Atmospheric and Oceanic Sciences

Abstract Extreme precipitation events can have severe impacts on society and the environment. Understanding what causes these events is a vital step towards better prediction and improved disaster preparedness. One research direction, is to answer the question: Where did the moisture that rained here come from? The moisture sources for precipitation (i.e., where the moisture originally evaporated) cannot be measured directly and, therefore, a variety of different moisture tracking methods have been developed and evolved over time. To better understand the uncertainty of these methods, we unite the community to advance common understanding and guidelines. As the first step, in this study, we quantify moisture sources of three extreme precipitation events, using methods obtained from 14 different research groups. These three events cover different meteorological conditions: monsoon precipitation in Pakistan, convective precipitation in Australia, and atmospheric river-associated precipitation over Scotland. We find that for the three cases the different moisture tracking methods qualitatively agree in moisture source patterns, but there are regional and quantitative differences. For example, for the Pakistan case, the recycling ratio shows a multi-method spread of 2-20%. We also find similar behavior across methods for the three different events, where methods consistently show either more recycling or more sources further away from the precipitation region. This coordinated model intercomparison facilitates the explanation and quantification of uncertainty, acting as a point of reference and inspiration for future work and literature on moisture tracking.

Under global warming, compound extreme events such as Drought–Flood Abrupt Alternation (DFAA) are becoming increasingly common, yet the structural asymmetry and spatial dynamic evolution between Drought to Flood (D–F) and Flood to Drought (F–D) processes remain under-researched. Using high-resolution daily precipitation data spanning 1961 to 2022 from nine major river basins in China, DFAA events were identified via the Standardized Weighted Average Precipitation (SWAP) index coupled with run theory, and their evolution was analyzed using multidimensional spatiotemporal metrics. Our results reveal a spatial frequency and severity mismatch, where southern basins exhibit high frequency occurrences dominated by slight to moderate events, whereas northern and inland basins experience lower overall frequency but a significantly higher proportion of severe events. Spatial polarity asymmetry is evident, with D–F events dominating nationwide and exceeding 74% in northern and inland basins, while southern humid basins exhibit a more balanced D–F/F–D structure. Temporally, D–F processes involve prolonged moisture accumulation, whereas F–D processes manifest as short-lived post-rainfall moisture deficits. Based on risk trajectories, basins were categorized into four impact patterns: highly oscillatory pattern, intensifying pattern, long-cycle accumulative pattern, and baseline pattern. Ultimately, regional DFAA risks are governed by polarity asymmetry and non-stationary evolution rather than absolute frequency alone, providing a critical scientific basis for basin-specific disaster mitigation strategies under climate change.

Reliable prediction of CO2 injectivity decline is essential for safe geological carbon storage, yet existing machine learning models often provide deterministic point predictions that lack uncertainty quantification. This paper presents a physics-informed Gaussian process regression (PC-GPR) framework for Relative Injectivity Change (RIC) prediction, embedding constraints derived from two independently grounded physical laws: the Derjaguin-Landau-Verwey-Overbeek (DLVO) colloidal monotonicity condition and the Civan-Kozeny-Carman permeability impairment model. Four GP variants are developed and benchmarked on a curated laboratory dataset (n = 44) under a three-tier validation protocol combining Leave-One-Out cross-validation, repeated k-fold cross-validation, and non-parametric bootstrap confidence intervals. Two complementary uncertainty quantification mechanisms are employed: GP posterior calibration via the Expected Calibration Error (ECE) and split-conformal prediction intervals. The GP-Base model achieves strong predictive performance (LOO R2 = 0.9401, 95% CI: [0.882, 0.978]) with well-calibrated uncertainty (ECE = 0.026) and reliable coverage (97.7% at the nominal 95% level). The PC-GPR-M variant reduces DLVO monotonicity violations to 1.5% across the input domain, demonstrating effective soft constraint enforcement. Operationally, the proposed framework translates predictive uncertainty into actionable injection scheduling guidance, identifying high-risk regions at salinity >30,000 ppm and jamming ratio >0.04. These results provide an uncertainty-aware baseline for future PIML research in subsurface carbon storage.

Abstract Arctic sea ice is disappearing rapidly, but the melt comes from two different pathways: direct surface warming and the movement of warm water into the Arctic. This study focuses on the latter, i.e., dynamic melting, defined specifically as sea-ice loss caused by circulation-driven ocean heat transport, not mechanical ice motion or thermodynamic surface heating. Using CESM2 and its slab ocean counterpart, which preserves surface warming but suppresses changes in ocean heat transport, we approximately separate these two pathways and examine how they respond to different climate forcing agents. We find that atmospheric circulation-driven ocean heat transport accounts for roughly 25.2% of the total sea-ice loss on the Pacific side of the Arctic. Aerosols strengthen a North Pacific anticyclone that pushes more warm Pacific water through the Bering Strait, accelerating sea-ice decline even though aerosols themselves cool the surface. Greenhouse gases, by contrast, generate atmospheric circulation patterns that partially offset their own thermodynamic warming. These results show that dynamic melting is strongly forcing-dependent and that models lacking realistic ocean-atmosphere circulation may systematically underestimate regional Arctic sea-ice vulnerability. Our framework separates a 12.2% thermodynamic sea-ice loss from a 16.3% coupled loss, implying that circulation-driven ocean heat transport explains about 25.2% of the total western Chukchi response. This dynamic contribution is forcing-dependent: aerosols enhance ocean-driven ice loss, offsetting 26.8% of their thermodynamic cooling effect, whereas greenhouse gases produce an opposite-signed circulation response that partly damps thermodynamic ice loss.

The Atlantic Meridional Overturning Circulation (AMOC) plays a critical role in regulating global climate patterns with its potential destabilization and collapse posing significant risks. In this study, we introduce two novel methods to assess AMOC stability, based on a Bayesian framework to estimate its sensitivity, and ensembles of parabolic approximations, respectively. They provide alternative indicators to detect early warning signals (EWSs) for a potential destabilization. By incorporating non-linear and physically motivated drivers, such as temperature and Greenland meltwater runoff, we obtain a more realistic representation of AMOC dynamics and address an important limitation of previous studies, which often rely on linear forcing assumptions. We detect significant EWS for a potential ongoing AMOC destabilization in our sensitivity-based indicator across many combinations of forcing scenarios and response models. However, it revealed large variations, dependent on the considered forcing scenario and methodological choices. While an AR(1) response to linear forcing, consistent with assumptions in previous EWS analyses of the AMOC, emerged as the “best” fit based on Bayes factor analysis, other evaluation criteria provided no clear support. Even though the second EWS indicator, obtained using parabolic approximations, revealed a recent significant peak under linear forcing, and appears to suggests that the AMOC might have passed a critical transition in the late 20th century assuming a temperature driver, we find no clear indication of a considerable destabilization in either case. Our findings highlight the sensitivity of AMOC stability assessments to assumptions about forcing scenarios, response models, and evaluation criteria, emphasizing the need for careful interpretation of EWSs for abrupt transitions in the Earth's climate. While our methods advance EWS analyses by incorporating non-linear forcing and alternative response functions to better represent AMOC dynamics, they also underscore the limitations of applying such tools to complex climate subsystems, represented by one-dimensional time series.

Abstract. Climate change is driving wildfires to higher elevations, yet the hazard cascades that follow the burning of pristine tropical mountain ecosystems remain largely unexplored. Here, we analyse the long-term cascade following a February 2012 wildfire that burned 31 km2 of forest and wetland in Uganda's Rwenzori Mountains National Park, including sections above 3800 m elevation with no major fire history in 12 000 years. Combining remote sensing, humanitarian records, field surveys and interviews, we document ten major floods since 2012, including two debris floods that required large-scale humanitarian responses. Post-fire increases in erosion and mass movement have widened the River Nyamwamba sevenfold since 2012, breaching copper-cobalt mine tailings and mobilising an estimated 744 000 t of waste into the river. Slow vegetation recovery at high altitudes and positive feedbacks between hazards have prolonged this high-risk state. These findings point to an urgent need to understand where emergent tropical mountain fires can occur, how their impacts cascade downstream, and where early interventions can reduce risk.

Abstract Diked marshes with decades of restricted tidal flow exhibit fundamentally altered hydrologic and ecologic dynamics that result in limited vertical accretion, land subsidence, and underdeveloped tidal channels. These changes complicate marsh restoration efforts and constrain the marsh's ability to build elevation in response to future sea levels. In this study, we developed an integrated hydrodynamic and marsh accretion model to conduct the first assessment of the restoration potential of a diked marsh in response to reintroduction of tidal flow and sea‐level rise (SLR). We applied our model to evaluate changes in biomass production in the Herring River estuary, Cape Cod, MA, USA—which has experienced significant ecological shifts due to more than a century of tidal restriction—in response to a range of SLR scenarios, tidal flow conditions, management interventions, and biomass production rates through 2100. We found that with restoration of tidal flow, the simulated marsh extent expanded in areas with efficient drainage, increasing biomass production by up to 4.4 times. However, when SLR was imposed, parts of the marsh gradually transitioned to open water, leading to a decline in marsh coverage from 19% to as low as 12%. Management interventions aimed at enhancing marsh elevation and improving the drainage capacity of tidal channels slowed marsh loss, supporting marsh area through 2100. Our findings provide insight into the key drivers of marsh resilience and loss, and our approach can be applied across tidally restricted and restored systems to develop targeted, adaptive restoration strategies that achieve desired ecological outcomes.

Abstract The incorporation of appropriate dynamic constraints in data assimilation (DA) is highly important for improving the forecasting of disastrous weather. It is often challenging to describe subgrid physical processes with strong nonlinearity and discontinuity in dynamic constraints because of difficulties in tangent linear and adjoint development. Based on the frictionless momentum equation weak constraint within the Weather Research and Forecasting ensemble‐three‐dimensional variational (En3DVar) DA, this study incorporates the subgrid planetary boundary layer (PBL) turbulent friction effects through a machine learning (ML) approach. The new constraint is aimed at improving the coupling between dynamic and thermodynamic variables. A deep neural network (DNN) is applied to emulate the model parameterized PBL horizontal momentum tendency. The tendency is introduced into the momentum equation as a turbulent friction term. The tangent linear and adjoint of the DNN are embedded into the variational framework to construct a ML‐improved DA scheme. Sensitivity experiments for tropical cyclones (TCs) reveal that including turbulent friction effectively compensates for the insufficient pressure adjustments during wind assimilation according to the original constraint framework. Three DA experiments that involve assimilating radar radial velocities are conducted using En3DVar (EVAR), En3DVar with the original constraint (EVAR–DC) and En3DVar with the ML‐improved constraint (EVAR–MLDC) for landfalling TCs Muifa (2022) and Doksuri (2023). In response to wind increments, compared with EVAR–DC and EVAR, EVAR–MLDC produces larger pressure adjustments around the eyewall. Overall, a better description of thermodynamic states is obtained by the new scheme, which plays a positive role in TC track and intensity forecasting.

Abstract China is a major source of anthropogenic emissions, with important implications for chemical composition and aerosol processes in the upper troposphere and lower stratosphere (UTLS), particularly within the Asian Summer Monsoon (ASM) anticyclone. Carbon monoxide (CO) is a robust tracer of anthropogenic influence and is routinely observed by remote platforms. In the ASM region, the GEOS‐FP forecast system predicts frequent, rapid convective transport of boundary‐layer CO into the UTLS. During the 2022 Asian Summer Monsoon Chemical and Climate Impact Project (ACCLIP), five in‐situ spectrometers aboard two coordinated research aircraft provided some of the first vertically resolved UTLS CO measurements in the ASM region, providing generally good agreement with forecast abundances. On 19 August 2022, exceptionally enhanced UTLS CO, exceeding 325 ppb, was recorded by both ACCLIP research aircraft over the Yellow Sea west of Korea, exceeding GEOS‐FP predictions and producing a pronounced “C‐shaped” vertical profile indicative of strong convective outflow. None of the satellite profile products examined (MLS, MOPITT, AIRS, CrIS, or IASI) captured the magnitude or vertical structure of this enhancement, highlighting limitations in remote‐sensing sensitivity to sharp UTLS gradients. Simulations from global models with sophisticated chemistry schemes, GEOS‐GOCART and CESM2‐WACCM, attribute the sampled plume to a local convective outbreak approximately 12 hr before in‐situ sampling. GEOS‐GOCART better reproduced the observed profile, while CESM2‐WACCM simulated weaker lofting; however, both underestimated the magnitude of CO in the UTLS. This case underscores challenges in validating localized deep‐convective transport and demonstrates the continued need for high‐resolution (spatial and temporal) in‐situ UTLS observations.

Study Region Europe (pan-European domain), with particular focus on northern, central, southern, eastern, and Mediterranean Europe, covering the period 1902–2024. Study Focus Warming is well established to transform droughts from precipitation-driven to 'hotter' events dominated by evaporative demand. Building on this foundation, we provide a century-scale pan-European observational synthesis of the compound hot–dry hazard and explicitly separate shifts in joint hot–dry probabilities into contributions from marginal warming and from changes in heat–dryness dependence, a decomposition not previously applied at this spatio-temporal scale. We identify a marked seasonal and regional reorganization of hydroclimate, with cold-season wetting concentrated in northern Europe and warm-season drying across the broader central, southern, and eastern European domain. A direct comparison of the Standardized Precipitation Evapotranspiration Index (SPEI) and the Standardized Precipitation Index (SPI) trends shows that across central and southern Europe, the late-summer drying signal exceeds what precipitation changes alone can explain, identifying rising evaporative demand as a substantial and spatially coherent driver of recent drying. New Hydrological Insights for the Region Since the mid-1990s, compound hot–dry months have routinely affected more than 25% of Europe, and joint hot–dry probabilities have increased significantly. An explicit probability decomposition reveals that this increase is driven primarily by the rise in marginal warm-months frequency, while changes in heat–dryness dependence are spatially heterogeneous and act as a secondary, regionally specific amplifier in water-limited southern European climates. These findings highlight the need for drought monitoring, hazard assessment, and water-resources planning frameworks that explicitly incorporate evaporative demand and compound-event behavior, rather than relying solely on precipitation-based indicators.

Abstract In a warming climate, the increasing frequency and severity of temperature extremes across the Arabian peninsula (AP) demand skillful prediction several weeks in advance for effective risk management. However, current coarse‐resolution models constrain subseasonal forecast skill to adequately resolve regional‐scale processes and extremes. Here, we examine the subseasonal predictability of daily maximum temperature ( T max ) and heat extremes over the AP by dynamically downscaling ensemble reforecasts from the European Centre for Medium‐Range Weather Forecasts (ECMWF) Integrated Forecasting System (IFS) using a convection‐permitting weather research and forecasting (WRF) model. An 11‐member WRF ensemble reforecast at 4‐km resolution was generated for the period 1998–2017 over the AP with lead times extending up to four weeks. We assessed the performance of WRF‐ downscaled T max relative to the ECMWF by comparing them against station observations. The WRF predictions exhibited small biases within ±1°C, whereas the ECMWF exhibited pronounced cold biases exceeding −2°C. Moreover, the WRF achieved consistently lower root‐mean‐square error and higher Nash–Sutcliffe efficiencies, whereas both models maintained strong correlations with the observations. Beyond week–2, the results revealed that the WRF experienced slower skill degradation than the ECMWF, highlighting its robustness at extended lead times. At weeks 3–4, WRF outperforms ECMWF by maintaining lower Continuous Ranked Probability Score and higher Brier Skill Score, demonstrating better probabilistic skill at subseasonal lead times. WRF also outperforms bias‐corrected ECMWF at subseasonal lead times. The ECMWF largely underestimated the hot‐day frequencies (˜30 days), while the WRF reduced this bias by a factor of nearly three (±10 days). In addition, WRF achieved higher hit rate and maintained greater reliability at longer leads (weeks 3–4) than ECMWF. WRF, however, show limitations in capturing T max and hot days along the coasts of the Red Sea. The findings of this study highlight the critical role of convection‐permitting downscaling in improving the local‐scale subseasonal predictability of temperature extremes in the AP.

Abstract Zooplankton play a key role in setting the particulate organic carbon (POC) distribution in the ocean, but this role remains poorly quantified on large scales due to the complexity of zooplankton ecosystems and the sparsity and variability of observations. We address this by applying boosted regression trees to a global in situ zooplankton image database from the Underwater Vision Profiler 5 and an Argo float and satellite‐derived POC product. We compute taxonomic, morphological, and trophic community metrics, then identify three metrics—abundance, mean gray level, taxonomic evenness—that explain almost half of the spatial variability in the vertical concentration gradients of small POC (<100 μm). Partial dependence analysis and the consistency of the identified relationships with known zooplankton‐mediated particle‐processing pathways suggest mechanistically interpretable ecological linkages. These results advance quantification of zooplankton communities' influence on ocean carbon cycling and indicate carbon‐cycle‐relevant zooplankton community properties for improving ocean biogeochemical models.

Abstract Global marine primary producers, phytoplankton, are the base of the marine food web and vary on short timescales, characterized by seasonal blooms. There is growing concern about the occurrence of short‐term extreme events in phytoplankton abundance, which may impact higher trophic levels and economically‐important species. Previous work has investigated the occurrence and impacts of extremes, but forecasting of large‐scale extremes has not been attempted. Here, we leverage the Community Earth System Model Seasonal‐to‐Multiyear Large Ensemble (CESM SMYLE) to assess the potential predictability of phytoplankton extremes. We find that low phytoplankton biomass extremes (LBX) are significantly predictable up to 6‐month in advance. LBX are closely related to enhanced upper ocean stratification, which impacts nutrient availability. We find that compound events (LBX with marine heatwave and low oxygen extremes) are also significantly predictable up to 6 months in advance. These results could inform future model development with impacts for marine resource managers.

Abstract Climate models project that the Atlantic Meridional Overturning Circulation (AMOC) will weaken in the 21st century, but the magnitude is highly uncertain. Some of this uncertainty is structural, as most climate models neglect increasing meltwater from the Greenland ice sheet and do not explicitly capture mesoscale ocean eddies. Here, we quantify the impact of Greenland meltwater on the AMOC until 2100 under SSP5‐8.5 forcing for the first time in a strongly eddying (1/10° horizontal resolution) ocean model. The meltwater‐induced additional AMOC weakening is small (0.6 0.2 Sv) compared to the weakening due to warming alone, and similar at high and low resolution. The same meltwater would cause a stronger AMOC weakening under present‐day climate conditions. We link both resolution‐independence and state‐dependence to large‐scale controls of the AMOC. Our results demonstrate that the background ocean state is more important than resolution in determining how Greenland meltwater affects the AMOC.

Summary Sediment transport in high mountain environments is affected by the interplay of hydro‐meteorological dynamics, erosion processes, and climate‐driven landscape change, such as glacier retreat and mountain greening. Mountain catchments in High Mountain Asia (HMA) are prone to debris flow hazard due to the steep terrain, intense monsoon precipitation, and rapid cryosphere degradation. Catchments susceptible to debris flow events have previously been classified as having either transport‐limited or supply‐limited sediment regimes. However, the contribution of glacier meltwater, precipitation extremes, changing vegetation cover, and sediment availability, on debris flow activity remains poorly quantified across environments in the region. Here, we modified the Sediment Cascade (SedCas) model to simulate water balance components and subsequent sediment transport and debris flow events of a conceptual debris‐flow catchment in two monsoon‐dominated regions of the Central Himalaya: Langtang (humid, monsoon‐dominated) and Mustang (semi‐arid, rain‐shadow), Nepal. We consider a total of 55 scenarios reflecting transitions from glaciated to vegetated terrain and a set of sediment supply scenarios and intra‐annual erosion patterns. Our results show that glacier melt more strongly affects runoff and debris flow magnitude in drier, continental climate, while increases in vegetation and a reduction in glacier cover substantially decrease potential sediment supply and shift debris flow events to later in the monsoon season. The number of annual debris flow events remains relatively stable across scenarios, but their magnitude and seasonality are altered by both sediment recharge (i.e., sediment supply recovery) timing and land cover evolution. The modeling experiments show that as the landscape changes (through glacier retreat and vegetation succession), transport‐limited and supply‐limited conditions become more similar, suggesting that future debris flows may become smaller. These findings highlight the importance of hydro‐geomorphic feedbacks under rapid environmental change to understand sediment‐related hazards in high‐relief Asian mountain regions.

Background As climate warming persists, heat-related health risks continue to escalate. This study aims to forecast the future community-level burden of heat-attributable hospital admissions and emergency department (ED) visits in Victoria, Australia, under multiple climate, population, and adaptation scenarios. Methods We estimated the associations between high temperature exposure and daily hospital admissions and ED visits during the hot seasons of 2014–2019 in Victoria. These estimated associations were then applied to projected future climate and population scenarios to quantify the future heat-related health burden. Results Under all future climate change scenarios, the heat-attributable burden of hospital admissions and ED visits was projected to rise substantially over time. Compared with 2020–2029, excess heat-related hospital admissions and ED visits during 2090–2099 were expected to increase by 74% and 67%, respectively, under a low-emission scenario (Shared Socioeconomic Pathway [SSP]1–2.6), and by 1177% and 423%, respectively, under the high-emission scenario (SSP5–8.5). The corresponding heat-attributable rates are projected to increase by 11% and 7% under SSP1–2.6, and by 453% and 126% under SSP5–8.5 for hospital admissions and ED visits, respectively. Although adaptation may partially alleviate the health burden of heat, pronounced increasing trends remained evident. Further analysis revealed consistently higher heat-related burdens of hospital admissions and ED visits in rural areas compared to urban areas, with the urban-rural disparity expected to widen over time. Conclusion Rising temperatures are projected to increase the burden of heat-related hospital admissions and ED visits, particularly in rural areas. This underscores the urgent need for climate change mitigation efforts and targeted public health interventions to reduce future health risks.

Earth's landscapes-from mountains to river deltas-are shaped by sediment erosion, transport, and deposition. However, most landscape evolution models represent only two processes: diffusive soil creep and bedrock river incision. How much of Earth's surface can be represented by these two processes alone? To address this question, we mapped the global abundance and distribution of eight major geomorphic process domains, finding that ~75% of Earth's land area consists of soil-mantled hillslopes and flatlands dominated by diffusive soil creep, while 0.5% is a bedrock river corridor. While other process domains (glaciers, lakes, alluvial river corridors, aeolian deserts, and bedrock hillslopes) make up only ~24% of Earth's land area, they transport most of the sediment from continents to oceans. Process domains align with different topographic, climatic, and tectonic conditions, indicating a pathway to improve coupled global landscape, climate, and tectonic evolution models. Overall, our results reveal the distribution of the eight major geomorphic process domains globally and highlight the spatial dominance of diffusive topography.

Abstract Accurate aerosol classification from spaceborne lidar is fundamental for understanding aerosol–radiation interactions and their climatic impacts. Physically consistent aerosol typing remains challenging due to overlapping optical signatures arising from particle non‐sphericity and internal inhomogeneity, as well as the limited labeled observational data. Traditional threshold‐ or clustering‐based approaches lack robustness across diverse aerosol regimes, while machine learning methods improve upon thresholding but are constrained by limited observational data or driven exclusively by synthetic data sets, reducing data utilization efficiency. These limitations hinder the development of classifiers that are physically consistent and robust. To overcome these challenges, we propose ST‐AeroClassifier, a simulation‐driven typing framework bridging comprehensive simulations and observations. Its core is a Transformer‐based kernel pre‐trained on a database generated using advanced super‐spheroidal models. This kernel learns intrinsic relationships between aerosol microphysical properties and lidar observables, enabling more efficient use of limited observational data and achieves balanced discrimination across aerosol types. The framework relies only on the particle depolarization ratio and lidar ratio, standard products of current and upcoming spaceborne lidar missions. Application to Atmospheric Lidar (ATLID) measurements aboard the Earth Cloud Aerosol and Radiation Explorer (EarthCARE) demonstrates that the framework retrieves physically consistent aerosol‐type groupings. Its extensibility is further confirmed by successful adaptation to 532 nm Advanced Carbon Dioxide Detection Lidar (ACDL) observations on Daqi‐1. Embedding aerosol simulations into the classification process, ST‐AeroClassifier provides a transferable, observation‐efficient pathway for spaceborne lidar aerosol typing, underscoring its scalability and broad applicability in future studies.

In this study, a hybrid deep learning and ensemble-based approach is introduced to predict typhoon intensity accurately by utilizing historical meteorological track data. The model is compounded by Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) to comprehensively learn spatial and temporal features of typhoon patterns. It first uses CNN to extract spatial features of typhoon track at-tributes (such as wind speed, pressure, latitude, and longitude) and time series attributes and the LSTM network captures time series features and sequential dynamics of typhoon motions. To mitigate the class imbalance problem due to the much lower number of severe typhoon instances, the Synthetic Minority Oversampling Technique (SMOTE) is utilized for balancing the dataset and for achieving better generalization of the model. In addition, a Random Forest (RF) classifier is adopted as ensemble component to further improve the robustness of classification and predictive performance. The model is tested on the Kaggle Typhoon Track Dataset, in which the typhoon intensity is divided into three ranks: Low, Moderate, and Severe. The expire-mental results show that the proposed hybrid CNN–LSTM–RF–SMOTE framework obtains an overall classification accuracy of 97.1% with F1-scores of 0.910, 0.955 and 0.992 for the Low, Moderate and Severe class, respectively. The results of experiments on real data demonstrate that the proposed HST-LSTM+ model can effectively capture the complex spatiotemporal dependencies of typhoon track data and subsequently improve the reliability of a prediction on the typhoon intensity. This prediction ability may contribute to the disaster preparedness, risk management, and early warning system for extreme weather.

Abstract Large submarine earthquakes introduce significant distortions into the ocean bottom pressure (OBP) data from the Gravity Recovery and Climate Experiment (GRACE) and its follow‐on mission (GRACE‐FO), leading to non‐closure of regional sea‐level budgets in seismically active oceans. In this study, we show that earthquake‐induced seafloor deformations can be effectively isolated using an iterative Empirical Orthogonal Function method that accounts for both co‐seismic and post‐seismic effects. The resulting seismic biases are systematic, driven by ongoing tectonic drift, and can (a) exceed total steric contributions to sea‐level change in some regions and (b) surpass Glacial Isostatic Adjustment (GIA) contributions in the global mean. These findings indicate that seismic deformation represents a non‐negligible component for separating natural and human‐driven contributions in climate assessments. Applying the seismic corrections to the OBP fields restores regional sea‐level budget closure and improves the physical consistency of oceanographic and climate analyses.

Abstract. Snow cover over the Tibetan Plateau (TP) is not only a key land forcing for the regional and global climate but also an important water resource for surround regions. However, state-of-the-art climate models still exhibit substantial biases in simulating winter snow cover over the TP, which constitutes one of the major sources of uncertainty in climate prediction. Using satellite-based snow cover datasets, this study reveals that the Community Land Model version 5 (CLM5) systematically overestimates the winter snow cover fraction (SCF) over the TP. This bias mainly arises because the original SCF parameterization scheme neglects the spatially varying probability distribution of snowfall accumulation and underestimates snow depletion over barren land during the melting period. By accounting for the effects of non-growing-season low vegetation (i.e., withered grass stems) and topographic relief, we parameterize the snow accumulation probability factor (kaccum) instead of prescribing it as a constant. In addition, a revised factor is introduced to modify the snow depletion curve shape parameter (Nmelt), thereby optimizing the SCF parameterization scheme. Preliminary validation indicates that the optimized scheme substantially reduces positive winter SCF biases over the entire TP by 63 %, and improves surface albedo simulations, thereby alleviating cold surface temperature biases by approximately 1–2 °C in snow-affected regions.

Abstract Accurate projection of future precipitation remains challenging due to uncertainties in reference data sets, bias correction and global climate models (GCMs). Here, we evaluated these uncertainties across 13 major cities of the U.S. Gulf Coast, a hurricane‐prone region, under 192 historical and future scenarios. Four reference gridded precipitation data sets, eight CMIP6 GCMs and one HighResMIP model (CMCC‐CM2‐VHR4) were first evaluated against in situ measurements. All GCMs were then bias corrected using four reference data sets and two statistical techniques, empirical quantile mapping (EQM) and a hybrid EQM with linear correction (EQM‐LIN). The bias corrected outputs were then evaluated against in situ measurements. All gridded data sets, HighResMIP and GCMs tended to overestimate light events but underestimate extremes. PRISM and CPC showed the strongest and weakest agreement, respectively, while AORC outperformed others across the Southwest Florida Peninsula, where land‐sea interactions and spatial heterogeneity challenge coarse‐resolution models. Bias correction substantially improved model performance up to the 90th percentile, reducing MAE and RMSE by more than 70% in some cases. However, the performance degraded beyond this percentile; very high percentiles (≥95th) remained underestimated. Future projections under SSP2‐4.5 and SSP5‐8.5 indicated that bias correction reduces inter‐model spread of extreme precipitation indices (Rx1day, SDII and R95p) by approximately 60%–80%, while HighResMIP projections generally remained within the CMIP6 ensemble range. These findings highlighted that credible projections of future precipitation depend more on the representativeness and quality of reference data sets and bias correction technique than the GCMs. The results provide guidance for improving future precipitation projections, updating intensity‐duration‐frequency curves and advancing resilience planning in hurricane‐prone regions.

Groundwater extraction can deplete streamflow in headwater catchments, but the complexity of subsurface hydrological processes make impacts difficult to detect. Using hydrograph-inferred hillslope groundwater storage and streamflow relationships, we propose a novel approach to estimate streamflow depletion from groundwater pumping that is well-suited to areas with limited groundwater monitoring infrastructure. We apply this method in two well-studied watersheds in California’s North Coast to quantify potential hydrologic impacts of cannabis agriculture, which is concentrated in the region and has been identified as a potential threat to salmon-bearing streams. We use a scenario-based approach to explore the relative effects of cannabis cultivation area, irrigation water source (groundwater pumping vs. surface diversion), irrigation efficiency, stream discharge at the onset of the growing season, and lithology on streamflow depletion risk. Our models show that Elder Creek , a perennial stream, could be de-watered by the late dry season with high levels (1% land cover) of cannabis irrigation from groundwater when dry season discharge is low at the start of the season (1 mm/day). In Dry Creek , a non-perennial stream, dry season flow cessation could be advanced by five weeks from similar levels of cannabis water demands. Streamflow impacts are more pronounced in drier years, and the impacts from well-water extraction exhibit a muted effect relative to surface water diversion of the same volume. Storage-discharge functions like those in our case study can estimate how groundwater extraction affects headwater streams wherever streamflow data exist

Abstract The structure and governing processes associated with fast and slow Madden–Julian Oscillation (MJO) propagation speed are examined. On the basis of spectral analysis and three‐dimensional normal‐mode decomposition, we find that slow MJOs exhibit a shorter zonal scale relative to faster MJOs, in agreement with previous work. Thermodynamic distinctions between the two MJO types are then considered by evaluating four criteria that so‐called moisture modes should satisfy, and by examining their column‐integrated moist static energy (MSE) budget. Slow MJO events align more closely with the moisture‐mode criteria than fast events, satisfying all four criteria throughout the broad Indo‐Pacific warm pool, whereas fast MJOs only meet these conditions over the Indian Ocean. These results imply that moisture governs the thermodynamics of slow MJOs, while temperature and moisture play comparable roles in fast MJO thermodynamics. When examining the column MSE budget, we find that the slow MJO MSE anomalies propagate eastward primarily via horizontal MSE advection, with zonal moisture advection accounting for up to 50% of the total MSE tendency over the Indo‐Pacific warm pool. In contrast, vertical MSE advection plays a larger role in fast MJOs, especially east of the Maritime Continent. Lastly, we compare the vorticity budget of fast and slow MJOs. The vorticity anomalies in both fast and slow MJOs propagate eastward via vortex stretching. However, advection of planetary vorticity opposes this vortex stretching and this process is stronger in slow MJOs. Collectively, these findings indicate that MJO governing processes vary with propagation speed and region.