Atmospheric and Oceanic Sciences

Abstract. Accurate and temporally continuous global observations of atmospheric carbon dioxide (XCO2) are essential for climate monitoring and emission assessment. However, satellite-based XCO2 observations are often spatially incomplete and temporally discontinuous, while existing products typically suffer from coarse spatial resolution, hindering the detection of fine-scale emission changes. Here, we developed a novel spatiotemporal Transformer–BiLSTM deep-learning network that combines the Transformer's ability to model long-range spatial dependencies through self-attention mechanisms with the BiLSTM's ability to capture temporal dynamics. The network assimilates multisource data from satellite observations, meteorological reanalysis, and precursor gases to reconstruct a global, daily, and seamless XCO2 dataset over land at 0.1° spatial resolution from 2003 to 2022. Independent validation of the data-fused XCO2 product against Total Carbon Column Observing Network (TCCON) measurements shows excellent agreement, with an R2 of 0.99, an RMSE of 1.10 ppm, and a mean bias of 0.01 ppm. After bias correction, cross-satellite consistency is further enhanced, achieving a sample-based CV-R2 of 0.99 and an RMSE of 0.36 ppm. The dataset provides accurate daily XCO2 estimates over global land surfaces, enabling investigations of spatial heterogeneity and regional-to-local XCO2 enhancement patterns linked to anthropogenic emissions and biomass-burning events. The record reveals a persistent global increase in atmospheric XCO2 over the past two decades, with a mean growth rate of 2.24 ppm yr−1 (p<0.001). It reliably resolves global XCO2 variability across a wide range of temporal scales, from day-to-day fluctuations to long-term trends. It consistently captures large-scale climate-driven signals, such as ENSO-related interannual variability, and short-lived XCO2 enhancements associated with major wildfire events, demonstrating its capability to represent both persistent and episodic emission signals. This high-resolution, daily global XCO2 (GlobalHighXCO2) product provides a valuable resource for carbon-cycle research, atmospheric model evaluation, and emission monitoring, and is publicly available at https://doi.org/10.5281/zenodo.18220962 (Qu and Wei, 2026).

Abstract Climate warming is accelerating abrupt permafrost thaw, driving substantial carbon emissions. Retrogressive thaw slumps (RTSs) represent the most severe instance of abrupt permafrost thaw, yet their carbon emissions remain poorly quantified due to limited observations. Here, by synthesizing 4728 RTS incidents and 1862 in-situ CO 2 and CH 4 measurements from RTS-affected zones across the Tibetan Plateau, we estimate that the area of RTS susceptibility will expand by 17–19% by 2100 relative to 2022, driven primarily by precipitation changes. Compared to control areas, the ecosystem respiration rate in collapsed areas decreases by 14.4%, while CH 4 release rate increases by 20.0%. The combined CO 2 and CH 4 release associated with RTS expansion increased 1.1-fold between 2016 and 2022. Under the intermediate Shared Socioeconomic Pathways scenario, carbon emissions from RTS-susceptible areas are projected to surge 2.7-fold by 2100. These findings highlight that abrupt thaw strengthens permafrost carbon-climate feedback in high-altitude regions, underscoring the urgent need for targeted permafrost protection strategies to achieve carbon neutrality goals.

Abstract Reconstructing past South Asian monsoon dynamics, which governs wind and rainfall patterns across the Indian subcontinent, is crucial for understanding low-latitude climate. Constraining monsoon drivers requires separating summer and winter components in palaeoclimate archives, but this remains challenging because proxy signals usually integrate seasonal signals. The northeastern Arabian Sea provides a unique setting in which sedimentary and geochemical proxies are driven by the summer monsoon, while sea surface temperature is primarily controlled by winter monsoon winds. Here we employed mass spectrometry and hyperspectral imaging on a sediment core off Pakistan to reconstruct subdecadal changes in sea surface temperature and marine primary production during the last deglaciation (~16,000–12,000 years ago). Atmospheric humidity and vegetation changes were additionally assessed by plant-wax isotope analyses on a subset of samples. We show that summer monsoon winds were driven by the Northern Hemisphere high-latitude climate on centennial-to-millennial timescales. Winter monsoon was primarily characterized by a millennial-scale decline in wind strength driven by increasing global temperatures, with superimposed centennial-scale variability. We identified an inverse relationship between winter monsoon wind strength and the amount of winter non-monsoonal precipitation. Mechanistic insights of seasonal monsoon dynamics improved interpretation of regional palaeoprecipitation records and may enhance climate model performances in low latitudes.

Tropical cyclones over land do not simply decay; they reorganize and interact with orography, producing extreme rainfall and flooding far from the center of the storm. In the Appalachian region, orographic enhancement results in areas with extreme rainfall amounts, leading to precipitation and flood peak distributions that do not exhibit evidence of an upper bound based on available observations and statistical analyses, although the existence of a physical upper limit cannot be ruled out. Here we use an empirical approach revolving around Hurricane Helene and long-term records to support the notion that design practices anchored in upper bounds (e.g., Probable Maximum Precipitation) are increasingly misaligned with observed extremes and projected risks. We argue for probabilistic frameworks that explicitly account for extremely low annual exceedance probabilities, mechanism-dependent tail behavior, and uncertainty beyond historical records. By integrating statistics reflective of the lack of an upper bound, high-resolution modeling, and by accounting for different processes in flood frequency analyses, the meteorological and hydrologic community can better inform resilient infrastructure and risk communication in an era of accelerating climate extremes.

Abstract. Ocean fronts are critical interfaces between different water masses and profoundly influence atmosphere–ocean interactions, weather systems, marine ecosystems, and climate regulation. Accurate and long-term observations of ocean fronts are essential for advancing studies in meteorology, oceanography, and climate science. However, no publicly available, long-term ocean front dataset currently exists, and the existing detection methods often rely on time-consuming manual labeling or traditional algorithms with limited accuracy in complex frontal regions. In this study, we release the first publicly available 30-year ocean front dataset (1993–2023) for the Northwest Pacific, which was generated by applying a deep learning framework (Mask R-CNN) to daily sea surface temperature (SST) fields with manually annotated samples for model training. The model was trained utilizing both L3 remote sensing satellite SST data and the GLORYS12V1 L4 reanalysis SST product. The dataset provides pixel-level frontal boundaries along with the associated attributes, including the position, intensity, and width, stored in NetCDF-4 format at a 1/12° spatial and daily temporal resolution. An accuracy evaluation shows that the mean average precision (mAP) exceeds 0.90, and compared with traditional gradient methods, the errors in front width and intensity are smaller. The dataset offers three main contributions: (1) It fills a critical gap by providing a standardized, long-term ocean front product; (2) it serves as a ready-to-use training resource for deep learning models that greatly reduces the need for manual labeling; and (3) it provides benchmark samples for validation and intercomparison of other ocean front detection products. This dataset supports robust investigations of the seasonal-to-interannual frontal variability and provides a valuable foundation for applications in meteorology, ecosystems management and climate change research. The ocean front dataset developed in this study is available at https://doi.org/10.5281/zenodo.16921277 (Niu, 2025a).

Abstract. The El Niño–Southern Oscillation (ENSO) is a dominant mode of interannual climate variability, yet accurately simulating ENSO in climate models remains a major challenge due to its complex coupled dynamics. In this study, we present a linear optimization framework and systematically adjust atmospheric parameters to improve ENSO fidelity in the Icosahedral Nonhydrostatic eXtended Predictions and Projections (ICON XPP) Earth System Model of the Max-Planck-Institute for Meteorology. The optimization approach is based on the superposition of parameter sensitivities and a Nelder–Mead algorithm that reduces the ENSO cost function. The cost function accounts for ENSO-related tropical climatology, variability, and feedbacks, which are estimated with the ENSO metric package. We first assess the sensitivity of ENSO metrics to 21 atmospheric parameters in atmosphere-only simulations. The optimization approach reduces the ENSO cost function by 30 % in the optimized atmosphere-only runs. Key improvements include reduced precipitation bias and strengthened atmospheric feedbacks such as the Bjerknes and thermal damping feedbacks. These results demonstrate the effectiveness of our method in improving ENSO metrics within the atmosphere-only configuration. Six parameters identified as most impactful from atmosphere-only tuning experiments are subsequently tuned in fully coupled simulations. The optimized fully coupled run yields moderate improvements in ENSO amplitude, cold tongue SST bias, seasonal phase-locking, ocean-atmosphere coupling and teleconnection patterns. However, isolated ENSO tuning introduces unrealistic global warming, which is further corrected by adjusting turbulence-related parameters without degrading ENSO skill. These results demonstrate that systematic ENSO tuning can yield performance gains but must be balanced with broader climate stability constraints. Our method offers a scalable, physically grounded optimization strategy, with strong potential for tuning ENSO in climate model configurations.

Abstract. This article presents the latest version of the Advanced Scatterometer (ASCAT) surface soil moisture (SSM) dataset provided by the Satellite Application Facility on Support to Operational Hydrology and Water Management (H SAF) lead by the European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT). This new release brings the operational near real-time (NRT) product up to date with the historical offline data record. For years, the H SAF ASCAT SSM data records have benefited from successive algorithmic improvements while the H SAF ASCAT SSM NRT product has only received minor updates until its discontinuation in July 2025. A new processing chain replaces the previous service and applies the latest soil moisture retrieval algorithm to both data streams, creating a unified offline/NRT dataset and representing a major advancement for the H SAF ASCAT SSM NRT product. The H SAF ASCAT SSM climate data record (CDR) covers the time period 1 January 2007 until 31 December 2024, which is extended offline by an interim climate data record (ICDR) as well as in NRT. The new release also introduces a high-resolution 6.25 km sampling H SAF ASCAT SSM dataset, alongside the standard 12.5 km sampling SSM dataset. This is achieved by customising the spatial resampling process of the ASCAT Level 1B full-resolution backscatter data. A new key development in the algorithm for ASCAT SSM concerns the estimation of the dry and wet backscatter references. Specifically, a moving-window approach is now applied to mitigate artificial trends caused by long-term land cover changes. Furthermore, a new monthly subsurface scattering flag has been added to filter out unreliable SSM measurements where backscatter and soil moisture indicate an inverted relationship. Quality control of the H SAF ASCAT SSM datasets is performed by using soil moisture estimates from Noah GLDAS-2.1 and the ESA CCI Passive Soil Moisture (SM) v09.1 product, as well as in-situ observations provided by the International Soil Moisture Network (ISMN). The validation results show that both H SAF ASCAT SSM datasets have a comparable performance in terms of the Pearson correlation coefficient (H SAF ASCAT SSM 6.25 km vs ESA CCI Passive SM: 17.9 % > 0.75 and 57.8 % > 0.5; H SAF ASCAT SSM 12.5 km vs ESA CCI Passiv SM: 19.6 % > 0.75 and 59.2 % > 0.5) and signal-to-noise ratio (SNR) derived using triple collocation analysis (H SAF ASCAT SSM 6.25 km SNR: 56.0 % > 0 dB, 35.6 % > 3 dB, H SAF ASCAT SSM 12.5 km SNR: 58.1 % > 0 dB, 38.6 % > 3 dB). The best agreement can be found in regions with strong seasonal variability, including monsoonal, savanna, Mediterranean, and tropical wet-and-dry zones. A weaker consistency can be found in areas characterised by limited soil moisture variability (such as deserts), dense vegetation, pronounced topographic complexity, wetland areas, or higher latitudes (> 60∘ N) experiencing longer periods of frozen soil and snow cover. The H SAF ASCAT SSM CDR and ICDR datasets (6.25 km sampling: https://doi.org/10.15770/EUM_SAF_H_0012, H SAF, 2025c, and 12.5 km sampling: https://doi.org/10.15770/EUM_SAF_H_0011, H SAF, 2025a) are publicly available online (H SAF, 2025a, c, d, e), while H SAF ASCAT SSM NRT datasets (H SAF, 2025b, f) are distributed via the broadcasting system EUMETCast.

Abstract Climate change is pushing many regions across Europe into heat conditions beyond local historical experience, challenging heat-mortality models primarily calibrated on local observations. These locally learned exposure-response relationships are often weakest precisely where extreme heat has been rare and risk is now emerging. Here, we show that jointly learning temperature-mortality relationships across regions using a shared European exposure space improves estimation of excess mortality under locally unprecedented heat events and enhances near-term, out-of-sample predictive performance. By borrowing strength across regions with heterogeneous heat exposure, this approach replaces local extrapolation beyond observed temperatures with interpolation across a broader range of observed exposures. Joint learning yields the largest gains in cooler climates with historically rare heat exposure (mean increase in pseudo-R2=0.09, IQR [0.07-0.13]) and in regions experiencing temperatures exceeding local past maxima. Importantly, the learned relationships remain epidemiologically interpretable and stable when incorporating concurrent environmental exposures, enabling assessment of modifiers such as humidity while preserving coherent temperature-mortality relationships. As climate change continues to push populations into locally unprecedented heat conditions, these strengths support the use of joint learning for the development of early warning systems and frameworks for the near-term forecasting of heat-related mortality.

Abstract. Robust estimates of methane emissions are critical for understanding their impacts on atmospheric warming and air quality, and for assessing methane mitigation strategies. Gridded inventories, such as the U.S. Environmental Protection Agency's Greenhouse Gas Inventory (EPA GHGI), the Emissions Database for Global Atmospheric Research (EDGAR 2024), and the National Oceanic and Atmospheric Administration's Fossil Fuel Oil and Gas inventory (NOAA FOG), are constructed to evaluate large-scale emission patterns and support identifying emission mitigation priorities and prioritizing future measurements. However, substantial differences across inventories complicate such assessments. We benchmark EPA GHGI, EDGAR 2024, and NOAA FOG against flux estimates from an atmospheric inversion of Greenhouse Gases Observing Satellite (GOSAT) data from 2012 to 2020 over the Contiguous United States (CONUS). A key technical challenge is the heterogeneous sensitivity of satellite-derived fluxes, which depends on measurement uncertainty, coverage, and inversion model configuration. We account for this heterogeneity by applying an inversion operator to each inventory prior to comparison with the GOSAT-based estimates. The GOSAT estimates are most sensitive to oil and gas and livestock emissions; oil and gas emissions are consistent with NOAA FOG (14.1 Tg CH4 yr1 in 2015), but exceed EPA GHGI and EDGAR, particularly across Texas, Oklahoma, and Louisiana. GOSAT-based livestock emissions exceed EPA GHGI and EDGAR by 1–2 Tg CH4 yr1, with the largest differences in the Midwest and California. Despite these discrepancies, both activity and satellite based estimates show no observable trends from 2012 to 2020 in fossil and livestock emissions.

China’s clean air actions have substantially improved air quality and reduced PM 2.5 -attributable mortality, but their equity impacts across life stages remain unclear. Here, we assessed spatiotemporal changes in PM 2.5 -attributable mortality among infants and adults in China from 2013 to 2023 using high-resolution PM 2.5 data, gridded population data sets, and exposure–response functions from the Global Exposure Mortality Model (GEMM) and previous meta-analysis. Socioeconomic inequality was evaluated using gridded gross domestic product data and quantified by the concentration index (CI), which is a classical metric of socioeconomic inequality and has been applied to assess the health impacts from air pollution exposure. From 2013 to 2023, population-weighted PM 2.5 exposure declined substantially in both groups, accompanied by marked reductions in attributable mortality. PM 2.5 -attributable infant deaths declined from 89.2 thousand (95% uncertainty interval: 36.3–110.8) in 2013 to 14.0 thousand (95% UI: 5.0–20.6) in 2023, representing an 84% reduction, compared with a 39% reduction among adults. Despite these overall improvements, PM 2.5 -attributable infant deaths were disproportionately concentrated in economically disadvantaged regions, with the concentration index worsening from −0.071 in 2013 to −0.096 in 2023, whereas inequality among adults narrowed toward zero. To reduce environmental inequality, increased focus on reducing air pollution exposure in economically disadvantaged regions is warranted.