances prediction reliability and delivers actionable insights for climate-resilient flood risk management in data-scarce environments.

Methane (CH 4 ) is a potent greenhouse gas, and tropical forests account for roughly one–third of global atmospheric CH 4 uptake by soils. Projected warming and more frequent hurricanes in these ecosystems may alter soil CH 4 sink strength, as warmer and wetter soils enhance methanogenesis activity. We measured soil CH 4 and CO 2 efflux during the calendar summer months of 2023 and 2024 alongside continuous records of soil moisture, soil and air temperature, and precipitation in an in–situ warming experiment (TRACE) located in a lowland tropical forest in Puerto Rico, six to seven years after Hurricanes Irma and Maria (2017). The realized warming (∼1.95°C) enhanced soil respiration only in summer 2023 ( p < 0.05), but net soil CH 4 uptake was invariant in both campaigns ( p > 0.05). Instead, sampling day and between–plot variability explained soil CH 4 dynamics much more than treatment contrasts. Importantly, CH 4 uptake was consistently coupled to CO 2 efflux, suggesting tight linkages between methanotrophic and heterotrophic activities. Between treatments, CH 4 and CO 2 responses to soil temperature variation were less sensitive in warmed plots, which may suggest weak metabolic upregulation under elevated temperatures. Together, these findings indicate that lowland tropical soils remain CH 4 sink even under warming and years after hurricane disturbance, with CH 4 dynamics driven more by spatial and temporal variability than experimental warming. Long–term, high–resolution monitoring integrating soil biogeochemistry and microbial processes will be critical to determine whether the observed net CH 4 uptake signal represents a sustainable or transient response under continued warming and disturbance.

Abstract The impactful 04 July 2025 Central Texas extreme rainfall event is examined to understand how surface conditions influence storm development. Utilizing convection‐permitting model simulations, we evaluate the sensitivity of this event to Gulf of Mexico sea surface temperature anomalies (SSTAs) and antecedent soil moisture distributions. The precursor wet soil conditions enhanced storm rainfall, whereas warm coastal and central Gulf SSTAs suppressed rainfall through perturbations of the low‐level circulation, including the Great Plains low‐level jet, which modified moisture transport and moisture convergence. When compared with climatological conditions, SST and soil moisture anomalies produced a rainfall reduction, indicating SST forcing dominated the combined response. These results suggest that this extreme storm would have produced higher rainfall totals had SSTs been closer to their recent climatological average.

Abstract River‐floodplain exchange occurs when high flows inundate the floodplain and exchange water bi‐directionally between substrates and communities. While well‐understood locally, the full spatial heterogeneity of exchange event duration (the residence time) and magnitude (the exchange flux, or discharge) remains poorly constrained, limiting our understanding of the role that underlying structural connectivity and hydrologic regime plays in basin‐scale mixing. Here, we use machine learning to predict river‐floodplain exchange duration and magnitude for over 1.8M rivers in the contiguous United States. We find residence time is on average 3.4 times longer in the floodplain than the river, with that difference decreasing as both event and river size increase. Further, more than 31% of basin water may exchange with the floodplain during large floods. We confirm that the cumulative effect of reach‐scale exchange influences basin‐scale mixing and subsequent biogeochemical processing, though the nature of that influence will be region, river, and constituent specific.

Abstract Rainfall structure is projected to shift toward heavier precipitation under global warming due to increased atmospheric moisture. Such a shift can heighten the risks of both floods and droughts. Stratospheric aerosol injection (SAI) has been proposed to counteract continued warming, but its effects on rainfall structure and potential to introduce new hydrological risks remain unclear. Here, using simulations from two models (CESM2 and UKESM1) participating in ARISE‐SAI‐1.5, we show that restoring global mean temperature increase to 1.5°C through SAI produces contrasting responses between monsoon and adjacent dry regions. Specifically, the global monsoon region is expected to experience a reduction in heavy rainfall, while the global dry region exhibits increased precipitation across a broad range of intensities. Inter‐model comparisons further reveal the sensitivity of rainfall structure to residual temperature anomalies over monsoon sub‐regions. Monsoon regions exhibit reduced heavy rainfall but increased drought risk in the hemisphere experiencing stronger cooling. In contrast, dry regions consistently experience enhanced precipitation and fewer dry days across models. These responses are primarily driven by dynamic processes rather than thermodynamic effects. The emergence of potential new regional “hotspots” of precipitation extremes underscores the need for caution when deploying SAI.

Abstract Atmospheric transport plays a crucial yet poorly quantified role in shaping atmospheric CO 2 seasonality across the Northern Hemisphere. This study used GEOS‐Chem simulations and a statistical decomposition to explicitly quantify the import and export effects of atmospheric transport on CO 2 seasonal amplitude and phase in the troposphere. We found that atmospheric transport explained 28%, the second to terrestrial ecosystems, of the boreal CO 2 seasonality and exceeded terrestrial ecosystems in the northern mid‐to‐low latitudes (&gt;50%). The import (export) enhanced (suppressed) the seasonal amplitude, with the largest enhancement (decrease) in the northern middle latitudes, and caused phase shifts, with the minimum phase showing the largest advances in the northern tropics. The net effects revealed that the northern middle latitudes served as the strongest sources of CO 2 seasonality, followed by the northern high latitudes, while the northern tropics acted as the sinks. These results provide new insights on how atmospheric circulation redistributes CO 2 seasonality, advancing our understanding of the coupling between surface carbon fluxes and atmospheric CO 2 .

Abstract Mountain snowpacks provide vital water resources for communities in the western U.S. (WUS), but high spatial variability challenges accurate measurement of snow water equivalent (SWE) from remote sensing platforms. Studies using repeat airborne L‐band (∼25 cm wavelength) Interferometric Synthetic Aperture Radar (InSAR) have demonstrated sensitivity to forest cover fraction (FCF), liquid water content, and incidence angle. We use these factors to map feasibility of L‐band InSAR for SWE change (ΔSWE) retrievals in major mountain ecoregions of WUS. We found feasibility declines from ∼65% on 1 February, to 58% on 1 March, and 30% on 1 April, corresponding to 73%, 70%, and 49% of total SWE volume, respectively. Thus, these feasibility maps provide groundwork for future InSAR snow studies using satellite data from missions such as NISAR (NASA‐ISRO Synthetic Aperture Radar) to refine work based on regional conditions and hence improve ΔSWE retrievals globally.

Study region Pearl River Basin, Southern China. Study focus This study conducted a comprehensive nonstationary risk analysis of Drought-Flood Abrupt Alternation (DFAA) events in the Pearl River Basin (1961–2020). A total of 38 wet-to-dry and 49 dry-to-wet events were identified using the Standardized Weighted Average of Precipitation index. The Generalized Additive Models for Location, Scale, and Shape (GAMLSS) framework was employed to model time-varying parameters for the marginal distributions of intensity and duration. Both stationary and nonstationary bivariate joint distributions were constructed using Clayton and Joe copulas for DFAA events, respectively, with dependence nonstationarity confirmed by Copula Likelihood-Ratio tests. New hydrological insights for the region DFAA properties and dependence structures are nonstationary, and ignoring this leads to systematic underestimation of compound risks. Under nonstationary conditions, 18.4% of wet-to-dry and 4.1% of dry-to-wet events exceeded the 50-year "AND" joint return period. wet-to-dry events were found to be generally more severe than dry-to-wet events. The time-varying copula parameters for the dependency structure decreased by 27% for wet-to-dry events and increased by 43% for dry-to-wet events by 2020 compared to stationary estimates. Multivariate design values substantially diverged from univariate estimates, emphasizing the necessity of multivariate risk assessment. The GAMLSS-time-varying copula framework provides a robust tool for nonstationary frequency analysis, offering critical guidance for adaptive water infrastructure design and climate resilience strategies.

Abstract Precipitation patterns in the Himalayas vary in complex ways with elevation but remain poorly understood due to limited in situ data. In this study, we present observations from two rain‐gauge networks spanning different elevations established in the Yarlung Tsangpo Grand Canyon and the Khumbu Valley. Using these ground‐based data, we provide the first comparative investigation of summer precipitation variation along elevation gradients in the Central and Eastern Himalayas. In both regions, precipitation increases with elevation up to a peak and then decreases between 1,500 and 4,300 m above sea level. A marked precipitation maximum occurs near 2,500 m above sea level, largely induced by frequent heavy precipitation events (≥1 mm hr −1 ). The diurnal cycle of mean precipitation generally displays a consistent unimodal pattern, with a daytime minimum and a nighttime maximum. However, the diurnal variation exhibits distinct afternoon differences with elevation in the Central and Eastern Himalayas: an afternoon secondary peak in precipitation is discernible at low‐ and mid‐elevation sites in the Central Himalaya, which is absent in the Eastern Himalaya. The former is primarily associated with ridge convection triggered by afternoon upslope flow, whereas the latter is suppressed by the descending branch of local mountain circulation driven by convection at high elevations.

Abstract Stromboli volcano (Italy) exhibits a persistent, low-intensity activity, but can occasionally produce more powerful eruptions. These paroxysms can generate pyroclastic density currents usually confined in the Sciara del Fuoco, a horseshoe-shaped depression about 2 km from inhabited areas. However, at least twice in modern times (1930 and 1944), PDCs occurred outside the Sciara del Fuoco, rising the hazard impact. This study focuses on a small valley adjacent to the Sciara del Fuoco in the Punta Labronzo area, crossed by a popular touristic path. There, recent precipitations unveiled a previously unknown, ≈1 m thick, massive PDC deposit. The collected samples contain older accretionary lapilli engulfed by the flow, angular clasts and non-abraded spatters all along the runout, and very low fine ash content. By performing depth-averaged flow modelling, we tested different assumptions on the initial volume, basal friction, and turbulent dissipation parameters and compared them to field data for a qualitative validation. The simulations showed that the Labronzo PDC had a very low volume in the order of 10 3 m 3 , generally low-moderate velocities (&lt; 10 m/s), a reduced basal friction and limited turbulent dissipation. These relatively slow dynamics limited abrasion processes, preserving accidental accretionary pellets and inhibiting fine ash generation. Major element chemistry revealed that the composition of the juvenile clasts is the same as juvenile clasts from a paroxysm in the sixteenth century, though radiocarbon analysis of charcoal pieces revealed an age closer to the eighteenth century. This makes the Labronzo PDC a new instance of a PDC outside the Sciara del Fuoco in modern times, different from 1930 and 1944.

Abstract Frequent dust storms in the Middle East cause major health and economic impacts. The Qatar Meteorology Department (QMD), Doha, in collaboration with the Indian Institute of Tropical Meteorology (IITM), Pune, and the Integrated Land Ecosystem‐Atmosphere Processes Study (iLEAPS) developed a dust forecast system in experimental mode to provide dust forecasts during the 2022 FIFA World Cup in Doha, Qatar. This system provided PM 10 (particulate matter with an aerodynamic diameter ≤10 μm) forecasts up to 3 days in advance from November–December 2022. It satisfactorily captured a dust storm on 9 November 2022 with observed and forecasted peaks of PM 10 at ∼1,600 μg/m 3 and ∼1,200 μg/m 3 , respectively. While the system exhibited reasonable performance (Mean Fractional Bias (MFB): −0.02, Mean Fractional Error: 0.71, Root Mean Square Error: 198 μg/m 3 , Normalized Mean Bias: 15%), occasional false alarms were noted. The analysis of false alarm events revealed that the combined effect of wind speed overestimation and uncertainties in the dust source representation over the broader Doha region, including northwestern Qatar, appears to be the primary cause of these false alarm events. Diagnostic sensitivity experiments over this region suggested that a 50% reduction in the erodibility reduced false alarms and closely matched the true dust storm event. The findings emphasize the need for higher resolution modeling, improved region specific dust source data set, and refined surface physics representation to enhance operational dust forecast reliability over the Middle East.

Abstract Climate change is intensifying Sri Lanka’s exposure to multiple hazards, including floods, droughts, landslides, cyclones, coastal risks and heat stress, with growing consequences for livelihoods, infrastructure, public health and national development. This analysis evaluates whether Sri Lanka’s contemporary climate policy architecture is sufficiently coherent and operationally robust to support effective climate resilience. A qualitative policy analysis was undertaken of eight core national climate-related instruments, generated by analytically combining thirteen policy and institutional documents published between 2021 and 2025. The study employed NVivo-based content analysis and a structured benchmarking framework based on an adapted Organisation for Economic Co-operation and Development Development Assistance Committee framework, which assessed relevance, coherence, effectiveness, efficiency, equity and sustainability. The pilot coding process achieved an intercoder reliability score of 0.81, indicating strong agreement between coders. The findings indicate that Sri Lanka’s policy architecture performs strongly on relevance and coherence, with mean scores of 4.1 and 3.9, respectively, reflecting substantial strategic alignment with the country’s multi-hazard climate risk profile. However, effectiveness and efficiency remained moderate, with both recording mean scores of 3.1, while equity and gender considerations emerged as the weakest dimension at 3.0. Sustainability varied across instruments depending on the extent to which they incorporated operational monitoring, statutory review cycles and multi-year financing arrangements. Instruments with clearer SMART targets, operational monitoring, reporting and verification systems, designated institutional responsibilities and costed financing pathways performed more strongly than those relying on broad commitments or annualised budget structures. Overall, Sri Lanka’s principal climate governance challenge lies not in the absence of policy ambition, but in the incomplete institutionalisation of delivery systems needed to convert strategic coherence into measurable resilience outcomes. Aim of the study To critically evaluate the coherence and implementation robustness of Sri Lanka’s contemporary national climate policy architecture in order to determine whether strategic policy commitments are supported by adequate institutional design, financing arrangements, monitoring systems, equity safeguards and long-term sustainability in the context of escalating multi-hazard climate risks.

Hydroxymethanesulfonate (HMS), formed from liquid-phase reactions of formaldehyde (HCHO) and SO2, is a key organosulfur component. While its importance in northern China’s sulfur chemistry is recognized, HMS dynamics in southern coastal regions remain poorly understood. This study analyzes eight months of hourly single-particle measurements (5000 data sets, 76,515 HMS particles, 1.17% of total) from coastal Hong Kong, revealing a distinct winter-to-summer decline. Crucially, higher HMS particle numbers occurred during nonfog episodes versus fog, specifically under elevated iron (Fe) levels. Integrated statistical and machine learning models confirmed differing influences of temperature, liquid water content, pH, Fe, and precursor concentrations (SO2 and HCHO), with Fe-driven enhancement occurring under the characteristically low particle pH (0–3) of the coastal environment. This acidity promotes Fe dissolution into soluble forms that stabilize organosulfur complexes. The broader pH range of the Fe-sensitive zone under nonfog conditions (0–2.4, compared with 0–1.4 in fog) likely contributes to the elevated HMS number concentrations. These findings highlight Fe-mediated pathways, crucially modulated by particle acidity, as a potentially major but overlooked source of atmospheric organosulfur. Elucidating the specific low-pH reaction mechanisms governing this Fe-HMS link is critical to advance fundamental understanding of coastal sulfur transformations and aerosol impacts.

Abstract Recent advances in solar physics increasingly rely on automated identification of coronal structures using machine learning. Yet most studies emphasize scientific performance without evaluating feasibility for onboard deployment to prioritize downlink observations. This work investigates the automated identification of active regions and coronal holes by applying segmentation and detection techniques to Solar Dynamics Observatory (SDO) data. We compare three approaches: SCSS‐Net, a deep learning model for semantic segmentation; YOLOv8n, a lightweight object detector; and a traditional pipeline based on basic computer vision operations (BCVO). Each method is assessed for its scientific accuracy and its suitability for deployment in future resource‐limited missions. While no direct hardware benchmarking is performed in this study, we assess the feasibility of onboard implementation based on the associated number of trainable parameters, architecture and hardware requirements. Training and evaluation are first conducted on well‐calibrated SDO images. We then extend the evaluation to raw and uncalibrated SDO images affected by instrumental artifacts. Performance is measured using the Intersection over Union (IoU) and Dice score. Results show that while SCSS‐Net achieves the highest segmentation quality, YOLOv8n offers a strong balance between accuracy and efficiency. The BCVO pipeline remains viable under strict hardware limitations. Interestingly, our models retain compatibility on Level‐0 observations. This is the first study comparing these widely used methods from the perspective of onboard deployment. Our findings provide a foundation for designing frameworks tailored to onboard hardware configurations.

Mn-based catalysts are among the most promising candidates for the ultralow-temperature (≤150 °C) selective catalytic reduction of NOx with NH3 (NH3–SCR). However, their narrow operating temperature window and insufficient resistance to SO2/H2O limit their broader practical application. In this work, Mn single atoms featuring a unique low-coordination configuration are uniformly anchored onto CeO2 nanoislands that were predeposited on H2Ti3O7 nanotubes (TNTs) via a newly developed in situ redox self-assembly strategy. The resulting catalyst, denoted as LC-Mn/CeO2, exhibits exceptional ultralow-temperature NH3–SCR activity and SO2 resistance. It achieves over 90% NOx conversion at 100 °C and maintains nearly 90% conversion for over 20 h at 140 °C in the presence of 50 ppm of SO2 and 10 vol % H2O. Experimental and DFT results reveal that the unique electronic modulation of the low-coordination Mn centers facilitates the formation of asymmetric oxygen vacancies. By promoting the oxidation of NO to NO2, these vacancies significantly enhance the low-temperature reaction rate of the NH3–SCR reaction. Simultaneously, the electron-rich environment of Mn sites suppresses the oxidation of SO2 by weakening the Mn–SO2 charge transfer, thereby improving SO2 resistance. Our work provides a novel strategy of modulating the coordination environment of single atoms to enhance ultralow-temperature activity and SO2/H2O resistance.