Extreme Arctic sea ice lows investigated with rare event sampling
(2026)
Files
Thesis_JeromeSauer_2026.pdf
Open Access - Adobe PDF
- 30.99 MB
Details
- Authors
- Supervisors
; Francesco Ragone- Abstract
- The Arctic sea ice cover and sea ice volume have been decreasing since at least the late 1970s, in large part due to anthropogenic emissions of greenhouse gases. Despite steadily increasing greenhouse gas concentrations, the sea ice decline is non-linear, and internal climate variability and feedback mechanisms modulate the downward trend of the sea ice. Accelerated sea ice decline occurred from the mid-2000s to 2012, including record reductions of the sea ice area compared to the trend line in 2007 and 2012. The decline and extreme fluctuations of Arctic sea ice cover can impact the integrity of the permafrost, coastal erosion, mid- and high-latitude weather and climate, and have implications for Arctic Ocean accessibility. The latter concerns trans-Arctic and local shipping, the exploitation of natural resources, fishing, polar ecotourism and daily activities of local communities. Being able to predict summer sea ice conditions on seasonal time scales, in particular on a regional level, is therefore of potential interest for different users. Various studies identified possible drivers of extremes of Arctic sea ice reduction, such as observed in the summers of 2007 and 2012, including sea ice-ocean preconditioning, oceanic heat transport, large-scale atmospheric circulation variability and synoptic-scale cyclones. However, a robust quantitative statistical analysis of extremes of sea ice reduction is hindered by the small number of rare events that can be sampled from observations and numerical simulations with computationally expensive climate models. A better understanding of the precursors of extreme sea ice lows and a more precise estimate of their probabilities are crucial to improve seasonal predictions of these events and to quantify their risk of occurrence under different climate change scenarios. The access to an improved statistics of extreme sea ice lows could also help to better assess their impacts. Recent studies tackled the problem of sampling climate extremes by using rare event algorithms, i.e., computational techniques developed in statistical physics to reduce the computational cost required to sample rare events in numerical simulations. While preserving the dynamical consistency of the model, these techniques allow to increase the number of simulated extremes by several orders of magnitude compared to direct numerical simulations for a given computational cost, and to generate ultra-rare events that are almost impossible to study using direct sampling. In this thesis, we apply a genealogical selection rare event algorithm to ensemble simulations with the coupled climate models Planet Simulator (PlaSim) and European community Earth-System-Model version 3 (EC-Earth3) to study extremes of summer pan-Arctic sea ice area reduction under present-day climate conditions. The specific algorithm used in this thesis is designed to improve the sampling of extremes whose dynamics is characterized by time-persistency. We apply this technique via two approaches. Firstly, we initialize ensembles with the algorithm from statistically independent initial conditions taken from a long control run with PlaSim. This approach serves to oversample the areas of the phase space where extreme events in an absolute sense are. We simulate extremely low sea ice summers with return times up to 100 000 years for a computational cost of 1000 years, and we compute statistically significant composite maps of dynamical quantities conditional on the occurrence of extremes with return times of more than 200 years. Using both PlaSim and EC-Earth3, the second approach corresponds to a seasonal climate prediction setup where ensembles are initialized from slightly perturbed identical initial conditions. For both models, we generate late summer sea ice lows with larger amplitudes compared to the deviation of the observed 2012 late summer sea ice area from the trend line. Using EC-Earth3, we also compute statistically significant composite maps of dynamical quantities during extremes of intra-seasonal pan-Arctic sea ice area reduction with estimates probabilities of less than 1% per year. We exploit the improved statistics of extremes of sea ice reduction to perform a quantitative statistical analysis of their precursors. This includes sea ice volume, sea ice area and surface energy budget analyses to disentangle the roles of dynamic vs. thermodynamic forcing for these events. Initialized ensemble simulations also allow us to assess the relative contributions of winter sea ice-ocean preconditioning vs. sub-seasonal weather variability to the amplitudes of extremely negative late summer sea ice area anomalies computed relative to a fixed baseline climatology. Our manuscript demonstrates that sources of probabilistic predictability of extremely low late summer sea ice area are given by 1) the late winter cumulative area with sea ice thickness larger than a certain threshold and 2) by systematically enhanced May-July sea ice volume loss prior to the annual sea ice minimum. We find that the latter is systematically associated with persistent negative mean sea level pressure anomalies over the central and eastern Arctic Ocean. The associated cyclonic atmospheric flow promotes enhanced sea ice loss both thermodynamically due to convergence of vertically integrated moisture and sensible heat transport over the Pacific-North American part of the Arctic Ocean and dynamically by favoring Ekman drift of sea ice out of the central Arctic Ocean towards its marginal seas. Finally, the methodological, technical and scientific achievements provided in this thesis open several future perspectives. These perspectives concern in particular attribution studies and the improvement of seasonal predictions of summer sea ice conditions using data-based statistical models, machine learning-based forecasts and climate model ensemble-based initialized forecasts. While this manuscript exclusively deals with the pan-Arctic sea ice area, the developed strategies can be adjusted to study anomalous sea ice conditions in specific regions of the Arctic and in Antarctica.
- Affiliations
Citations
Sauer, J. (2026). Extreme Arctic sea ice lows investigated with rare event sampling. https://hdl.handle.net/2078.5/278983