Introduction

This research addresses the drivers of global climate change, focusing on anthropogenic emissions of greenhouse gases and aerosols. The study highlights that warming is most significant in the Arctic, where temperatures are increasing more than three times as fast as the global mean. This trend is closely linked to a substantial decline in Arctic sea ice, with the annual mean area decreasing by approximately 2 million km² since 1980. Climate models within the Coupled Model Intercomparison Project Phase 6 (CMIP6) agree that sea ice will continue to decline, though the exact timing of a summer ice-free Arctic remains uncertain.

Large-scale variations in North Atlantic temperatures are intrinsically tied to the Atlantic Meridional Overturning Circulation (AMOC). While climate models suggest the AMOC will weaken in the future, they disagree on the rate of decline. This study utilizes a high-resolution version of the EC-Earth3 global coupled climate model (EC-Earth3-HR) to provide a more realistic representation of heat fluxes, topography, and the observed pathways of warm Atlantic water entering the Arctic.

Model, tuning and simulations

The EC-Earth3-HR model features a horizontal resolution of approximately 40 km (T511 spectral resolution) in the atmosphere with 91 vertical levels, and a 0.25° resolution in the ocean with 75 layers. This is significantly more detailed than the standard-resolution version (EC-Earth3-SR), which uses approximately 80 km in the atmosphere and a 1° ocean grid. The model underwent a rigorous tuning process targeting top-of-atmosphere (TOA) radiative imbalance and reduced global-mean biases in net radiative flux. Five specific atmospheric parameters were adjusted, including the rate of conversion of cloud water to rain (RPRCON) and the fall speed of ice particles (RVICE).

Ocean tuning involved testing advection schemes, with the Total Variations Diminishing (TVD) approach selected for its long-term stability. The tuning target for AMOC strength was set in the range of 17–18 Sv, consistent with observational estimates. The researchers performed a 350-year pre-industrial control simulation, followed by a historical simulation (1850–2014) and a future projection (2015–2100) under the SSP2-4.5 scenario.

General evaluation of the model

Evaluation against ERA5 reanalysis data shows that the EC-Earth3-HR model accurately depicts global mean near-surface temperatures and sea surface temperatures (SST). The high-resolution model better captures observed warming since 1950 compared to the standard-resolution ensemble. However, EC-Earth3-HR maintains a cold bias over land, with temperatures approximately 2 K lower than observed values. Regional differences in sea ice concentration are noted; the model tends to overestimate ice in the Greenland-Iceland-Norway (GIN) and Barents Seas while underestimating it in the Labrador and Bering Seas.

The model captures the AMOC variability of the early 2000s and the rapid loss of Arctic sea ice. It also improves the representation of climate variability modes, such as the North Atlantic Oscillation (NAO) and the El Niño–Southern Oscillation (ENSO), although it still underestimates the intensity of pressure centres in the subtropical and North Atlantic regions.

Projected changes in the Arctic and North Atlantic Oceans under the SSP2-4.5 scenario

Under the SSP2-4.5 scenario, EC-Earth3-HR projects a nearly ice-free Arctic by 2040 in September, defined as sea ice area falling below 1 million km². This occurs earlier than projections from standard-resolution models. By 2050, fully ice-free conditions are expected. The model also simulates a significant weakening of the AMOC, projecting a reduction of approximately 39% (from 21 Sv to 13 Sv) by the end of the 21st century. This is more pronounced than the 25% weakening projected by EC-Earth3-SR.

Surface density changes indicate increased stratification in the North Atlantic, particularly in the Labrador, Irminger, and Greenland Seas. Salinity changes are projected to be non-uniform; while mid and high latitudes generally show freshening, salinification is projected for parts of the Greenland Sea and the eastern Arctic.

Discussion and conclusion

The study introduces a volume-budget framework to diagnose deep-water formation (DWF) and its regional contributions to the AMOC. The results indicate that DWF weakens across all analysed regions. The Labrador Sea is identified as the primary contributor to long-term AMOC weakening, with diagnosed DWF projected to cease entirely in this region by 2100. In comparison, DWF is expected to decline by 62% in the Greenland Sea and 13% in the Irminger Sea.

The researchers conclude that high-resolution modelling is essential for capturing key climate processes in the Arctic–North Atlantic sector. While some biases remain due to parameterisation limitations, the EC-Earth3-HR model provides valuable insights into the trajectory of global climate dynamics and the potential impacts of moderate mitigation strategies.