The decline of the Classic Maya civilisation (∼8th–10th centuries CE) occurred in part from drought stress induced by natural climate variability. Using the EC-Earth3 8K simulation, driven by time-varying orbital and greenhouse gas forcing, we show that the convergence of internal hydroclimatic rhythms alone, generated severe, prolonged droughts comparable to, or even greater than, observed extreme events. We identify the interaction of multi-centennial oscillations (∼600 years) and centennial (∼160 years) cycles, superimposed with robust sub-centennial oscillations (∼60–90 years) and persistent multi-decadal and inter-annual variability, created a shifting wet/dry regime over the Yucatán Peninsula. Severe droughts arose when different frequency modulations aligned in their dry phases. Our results provide the first model-based demonstration of how internal climate variability alone can trigger extreme events capable of reshaping societies, revealing key climatic drivers of societal vulnerability, with direct implications for how internal variability may interact with anthropogenic forcing to amplify future climate risks.

The Late Pliocene, particularly the Marine Isotope Stage KM5c (3.205 Ma BP) has been increasingly proposed as an analog to future climate change, especially considering changes in the hydrological cycle, monsoon systems, and atmospheric and ocean warming above Pre-Industrial (1850 CE) and historical levels. The Pliocene Modeling Intercomparison Project (PlioMIP), now in its third phase (PlioMIP3), seeks to explore climate of the Pliocene based on a combination of climate model simulations and proxy data reconstructions. One of its goals is also to assess the analogy between past and future climates and to quantify climate sensitivity to Pliocene boundary conditions. This work shall help to improve climate models and their application for both past and future warm climates and to provide a paleoclimate-informed assessment of uncertainties in modeled warm climates. With this manuscript we present the PlioMIP3 core simulations for the pre-industrial control (PI) and the Late Pliocene (LP) based on the AWI Climate Model, Version 3 (AWI-CM3). This represents the first application of AWI-CM3 at tectonic timescales which necessitates more extensive adjustment of model setups than the application for recent climate. We therefore take advantage of the opportunity to also document more generally the methods we devised to generate AWI-CM3 model setups for paleoclimate research under geographies that differ from the modern reference state. AWI-CM3 simulates a Late Pliocene climate that is about 4 °C warmer than the pre-industrial reference, with land warming exceeding ocean warming by a factor of 1.2. Polar amplification is particularly pronounced, with Antarctic surface air temperature anomalies exceeding 6 °C while Arctic anomalies reach 4 °C to 5 °C. In comparison to the previous PlioMIP2, this places AWI-CM3 among the warmer ensemble members, consistent with a relatively high equilibrium climate sensitivity of ~4 °C. Our simulations also display an intensified hydrological cycle, with global mean precipitation increasing by 0.31 mm d⁻¹. The ocean surface warms globally to about 3.06 °C, accompanied by contrasting salinity trends, with salinization of the North Atlantic (+3 PSU) and freshening of the Arctic (–2.5 PSU) and Indian (–1 PSU) Oceans. Additionally, the meridional overturning circulation (MOC) reorganizes, with the Atlantic MOC strengthening by about 8 Sv, the Pacific MOC remaining inactive, and the global Antarctic Bottom Water cell being substantially reduced (11 Sv weaker relative to PI). We find reduced global sea-ice extent, that is halved with respect to PI in the Southern Hemisphere, and enhanced northward ocean heat transport in the North Atlantic. Overall, AWI-CM3 reproduces the large-scale climate features of the Late Pliocene inferred from proxy records and the PlioMIP2 ensemble, while highlighting key ocean–atmosphere feedbacks shaping this warm climate.

The Earth's ice sheets, including the Antarctic Ice Sheet (AIS), are critical tipping points in the climate system. In recent years, the potential future collapse has garnered increased attention due to its cascading effects, which could significantly alter global climate patterns and cause large-scale, long-lasting, and potentially irreversible changes within human timescales. This study investigates the large-scale response of the polar Southern Hemisphere (pSH; comprising the Southern Ocean and Antarctica (60–90° S)) to the geometric reduction in ice sheets to a reconstructed Late Pliocene (LP) extent and imposing increased greenhouse gas (GHG) forcing in the Earth System. Using the PRISM4D reconstruction, where ice sheets such as the West Antarctic Ice Sheet (WAIS) were significantly diminished, we conducted multi-centennial simulations with the EC-Earth3 model at atmospheric CO₂ concentrations of 280 and 400 ppmv. The simulation performed with LP ice sheet extent leads to a 9.5 °C rise in surface air temperature, approximately a 16 % reduction in sea ice concentration (SIC) over Antarctica and the Southern Ocean. These changes far exceed those driven by CO₂ increase alone, which result in a 2.5 °C warming and a 9.3 % sea ice decline. Additionally, both experiments deduce there is a reversal in sea level pressure (SLP) polarity with respect to pre-industrial (PI) patterns. Higher-than-normal SLP is present over Antarctica, and lower-than-normal SLP is present in the mid-latitudes, indicative of a negative phase of the Southern Annular Mode (SAM). This is supported by a weakening of the westerly jet, which in turn contributes to the formation of a fresh cap in the upper ocean, induced by the imposed climatic impacts of our sensitivity experiments. This overall freshening of the upper ocean increases stratification in the water column and prevents deep convection in the Southern Ocean, thus leading to the formation of the Antarctic Bottom Water (AABW), which is paramount for the ventilation of the global ocean. Overall, our findings suggest that, by increasing the atmospheric concentration of CO₂, the AABW is suppressed at a multi-centennial timescale; however, by reducing the ice sheet extent, compensatory mechanisms, involving an extensive salinisation of the ocean interior, trigger partial recovery of this water mass. This emphasises the non-linearity of the climate system, since consequences of reducing the ice sheets induce an amplified warming and freshening in the near-surface, whereas they induce opposing mechanisms in the deep ocean that significantly alter the dynamics of water masses that feed the AABW. By isolating the climatic response to ice sheet extent reduction, whilst holding other parameters fixed, this study offers critical insights into the mechanisms driving atmospheric and oceanic variability around Antarctica and their broader implications for global climate dynamics. Here we provide a unique, targeted approach, specifically focusing on the direct impact of ice sheet retreat on regional climate.

With polar amplification warming the northern high latitudes at an unprecedented rate, understanding the future dynamics of vegetation and the associated carbon-nitrogen cycle is increasingly critical. This study uses the dynamic vegetation model LPJ-GUESS 4.1 to simulate vegetation changes for a future climate scenario, generated by the EC-Earth3.3.1 Earth System model, with the forcing of a 560 ppm CO₂ level. Using climate output from an earth system model without coupled dynamic vegetation, to run a higher resolution dynamic vegetation standalone model, allows for a more in depth exploration of vegetation changes. Plus, with this approach, the drivers of high latitude vegetation changes are isolated, but there is still a complete understanding of the climate system and the feedback mechanisms that contributed to it. Our simulations reveal an uneven greening response. The already vegetated Southern Scandinavia and western Russia undergo a shift in species composition as boreal species decline and temperate species expand. This is accompanied by a shift to a carbon sink, despite higher litterfall, root turnover and soil respiration rates, suggesting productivity increases are outpacing decomposition. The previously barren or marginal landscapes of Siberia and interior Alaska/Western Canada, undergo significant vegetation expansion, transitioning towards more stable, forested systems with enhanced carbon uptake. Yet, in the previously sparsely vegetated northern Scandinavia, under elevated CO₂ temperate species quickly establish, bypassing the expected boreal progression due to surpassed climate thresholds. Here, despite rising productivity, there is a shift to a carbon source. The deeply frozen soils in central Siberia resist colonisation, underscoring the role of continuous permafrost in buffering ecological change. Together, these results highlight that CO₂ induced greening does not always equate to enhanced carbon sequestration. The interplay of warming, nutrient constraints, permafrost dynamics and disturbance regimes creates divergent ecosystem trajectories across the northern high latitudes. These findings illustrate a strong need for regional differentiation in climate projections and carbon budget assessments, as the Arctic’s role as a carbon sink may be more heterogeneous and vulnerable than previously assumed.

The impact of mid-Pliocene boundary conditions on Afro-Asian summer monsoon (AfroASM) rainfall is examined using the fully coupled Earth System Model EC-Earth3-LR. Our focus lies on the effects of varying CO₂ concentration, diminished ice sheets and vegetation dynamics. We find that the enhanced AfroASM rainfall is predominantly caused by the “warmer-gets-wetter” mechanism due to elevated CO₂ levels. Additionally, the ice sheet, similar in size to that of the mid-Pliocene era, creates several indirect effects. These include sea ice-albedo feedback and inter-hemispheric atmosphere energy transport. Such influences result in the southward shift of Hadley circulation and formation of Pacific-Japan pattern, leading to reduced rainfall in North African and South Asian monsoon regions but increased rainfall in East Asian monsoon region. Interestingly, while dynamic vegetation feedback has a minimal direct effect on AfroASM rainfall, it significantly influences rainfall in the mid-high latitudes of the North Hemisphere by enhancing water vapor feedback.

Sea ice is crucial in regulating the heat balance between the ocean and atmosphere and quintessential for supporting the prevailing Arctic food web. Due to limited and often local data availability back in time, the sensitivity of sea-ice proxies to long-term climate changes is not well constrained, which renders any comparison with palaeoclimate model simulations difficult. Here we compiled a set of marine sea-ice proxy records with a relatively high temporal resolution of at least 100 years, covering the Common Era (past 2k years) in the Greenland–North Atlantic sector of the Arctic to explore the presence of coherent long-term trends and common low-frequency variability, and we compared those data with transient climate model simulations. We used cluster analysis and empirical orthogonal functions to extract leading modes of sea-ice variability, which efficiently filtered out local variations and improved comparison between proxy records and model simulations. We find that a compilation of multiple proxy-based sea-ice reconstructions accurately reflects general long-term changes in sea-ice history, consistent with simulations from two transient climate models. Although sea-ice proxies have varying mechanistic relationships to sea-ice cover, typically differing in habitat or seasonal representation, the long-term trend recorded by proxy-based reconstructions showed a good agreement with summer minimum sea-ice area from the model simulations. The short-term variability was not as coherent between proxy-based reconstructions and model simulations. The leading mode of simulated sea ice associated with the multidecadal to centennial timescale presented a relatively low explained variance and might be explained by changes in solar radiation and/or inflow of warm Atlantic waters to the Arctic Ocean. Short variations in proxy-based reconstructions, however, are mainly associated with local factors and the ecological nature of the proxies. Therefore, a regional or large-scale view of sea-ice trends necessitates multiple spatially spread sea-ice proxy-based reconstructions, avoiding confusion between long-term regional trends and short-term local variability. Local-scale sea-ice studies, in turn, benefit from reconstructions from well-understood individual research sites.

This study investigates the climatic response of the Arctic to key factors that could shape future climate scenarios: significantly reduced ice sheets, changes in vegetation, and elevated CO2 levels. Using the EC-Earth3.3 Earth system model (ESM), we explore the effects of these forcings under conditions reminiscent of the mid-Pliocene, a key reference for potential future warm climates. Our results reveal that the Arctic climate response varies significantly with different CO2 levels, primarily due to feedbacks involving sea ice and surface albedo. The effects of the reduced ice sheet are, in a pre-industrial CO2 environment (280 ppm), an Arctic warming of 2.4°C. This is driven by substantial sea ice loss in the Barents Sea, which reduces surface albedo. Surprisingly, at 400 ppm CO2, Arctic warming incurred from the ice sheet reduction is lower than expected, at 1.9°C, because sea ice loss is less pronounced compared to pre-industrial conditions, leading to smaller albedo changes. At 560 ppm, the warming is more substantial (2.9°C) but still less than expected, largely due to the already reduced sea-ice extent at this high CO2 level. Vegetation changes further modulate Arctic climate dynamics. At 400 ppm CO2, the expansion of needleleaf evergreen trees decreases surface albedo, adding an additional 0.5°C of warming. However, at 560 ppm CO2, the warming effect of vegetation growth was muted (0.3°C) due to the development of a more diverse canopy with brighter deciduous species, which mitigates the albedo-driven warming. Our findings underscore the complex interplay between CO2 levels, sea ice, and vegetation in determining in Arctic climate dynamics. They highlight that the importance of maintaining CO2 levels at or below 400 ppm to moderate Arctic warming effectively. This study emphasizes the value of integrating paleoclimate insights into future climate projections and underscores the need for a more detailed examination of feedback mechanisms to enhance the robustness of climate models.

Amassing the available solar energy over the Sahara desert, through the installation of a large-scale solar farm, would satisfy the world's current electricity needs. However, such land use changes may affect the global carbon cycle, possibly offsetting mitigation efforts. Here a fully coupled Earth System model EC-Earth was used to investigate the impact of a Saharan solar farm on the terrestrial carbon cycle, simulated with prescribed reduced surface albedo approximating the albedo effect of photovoltaic solar panels over the Sahara desert. The resulting changes to the carbon cycle were an enhancement of the carbon sink across Northern Africa, particularly around the Sahel but a simultaneous weakening of the carbon sink in the Amazon basin. This is observed through spatial pattern changes to the values of net biome production (NBP), more evident during Northern Hemisphere summer season. NBP changes are contributed by competing responses in the net primary production and heterotrophic respiration rates. These changes to carbon exchange correspond to a wetter and warmer climate occurring in Northern Africa and a drier and warmer climate in the Amazon, with stronger driving effects of precipitation. Due to these coupled responses and complex teleconnections, thorough investigation of remote impacts of solar farms are needed to avoid unintended consequences on the terrestrial carbon cycle.

A significant multi-centennial climate variability with a distinct peak at approximately 200 years is observed in a pre-industrial (PI) control simulation using the EC-Earth3-LR climate model. This oscillation originates predominately from the North Atlantic and displays a strong association with the Atlantic Meridional Overturning Circulation (AMOC). Our study identifies the interplay between salinity advection feedback and vertical mixing in the subpolar North Atlantic as key roles in providing the continues internal energy source to maintain this multi-centennial oscillation. The perturbation flow of mean subtropical-subpolar salinity gradients serves as positive feedback to sustain the AMOC anomaly, while the mean advection of salinity anomalies and the vertical mixing or convection acts as negative feedback, constraining the AMOC anomaly. Notably, this low-frequency variability persists even in a warmer climate with weakened AMOC, emphasizing the robustness of the salinity advection feedback mechanism.

The Indian Summer Monsoon (ISM) plays a vital role in the livelihoods and economy of those living on the Indian subcontinent, including the small, mountainous country of Bhutan. The ISM fluctuates over varying temporal scales and its variability is related to many internal and external factors including the El Nino Southern Oscillation (ENSO) and the Indian Ocean Dipole (IOD). In 2015, a Super El Nino occurred in the tropical Pacific alongside a positive IOD in the Indian Ocean and was followed in 2016 by a simultaneous La Nina and negative IOD. These events had worldwide repercussions. However, it is unclear how the ISM was affected during this time, both at a regional scale over the whole ISM area and at a local scale over Bhutan. First, an evaluation of data products comparing ERA5 reanalysis, TRMM and GPM satellite, and GPCC precipitation products against weather station measurements from Bhutan, indicated that ERA5 reanalysis was suitable to investigate ISM change in these two years. The reanalysis datasets showed that there was disruption to the ISM during this period, with a late onset of the monsoon in 2015, a shifted monsoon flow in July 2015 and in August 2016, and a late withdrawal in 2016. However, this resulted in neither a monsoon surplus nor a deficit across both years but instead large spatial-temporal variability. It is possible to attribute some of the regional scale changes to the ENSO and IOD events, but the expected impact of a simultaneous ENSO and IOD events are not recognizable. It is likely that 2015/16 monsoon disruption was driven by a combination of factors alongside ENSO and the IOD, including varying boundary conditions, the Pacific Decadal Oscillation, the Atlantic Multi-decadal Oscillation, and more. At a local scale, the intricate topography and orographic processes ongoing within Bhutan further amplified or dampened the already altered ISM.