#1
Pioneering integration of quantum computing and machine learning for hurricane prediction, high novelty and impact.
Abstract
Introduction Accurate prediction of hurricane tracks, intensity, and rapid intensification remains a major challenge in meteorological forecasting. This study investigates whether a hybrid quantum-classical approach, called the Quantum-Enhanced Neural Network (QENN), can improve tropical cyclone prediction. Methods QENN combines atmospheric features from multiple sources with a four-layer, 50-qubit variational circuit, grouped Pauli measurements, and a classical temporal prediction model. The evaluation included 847 Atlantic tropical cyclones recorded between 1980 and 2024. Separate hurricane-disjoint partitions were used for model development, validation, testing, and the 2024 holdout evaluation. Model reliability was further examined through ten random-seed runs, storm-level bootstrap confidence intervals, calibration analysis, comparisons with recent forecasting models, and a width-matched classical ablation study. Results On the locked 2016–2023 test set, QENN achieved a 120-h track mean absolute error of 83.7 km (95% confidence interval [CI]: 78.9–88.6 km), compared with 312.0 km (95% CI: 299.6–325.1 km) for the operational classical ensemble, representing a 73.2% reduction in relative error. For rapid-intensification prediction, QENN achieved a balanced accuracy of 94.7% (95% CI: 91.2%–97.1%), while the classical baseline achieved 67.3%. The median wall-clock time was 50.0 min for QENN and 210.0 min for the classical ensemble, equivalent to a 4.20-fold speed-up and a 76.2% reduction in runtime. In the 2024 holdout evaluation, which included six storms, the mean landfall-position error was 45 km (95% CI: 39–51 km). Discussion The results indicate that quantum-enhanced correlation features may improve forecasting accuracy and computational efficiency within the proposed experimental framework. However, the runtime comparison does not establish a formal quantum computational advantage. Further testing under operational forecasting conditions and validation on quantum hardware are needed to assess the reliability, generalizability, and practical value of the approach.
Introduction Accurate prediction of hurricane tracks, intensity, and rapid intensification remains a major challenge in meteorological forecasting. This study investigates whether a hybrid quantum-classical approach, called the Quantum-Enhanced Neural Network (QENN), can improve tropical cyclone prediction. Methods QENN combines atmospheric features from multiple sources with a four-layer, 50-qubit variational circuit, grouped Pauli measurements, and a classical temporal prediction model. The evaluation included 847 Atlantic tropical cyclones recorded between 1980 and 2024. Separate hurricane-disjoint partitions were used for model development, validation, testing, and the 2024 holdout evaluation. Model reliability was further examined through ten random-seed runs, storm-level bootstrap confidence intervals, calibration analysis, comparisons with recent forecasting models, and a width-matched classical ablation study. Results On the locked 2016–2023 test set, QENN achieved a 120-h track mean absolute error of 83.7 km (95% confidence interval [CI]: 78.9–88.6 km), compared with 312.0 km (95% CI: 299.6–325.1 km) for the operational classical ensemble, representing a 73.2% reduction in relative error. For rapid-intensification prediction, QENN achieved a balanced accuracy of 94.7% (95% CI: 91.2%–97.1%), while the classical baseline achieved 67.3%. The median wall-clock time was 50.0 min for QENN and 210.0 min for the classical ensemble, equivalent to a 4.20-fold speed-up and a 76.2% reduction in runtime. In the 2024 holdout evaluation, which included six storms, the mean landfall-position error was 45 km (95% CI: 39–51 km). Discussion The results indicate that quantum-enhanced correlation features may improve forecasting accuracy and computational efficiency within the proposed experimental framework. However, the runtime comparison does not establish a formal quantum computational advantage. Further testing under operational forecasting conditions and validation on quantum hardware are needed to assess the reliability, generalizability, and practical value of the approach.
#2
Major advance in Earth system data curation, enabling improved reanalysis and model development.
Abstract
Abstract The NOAA-NASA Joint Archive (NNJA pronounced like the word ninja) is a publicly accessible, curated collection of Earth system observations spanning 1979 to present, created to accelerate reanalysis production, support development of artificial intelligence (AI) applications, and enable access to the observational record underlying modern reanalyses. The archive encompasses atmospheric satellite radiances and conventional measurements, in-situ ocean profiles, sea ice concentrations and freeboard retrievals, and land surface observations from over 60 sensors provided by NOAA, NASA, and their partners. The number of atmospheric records increases from around 10 6 observations per day in 1979 to around 10 8 in the modern era. The dataset extends into near-real time through continuous archival of the NOAA operational data. NNJA observations are distributed through Amazon Web Services in formats compatible with established data assimilation systems and are accompanied by channel-level black and white lists that capture the quality history of the satellite record. Data integrity is verified through a three-tier process: file-level metadata checks, cross-validation against existing black and white lists from prior NOAA and NASA reanalyses, and a comparison of assimilation statistics with reference reanalysis and operational products. To lower barriers for the machine learning community, a companion dataset (NNJA-AI) transforms NNJA observations from legacy formats to column-oriented Apache Parquet, enabling sufficient large-scale queries and direct integration with AI training workflows. This article describes the creation, verification, and curation of the NNJA dataset, document its current scope and known limitations, and discusses its role as a shared observational infrastructure for the next generation of U.S. reanalyses.
Abstract The NOAA-NASA Joint Archive (NNJA pronounced like the word ninja) is a publicly accessible, curated collection of Earth system observations spanning 1979 to present, created to accelerate reanalysis production, support development of artificial intelligence (AI) applications, and enable access to the observational record underlying modern reanalyses. The archive encompasses atmospheric satellite radiances and conventional measurements, in-situ ocean profiles, sea ice concentrations and freeboard retrievals, and land surface observations from over 60 sensors provided by NOAA, NASA, and their partners. The number of atmospheric records increases from around 10 6 observations per day in 1979 to around 10 8 in the modern era. The dataset extends into near-real time through continuous archival of the NOAA operational data. NNJA observations are distributed through Amazon Web Services in formats compatible with established data assimilation systems and are accompanied by channel-level black and white lists that capture the quality history of the satellite record. Data integrity is verified through a three-tier process: file-level metadata checks, cross-validation against existing black and white lists from prior NOAA and NASA reanalyses, and a comparison of assimilation statistics with reference reanalysis and operational products. To lower barriers for the machine learning community, a companion dataset (NNJA-AI) transforms NNJA observations from legacy formats to column-oriented Apache Parquet, enabling sufficient large-scale queries and direct integration with AI training workflows. This article describes the creation, verification, and curation of the NNJA dataset, document its current scope and known limitations, and discusses its role as a shared observational infrastructure for the next generation of U.S. reanalyses.
#3
Demonstrates AI's transformative potential for subseasonal weather and climate prediction.
Abstract
An innovative artificial-intelligence application enhances weather forecasting in the difficult subseasonal range, providing accurate forecasts of precipitation, temperature and pressure 3–5 weeks in advance. The tool refines both physics- and AI-based weather models and improves the early prediction of extremes of heat, cold, rainfall and flooding. The tool boosts the ability of physics- and AI-based forecasting models to predict typical and extreme weather events.
An innovative artificial-intelligence application enhances weather forecasting in the difficult subseasonal range, providing accurate forecasts of precipitation, temperature and pressure 3–5 weeks in advance. The tool refines both physics- and AI-based weather models and improves the early prediction of extremes of heat, cold, rainfall and flooding. The tool boosts the ability of physics- and AI-based forecasting models to predict typical and extreme weather events.
#4
Links Tibetan Plateau uplift to increased tropical cyclone frequency, advancing paleoclimate understanding.
Abstract
Nearly one-third of global tropical cyclones (TCs) form over the western North Pacific (WNP), with severe hazards impacting Southeast and East Asia. TC genesis over the WNP underwent substantial shifts during the Cenozoic, yet the underlying mechanisms remain unclear. Here we show that Tibetan Plateau uplift was crucial in shaping the present-day WNP TC genesis pattern. With the present-day land-sea configuration and sea surface temperature prescribed, Tibetan Plateau uplift from early Eocene to modern elevations accounts for nearly half of the simulated increase in TC frequency between the early Eocene and present day. Mechanistically, Tibetan Plateau uplift generates an eastward-propagating Rossby wave train, enhancing the WNP subtropical high and thereby intensifying the monsoon trough. The amplified moisture supply and reinforced upward motion within the monsoon trough collectively favor TC genesis and intensification. Our findings underscore the pivotal role of High Mountain Asia in the Cenozoic evolution of TC frequency over the WNP. Tropical cyclogenesis over the western North Pacific was largely absent in the Eocene (~55 million years ago) but has increased over time, with Tibetan Plateau uplift acting as a key driver in this evolutionary process via reshaping large-scale atmospheric circulation.
Nearly one-third of global tropical cyclones (TCs) form over the western North Pacific (WNP), with severe hazards impacting Southeast and East Asia. TC genesis over the WNP underwent substantial shifts during the Cenozoic, yet the underlying mechanisms remain unclear. Here we show that Tibetan Plateau uplift was crucial in shaping the present-day WNP TC genesis pattern. With the present-day land-sea configuration and sea surface temperature prescribed, Tibetan Plateau uplift from early Eocene to modern elevations accounts for nearly half of the simulated increase in TC frequency between the early Eocene and present day. Mechanistically, Tibetan Plateau uplift generates an eastward-propagating Rossby wave train, enhancing the WNP subtropical high and thereby intensifying the monsoon trough. The amplified moisture supply and reinforced upward motion within the monsoon trough collectively favor TC genesis and intensification. Our findings underscore the pivotal role of High Mountain Asia in the Cenozoic evolution of TC frequency over the WNP. Tropical cyclogenesis over the western North Pacific was largely absent in the Eocene (~55 million years ago) but has increased over time, with Tibetan Plateau uplift acting as a key driver in this evolutionary process via reshaping large-scale atmospheric circulation.
#5
Establishes new standards for aerosol model benchmarking, crucial for atmospheric chemistry and climate models.
Abstract
Abstract Aerosol processes are a leading source of uncertainty in predicting human impacts on Earth’s energy balance and air quality. Aerosol effects depend critically on the characteristics of individual particles, including size, chemical composition, and morphology, but these particle-level attributes cannot be tracked explicitly in large-scale, long-running Earth system models. Instead, aerosol processes are simplified, with different modeling groups applying different approximations. These approximations introduce structural errors in model predictions that have not been well quantified. As a path forward, we introduce the Aerosol Model Benchmarking Repository and Standards (AMBRS) Initiative, a community-based platform for building, running, and evaluating aerosol chemical and microphysical schemes at the box-model level. By isolating the errors introduced by simplifications in current and proposed aerosol models, AMBRS enables transparent, reproducible comparison of process-level performance. The effort aims to support community standards for model benchmarking and promote iterative development of aerosol modules, ultimately helping reduce structural uncertainty in simulations of aerosol effects.
Abstract Aerosol processes are a leading source of uncertainty in predicting human impacts on Earth’s energy balance and air quality. Aerosol effects depend critically on the characteristics of individual particles, including size, chemical composition, and morphology, but these particle-level attributes cannot be tracked explicitly in large-scale, long-running Earth system models. Instead, aerosol processes are simplified, with different modeling groups applying different approximations. These approximations introduce structural errors in model predictions that have not been well quantified. As a path forward, we introduce the Aerosol Model Benchmarking Repository and Standards (AMBRS) Initiative, a community-based platform for building, running, and evaluating aerosol chemical and microphysical schemes at the box-model level. By isolating the errors introduced by simplifications in current and proposed aerosol models, AMBRS enables transparent, reproducible comparison of process-level performance. The effort aims to support community standards for model benchmarking and promote iterative development of aerosol modules, ultimately helping reduce structural uncertainty in simulations of aerosol effects.
#6
Utilizes remote sensing to quantify methane emissions from oil and gas, key for climate impact assessment.
Abstract
Abstract The Permian Basin is one of the largest methane sources in the United States, yet quantifying emissions at scales needed to attribute them to specific infrastructure remains challenging. Here we present a methane flux inversion from 2018 through 2025 using TROPOMI retrievals and a machine learning surrogate for atmospheric transport (FootNet). We estimate emissions at 1 × 1 spatial resolution and weekly time resolution over the Permian Basin, jointly solving for background concentrations. We find total Permian Basin emissions of 2.6 0.5 Tg/yr (structural uncertainty range: 1.7–3.2 Tg/yr), a factor of 2.4 higher than the EPA inventory, with strong spatial correspondence to gas wells, pipelines, storage tanks, compressor stations, and processing facilities. Permian sub‐basins have distinct drivers of methane emissions: gas wells in the Delaware, midstream infrastructure in the Midland. Despite a 2–3× increase in production, total emissions remained stable, driving a decline in methane intensity from 5% to 2%.
Abstract The Permian Basin is one of the largest methane sources in the United States, yet quantifying emissions at scales needed to attribute them to specific infrastructure remains challenging. Here we present a methane flux inversion from 2018 through 2025 using TROPOMI retrievals and a machine learning surrogate for atmospheric transport (FootNet). We estimate emissions at 1 × 1 spatial resolution and weekly time resolution over the Permian Basin, jointly solving for background concentrations. We find total Permian Basin emissions of 2.6 0.5 Tg/yr (structural uncertainty range: 1.7–3.2 Tg/yr), a factor of 2.4 higher than the EPA inventory, with strong spatial correspondence to gas wells, pipelines, storage tanks, compressor stations, and processing facilities. Permian sub‐basins have distinct drivers of methane emissions: gas wells in the Delaware, midstream infrastructure in the Midland. Despite a 2–3× increase in production, total emissions remained stable, driving a decline in methane intensity from 5% to 2%.
#7
Provides a high-resolution global groundwater dataset, supporting hydrology and water resource studies.
Abstract
Abstract. The global depletion of groundwater poses a significant challenge to water security. However, the coarse spatial resolution of GRACE satellite observations obscures fine-scale hydrological dynamics and limits the separation of localized anthropogenic extraction signals from large-scale climatic forcing. To address this limitation, we present a high-resolution (1 km), continuous monthly dataset of global groundwater storage anomalies (GWSA) covering the period from 2002 to 2020. The dataset is generated using a production framework that integrates Singular Spectrum Analysis (SSA) for temporal gap filling and an aquifer-stratified machine learning approach driven by 19 high-resolution hydroclimatic and geophysical predictors. To ensure robustness and spatial consistency, multiple predictive approaches were evaluated as part of a quality control procedure, and the most stable model was selected for final production based on multi-scale validation. Cross-scale evaluation shows that the downscaled dataset preserves the large-scale spatial patterns of the original GRACE observations with high agreement (R2 = 0.972, RMSE = 2.10 cm). Independent validation using 1,518 in situ monitoring wells, combined with a geographically stratified specific yield matrix for dimensional conversion, further demonstrates the ability of the dataset to capture long-term groundwater variability across diverse hydrogeological conditions (R2 = 0.44, p < 0.01). The resulting 1 km dataset provides enhanced spatial detail and enables the identification of sharp nonlinear boundaries associated with intensive human pumping, as well as spatial polarization patterns in groundwater storage changes. This dataset offers a reliable, observation-constrained resource for water resource assessment, hydrological modeling, and studies of coupled climate and human influences on groundwater systems.
Abstract. The global depletion of groundwater poses a significant challenge to water security. However, the coarse spatial resolution of GRACE satellite observations obscures fine-scale hydrological dynamics and limits the separation of localized anthropogenic extraction signals from large-scale climatic forcing. To address this limitation, we present a high-resolution (1 km), continuous monthly dataset of global groundwater storage anomalies (GWSA) covering the period from 2002 to 2020. The dataset is generated using a production framework that integrates Singular Spectrum Analysis (SSA) for temporal gap filling and an aquifer-stratified machine learning approach driven by 19 high-resolution hydroclimatic and geophysical predictors. To ensure robustness and spatial consistency, multiple predictive approaches were evaluated as part of a quality control procedure, and the most stable model was selected for final production based on multi-scale validation. Cross-scale evaluation shows that the downscaled dataset preserves the large-scale spatial patterns of the original GRACE observations with high agreement (R2 = 0.972, RMSE = 2.10 cm). Independent validation using 1,518 in situ monitoring wells, combined with a geographically stratified specific yield matrix for dimensional conversion, further demonstrates the ability of the dataset to capture long-term groundwater variability across diverse hydrogeological conditions (R2 = 0.44, p < 0.01). The resulting 1 km dataset provides enhanced spatial detail and enables the identification of sharp nonlinear boundaries associated with intensive human pumping, as well as spatial polarization patterns in groundwater storage changes. This dataset offers a reliable, observation-constrained resource for water resource assessment, hydrological modeling, and studies of coupled climate and human influences on groundwater systems.
#8
Reveals late Holocene Laurentide Ice Sheet reexpansion, offering new insights into glaciation and climate change.
Abstract
During interglaciations, the Laurentide Ice Sheet (LIS), which is the largest of the Pleistocene ice sheets, retreats to Arctic Canada, where it achieves a minimum configuration. The reexpansion of the LIS following this minimum marks the start of the next glaciation. Here, we show that the last remaining remnant of the most recent LIS, Barnes Ice Cap (Baffin Island, Canada), achieved its minimum interglacial dimensions after ~3.6 thousand years ago, before reexpanding ~2000 years ago, marking the onset of the next glaciation. The new LIS reached maximum dimensions in ~1960 CE, after which anthropogenic summer warming resulted in a negative mass balance, ice margin recession, and the termination of the next glaciation. The demise of the next glaciation, despite the ongoing decline in summer insolation, illustrates the influence of anomalous greenhouse gas forcing on climate.
During interglaciations, the Laurentide Ice Sheet (LIS), which is the largest of the Pleistocene ice sheets, retreats to Arctic Canada, where it achieves a minimum configuration. The reexpansion of the LIS following this minimum marks the start of the next glaciation. Here, we show that the last remaining remnant of the most recent LIS, Barnes Ice Cap (Baffin Island, Canada), achieved its minimum interglacial dimensions after ~3.6 thousand years ago, before reexpanding ~2000 years ago, marking the onset of the next glaciation. The new LIS reached maximum dimensions in ~1960 CE, after which anthropogenic summer warming resulted in a negative mass balance, ice margin recession, and the termination of the next glaciation. The demise of the next glaciation, despite the ongoing decline in summer insolation, illustrates the influence of anomalous greenhouse gas forcing on climate.
#9
Explores ocean carbon sequestration mechanisms during glacial climates, relevant for past and future carbon cycles.
Abstract
Glacial-interglacial transitions punctuate Earth's Quaternary climate, accompanied by substantial atmospheric carbon dioxide and ocean carbon storage changes. The cause of atmospheric carbon drawdown during glaciations is uncertain, involving many mechanisms whose interactions are poorly understood. Here, we show that the combined expansion of Antarctic sea ice and associated changes in the meridional overturning efficiently sequester carbon in the ocean abyss, leading to a decline in atmospheric carbon dioxide. In contrast, changes in sea ice and circulation individually result in ocean carbon dioxide outgassing. When sea ice expansion and circulation changes act together, they enhance biological carbon storage by reducing ocean-atmosphere fluxes in the Southern Ocean, without invoking external drivers of the biological pump. This reduction inhibits the outgassing of upwelled remineralized carbon and promotes its resubduction into the abyss. Together, these results reveal a reinforcing feedback linking atmospheric cooling, Antarctic sea ice expansion, and ocean circulation changes to enhanced biological carbon storage and atmospheric carbon dioxide drawdown.
Glacial-interglacial transitions punctuate Earth's Quaternary climate, accompanied by substantial atmospheric carbon dioxide and ocean carbon storage changes. The cause of atmospheric carbon drawdown during glaciations is uncertain, involving many mechanisms whose interactions are poorly understood. Here, we show that the combined expansion of Antarctic sea ice and associated changes in the meridional overturning efficiently sequester carbon in the ocean abyss, leading to a decline in atmospheric carbon dioxide. In contrast, changes in sea ice and circulation individually result in ocean carbon dioxide outgassing. When sea ice expansion and circulation changes act together, they enhance biological carbon storage by reducing ocean-atmosphere fluxes in the Southern Ocean, without invoking external drivers of the biological pump. This reduction inhibits the outgassing of upwelled remineralized carbon and promotes its resubduction into the abyss. Together, these results reveal a reinforcing feedback linking atmospheric cooling, Antarctic sea ice expansion, and ocean circulation changes to enhanced biological carbon storage and atmospheric carbon dioxide drawdown.
#10
Connects marine heatwaves and atmospheric jets to enhanced coastal precipitation, highlighting extreme weather interactions.
Abstract
Heavy coastal precipitation is traditionally attributed to atmospheric dynamics, but the influence of co-occurring oceanic thermal extremes remains largely unexplored. Using satellite observations and reanalysis data from 1998 to 2023 over the South China Sea (SCS), we investigate the synergistic impacts of marine heatwaves (MHWs) and atmospheric boundary layer jets (BLJs). We reveal a compound effect: when MHWs and BLJs co-occur, regional coastal precipitation increases by 51.9% relative to climatology, more than doubling the 21.0% increase during pure BLJ events. Moist static energy (MSE) diagnostics demonstrate that MHWs act as a thermodynamic catalyst, enhancing surface evaporation by 20%-30% and boosting low-level atmospheric latent energy. The concurrent BLJ functions as a dynamic conveyor, rapidly transporting this excess moisture landward to fuel heavy coastal rainfall. These results imply that assessing coastal precipitation from atmospheric dynamics alone may largely underestimate the rainfall response when marine heatwaves are present, a coupling set to strengthen as the ocean warms.
Heavy coastal precipitation is traditionally attributed to atmospheric dynamics, but the influence of co-occurring oceanic thermal extremes remains largely unexplored. Using satellite observations and reanalysis data from 1998 to 2023 over the South China Sea (SCS), we investigate the synergistic impacts of marine heatwaves (MHWs) and atmospheric boundary layer jets (BLJs). We reveal a compound effect: when MHWs and BLJs co-occur, regional coastal precipitation increases by 51.9% relative to climatology, more than doubling the 21.0% increase during pure BLJ events. Moist static energy (MSE) diagnostics demonstrate that MHWs act as a thermodynamic catalyst, enhancing surface evaporation by 20%-30% and boosting low-level atmospheric latent energy. The concurrent BLJ functions as a dynamic conveyor, rapidly transporting this excess moisture landward to fuel heavy coastal rainfall. These results imply that assessing coastal precipitation from atmospheric dynamics alone may largely underestimate the rainfall response when marine heatwaves are present, a coupling set to strengthen as the ocean warms.