As sea ice disappears, the emergence of open ocean deep convection in the Arctic has been suggested. Here, using 36 state-of-the-art climate models and up to 50 ensemble members per model, we show that Arctic deep convection is rare even under the strongest warming scenario. Only 5 models have somewhat permanent convection by 2100, while 11 have had convection by the middle of the run. For all, the deepest mixed layers are in the Eurasian basin, by St Anna Trough. When the models convect, that region undergoes a salinification and increasing wind speeds; it is freshening otherwise. We discuss the causality and potential reasons for the opposite trends. Given the model’s different parameterisations, and given that the ensemble members that convect the deepest, most often, are those with the strongest sensitivity, we conclude that differences in deep convection are most likely linked to the model formulation.

A document by Daniel Michael Gilford, PhD. Click on the document to view its contents.

Cloud microphysics is a critical aspect of the Earth’s climate system, which involves processes at the nano- and micrometer scales of droplets and ice particles. In climate modeling, cloud microphysics is commonly represented by bulk models, which contain simplified process rates that require calibration. This study presents a framework for calibrating warm-rain bulk schemes using high-fidelity super-droplet simulations that provide a more accurate and physically based representation of cloud and precipitation processes. The calibration framework employs ensemble Kalman methods including ensemble Kalman inversion (EKI) and unscented Kalman inversion (UKI) to calibrate bulk microphysics schemes with probabilistic super-droplet simulations. We demonstrate the framework’s effectiveness by calibrating a single-moment bulk scheme, resulting in a reduction of data-model mismatch by more than $75\%$ compared to the model with initial parameters. Thus, this study demonstrates a powerful tool for enhancing the accuracy of bulk microphysics schemes in atmospheric models and improving climate modeling.

Irrigated cultivation, as a prevalent anthropogenic activity, exerts a significant influence on land use and land cover, resulting in notable modifications to land-atmosphere interaction and the hydrological cycle. Given the extensive cropland, high productivity, compact rotation, semi-arid climate, intense irrigation, and groundwater depletion in the North China Plain (NCP), the development of a comprehensive crop-irrigation-groundwater model becomes imperative for understanding agricultural-induced climate response in this region. This study presents an integrated crop model explicitly tailored to the NCP, which incorporates double-cropping rotation, irrigation practice, and groundwater interactions into the regional climate model. The modifications are implemented to: (1) enable a seamless transition from field scale application to regional scale application, facilitating the incorporation of spatial variability, (2) capture the distinctive attributes of the NCP region, ensuring the model accurately reflects its unique characteristics, and (3) reinforce the direct interaction among crop-related variables, thereby enhancing the model’s capacity to simulate their dynamic behaviors. The integrated crop modeling system demonstrates a commendable performance in crop simulations using climatic conditions, which is substantiated by its identification of crop stages, estimation of field biomass, prediction of crop yield, and finally the projection of monthly leaf area index. In our next phase, this integrated crop modeling system will be employed in long-term simulations to enhance our understanding of the intricate relationship between agricultural development and climate change.

The study and simulation of the Urban Heat Island (UHI) and Heat Index (HI) effects in the Houston-Galveston metropolitan area demand special attention, particularly in considering moist processes aloft. During the warm season, the afternoon sea breeze phenomenon in this coastal city acts as a natural air conditioner for city residents, facilitating the dispersion of pollutants, moisture, and heat. To delve into the intricate relationships among urbanization, clouds, and land-sea interaction, we conducted cloud- and urban-resolving simulations at a 900 m grid resolution. Results show that urbanization correlates with the presence of shallower cumulus clouds, cloud bases at higher altitudes, and increased cloud duration over the Galveston-Houston region compared to rural areas. These urban clouds benefit from the enhanced sensible heat and dynamic drag imparted by the urban landscape, thereby intensifying vertical mixing and moisture flux convergence. This dynamic interplay uplifts heat and moisture convergence, contributing to the enhancement of moist static energy that sustains the additional urban convection. Interestingly, our findings suggest that urbanization augments the mean HI while mitigating its afternoon high. An urban circulation dome emerges, overpowering the influence of land-sea circulations. Contrary to expectations, urbanization doesn’t seem to promote a stronger sea breeze that would favor moist and cooler air mass to the city. Instead, the influence of urbanization on cloud enhancement emerges as a crucial pathway responsible for reducing the afternoon HI values. Moreover, uncertainties in SSTs are closely linked to the sensitivities of land-sea circulations, which in turn modulate UHI and extreme heat indicators.

The water-energy-food nexus has emerged as a critical research interest to support integrated resource planning, management, and security. For this reason, many tools have been developed recently to evaluate the WEF nexus security and monitor progress towards the WEF-related sustainable development goals. Among these, the calculation of the WEF composite index model is critical since it can provide a quantitative approach to demonstrate the WEF nexus security status. However, the current WEF nexus index model framework needs to include the incorporation of governance indicators, neglecting the importance of governance in the WEF nexus framework. Thus, this article develops a new WEF nexus composite index model that incorporates governance indicators in each subpillar is developed. The principal component analysis (PCA) is adopted to reduce the variables’ collinearity and the model’s dimensionality. A quasi-Monte Carlo based uncertainty and global sensitivity analysis are applied to the index model to assess its effectiveness. Finally, the new WEF index model is applied on the 16 South African Development Community (SADC) countries as a case study. A critical synergy effect within the WEF nexus framework is identified that nations with better WEF governance ability tend to perform better in improving the WEF accessibility capability, suggesting the importance of the governance in the WEF nexus security framework.

The Atlantic meridional overturning circulation (AMOC) in many CMIP6 models has been shown to be overly-sensitive to anthropogenic aerosol forcing, and it has been speculated that this is due to the inclusion of aerosol indirect effects for the first time in many models of that generation. We analyze the AMOC response in a newly-released ensemble of historic simulations performed with CESM2 and forced by the older CMIP5 input datasets (CESM2-CMIP5). This AMOC response is then compared to the CESM1 large ensemble (CESM1-LE, forced by the older CMIP5 inputs) and the CESM2 large ensemble (CESM2-LE, forced by the newer CMIP6 inputs). A key conclusion, only made possible by this experimental setup, is that changes in modeled aerosol-indirect effects cannot explain the differences in turbulent fluxes between CESM1-LE and CESM2-LE. Instead, differences in surface turbulent heat fluxes from changes in model inputs likely drive the different AMOC responses.

Despite increasing exposure to flooding and associated financial damages, estimates suggest more than two-thirds of flood-exposed properties are currently uninsured. This low adoption rate could undermine the climate resilience of communities and weaken the financial solvency of the United States National Flood Insurance Program (NFIP). We study whether repeated exposure to flood events, especially disaster-scale floods expected to become more frequent in a warming climate, could spur insurance adoption. Using improved estimates of residential insurance take-up in locations where such insurance is voluntary, and exploiting variation in the frequency and severity of flood events over time, we quantify how flood events impact local insurance demand. We find that a flood disaster declaration in a given year increases the take-up rate of insurance by 7% in the following year, but the effect diminishes in subsequent years and is gone after five years. This effect is more short-lived in counties in inland states that do not border the Gulf and Atlantic coasts. The effect of a flood on takeup is substantially larger if there was also a flood in the previous year. We also find that recent disasters are more salient for homeowners whose primary residences are exposed to a disaster declaration compared to non-primary residences. Our results provide a more comprehensive understanding of the salience effect of flooding on insurance demand compared to previous studies. Overall, these findings suggest that relying on households to self-adapt to increasing flood risks in a changing climate is insufficient for closing the insurance protection gap.

North Atlantic sea surface temperatures (NASST), particularly in the subpolar region, are among the most predictable locations in the world’s oceans. However, the relative importance of atmospheric and oceanic controls on their variability at multidecadal timescales remain uncertain. Neural networks (NNs) are trained to examine the relative importance of oceanic and atmospheric predictors in predicting the NASST state in the Community Earth System Model 1 (CESM1). In the presence of external forcings, oceanic predictors outperform atmospheric predictors, persistence, and random chance baselines out to 25-year leadtimes. Layer-wise relevance propagation is used to unveil the sources of predictability, and reveal that NNs consistently rely upon the Gulf Stream-North Atlantic Current region for accurate predictions. Additionally, CESM1-trained NNs do not need additional transfer learning to successfully predict the phasing of multidecadal variability in an observational dataset, suggesting consistency in physical processes driving NASST variability between CESM1 and observations.

High-resolution climate projections provide crucial insights into assessing climate risk and developing climate resilience strategies. The Seasonal Trends and Analysis of Residuals empirical statistical downscaling model (STAR-ESDM) is a computationally-efficient and flexible approach to generating high-resolution climate projections that can be applied globally using a broad range of predictands and predictors that can be sourced from weather stations, gridded datasets, satellites, reanalysis, and global or regional climate models. It uses signal processing combined with Fourier filtering and kernal density estimation techniques to decompose and smooth any quasi-Gaussian time series, gridded or point-based, into multi-decadal long-term means and/or trends; static and dynamic annual cycles; and probability distributions of high-frequency variability. Long-term predictor trends are bias-corrected and predictor components are used to map remaining predictand components to future conditions. Components are then recombined for each station or grid cell to produce a continuous, high-resolution bias-corrected and downscaled time series at the spatial and temporal scale of the original time series. Comparing STAR-ESDM output with high-resolution daily temperature and precipitation projections generated by a fully dynamical global model demonstrates that the method is extremely robust, capable of accurately reproducing projected changes for all but the most extreme temperature and precipitation values. For most continental areas, biases in 1-in-1000 hottest and coldest temperatures are less than 0.5°C and biases in the 1-in-1000 wet day precipitation amounts are less than 5 mm/day. As climate impacts intensify, STAR-ESDM represents a significant advance in generating consistent high-resolution projections to comprehensively assess risk and optimize resilience.

The northern permafrost region has been projected to shift from a net sink to a net source of carbon under global warming. However, estimates of the contemporary net greenhouse gas (GHG) balance and budgets of the permafrost region remain highly uncertain. Here we construct the first comprehensive bottom-up budgets of CO2, CH4, and N2O across the terrestrial permafrost region using databases of more than 1000 in-situ flux measurements and a land cover-based ecosystem flux upscaling approach for the period 2000-2020. Estimates indicate that the permafrost region emitted a mean annual flux of 0.36 (-620, 652) Tg CO2-C y-1, 38 (21, 53) Tg CH4-C y-1, and 0.62 (0.03, 1.2) Tg N2O-N y-1 to the atmosphere throughout the period. While the region was a net source of CH4 and N2O, the CO2 budget was near neutral with large uncertainties. Terrestrial ecosystems remained a CO2 sink, but emissions from fire disturbances and inland waters largely offset the sink in vegetated ecosystems. Including lateral fluxes, the permafrost region was a net source of C and N, releasing 136 (-517, 821) Tg C y-1 and 3.2 (1.9, 4.8) Tg N y-1. Large uncertainty ranges in these estimates point to a need for further expansion of monitoring networks, continued data synthesis efforts, and better integration of field observations, remote sensing data, and ecosystem models to constrain the contemporary net GHG budgets of the permafrost region and track their future trajectory.

Precipitation forecasts, particularly at subseasonal-to-seasonal (S2S) time scale, are essential for informed and proactive water resources management. Although S2S precipitation forecasts have been evaluated, no systematic decomposition of the skill, Nash-Sutcliffe Efficiency (NSE) coefficient, has been analyzed towards understanding the forecast accuracy. We decompose the NSE of S2S precipitation forecast into its three components – correlation, conditional bias, and unconditional bias – by four seasons, three lead times (1–12-day, 1-22 day, and 1-32 day), and three models (ECMWF, CFS, NCEP) over the Conterminous United States (CONUS). Application of dry mask is critical as the NSE and correlation are lower across all seasons after masking areas with low precipitation values. Further, a west-to-east gradient in S2S forecast skill exists and forecast skill was better during the winter months and for areas closer to the coast. Overall, ECMWF’s model performance was stronger than both ECCC and NCEP CFS’s performance, mainly for the forecasts issued during fall and winter months. However, ECCC and NCEP CFS performed better for the forecast issued during the spring months, and also performed better in in-land areas. Post-processing using simple Model Output Statistics could reduce both unconditional and conditional bias to zero, thereby offering better skill for regimes with high correlation. Our decomposition results also show efforts should focus on improving model parametrization and initialization schemes for climate regimes with low correlation values.

The long-term net sink of carbon (C), nitrogen (N) and greenhouse gases (GHGs) in the northern permafrost region is projected to weaken or shift under climate change. But large uncertainties remain, even on present-day GHG budgets. We compare bottom-up (data-driven upscaling, process-based models) and top-down budgets (atmospheric inversion models) of the main GHGs (CO2, CH4, and N2O) and lateral fluxes of C and N across the region over 2000-2020. Bottom-up approaches estimate higher land to atmosphere fluxes for all GHGs compared to top-down atmospheric inversions. Both bottom-up and top-down approaches respectively show a net sink of CO2 in natural ecosystems (-31 (-667, 559) and -587 (-862, -312), respectively) but sources of CH4 (38 (23, 53) and 15 (11, 18) Tg CH4-C yr-1) and N2O (0.6 (0.03, 1.2) and 0.09 (-0.19, 0.37) Tg N2O-N yr-1) in natural ecosystems. Assuming equal weight to bottom-up and top-down budgets and including anthropogenic emissions, the combined GHG budget is a source of 147 (-492, 759) Tg CO2-Ceq yr-1 (GWP100). A net CO2 sink in boreal forests and wetlands is offset by CO2 emissions from inland waters and CH4 emissions from wetlands and inland waters, with a smaller additional warming from N2O emissions. Priorities for future research include representation of inland waters in process-based models and compilation of process-model ensembles for CH4 and N2O. Discrepancies between bottom-up and top-down methods call for analyses of how prior flux ensembles impact inversion budgets, more in-situ flux observations and improved resolution in upscaling.

We explore the sensitivity of Southern Ocean surface and deep ocean temperature and salinity biases in the FOCI coupled climate model to atmosphere-ocean coupling time step and to lateral diffusion in the ocean with the goal to reduce biases common to climate models. The reference simulation suffers from a warm bias at the sea surface which also extends down to the seafloor in the Southern Ocean and is accompanied by a too fresh surface, in particular along the Antarctic coast. Reducing the atmosphere-ocean coupling time step from 3 hours to 1 hour results in increased sea-ice production on the shelf and enhanced melting to the north which reduces the fresh bias of the shelf water while also strengthening the meridional density gradient favouring a stronger Antarctic Circumpolar Current (ACC). With the shorter coupling step we also find a stronger meridional overturning circulation with more upwelling and downwelling south and north of the ACC respectively, as well as a reduced warm bias at almost all depths. Tuning the lateral ocean mixing has only a small effect on the model biases, which contradicts previous studies using a similar model configuration. We note that the latitude of the surface westerly wind maximum has a northward bias in the reference simulation and that this bias is unchanged as the surface temperature and sea-ice biases are reduced in the coupled simulations. Hence, the surface wind biases over the Southern Hemisphere midlatitudes appear to be unrelated to biases in sea-surface conditions.

The hyporheic zone is a hotspot for biogeochemical cycling where interactions with mineral metals preserve the release and biodegradation of organic matter (OM). A small fraction of OM can still be exchanged between localized sediments and the overlying water column, and recent evidence suggest there exists a longitudinal structuring in sediment dissolved OM (DOM) chemistry across the continental United States (CONUS). In this study, we tested a hypothesis that water extractable sediment DOM chemistry could be explained by sediment metal contents and integrative watershed scale features at the CONUS scale. Crowdsourced samples were characterized for high resolution mass spectrometry and coupled with sediment metals determined via x-ray fluorescence as well as with land cover and soil elemental information obtained from national databases. Our results highlight weak relationships between DOM chemistry and elemental composition at the CONUS scale indicating limited transferability of organo-metal linkages into multi-scale hydrobiogeochemical models.

During atmospheric river (AR) landfalls on the Antarctic ice sheet, the high waviness of the circumpolar polar jet stream allows for sub-tropical air masses to be advected towards the Antarctic coastline. These rare but high-impact AR events are highly consequential for the Antarctic mass balance; yet little is known about the various atmospheric dynamical components determining their life cycle. By using an AR detection algorithm to retrieve AR landfalls at Dumont d’Urville and non-AR analogues based on 700 hPa geopotential height, we examined what makes AR landfalls unique and studied the complete life cycle of ARs to affect Dumont d’Urville. ARs form in the mid-latitudes/sub-tropics in areas of high surface evaporation, likely in response to tropical deep convection anomalies. These convection anomalies likely lead to Rossby wave trains that help amplify the upper-tropospheric flow pattern. As the AR approaches Antarctica, condensation of isentropically lifted moisture causes latent heat release that – in conjunction with poleward warm air advection – induces geopotential height rises and anticyclonic upper-level potential vorticity tendencies downstream. As evidenced by a blocking index, these tendencies lead to enhanced ridging/blocking that persist beyond the AR landfall time, sustaining warm air advection onto the ice sheet. Finally, we demonstrate a connection between tropopause polar vortices and mid-latitude cyclogenesis in an AR case study. Overall, the non-AR analogues reveal that the amplified jet pattern observed during AR landfalls is a result of enhanced poleward moisture transport and associated diabatic heating which is likely impossible to replicate without strong moisture transport.

A new detection and attribution method is presented and applied to the global mean surface air temperature (GSAT) from 1900 to 2014. The method aims at attributing the climate changes to the variations of greenhouse gases, anthropogenic aerosols, and natural forcings. A convolutional neural network (CNN) is trained using the simulated GSAT from historical and single-forcing simulations of twelve climate models. Then, we perform a backward optimization with the CNN to estimate the attributable GSAT changes. Such a method does not assume additivity in the effects of the forcings. The uncertainty in the attributable GSAT is estimated by sampling different starting points from single-forcing simulations and repeating the backward optimization. To evaluate this new method, the attributable GSAT changes are also calculated using the regularized optimal fingerprinting (ROF) method. Using synthetic non-additive data, we first find that the neural network-based method estimates attribuable changes better than ROF. When using GSAT data from climate model, the attribuable anomalies are similar for both methods, which might reflect that the influence of forcing is mainy additive for the GSAT. However, we found that the uncertainties given both methods are different. The new method presented here can be adapted and extended in future work, to investigate the non-additive changes found at the local scale or on other physical variables.

Long-term carbon dioxide (CO2) flux measurements between the atmosphere and the ecosystem have been made since 1993 at a cool-temperate deciduous forest site (Takayama) in Japan influenced by Asian Monsoon, constituting the longest dataset among all the AsiaFlux sites. Interannual variations (IAVs) and trends of the annual carbon budget components and their environmental factors were examined. Annual net ecosystem production (NEP) (mean ± 1σ) during the period of eddy covariance measurement in 1999-2021 was 265 ± 86 gC m-2 yr-1, and its IAV was dependent more on gross primary production (GPP) than on ecosystem respiration. IAVs in annual NEP and GPP were correlated with the IAVs of the monthly mean NEP, GPP and leaf area index (LAI) from June to September, as well as with that of the length of the net carbon uptake period. Significant increasing and decreasing trends in the annual NEP and GPP were detected during 2004-2013 and 2013-2021, respectively; the increasing trends were mainly caused by the vegetation recovery from typhoon disturbances while the decreasing trends were partly influenced by recent extreme weather events. Significant positive correlations of the IAVs between the start and the end of the net carbon uptake period, and between the leaf expansion and leaf fall were found. These may be attributed to biological functions and interseasonal relationship of meteorological parameters associated with ENSO events that can also influence IAVs in annual NEP and GPP.

The effects of contemporary increases in riverine freshwater into the Arctic Ocean are estimated from ocean model simulations, using two runoff data sets. One runoff data set which is based on older climatological data, which has no inter-annual variability after 2007 and as such does not represent the observed increases in river runoff into the Arctic. The other data set comes from a hydrological model developed for the Arctic drainage basin, which includes contemporary changes in the climate. In the pan-Arctic this new data set represents an approximately 11% increase in runoff, compared with the older climatological data. Comparing two ocean model runs forced with the different runoff data sets, overall changes in different freshwater markers across the basin were found to be between 5-10%, depending on the area investigated. The strongest increases were seen from the Siberian rivers, which in turn caused the strongest freshening in the Eastern Arctic.

The northern permafrost region has been projected to shift from a net sink to a net source of carbon under global warming. However, estimates of the contemporary net greenhouse gas (GHG) balance and budgets of the permafrost region remain highly uncertain. Here we construct the first comprehensive bottom-up budgets of CO2, CH4, and N2O across the terrestrial permafrost region using databases of more than 1000 in-situ flux measurements and a land cover-based ecosystem flux upscaling approach for the period 2000-2020. Estimates indicate that the permafrost region emitted a mean annual flux of 0.36 (-620, 652) Tg CO2-C y-1, 38 (21, 53) Tg CH4-C y-1, and 0.62 (0.03, 1.2) Tg N2O-N y-1 to the atmosphere throughout the period. While the region was a net source of CH4 and N2O, the CO2 budget was near neutral with large uncertainties. Terrestrial ecosystems remained a CO2 sink, but emissions from fire disturbances and inland waters largely offset the sink in vegetated ecosystems. Including lateral fluxes, the permafrost region was a net source of C and N, releasing 136 (-517, 821) Tg C y-1 and 3.2 (1.9, 4.8) Tg N y-1. Large uncertainty ranges in these estimates point to a need for further expansion of monitoring networks, continued data synthesis efforts, and better integration of field observations, remote sensing data, and ecosystem models to constrain the contemporary net GHG budgets of the permafrost region and track their future trajectory.

Given the possibility of irreversible changes to the Earth system, technological interventions such as solar radiation management (SRM) are sometimes framed as possible climate emergency brakes. However, little knowledge exists on the efficacy of such disruptive interventions. To fill in this gap, we perform Community Earth System Model 2 (CESM 2) simulations of a SSP5-8.5 scenario on which we impose either gradual early-century SRM to stabilise surface temperatures or a rapid late-century cooling, both realised via stratospheric aerosol injection (SAI). While both scenarios cool Earth’s surface, we find that ocean conditions differ drastically. The rapid-cooling scenario fails to dissipate sub-surface ocean heat content (OHC), ends up in a weaker AMOC state and does not restore an ailing North Atlantic deep convection. Furthermore, the weakened AMOC state mediates the climate response to rapid SAI, thus inducing an interhemispheric temperature asymmetry. Our results advise caution when considering SAI as an emergency intervention.

Coarse-grid weather and climate models rely particularly on parameterizations of cloud fields, and coarse-grained cloud fields from a fine-grid reference model are a natural target for a machine-learned parameterization. We machine-learn the coarsened-fine cloud properties as a function of coarse-grid model state in each grid cell of NOAA’s FV3GFS global atmosphere model with 200 km grid spacing, trained using a 3 km fine-grid reference simulation with a modified version of FV3GFS. The ML outputs are coarsened fine fractional cloud cover and liquid and ice cloud condensate mixing ratios, and the inputs are coarse model temperature, pressure, relative humidity, and ice cloud condensate. The predicted fields are skillful and unbiased, but somewhat under-dispersed, resulting in too many partially-cloudy model columns. When the predicted fields are applied diagnostically (offline) in FV3GFS’s radiation scheme, they lead to small biases in global-mean top-of-atmosphere (TOA) and surface radiative fluxes. An unbiased global mean TOA net radiative flux is obtained by setting to zero any predicted cloud with grid cell mean cloud fraction less than a threshold of 6.5%; this does not significantly degrade the ML prediction of cloud properties. The diagnostic, ML-derived radiative fluxes are  far more accurate than those obtained with the existing cloud parameterization in the nudged coarse-grid model, as they leverage the accuracy of the fine-grid reference simulation’s cloud properties. 

Antarctic ice core records suggest that atmospheric CO2 increased by 15 to 20 ppm during Heinrich stadials (HS). These periods of abrupt CO2 increase are associated with a significant weakening of the Atlantic meridional overturning circulation (AMOC), and a warming at high southern latitudes. As such, modelling studies have explored the link between changes in AMOC, high southern latitude climate and atmospheric CO2. While proxy records suggest that the aeolian iron input to the Southern Ocean decreased significantly during HS, the potential impact on CO2 of reduced iron input combined with oceanic circulation changes has not been studied in detail. Here, we quantify the respective and combined impacts of reduced iron fertilisation and AMOC weakening on CO2 by performing numerical experiments with an Earth system model under boundary conditions representing 40,000 years before present (ka). Our study indicates that reduced iron input can contribute up to 6 ppm rise in CO2 during an idealized Heinrich stadial. This is caused by a 5% reduction in nutrient utilisation in the Southern Ocean, leading to reduced export production and increased carbon outgassing from the Southern Ocean. An AMOC weakening under 40ka conditions and without changes in surface winds leads to a ~0.5 ppm CO2 increase. The combined impact of AMOC shutdown and weakened iron fertilisation is almost linear, leading to a total CO2 increase of 7 ppm. Therefore, this study highlights the need of including changes in aeolian iron input when studying the processes leading to changes in atmospheric CO2 concentration during HS.

Over recent decades, the Southern Ocean (SO) has experienced multi-decadal surface cooling despite global warming. Earlier studies have proposed that recent SO cooling has been caused by the strengthening of surface westerlies associated with a positive trend of the Southern Annular Mode (SAM) forced by ozone depletion. Here we revisit this hypothesis by examining the relationships between the SAM, zonal winds and SO sea-surface temperature (SST). Using a low-frequency component analysis, we show that while positive SAM anomalies can induce SST cooling as previously found, this seasonal-to-interannual modulation makes only a small contribution to the observed long-term SO cooling. Global climate models well capture the observed interannual SAM-SST relationship, and yet generally fail to simulate the observed multi-decadal SO cooling. The forced SAM trend in recent decades is thus unlikely the main cause of the observed SO cooling, pointing to a limited role of the Antarctic ozone hole.