Atmospheric and Oceanic Sciences

Abstract Reliable typhoon forecasting demands both predictive accuracy and physical transparency. This study develops a tree-based framework for simultaneous multi-step prediction of typhoon trajectory and intensity in Western North Pacific (WNP), and applies SHapley additive explanations (SHAP) to quantify the contribution of individual environmental drivers to each forecast. The framework integrates Japan Meteorological Agency Best Track data (1977–2024) with ERA5 and ORAS5 reanalyses, incorporating mean potential temperature over 0–100 m (T100), a predictor largely overlooked in favour of sea surface temperature, together with upper ocean heat content (OHC) and thermocline depth, within a 780-dimensional 72 h lagged feature space. Among three tree-based architectures evaluated under identical conditions, XGBoost performs best across all targets and horizons: at +24 h on the independent 2020–2024 test set it attains RMSE of 1.90◦ and 2.92◦ for latitude and longitude and 11.9 hPa and 16.2 kt for central pressure (CP) and maximum wind speed (MW). This has effect of reducing RMSE by about 45% (track) and 24–25% (intensity) relative to persistence, with bootstrap 95% confidence intervals excluding zero in every case, and by 9% and 13% relative to Random Forest for CP and MW. Elbow-based feature selection shows trajectory converges on a compact predictor subset whereas intensity demands a richer feature space. SHAP analysis reveals a physically coherent horizon-dependent shift: at +6 h kinematic persistence governs both track and intensity, while at +24 h large-scale steering flow dominates track displacement and ocean–atmosphere thermodynamic forcing progressively takes over intensity. T100 exhibits a sharp nonlinear threshold near 26 ◦C that is robust across the test period and encodes upper OHC dynamics without explicit parameterisation. By coupling statistically significant forecast skill with transparent, physics-consistent attribution, the framework offers a physically interpretable complement to black-box systems for typhoon guidance in the WNP.

Abstract Understanding the consequences of continued greenhouse‐gas emissions requires determining whether recent emissions can be robustly linked to changes in extreme weather events. However, current approaches have a limited ability to quantify changes in extreme weather caused by small subsets of anthropogenic emissions, particularly for individual events. To address these limitations, we train a generative machine‐learning model to predict spatially‐explicit changes in daily surface temperature at different levels of cumulative emissions, conditional on the event's observed meteorological conditions. Using this approach, we find strong evidence (>95% probability) that the combined effects of anthropogenic emissions released since the 2015 UN Paris Agreement have increased the intensity of Europe's summer temperature extremes since 2021. Additionally, our results suggest >99% probability that regional‐mean temperature during Europe's high‐impact June 2025 heatwave would have been lower (median estimate of 0.34°C) if the same large‐scale meteorological conditions had occurred under 2015 levels of cumulative emissions.

The Miocene Climatic Optimum (MCO, ~ 15 Ma) offers insights into warm-climate dynamics and future climate change. However, its global warmth magnitude remains uncertain due to limitations in surface and benthic proxy records. Here, we develop long-run Miocene simulations featuring deep-ocean equilibration and water isotope capability, and present a probabilistic inference framework integrating them with a global compilation of benthic foraminiferal δ18O to better constrain the MCO warmth. Our approach yields a maximum likelihood estimate of MCO global mean surface temperature of 7.5∘C above preindustrial, significantly warmer than previous benthic δ18O-based reconstructions, implying a higher MCO-derived estimate of Earth system sensitivity. The corresponding surface temperature field shows among the best agreements with independent surface temperature proxies. This study highlights the importance of deep-ocean equilibration and proxy-model integration for accurately estimating both deep-ocean and surface temperatures, and offers a method applicable to improving global climate reconstructions across other time intervals. Combining long climate-isotope simulations with deep-sea oxygen isotope records, the Miocene Climatic Optimum (15 Ma) is estimated to have been 7.5∘C warmer than preindustrial -- 30% above earlier estimates.

Long-term monitoring of global river discharge has been hindered by the uneven distribution of gauging stations and limited data accessibility, a challenge that is particularly acute in China. Although China contains one of the world's densest river networks, high-frequency in situ discharge observations remain largely unavailable in the international public domain. To address this gap, we compiled daily discharge records from 1196 gauges across China, comprising approximately 2.33 ×10 6 observations – 39 times as many gauges as are currently available for the region in the Global Runoff Data Centre (GRDC). Leveraging this unprecedented collection of in situ discharge records, along with river width time series derived from Landsat and Sentinel-2 imagery and gauge-specific hydraulic geometry relationships, we reconstructed and extended daily river discharge observations for 310 gauges from 1990 to 2024, resulting in the China Daily River Discharge Records (CDR 2 ) dataset. Compared with existing global satellite-derived discharge products, CDR 2 expands the number of available gauges in China by more than fivefold while delivering substantially improved performance, achieving a median Kling–Gupta efficiency of 0.66 during validation. Moreover, uncertainties propagated from the fitted hydraulic geometry parameters remain low, with a median relative uncertainty of only 11.89 % in the reconstructed discharge estimates. Sensitivity analyses further indicate that discharge estimation accuracy increases markedly with greater river width variability and stronger hydraulic sensitivity. Trend analysis reveals that nearly 65 % of gauges exhibit declining discharge over 1990–2024, with a median relative trend of −0.21 % yr −1 , most pronounced in the Haihe, Liaohe, Yellow River, and middle Yangtze River basins. As the most extensive satellite-derived, gauge-constrained river discharge dataset currently available for China at daily resolution, CDR 2 bridges a critical geographic gap in global river monitoring and provides a valuable benchmark for future discharge estimation, hydrological research, water resources management, and the calibration and evaluation of satellite missions. The CDR 2 dataset is publicly available on Zenodo at https://doi.org/10.5281/zenodo.22231453 (Wang and Li, 2026).

Abstract Climate change increases fire weather globally. Hot, dry and windy conditions raise the likelihood of fires igniting and spreading and make suppression more challenging. With further warming, fire weather is projected to intensify across Europe, yet implications for today’s young generations remain unclear. Here, we analyse lifetime exposure to extreme fire weather across Europe using an ensemble of bias-adjusted and downscaled global climate models. To this end, we developed dem4cli , an open-source Python package that integrates demographic data into climate hazard assessments. We find younger generations are projected to be disproportionately exposed compared to older generations across all warming pathways, while also benefitting most from ambitious mitigation. Across Europe, individuals born in 2025 are projected to experience, on average, 3.2 additional years of exposure to extreme fire weather compared to those born in 1950, corresponding to an 84% increase, with larger increases observed in Southern and Eastern Europe. Focusing on Portugal, under current policies, people born in 2025 are projected to experience nearly twice the exposure compared with those born in 1950, and up to 2.5 times more in the most exposed regions of northeastern Portugal. Under a 1.5 ∘ C pathway, lifetime exposure is substantially reduced. Each additional degree of global warming by 2100 adds approximately 270 days of extreme fire weather to the lifetime exposure of a person born in Portugal in 2025. These results reveal pronounced intergenerational inequalities in exposure to fire weather and underscore the urgency of ambitious mitigation to limit cumulative exposure for younger generations.

High-resolution gridded climate datasets are essential for Earth system modelling and impact assessments, yet generating them from sparse, irregularly distributed station networks remains a significant challenge, particularly in regions with complex topography. This study evaluates the Spatial Multi-Attention Conditional Neural Process (SMACNP), a probabilistic deep learning framework, for the daily spatial interpolation of air temperature and precipitation, marking the first application of its localized encoder variant to the challenge of gridding climate data from a sparse station network. We investigate two distinct encoder configurations – Global and Localized – to determine the optimal structural prior for capturing spatial dependencies in data-scarce regimes. The models were developed and evaluated using data from a sparse network of meteorological stations in Romania from 2020 to 2023. To ensure applicability for long-term historical reconstruction, the input features were restricted to static topographic predictors derived from a Digital Elevation Model (DEM). Performance was benchmarked against Regression Kriging (RK), a standard geostatistical baseline that incorporates these same topographic covariates. Results demonstrate that the SMACNP architectures substantially outperform the RK baseline for both variables. The SMACNP (Localized) configuration, which utilizes an attention mechanism, emerged as the most robust model, achieving the lowest Mean Absolute Error (MAE) and the highest correlation across the majority of seasons. The performance gains were particularly pronounced for precipitation, where the deep learning models effectively captured fine-scale spatial heterogeneity and non-linearities that traditional methods tended to over-smooth. Furthermore, the SMACNP framework demonstrated superior uncertainty quantification; while RK exhibited significant overconfidence in precipitation estimates, the SMACNP (Localized) model produced well-calibrated probabilistic predictions with near-ideal empirical coverage. These findings indicate that localized neural process-based models offer a powerful, scalable, and physically plausible alternative to geostatistical methods for generating high-quality gridded climate datasets in complex, data-sparse environments.

Accidental oil spills in the Alboran Sea represent a high-risk hazard because intense maritime traffic, semi-enclosed circulation, and sensitive Moroccan coastal ecosystems converge within a narrow coastal corridor. Operational-scale simulations for the Nador-Saïdia sector remain scarce, particularly for persistent heavy fuel oil and for the combined effects of mesoscale circulation, wind drift, Stokes drift, vertical mixing, and weathering. This study applies the OpenOil module of the OpenDrift Lagrangian framework to simulate a hypothetical release of 100 tons of heavy fuel oil (Generic Bunker C) offshore the Driouch coast (35.45° N, 2.95° W) over a seven-day (168 h) period, beginning on 02 July 2024 at 12:00 UTC. The model was forced with CMEMS Mediterranean physics fields, CMEMS Mediterranean wave data, and NOAA-GFS 10 m winds. The selected summer scenario represents a high-impact vulnerability case characterized by strong stratification, sea-breeze influence, peak beach-tourism exposure, and high ecological sensitivity of the Marchica Lagoon, Moulouya River estuary, and Cap des Trois Fourches. The simulations incorporated high-resolution meteorological and oceanographic forcing from CMEMS and NOAA-GFS. Results highlight the strong influence of mesoscale circulation patterns in the Alboran Sea, particularly the interaction between regional gyres, the Algerian Current, and local wind-driven drift. Lagrangian trajectories indicate a dominant east–southeastward transport, posing a direct threat to environmentally sensitive coastal areas. By the end of the simulation, 4,253 particles were stranded, corresponding to approximately 85.1 t of oil, while 747 particles remained active at sea, corresponding to approximately 14.9 t. Weathering was dominated by emulsification and viscosity increase, whereas evaporation remained limited to approximately 10–15% of the released mass and natural dispersion remained below 10%. The revised analysis indicates that high viscosity and density lower than seawater explain the long surface residence time of fresh Bunker C, while sedimentation risk becomes critical mainly in shallow, sediment-rich environments through oil-mineral aggregation. The study identifies priority coastal exposure pathways and provides management-relevant guidance for preparedness planning, including pre-positioning of response equipment adapted to viscous oils, rapid protection of lagoon inlets where feasible, and transboundary coordination between Morocco and Algeria.

Late spring frost remains a major threat to winter wheat even as the climate warms, because accelerated crop development can increase the overlap between frost-sensitive stages and episodic spring cold events. Yet large-scale yield risk remains difficult to quantify because this overlap changes over time and frost damage differs among cultivars. We developed a phenology-constrained, event-based framework that integrates controlled low-temperature experiments, phenology reconstruction constrained by agrometeorological observations, machine-learning-based yield-reduction modelling, and bias-corrected hourly temperature projections. We assessed three cultivar resistance groups across China’s major winter wheat regions during 1980–2024 and under SSP2–4.5 and SSP5–8.5 during 2030–2100. Warming did not uniformly reduce frost damage but shifted its geographic center. For cold-sensitive cultivars, historical mean yield reduction was greatest in the Middle–Lower Reaches of the Yangtze River region (10.02%), whereas the Huang–Huai region became the principal future hotspot, reaching 6.49% under SSP5–8.5. This redistribution arose from phenological advancement and changes in the frequency and persistence of cold exposure during sensitive stages, while changes in minimum temperature were comparatively small. Counterfactual simulations using fixed historical phenology showed that dynamic phenological shifts increased future losses, mainly during jointing, the stage that dominated losses across both periods. Cold-tolerant cultivars reduced losses by approximately half in the most vulnerable regions. These findings show that warming reshapes rather than removes late spring frost risk by altering phenology–climate overlap, underscoring the need for phenology-aware monitoring and region-specific deployment of cold-tolerant cultivars.

Abstract Concurrent heatwaves over Europe and North China are the most frequent compound heat extremes across the Northern Hemisphere. They are systematically organized by a recurrent teleconnection along the polar front jet over Eurasia. Using observations and large‐ensemble simulations, we show that even after removing the forced warming trend, future concurrent heatwave events (2070–2100) intensify substantially relative to the historical period (1920–2005), with excess warming over both regions exceeding 0.5°C locally and 0.3°C regionally. This intensification is driven by fundamentally different mechanisms in the two regions. Over North China, a strengthened ridge promotes subsidence and clear‐sky conditions that enhance radiative heating, whereas over Europe, progressive soil drying under climate change reduces evaporative cooling, producing a net energy surplus even as the circulation anomaly weakens. These findings demonstrate that the concurrent structure of future heat extremes intensifies beyond the mean climate response, a compound risk not captured by mean‐state projections.

Abstract Earth system models exhibit persistent biases in simulating cloud condensation nuclei (CCN) over the Southern Ocean and Antarctica, largely due to inadequate representation of aerosol sources. Here, we present year‐round (January 2023–February 2024) measurements of aerosol hygroscopicity (HTDMA), particle number size distribution (SMPS), and CCN concentration from King Sejong Station (KSJ), Antarctica, to constrain CCN sources in this pristine environment. Size‐resolved hygroscopicity‐based source apportionment shows that sea salt aerosol (SSA) contributes 43%–54% of CCN at cloud‐relevant supersaturations (Sc = 0.2%–1.0%). Compared against the Copernicus Atmosphere Monitoring Service reanalysis (CAMSRA), our observations demonstrate that models underestimate SSA‐derived CCN at KSJ by a factor of 30 at Sc = 0.2%, providing critical constraints for aerosol–cloud interactions.