Nils Bochow

Historical records of climate fields are often sparse because of missing measurements, especially before the introduction of large-scale satellite missions. Several statistical and model-based methods have been introduced to fill gaps and reconstruct historical records. Here, we use a recently introduced deep learning approach based on Fourier convolutions, trained on numerical climate model output, to reconstruct historical climate fields. Using this approach, we are able to realistically reconstruct large and irregular areas of missing data and to reproduce known historical events, such as strong El Niño or La Niña events, with very little given information. Our method outperforms the widely used statistical kriging method, as well as other recent machine learning approaches. The model generalizes to higher resolutions than the ones it was trained on and can be used on a variety of climate fields. Moreover, it allows inpainting of masks never seen before during the model training.

Melting of the Greenland ice sheet (GrIS) in response to anthropogenic global warming poses a severe threat in terms of global sea-level rise (SLR)1. Modelling and palaeoclimate evidence suggest that rapidly increasing temperatures in the Arctic can trigger positive feedback mechanisms for the GrIS, leading to self-sustained melting2–4, and the GrIS has been shown to permit several stable states5. Critical transitions are expected when the global mean temperature (GMT) crosses specific thresholds, with substantial hysteresis between the stable states6. Here we use two independent ice-sheet models to investigate the impact of different overshoot scenarios with varying peak and convergence temperatures for a broad range of warming and subsequent cooling rates. Our results show that the maximum GMT and the time span of overshooting given GMT targets are critical in determining GrIS stability. We find a threshold GMT between 1.7 °C and 2.3 °C above preindustrial levels for an abrupt ice-sheet loss. GrIS loss can be substantially mitigated, even for maximum GMTs of 6 °C or more above preindustrial levels, if the GMT is subsequently reduced to less than 1.5 °C above preindustrial levels within a few centuries. However, our results also show that even temporarily overshooting the temperature threshold, without a transition to a new ice-sheet state, still leads to a peak in SLR of up to several metres.

The Amazon rainforest is threatened by land-use change and increasing drought and fire frequency. Studies suggest an abrupt dieback of large parts of the rainforest after partial forest loss, but the critical threshold, underlying mechanisms, and possible impacts of forest degradation on the monsoon circulation remain uncertain. Here, we use a nonlinear dynamical model of the moisture transport and recycling across the Amazon to identify several precursor signals for a critical transition in the coupled atmosphere-vegetation dynamics. Guided by our simulations, we reveal both statistical and physical precursor signals of an approaching critical transition in reanalysis and observational data. In accordance with our model results, we attribute these characteristic precursor signals to the nearing of a critical transition of the coupled Amazon atmosphere-vegetation system induced by forest loss due to deforestation, droughts, and fires. The transition would lead to substantially drier conditions, under which the rainforest could likely not be maintained.