Heatwaves are among the deadliest natural hazards, yet their evolving atmospheric drivers remain unclear. We apply an unsupervised machine learning model to reanalysis data to identify multivariate heatwave patterns over Europe. The model captures seasonally distinct regimes and reveals significant shifts between historical and recent events across all seasons, highlighting changes in heatwave dynamics over time.

The El Niño-Southern Oscillation (ENSO) drives short-term global plant changes, but its influence on decadal changes is not fully known. Using global climate models, we found that plants show pronounced ENSO-driven changes on decadal and longer timescales. Slow plant responses are influenced by soil water content and plant dynamics but only have a weak effect on the plant’s net carbon uptake. Our work shows that vegetation memory could be an important source of decadal climate predictability.

Decadal forecasts use models to predict near-term natural and human-induced climate changes. Large scale pressure patterns affect surface impacts, but in summer they are hard to predict. North Atlantic surface temperatures are known to influence these patterns in summer. We find that a large set of retrospective decadal forecasts can predict this ocean-atmosphere interaction but also underestimates it, consistent with other predictable North Atlantic pressure patterns in climate forecasts.

This research investigates the interactions between the Atlantic Meridional Overturning Circulation (AMOC) and the Antarctic Ice Sheet, which are both recognized as critical climate tipping elements in the Earth system. We conduct a computer simulation with artificial freshwater input in the North Atlantic to collapse the AMOC for 1500 years. Our findings show no destabilization of the Antarctic Ice Sheet induced by changes in the Southern Ocean during this period.

We use artificial intelligence to learn simple equations from historical climate data that describe how North Atlantic ocean temperature, air pressure and rainfall vary and influence each other over decades. Analysing the model's behaviour and equation terms, we find rainfall strongly feeds back on both the ocean and the atmosphere. These interactions are well captured by the models and allow rainfall to be predicted over the ocean, and nearby regions such as Europe over coming decades.

We present a first catalogue of abrupt changes/state transitions in ocean, sea-ice, and atmospheric variables in future Coupled Model Intercomparison Project Phase 6 climate scenarios. We find many models that show Arctic sea-ice disappearance, North Atlantic winter mixed-layer collapse and/or subsequent transition of the Atlantic Meridional Overturning Circulation to a weak state. We find more abrupt changes/state transitions than in previous work at lower global warming levels.

We present climate model simulation of stratospheric aerosol injection, a proposed method of climate intervention where reflective particles are placed in the stratosphere to cool the planet. We design our simulated interventions to maintain certain surface temperature patterns, and show that the same temperatures can be maintained in different ways because of how the ocean transports heat. This has implications for how climate intervention model experiments are designed in the future.

We investigated a response of permafrost using an Earth system model driven by an idealized overshooting carbon emission scenario, and found that the permafrost response to warming and cooling is reversible in the area but irreversible in its property. The permafrost response shows hysteresis, arising from a slow soil response tied to heat conductivity and specific heat of water phase change. A carbon release from thawed permafrost accounts for 0.6–41% of the cumulative carbon emission.