#1
Top score, advances subseasonal forecasting with multi-sphere deep learning.
Abstract
Accurate subseasonal-to-seasonal (S2S) prediction is critical for disaster mitigation under accelerating climate change, yet remains challenging due to complex multi-sphere interactions and intrinsic atmospheric uncertainty. Here we present TianXing-S2S, a multi-sphere coupled probabilistic deep learning model for global S2S ensemble forecasting. TianXing-S2S first encodes diverse multi-sphere predictors into a compact latent space, then employs a diffusion model to generate daily ensemble forecasts. A novel coupling module based on optimal transport (OT) is incorporated in the denoiser to optimize the interactions between atmospheric and multi-sphere boundary conditions. Across key atmospheric variables, TianXing-S2S outperforms both the European Centre for Medium-Range Weather Forecasts (ECMWF) S2S system and FuXi-S2S in 45-day daily-mean ensemble forecasts at 1.5° × 1.5° resolution. Our model improves the prediction of extreme events including heat waves and anomalous precipitation, with multi-sphere coupling identified as a key contributor. Furthermore, we demonstrate that TianXing-S2S can generate stable rollout forecasts up to 180 days, suggesting its potential as a useful framework for probabilistic S2S prediction under a changing climate.
Accurate subseasonal-to-seasonal (S2S) prediction is critical for disaster mitigation under accelerating climate change, yet remains challenging due to complex multi-sphere interactions and intrinsic atmospheric uncertainty. Here we present TianXing-S2S, a multi-sphere coupled probabilistic deep learning model for global S2S ensemble forecasting. TianXing-S2S first encodes diverse multi-sphere predictors into a compact latent space, then employs a diffusion model to generate daily ensemble forecasts. A novel coupling module based on optimal transport (OT) is incorporated in the denoiser to optimize the interactions between atmospheric and multi-sphere boundary conditions. Across key atmospheric variables, TianXing-S2S outperforms both the European Centre for Medium-Range Weather Forecasts (ECMWF) S2S system and FuXi-S2S in 45-day daily-mean ensemble forecasts at 1.5° × 1.5° resolution. Our model improves the prediction of extreme events including heat waves and anomalous precipitation, with multi-sphere coupling identified as a key contributor. Furthermore, we demonstrate that TianXing-S2S can generate stable rollout forecasts up to 180 days, suggesting its potential as a useful framework for probabilistic S2S prediction under a changing climate.
#2
High-impact attribution study linking climate change to flood amplification.
Abstract
Anthropogenic climate change intensifies extreme precipitation, but how it affects flood hazards remains uncertain. In September 2024, Storm Boris triggered catastrophic flooding across central Europe, underscoring the need to quantify the contribution of human-caused climate change to flood intensity for effective risk management. Using a storyline approach, we show that precipitation associated with Storm Boris would have been 6 to 9% lower in the absence of anthropogenic climate change, whereas flood peaks would, on average, have been reduced by 14 to 22%. Our estimates are based on nudged simulations from two climate models coupled with two hydrological models across hundreds of affected catchments. The results imply that anthropogenic forcing amplified flood peaks by a factor of 2.1 to 2.9 relative to the increase in event precipitation. Overall, we demonstrate that attributing precipitation alone would substantially underestimate flood hazard amplification from climate change-an effect that should inform future risk assessments and loss estimates.
Anthropogenic climate change intensifies extreme precipitation, but how it affects flood hazards remains uncertain. In September 2024, Storm Boris triggered catastrophic flooding across central Europe, underscoring the need to quantify the contribution of human-caused climate change to flood intensity for effective risk management. Using a storyline approach, we show that precipitation associated with Storm Boris would have been 6 to 9% lower in the absence of anthropogenic climate change, whereas flood peaks would, on average, have been reduced by 14 to 22%. Our estimates are based on nudged simulations from two climate models coupled with two hydrological models across hundreds of affected catchments. The results imply that anthropogenic forcing amplified flood peaks by a factor of 2.1 to 2.9 relative to the increase in event precipitation. Overall, we demonstrate that attributing precipitation alone would substantially underestimate flood hazard amplification from climate change-an effect that should inform future risk assessments and loss estimates.
#3
Novel insight into vegetation-climate feedbacks using remote sensing.
Abstract
Abstract Terrestrial vegetation regulates global carbon cycling and climate through responses to water availability. Yet, it remains unclear whether soil moisture (SM) or atmospheric dryness (vapour pressure deficit, VPD) dominates vegetation responses and, more critically, how their relative influence changes over time, limiting predictions of ecosystem–climate feedback under warming. Here we analyse satellite-derived vegetation variability over the past four decades using an explainable artificial intelligence framework that disentangles nonlinear hydroclimatic driver contributions. We show SM dominates vegetation variability in most regions. However, VPD influence has strengthened while SM influence has weakened over time. These changes highlight an asymmetric transition with more SM-dominated ecosystems shifting towards increasing VPD control than vice versa. Changes in the mean and variability of SM and VPD drive this transition by altering plant water-stress responses. Continued warming will probably intensify this pattern as rising VPD and more frequent extreme events strengthen atmospheric constraints on vegetation.
Abstract Terrestrial vegetation regulates global carbon cycling and climate through responses to water availability. Yet, it remains unclear whether soil moisture (SM) or atmospheric dryness (vapour pressure deficit, VPD) dominates vegetation responses and, more critically, how their relative influence changes over time, limiting predictions of ecosystem–climate feedback under warming. Here we analyse satellite-derived vegetation variability over the past four decades using an explainable artificial intelligence framework that disentangles nonlinear hydroclimatic driver contributions. We show SM dominates vegetation variability in most regions. However, VPD influence has strengthened while SM influence has weakened over time. These changes highlight an asymmetric transition with more SM-dominated ecosystems shifting towards increasing VPD control than vice versa. Changes in the mean and variability of SM and VPD drive this transition by altering plant water-stress responses. Continued warming will probably intensify this pattern as rising VPD and more frequent extreme events strengthen atmospheric constraints on vegetation.
#4
Highlights emerging air quality risks from amine-based carbon capture.
Abstract
Abstract Carbon capture (CC) is indispensable to carbon neutrality, and amine-based absorption using monoethanolamine (MEA) is among the most mature CC technologies. MEA can enhance atmospheric new particle formation (NPF), the primary source of atmospheric particles. However, the atmospheric impacts of MEA escaping from future large-scale CC deployment remain unknown. Here, we develop a three-dimensional model that represents MEA sources, sinks, and H2SO4-MEA nucleation based on quantum chemical and cluster dynamics calculations, to evaluate atmospheric MEA and its impacts from CC deployment at China’s coal-fired power plants under carbon-neutrality scenarios. By 2060, even moderate CC deployment raises atmospheric MEA concentrations to tens of pptv, exceeding dimethylamine (DMA), the canonical urban NPF agent. H2SO4-MEA nucleation consequently overtakes H2SO4-DMA nucleation across extensive CC-deployment regions (∼1 million km2) and is more than 10 times stronger within the CC-intensive core (∼0.2 million km2), raising particle number concentrations by 50–187% and exposing ∼430 million people to levels exceeding the WHO high-exposure benchmark. The associated rise in ultrafine particles is estimated to cause ∼34,800 additional deaths across China in 2060. We reveal that CC deployment may deliver climate benefits while inadvertently worsening particulate pollution and public-health burdens, warranting proactive and early preventive action.
Abstract Carbon capture (CC) is indispensable to carbon neutrality, and amine-based absorption using monoethanolamine (MEA) is among the most mature CC technologies. MEA can enhance atmospheric new particle formation (NPF), the primary source of atmospheric particles. However, the atmospheric impacts of MEA escaping from future large-scale CC deployment remain unknown. Here, we develop a three-dimensional model that represents MEA sources, sinks, and H2SO4-MEA nucleation based on quantum chemical and cluster dynamics calculations, to evaluate atmospheric MEA and its impacts from CC deployment at China’s coal-fired power plants under carbon-neutrality scenarios. By 2060, even moderate CC deployment raises atmospheric MEA concentrations to tens of pptv, exceeding dimethylamine (DMA), the canonical urban NPF agent. H2SO4-MEA nucleation consequently overtakes H2SO4-DMA nucleation across extensive CC-deployment regions (∼1 million km2) and is more than 10 times stronger within the CC-intensive core (∼0.2 million km2), raising particle number concentrations by 50–187% and exposing ∼430 million people to levels exceeding the WHO high-exposure benchmark. The associated rise in ultrafine particles is estimated to cause ∼34,800 additional deaths across China in 2060. We reveal that CC deployment may deliver climate benefits while inadvertently worsening particulate pollution and public-health burdens, warranting proactive and early preventive action.
#5
Provides a unique, high-resolution NO2 dataset for Asian air quality studies.
Abstract
Abstract. High-frequency continuous monitoring of tropospheric nitrogen dioxide (NO2) is crucial for assessing regional air quality and investigating photochemical dynamic evolution. Although new-generation geostationary orbit (GEO) satellites (e.g., GEMS) provide hourly observations, their spatial coverage is often severely compromised by cloud obstruction and inherent retrieval limitations. Furthermore, their short observation history hinders long-term, cross-regional environmental assessments. To address this, this study presents a gapless, 0.05° hourly tropospheric NO2 vertical column density (VCD) dataset for representative hotspots in Asia spanning 2019 to 2024, reconstructed using a physics-aware deep learning framework. The framework integrates partial convolutions to handle irregularly missing satellite observations and proposes a Physics-aware Normalization (PhysNorm) module. PhysNorm dynamically modulates 0.05° high-resolution feature maps using 0.25° low-resolution physical features such as ERA5 meteorological fields and EAC4 chemical priors, thereby ensuring rigorous physical continuity while filling data gaps. Validation results show that the model performs exceptionally on the test set (R2 = 0.889) and maintains high generalization stability in an independent validation on 2024 data, which was not used for training. Cross-validation against an independent polar-orbiting satellite (GOME-2C) confirms the reliability of the dataset in reconstructing pollution hotspots and its applicability to periods without GEMS observations (2019–2022). Using this reconstructed dataset, this study finely delineates the periodic diurnal characteristics of NO2 in typical Asian regions and accurately captures the inter-annual concentration gradients driven by public health events and emission reduction policies over the past six years. This newly generated dataset provides an unprecedented spatiotemporally continuous record, overcoming the limitations of cloud cover and short observational histories, thereby facilitating long-term, high-resolution air quality assessments and epidemiological studies. This dataset is available at https://doi.org/10.5281/zenodo.20427767 (Gao et al., 2026).
Abstract. High-frequency continuous monitoring of tropospheric nitrogen dioxide (NO2) is crucial for assessing regional air quality and investigating photochemical dynamic evolution. Although new-generation geostationary orbit (GEO) satellites (e.g., GEMS) provide hourly observations, their spatial coverage is often severely compromised by cloud obstruction and inherent retrieval limitations. Furthermore, their short observation history hinders long-term, cross-regional environmental assessments. To address this, this study presents a gapless, 0.05° hourly tropospheric NO2 vertical column density (VCD) dataset for representative hotspots in Asia spanning 2019 to 2024, reconstructed using a physics-aware deep learning framework. The framework integrates partial convolutions to handle irregularly missing satellite observations and proposes a Physics-aware Normalization (PhysNorm) module. PhysNorm dynamically modulates 0.05° high-resolution feature maps using 0.25° low-resolution physical features such as ERA5 meteorological fields and EAC4 chemical priors, thereby ensuring rigorous physical continuity while filling data gaps. Validation results show that the model performs exceptionally on the test set (R2 = 0.889) and maintains high generalization stability in an independent validation on 2024 data, which was not used for training. Cross-validation against an independent polar-orbiting satellite (GOME-2C) confirms the reliability of the dataset in reconstructing pollution hotspots and its applicability to periods without GEMS observations (2019–2022). Using this reconstructed dataset, this study finely delineates the periodic diurnal characteristics of NO2 in typical Asian regions and accurately captures the inter-annual concentration gradients driven by public health events and emission reduction policies over the past six years. This newly generated dataset provides an unprecedented spatiotemporally continuous record, overcoming the limitations of cloud cover and short observational histories, thereby facilitating long-term, high-resolution air quality assessments and epidemiological studies. This dataset is available at https://doi.org/10.5281/zenodo.20427767 (Gao et al., 2026).
#6
Reveals underestimation of coastal humid-heat extremes in climate models.
Abstract
Humid-heat extremes reach intensities close to the limit of human survivability around the Arabian Peninsula. Here, we analyze fully coupled kilometer-scale Icosahedral Nonhydrostatic (ICON) and Integrated Forecasting System–FinitevolumE Sea ice-Ocean Model (IFS-FESOM) simulations from the nextGEMS project, with original resolutions of around 9 kilometers, and compare them with station observations, ERA5, and CMIP6 projections. IFS-FESOM reproduces well the observed upper-tail wet-bulb temperature (TW) and lies near the high end of the CMIP6 distribution over coastal regions, while ICON simulates weaker extremes. Spatial smoothing suppresses localized coastal maxima, reducing daily-maximum TW extremes by up to 8°C and lowering exceedance frequencies across multiple thresholds. Native resolution fields also show stronger mid-century increases in exposure under SSP3-7.0 than their coarse-gridded counterparts. These results suggest that conventional coarse resolution projections may underestimate localized coastal humid-heat risk, although the magnitude of this effect remains model dependent.
Humid-heat extremes reach intensities close to the limit of human survivability around the Arabian Peninsula. Here, we analyze fully coupled kilometer-scale Icosahedral Nonhydrostatic (ICON) and Integrated Forecasting System–FinitevolumE Sea ice-Ocean Model (IFS-FESOM) simulations from the nextGEMS project, with original resolutions of around 9 kilometers, and compare them with station observations, ERA5, and CMIP6 projections. IFS-FESOM reproduces well the observed upper-tail wet-bulb temperature (TW) and lies near the high end of the CMIP6 distribution over coastal regions, while ICON simulates weaker extremes. Spatial smoothing suppresses localized coastal maxima, reducing daily-maximum TW extremes by up to 8°C and lowering exceedance frequencies across multiple thresholds. Native resolution fields also show stronger mid-century increases in exposure under SSP3-7.0 than their coarse-gridded counterparts. These results suggest that conventional coarse resolution projections may underestimate localized coastal humid-heat risk, although the magnitude of this effect remains model dependent.
#7
Presents a new global groundwater storage dataset, key for hydrology.
Abstract
Abstract. Groundwater is one of the most important freshwater resources for ecosystems and mankind. Because of its fundamental role in the Earth's water and energy cycles, groundwater has been declared an essential climate variable by GCOS, the Global Climate Observing System. Similar to other subsurface states and fluxes, groundwater is difficult to monitor at the global scale, with sufficient spatial coverage and over climate-relevant time scales. The Global Gravity-based Groundwater Product (G3P) is a global observation-based data set of large-scale groundwater storage variations. G3P capitalizes on the unique capability of GRACE and GRACE-FO satellite gravimetry as the only remote sensing technology to monitor subsurface mass variations. In a mass balance approach, satellite-based, in situ observation-based and model-based water storage variations of snow water equivalent, root-zone soil moisture, glacier mass, and surface water storage are subtracted from GRACE/-FO terrestrial water storage anomalies to result in monthly variations of groundwater storage. For this combination, the individual compartmental storage data are spatially filtered to be consistent with the spatial resolution of terrestrial water storage from satellite gravimetry. The G3P data set presented here covers the period 2002 to 2023 with monthly resolution on a 0.5° global grid and includes propagated uncertainty information. We describe the details of the G3P data processing chain and of each contributing data stream, provide examples of spatial and temporal groundwater storage variations represented by the G3P data set, and present exemplary evaluation results against in situ groundwater observations for three large aquifer systems. G3P is a prototype for an operational global groundwater service, under development as a cross-cutting extension of the existing portfolio of the Copernicus Climate Change Service C3S. The G3P data set is available via GFZ Data Services at https://doi.org/10.5880/G3P.2024.001 (Güntner et al., 2024).
Abstract. Groundwater is one of the most important freshwater resources for ecosystems and mankind. Because of its fundamental role in the Earth's water and energy cycles, groundwater has been declared an essential climate variable by GCOS, the Global Climate Observing System. Similar to other subsurface states and fluxes, groundwater is difficult to monitor at the global scale, with sufficient spatial coverage and over climate-relevant time scales. The Global Gravity-based Groundwater Product (G3P) is a global observation-based data set of large-scale groundwater storage variations. G3P capitalizes on the unique capability of GRACE and GRACE-FO satellite gravimetry as the only remote sensing technology to monitor subsurface mass variations. In a mass balance approach, satellite-based, in situ observation-based and model-based water storage variations of snow water equivalent, root-zone soil moisture, glacier mass, and surface water storage are subtracted from GRACE/-FO terrestrial water storage anomalies to result in monthly variations of groundwater storage. For this combination, the individual compartmental storage data are spatially filtered to be consistent with the spatial resolution of terrestrial water storage from satellite gravimetry. The G3P data set presented here covers the period 2002 to 2023 with monthly resolution on a 0.5° global grid and includes propagated uncertainty information. We describe the details of the G3P data processing chain and of each contributing data stream, provide examples of spatial and temporal groundwater storage variations represented by the G3P data set, and present exemplary evaluation results against in situ groundwater observations for three large aquifer systems. G3P is a prototype for an operational global groundwater service, under development as a cross-cutting extension of the existing portfolio of the Copernicus Climate Change Service C3S. The G3P data set is available via GFZ Data Services at https://doi.org/10.5880/G3P.2024.001 (Güntner et al., 2024).
#8
Global mapping of peatland disturbance and GHG emissions, relevant for mitigation.
Abstract
Peatlands and other organic soils occupy a small fraction of the land surface but store a large share of global soil carbon, and drainage and fire in these systems are major, poorly mapped sources of greenhouse gas (GHG) emissions. Remote sensing enables global monitoring of organic-soil disturbance, but few globally consistent systems link organic-soil extent, disturbance activity, and emissions at high spatial resolution. Here we develop a nominal 30 m (0.00025°) global framework that overlays an organic-soil extent map with multi-temporal land-cover, infrastructure, plantation, peat-extraction, and burned-area data to delineate disturbed organic soils. Using IPCC Wetlands Supplement (Tier 1) methods, we estimate CO 2 , CH 4 , N 2 O, and CO emissions over 2001–2024. The baseline uses an operational organic-soil extent from an ensemble probability layer, thresholded separately within tropical, boreal, and temperate IPCC climate domains. Because mapped organic-soil extent is a primary determinant of disturbed area and emissions, we interpret this baseline as a spatially explicit monitoring estimate rather than as a definitive global total. In the baseline configuration, disturbed organic soils emitted about 2.6 Gt CO 2 e yr − 1 (2.5–2.7 Gt CO 2 e yr − 1 across five inventory periods), with roughly three-quarters from drainage and one-quarter from fires. Emissions are heavily concentrated in a few countries and land uses, led by Indonesia and Russia, with the tropics contributing the largest share. The leading drained land-use sources are settlement/built-up pixels, oil palm, and forest inferred as drained from nearby infrastructure, followed by cropland on organic soils. For 2021–2024, alternative extent products yield 2.6–3.3 Gt CO 2 e yr − 1 , and jointly varying the classification threshold and Tier 1 emission factors—the two dominant sources of tested variation—spans an overall scenario envelope of approximately 0.7–4.5 Gt CO 2 e yr − 1 . These estimates should be read not as a correction to existing peatland-specific estimates but as complementary, more comprehensive monitoring of disturbed organic-soil systems under a harmonized, globally consistent framework. The resulting 30 m maps of organic-soil state, disturbance, and emissions show how multi-temporal Earth observation can be combined with GHG inventory methods to identify high-emitting hotspots and support inventories, nationally determined contributions, and peatland conservation and restoration planning.
Peatlands and other organic soils occupy a small fraction of the land surface but store a large share of global soil carbon, and drainage and fire in these systems are major, poorly mapped sources of greenhouse gas (GHG) emissions. Remote sensing enables global monitoring of organic-soil disturbance, but few globally consistent systems link organic-soil extent, disturbance activity, and emissions at high spatial resolution. Here we develop a nominal 30 m (0.00025°) global framework that overlays an organic-soil extent map with multi-temporal land-cover, infrastructure, plantation, peat-extraction, and burned-area data to delineate disturbed organic soils. Using IPCC Wetlands Supplement (Tier 1) methods, we estimate CO 2 , CH 4 , N 2 O, and CO emissions over 2001–2024. The baseline uses an operational organic-soil extent from an ensemble probability layer, thresholded separately within tropical, boreal, and temperate IPCC climate domains. Because mapped organic-soil extent is a primary determinant of disturbed area and emissions, we interpret this baseline as a spatially explicit monitoring estimate rather than as a definitive global total. In the baseline configuration, disturbed organic soils emitted about 2.6 Gt CO 2 e yr − 1 (2.5–2.7 Gt CO 2 e yr − 1 across five inventory periods), with roughly three-quarters from drainage and one-quarter from fires. Emissions are heavily concentrated in a few countries and land uses, led by Indonesia and Russia, with the tropics contributing the largest share. The leading drained land-use sources are settlement/built-up pixels, oil palm, and forest inferred as drained from nearby infrastructure, followed by cropland on organic soils. For 2021–2024, alternative extent products yield 2.6–3.3 Gt CO 2 e yr − 1 , and jointly varying the classification threshold and Tier 1 emission factors—the two dominant sources of tested variation—spans an overall scenario envelope of approximately 0.7–4.5 Gt CO 2 e yr − 1 . These estimates should be read not as a correction to existing peatland-specific estimates but as complementary, more comprehensive monitoring of disturbed organic-soil systems under a harmonized, globally consistent framework. The resulting 30 m maps of organic-soil state, disturbance, and emissions show how multi-temporal Earth observation can be combined with GHG inventory methods to identify high-emitting hotspots and support inventories, nationally determined contributions, and peatland conservation and restoration planning.
#9
Explores ocean memory's role in marine heatwave variability.
Abstract
Abstract The duration of marine heatwaves (MHWs) determines the cumulative thermal stress imposed on marine organisms, thereby strongly influencing the severity of ecological consequences. Recently, day-to-day ocean memory, quantified by a first-order sea surface temperature (SST) autocorrelation model, has been found as an effective metric to depict the duration of MHWs, due to its consistent climatological distribution and trends with MHW duration in the observations. However, the relative importance of external forcing and internal variability in modulating the day-to-day ocean memory and, more importantly, its relationship with MHW duration, remain unclear. In this study, we employ 100 ensemble members of CESM2 and 32 CMIP6 models to analyze the roles of external forcing and internal variability in the day-to-day ocean memory and MHW duration. We find that internal variability dominates the trends of day-to-day ocean memory in the recent decades and further modulates the trends of MHW duration, particularly in the tropical oceans. The zonal gradient of SST trends over the tropical Pacific emerges as a key feature of the internal variability. By combining a stochastic climate model framework with a mixed-layer heat budget, we further reveal that ocean mixing and dynamics are the primary processes through which internal variability modulates trends in day-to-day ocean memory, whereas SST-surface heat flux feedback provides a secondary contribution and variations in mixed-layer depth play negligible roles. This study demonstrates that internal variability exerts a considerable influence on MHW duration through its modulation on day-to-day ocean memory, providing crucial evidence and a new perspective for understanding and predicting MHW duration in a warming climate.
Abstract The duration of marine heatwaves (MHWs) determines the cumulative thermal stress imposed on marine organisms, thereby strongly influencing the severity of ecological consequences. Recently, day-to-day ocean memory, quantified by a first-order sea surface temperature (SST) autocorrelation model, has been found as an effective metric to depict the duration of MHWs, due to its consistent climatological distribution and trends with MHW duration in the observations. However, the relative importance of external forcing and internal variability in modulating the day-to-day ocean memory and, more importantly, its relationship with MHW duration, remain unclear. In this study, we employ 100 ensemble members of CESM2 and 32 CMIP6 models to analyze the roles of external forcing and internal variability in the day-to-day ocean memory and MHW duration. We find that internal variability dominates the trends of day-to-day ocean memory in the recent decades and further modulates the trends of MHW duration, particularly in the tropical oceans. The zonal gradient of SST trends over the tropical Pacific emerges as a key feature of the internal variability. By combining a stochastic climate model framework with a mixed-layer heat budget, we further reveal that ocean mixing and dynamics are the primary processes through which internal variability modulates trends in day-to-day ocean memory, whereas SST-surface heat flux feedback provides a secondary contribution and variations in mixed-layer depth play negligible roles. This study demonstrates that internal variability exerts a considerable influence on MHW duration through its modulation on day-to-day ocean memory, providing crucial evidence and a new perspective for understanding and predicting MHW duration in a warming climate.
#10
Links paleomonsoon dynamics to North African aridification, broad Earth system relevance.
Abstract
North Africa experienced a pronounced transition from humid to arid conditions during the early Cenozoic. Meanwhile, the South Asian monsoon underwent substantial intensification, largely driven by the uplift of the Himalayan-Tibetan Plateau. Here, we demonstrate that the two events, which greatly shaped the paleoenvironment and modern regional climate, are not independent responses to paleogeography changes; rather, the former is primarily driven by the latter through the monsoon-desert mechanism, in which South Asian monsoon diabatic heating produces large-scale subsidence and aridity over North Africa. Using multiple ensembles of paleoclimate simulations and a model for the atmospheric response to imposed diabatic heating, we show that the intensification and poleward shift of Asian monsoon rainfall each make comparable contributions to the aridification of North Africa. Our findings identify a persistent dynamical pathway linking the South Asian and North African monsoons and suggest that it played a major role in early Cenozoic North African aridification, while its climatic expression at other times depended on the relative strengths of competing forcings.
North Africa experienced a pronounced transition from humid to arid conditions during the early Cenozoic. Meanwhile, the South Asian monsoon underwent substantial intensification, largely driven by the uplift of the Himalayan-Tibetan Plateau. Here, we demonstrate that the two events, which greatly shaped the paleoenvironment and modern regional climate, are not independent responses to paleogeography changes; rather, the former is primarily driven by the latter through the monsoon-desert mechanism, in which South Asian monsoon diabatic heating produces large-scale subsidence and aridity over North Africa. Using multiple ensembles of paleoclimate simulations and a model for the atmospheric response to imposed diabatic heating, we show that the intensification and poleward shift of Asian monsoon rainfall each make comparable contributions to the aridification of North Africa. Our findings identify a persistent dynamical pathway linking the South Asian and North African monsoons and suggest that it played a major role in early Cenozoic North African aridification, while its climatic expression at other times depended on the relative strengths of competing forcings.