#1
Exceptional relevance and novelty on record-setting precipitation and climate extremes.
Abstract
Precipitation extremes have increased globally in recent decades, consistent with a warming climate. We find that 2024 experienced a record-high frequency of heavy (≥99th percentile; F P≥99 ) precipitation events globally based on MSWEP (Multi-Source Weighted-Ensemble Precipitation) observations from 1979 to 2025. This extends a long-term trend of ∼0.03%/decade in F P≥99 , with similar but more uncertain increases for the 99.9th percentile. Unlike prior high F P≥99 years, 2024’s record was more globally distributed and emerged from a distinct seasonal evolution. El Niño and rare concurrent multi-basin tropical ocean warmth drove elevated extremes early in the year, before lagged global tropospheric responses and long-term warming sustained and amplified extremes through boreal summer. Climate models indicate that years with 2024-level F P≥99 typically require either stronger El Niño forcing or concurrent multi-basin sea surface temperature (SST) anomalies. Diagnostic projections within constrained ensemble members suggest that ∼40% of the 2024 anomaly aligns with global mean temperatures and ∼20% with El Niño variability, with some of the remainder linked to basin-scale SST structure.
Precipitation extremes have increased globally in recent decades, consistent with a warming climate. We find that 2024 experienced a record-high frequency of heavy (≥99th percentile; F P≥99 ) precipitation events globally based on MSWEP (Multi-Source Weighted-Ensemble Precipitation) observations from 1979 to 2025. This extends a long-term trend of ∼0.03%/decade in F P≥99 , with similar but more uncertain increases for the 99.9th percentile. Unlike prior high F P≥99 years, 2024’s record was more globally distributed and emerged from a distinct seasonal evolution. El Niño and rare concurrent multi-basin tropical ocean warmth drove elevated extremes early in the year, before lagged global tropospheric responses and long-term warming sustained and amplified extremes through boreal summer. Climate models indicate that years with 2024-level F P≥99 typically require either stronger El Niño forcing or concurrent multi-basin sea surface temperature (SST) anomalies. Diagnostic projections within constrained ensemble members suggest that ∼40% of the 2024 anomaly aligns with global mean temperatures and ∼20% with El Niño variability, with some of the remainder linked to basin-scale SST structure.
#2
Timely analysis of global warming acceleration and Paris Agreement implications.
Abstract
Timely and accurate projections of when global warming will exceed the Paris Agreement’s 1.5 °C threshold are essential for effective climate policy. Existing estimates place this exceedance in the early 2030 s but carry substantial uncertainty. Here we analyze the latest observational temperature records through 2025 and identify a statistically significant doubling of the global warming rate during 2011–2025 (0.40 °C per decade) compared to 1971–2025 (0.20 °C per decade). Extrapolating this accelerated rate projects a crossing of the 1.5 °C threshold by 2028, nearly a decade earlier than previous estimates. Regional analyses attribute this acceleration primarily to intensified warming over East Asia–the North Pacific, mid-latitude Eurasia, and the South Atlantic–Antarctic regions. The ensemble mean of 26 climate models fails to capture the observed doubling of the warming rate, simulating a rate of only 0.29 °C per decade. These findings suggest a narrower window remains to limit warming to 1.5 °C, underscoring the urgency for rapid and substantial reductions in greenhouse gas emissions.
Timely and accurate projections of when global warming will exceed the Paris Agreement’s 1.5 °C threshold are essential for effective climate policy. Existing estimates place this exceedance in the early 2030 s but carry substantial uncertainty. Here we analyze the latest observational temperature records through 2025 and identify a statistically significant doubling of the global warming rate during 2011–2025 (0.40 °C per decade) compared to 1971–2025 (0.20 °C per decade). Extrapolating this accelerated rate projects a crossing of the 1.5 °C threshold by 2028, nearly a decade earlier than previous estimates. Regional analyses attribute this acceleration primarily to intensified warming over East Asia–the North Pacific, mid-latitude Eurasia, and the South Atlantic–Antarctic regions. The ensemble mean of 26 climate models fails to capture the observed doubling of the warming rate, simulating a rate of only 0.29 °C per decade. These findings suggest a narrower window remains to limit warming to 1.5 °C, underscoring the urgency for rapid and substantial reductions in greenhouse gas emissions.
#3
Innovative AI-driven approach for metropolitan flood forecasting and hydrodynamic modeling.
Abstract
AI-driven flood digital twins demand fast hydrodynamic surrogates for ensemble forecasting and observation assimilation. Yet even GPU-accelerated two-dimensional shallow water equation (SWE) solvers still require minutes per -hour run on a -million-active-cell metropolitan basin (the Des Plaines River basin at resolution), making such workloads prohibitive at native resolution. We present the Conditional Latent Dynamics Network (CLDNet): a low-dimensional latent neural ODE driven by rainfall, paired with a coordinate-based decoder conditioned on static terrain (elevation, slope, Manning roughness) that reconstructs depth and discharge at arbitrary query points. Pointwise decoding decouples memory from grid size and handles irregular watersheds natively, enabling metropolitan-scale training on a single compute node and direct queries at exact gauge coordinates without raster snapping. We evaluate CLDNet on a synthetic -cell Texas benchmark and on a new Des Plaines case study of real-rainfall Stage IV storms whose reference simulator we validate against United States Geological Survey (USGS) gauges at the April 2013 flood-of-record (Nash–Sutcliffe efficiency on mean-recentered water-surface elevation). CLDNet reduces the relative root-mean-squared error of an unconditional baseline by , outperforms regular-grid VAE–ConvLSTM and FNO baselines on the Texas benchmark (both presuppose a Cartesian grid and do not apply to the irregular Des Plaines watershed), reaches a critical success index of at the inundation threshold, and produces a full -hour basin-wide forecast in seconds—a speedup. Finally, we demonstrate that CLDNet’s compact latent state supports online observation coupling: using the Latent Dynamics Ensemble Score Filter (LD-EnSF), we assimilate water-depth observations at gauge locations into the surrogate’s latent dynamics and recover a mean Nash–Sutcliffe efficiency of against USGS observations at six held-out gauges, from for the rainfall-perturbed open-loop forecast, realizing the observation-update step required of an AI-driven flood digital twin.
AI-driven flood digital twins demand fast hydrodynamic surrogates for ensemble forecasting and observation assimilation. Yet even GPU-accelerated two-dimensional shallow water equation (SWE) solvers still require minutes per -hour run on a -million-active-cell metropolitan basin (the Des Plaines River basin at resolution), making such workloads prohibitive at native resolution. We present the Conditional Latent Dynamics Network (CLDNet): a low-dimensional latent neural ODE driven by rainfall, paired with a coordinate-based decoder conditioned on static terrain (elevation, slope, Manning roughness) that reconstructs depth and discharge at arbitrary query points. Pointwise decoding decouples memory from grid size and handles irregular watersheds natively, enabling metropolitan-scale training on a single compute node and direct queries at exact gauge coordinates without raster snapping. We evaluate CLDNet on a synthetic -cell Texas benchmark and on a new Des Plaines case study of real-rainfall Stage IV storms whose reference simulator we validate against United States Geological Survey (USGS) gauges at the April 2013 flood-of-record (Nash–Sutcliffe efficiency on mean-recentered water-surface elevation). CLDNet reduces the relative root-mean-squared error of an unconditional baseline by , outperforms regular-grid VAE–ConvLSTM and FNO baselines on the Texas benchmark (both presuppose a Cartesian grid and do not apply to the irregular Des Plaines watershed), reaches a critical success index of at the inundation threshold, and produces a full -hour basin-wide forecast in seconds—a speedup. Finally, we demonstrate that CLDNet’s compact latent state supports online observation coupling: using the Latent Dynamics Ensemble Score Filter (LD-EnSF), we assimilate water-depth observations at gauge locations into the surrogate’s latent dynamics and recover a mean Nash–Sutcliffe efficiency of against USGS observations at six held-out gauges, from for the rainfall-perturbed open-loop forecast, realizing the observation-update step required of an AI-driven flood digital twin.
#4
Key insights into nitrous acid evolution and its impact on urban air quality and secondary aerosols.
Abstract
Nitrous acid (HONO) is a crucial urban source of HOx radicals, promoting surface ozone (O3) and secondary fine particle (PM2.5) formation, yet its long-term evolution and air-quality impacts remain poorly constrained. Here, using a deep neural network established on worldwide field campaigns, we reconstruct global surface HONO distributions. Typical megacities exhibited contrasting trends in 2005–2021, marked by a sharp decline in Beijing (−47.4% in 2013–2021), an increase in Delhi (+10.0% in 2005–2018), and slight decreases in London, New York and Los Angeles. These HONO changes are identified as major driver of O3 and particulate nitrate variations, outweighing contributions of conventional NOx and volatile organic compounds abatement in Beijing and Delhi, while showing comparable impacts in the remaining cities. A fused data-driven framework further uncovers regime-dependent HONO reduction strategies, prioritizing NOx control in summer and co-reduction of NOx, PM2.5 and direct combustion emissions in winter, to accelerate O3 and PM2.5 joint mitigation. Deep learning reveals a sharp decline of HONO in China, modest drops in EU and US but an increase in India. HONO changes drive O3 and PM2.5 variations, outweighing traditional NOx and VOC control, calling for HONO-targeted air quality management.
Nitrous acid (HONO) is a crucial urban source of HOx radicals, promoting surface ozone (O3) and secondary fine particle (PM2.5) formation, yet its long-term evolution and air-quality impacts remain poorly constrained. Here, using a deep neural network established on worldwide field campaigns, we reconstruct global surface HONO distributions. Typical megacities exhibited contrasting trends in 2005–2021, marked by a sharp decline in Beijing (−47.4% in 2013–2021), an increase in Delhi (+10.0% in 2005–2018), and slight decreases in London, New York and Los Angeles. These HONO changes are identified as major driver of O3 and particulate nitrate variations, outweighing contributions of conventional NOx and volatile organic compounds abatement in Beijing and Delhi, while showing comparable impacts in the remaining cities. A fused data-driven framework further uncovers regime-dependent HONO reduction strategies, prioritizing NOx control in summer and co-reduction of NOx, PM2.5 and direct combustion emissions in winter, to accelerate O3 and PM2.5 joint mitigation. Deep learning reveals a sharp decline of HONO in China, modest drops in EU and US but an increase in India. HONO changes drive O3 and PM2.5 variations, outweighing traditional NOx and VOC control, calling for HONO-targeted air quality management.
#5
Novel 3D reconstruction of glacier freshwater in East Antarctica with climate change implications.
Abstract
The melting of the Antarctic Ice Sheet contributes to global sea-level rise and widespread ocean freshening. A major driver of this process is the intrusion of warm deep water onto the continental shelf, which enhances basal melting of ice shelves and weakens their buttressing of inland ice. However, the three-dimensional distribution of glacier-derived freshwater remains poorly constrained in East Antarctica. Here we present a hybrid multiple linear regression-neural network (MLR-NN) framework that reconstructs glacier-derived freshwater from dissolved inorganic carbon (DIC) residuals without prescribing fixed freshwater end-member compositions. Applying this framework to hydrographic observations around the Totten and Moscow University Ice Shelves, we estimate freshwater discharge rates of 70 ± 7 and 53 ± 9 Gt yr⁻¹, respectively. The reconstructed freshwater signal extends across the continental shelf to approximately 64°S. We also identify differences in freshwater discharge between observations collected during the 2010s and the 2020 s, although sparse temporal sampling prevents these differences from being interpreted as a robust long-term trend. Our results provide a three-dimensional observational reconstruction of glacier-derived freshwater in this sector of East Antarctica and demonstrate the potential of an end-member-independent hydrographic parameterization for investigating glacier-ocean interactions in data-limited polar environments.
The melting of the Antarctic Ice Sheet contributes to global sea-level rise and widespread ocean freshening. A major driver of this process is the intrusion of warm deep water onto the continental shelf, which enhances basal melting of ice shelves and weakens their buttressing of inland ice. However, the three-dimensional distribution of glacier-derived freshwater remains poorly constrained in East Antarctica. Here we present a hybrid multiple linear regression-neural network (MLR-NN) framework that reconstructs glacier-derived freshwater from dissolved inorganic carbon (DIC) residuals without prescribing fixed freshwater end-member compositions. Applying this framework to hydrographic observations around the Totten and Moscow University Ice Shelves, we estimate freshwater discharge rates of 70 ± 7 and 53 ± 9 Gt yr⁻¹, respectively. The reconstructed freshwater signal extends across the continental shelf to approximately 64°S. We also identify differences in freshwater discharge between observations collected during the 2010s and the 2020 s, although sparse temporal sampling prevents these differences from being interpreted as a robust long-term trend. Our results provide a three-dimensional observational reconstruction of glacier-derived freshwater in this sector of East Antarctica and demonstrate the potential of an end-member-independent hydrographic parameterization for investigating glacier-ocean interactions in data-limited polar environments.
#6
Highlights the amplification of food insecurity by compound climate extremes.
Abstract
The increasing co-occurrence of climatic hazards poses growing risks to crop production, yet their implications for rice supply and self-sufficiency remain insufficiently quantified. Here, we combine a vine-copula framework with calibrated ORYZA (v3) model to characterize nonlinear dependencies among climate indicators, estimate the probability of compound climatic events, and assess their impacts on rice yield potential, yield gaps, and national self-sufficiency across major rice-growing regions of China. We show that (1) intensification of the global water cycle increases the frequency of extreme heat from 25 to 50% in the late 20th century to over 90% by 2050, while the risk of compound events involving extreme heat, rainfall and/or solar radiation is projected to change by approximately 3–50% in single-rice systems and 1–60% in double-rice systems relative to 1981–2017; (2) increasing probability of compound climatic extreme events apparently affect potential production for single rice (−2.8 to 0.2 t ha −1 ) and double-rice growing systems (−1.3 to 0.8 t ha −1 ), with downside risk increasing progressively over time; (3) the negative impacts on rice yield potential outweigh the benefits (0.5 to 1.5 t ha −1 ) gained from increased atmospheric CO 2 , leading to reduction in yield gaps. Assuming production remains near 70% of yield potential, projected national rice supply in 2030–2050 (∼216 Mt. with CO 2 effects; ∼195 Mt. without) may fall below projected demand (∼220 Mt), implying increasing pressure on China's rice self-sufficiency. Our results highlight the need for targeted adaptation and resilience investment, particularly in single-cropping regions, under intensifying compound climate risk.
The increasing co-occurrence of climatic hazards poses growing risks to crop production, yet their implications for rice supply and self-sufficiency remain insufficiently quantified. Here, we combine a vine-copula framework with calibrated ORYZA (v3) model to characterize nonlinear dependencies among climate indicators, estimate the probability of compound climatic events, and assess their impacts on rice yield potential, yield gaps, and national self-sufficiency across major rice-growing regions of China. We show that (1) intensification of the global water cycle increases the frequency of extreme heat from 25 to 50% in the late 20th century to over 90% by 2050, while the risk of compound events involving extreme heat, rainfall and/or solar radiation is projected to change by approximately 3–50% in single-rice systems and 1–60% in double-rice systems relative to 1981–2017; (2) increasing probability of compound climatic extreme events apparently affect potential production for single rice (−2.8 to 0.2 t ha −1 ) and double-rice growing systems (−1.3 to 0.8 t ha −1 ), with downside risk increasing progressively over time; (3) the negative impacts on rice yield potential outweigh the benefits (0.5 to 1.5 t ha −1 ) gained from increased atmospheric CO 2 , leading to reduction in yield gaps. Assuming production remains near 70% of yield potential, projected national rice supply in 2030–2050 (∼216 Mt. with CO 2 effects; ∼195 Mt. without) may fall below projected demand (∼220 Mt), implying increasing pressure on China's rice self-sufficiency. Our results highlight the need for targeted adaptation and resilience investment, particularly in single-cropping regions, under intensifying compound climate risk.
#7
Broad review of marine ecosystem ecology and biodiversity under climate change.
Abstract
Research in marine ecosystem ecology increasingly integrates biodiversity status and trends, ecosystem functioning, and human pressures. Understanding how marine biodiversity supports ecosystem functioning remains central to the field, while growing concern over ecosystem condition has shifted attention towards relationships between human pressures and ecological responses. These established questions are increasingly complemented by the need to anticipate climate-driven ecosystem transformations and translate ecological knowledge into effective conservation and management. Resilience – the capacity of ecosystems to persist and maintain their state and functions through resistance to and recovery from disturbance – has therefore become a central scientific priority under intensifying cumulative pressures. Major challenges for the coming decade include ecological non-stationarity, rapidly expanding data volumes, erosion of traditional expertise, uneven uptake of technological advances, and restoration under changing environmental conditions. Meeting these challenges requires marine ecosystem ecology to keep pace with global change while improving the transfer of knowledge into policy and management. Here, we identify these priorities and outline pathways towards more predictive, integrated, and decision-relevant marine ecosystem science.
Research in marine ecosystem ecology increasingly integrates biodiversity status and trends, ecosystem functioning, and human pressures. Understanding how marine biodiversity supports ecosystem functioning remains central to the field, while growing concern over ecosystem condition has shifted attention towards relationships between human pressures and ecological responses. These established questions are increasingly complemented by the need to anticipate climate-driven ecosystem transformations and translate ecological knowledge into effective conservation and management. Resilience – the capacity of ecosystems to persist and maintain their state and functions through resistance to and recovery from disturbance – has therefore become a central scientific priority under intensifying cumulative pressures. Major challenges for the coming decade include ecological non-stationarity, rapidly expanding data volumes, erosion of traditional expertise, uneven uptake of technological advances, and restoration under changing environmental conditions. Meeting these challenges requires marine ecosystem ecology to keep pace with global change while improving the transfer of knowledge into policy and management. Here, we identify these priorities and outline pathways towards more predictive, integrated, and decision-relevant marine ecosystem science.
#8
Advances physics-informed machine learning for streamflow prediction in river networks.
Abstract
The basin in high-altitude mountainous areas is constrained by sparse station networks, complex river networks, and significant non-stationary hydrological processes, making it difficult for traditional time series models to simultaneously consider spatial connectivity and physical consistency. To address this issue, we propose Physics-Informed Hydrological Graph Neural Network (PI-HydroGNN) for distributed daily streamflow prediction in directed river networks. It combines temporal feature extraction, upstream connectivity and topological inflow aggregation, graph-based topology message passing, and physical constraints in a unified framework to represent temporal dynamics, river-network connectivity, and local mass conservation. The model was evaluated through comparisons with independent station-wise LSTM, regional multi-output LSTM, GCN and SWAT, together with ablation experiments, multi-step forecasting, analysis of Quality Balance Violations and ungauged basin extrapolation. The results showed that PI-HydroGNN outperformed the benchmark models in the training, validation, and testing periods, with an average NSE of 0.84 during the testing period and a RMSE reduction of about 16% compared to station-wise LSTM. The ablation experiment results showed that all the modules we considered made significant contributions to the predictive performance and exhibited significant spatial heterogeneity. At Yangbajing, the model still maintains good simulation ability, with an NSE of 0.79 and an R 2 of 0.80, providing preliminary evidence of its spatial extrapolation capability. Overall, this study demonstrates that combining prior knowledge of hydrological physics with graph learning offers a promising approach for streamflow prediction in complex river networks, and suggests potential for application in ungauged basins that should be further validated under broader hydro-climatic conditions.
The basin in high-altitude mountainous areas is constrained by sparse station networks, complex river networks, and significant non-stationary hydrological processes, making it difficult for traditional time series models to simultaneously consider spatial connectivity and physical consistency. To address this issue, we propose Physics-Informed Hydrological Graph Neural Network (PI-HydroGNN) for distributed daily streamflow prediction in directed river networks. It combines temporal feature extraction, upstream connectivity and topological inflow aggregation, graph-based topology message passing, and physical constraints in a unified framework to represent temporal dynamics, river-network connectivity, and local mass conservation. The model was evaluated through comparisons with independent station-wise LSTM, regional multi-output LSTM, GCN and SWAT, together with ablation experiments, multi-step forecasting, analysis of Quality Balance Violations and ungauged basin extrapolation. The results showed that PI-HydroGNN outperformed the benchmark models in the training, validation, and testing periods, with an average NSE of 0.84 during the testing period and a RMSE reduction of about 16% compared to station-wise LSTM. The ablation experiment results showed that all the modules we considered made significant contributions to the predictive performance and exhibited significant spatial heterogeneity. At Yangbajing, the model still maintains good simulation ability, with an NSE of 0.79 and an R 2 of 0.80, providing preliminary evidence of its spatial extrapolation capability. Overall, this study demonstrates that combining prior knowledge of hydrological physics with graph learning offers a promising approach for streamflow prediction in complex river networks, and suggests potential for application in ungauged basins that should be further validated under broader hydro-climatic conditions.
#9
Paleoclimate study of simultaneous megapluvials in the Americas, enhancing ENSO understanding.
Abstract
Abstract Earth's climate has experienced prolonged hydroclimatic extreme events that can manifest decadal or longer anomalies. Such events can shape ecological systems and human societies. Decadal‐scale droughts, megadroughts, occurred in southwestern North and South America (North American Southwest, NASW; South American Southwest, SASW) over the past millennium, including simultaneous events. Similarly, these regions experienced prolonged wet periods, megapluvials, including well‐documented episodes over the 20th century. Using a paleoclimate data assimilation product, we identify 18 megapluvials in each region (12 overlapping), 13 NASW and 15 SASW megadroughts (9 overlapping). Both phenomena show similar duration and severity. We find that megapluvials in both regions are driven by a reduction in drying La Niña‐like states, not solely by an increase in wetting El Niño‐like states; while both changes are statistically significant, the decrease in La Niña‐like conditions is stronger than the increase in El Niño‐like conditions. Megadroughts exhibit an analogous asymmetric mechanism: an increase in La Niña‐like states and a stronger reduction in El Niño‐like states. We also find volcanic forcing influences these events through the El Niño‐Southern Oscillation (ENSO): large eruptions reduce the frequency of La Niña‐like states, causing overall wetting. This mechanism is most clearly seen in the SASW where ENSO teleconnections are stronger in the paleoclimate reconstruction. These findings demonstrate that megapluvials exhibit interhemispheric synchronization, similar to megadroughts, and are similarly influenced by Pacific variability on decadal timescales. Our results highlight the need for better understanding and representation of ENSO's response to external forcing to improve projections of decadal hydroclimate variability in the NASW and SASW.
Abstract Earth's climate has experienced prolonged hydroclimatic extreme events that can manifest decadal or longer anomalies. Such events can shape ecological systems and human societies. Decadal‐scale droughts, megadroughts, occurred in southwestern North and South America (North American Southwest, NASW; South American Southwest, SASW) over the past millennium, including simultaneous events. Similarly, these regions experienced prolonged wet periods, megapluvials, including well‐documented episodes over the 20th century. Using a paleoclimate data assimilation product, we identify 18 megapluvials in each region (12 overlapping), 13 NASW and 15 SASW megadroughts (9 overlapping). Both phenomena show similar duration and severity. We find that megapluvials in both regions are driven by a reduction in drying La Niña‐like states, not solely by an increase in wetting El Niño‐like states; while both changes are statistically significant, the decrease in La Niña‐like conditions is stronger than the increase in El Niño‐like conditions. Megadroughts exhibit an analogous asymmetric mechanism: an increase in La Niña‐like states and a stronger reduction in El Niño‐like states. We also find volcanic forcing influences these events through the El Niño‐Southern Oscillation (ENSO): large eruptions reduce the frequency of La Niña‐like states, causing overall wetting. This mechanism is most clearly seen in the SASW where ENSO teleconnections are stronger in the paleoclimate reconstruction. These findings demonstrate that megapluvials exhibit interhemispheric synchronization, similar to megadroughts, and are similarly influenced by Pacific variability on decadal timescales. Our results highlight the need for better understanding and representation of ENSO's response to external forcing to improve projections of decadal hydroclimate variability in the NASW and SASW.
#10
Comprehensive review of Antarctic sea ice's role in Earth system interactions over 130,000 years.
Abstract
Abstract. Antarctic sea-ice cover reached historically low levels in 2023, consistent with the simulated decrease in sea-ice extent in response to anthropogenic warming. Antarctic sea ice is closely linked to multiple components of the Earth system, thus its demise could precipitate widespread, cascading changes across the cryosphere, atmosphere, and ocean. However, the nature and strength of these interconnections are poorly understood, and they are often inadequately represented in models. In this review paper we use modern observations, models and paleoclimate archives covering the last glacial cycle to gain insights into how reductions in sea ice may affect other components of the Earth system. We review how Antarctic sea ice interacts with ocean and atmosphere circulation, ice sheets and ice shelves, marine productivity, and the carbon cycle over the last glacial cycle, for which we have the most robust sea-ice reconstructions. The review finds strong evidence from theory and models for impacts of Antarctic sea ice on the Earth system. Paleo-proxy reconstructions provide examples where changes in sea ice co-occur with changes in the carbon cycle, marine productivity, and ocean circulation. However, challenges remain in isolating the impact of sea ice in a highly interconnected system.
Abstract. Antarctic sea-ice cover reached historically low levels in 2023, consistent with the simulated decrease in sea-ice extent in response to anthropogenic warming. Antarctic sea ice is closely linked to multiple components of the Earth system, thus its demise could precipitate widespread, cascading changes across the cryosphere, atmosphere, and ocean. However, the nature and strength of these interconnections are poorly understood, and they are often inadequately represented in models. In this review paper we use modern observations, models and paleoclimate archives covering the last glacial cycle to gain insights into how reductions in sea ice may affect other components of the Earth system. We review how Antarctic sea ice interacts with ocean and atmosphere circulation, ice sheets and ice shelves, marine productivity, and the carbon cycle over the last glacial cycle, for which we have the most robust sea-ice reconstructions. The review finds strong evidence from theory and models for impacts of Antarctic sea ice on the Earth system. Paleo-proxy reconstructions provide examples where changes in sea ice co-occur with changes in the carbon cycle, marine productivity, and ocean circulation. However, challenges remain in isolating the impact of sea ice in a highly interconnected system.