Since 2016, Antarctic sea ice area (SIA) has set three record summertime minimums occurring in 2017, 2022, and 2023. These recent extremes culminated in a record low SIA anomaly in austral winter 2023, when SIA was over 2 million km^2 below climatology, resulting from negative sea ice concentration (SIC) anomalies centered in the Ross, eastern Weddell, and East Antarctic Seas. We show that a fully-coupled Earth System Model run with historic and anthropogenic forcing and nudged to observed winds over 1950-2023 reproduces the observed austral winter 2023 SIC anomalies. In a sensitivity test that had the impact of 2023 ENSO conditions removed via regression, we show that the 2023 transition from La Niña to El Niño had a negligible impact. Next, using an ensemble initialized on January 1st 2023 nudged to past years' winds, we demonstrate that ~70% of the total Antarctic SIA anomaly was predictable six months in advance and driven by warm Southern Ocean sea surface temperatures that developed prior to 2023, with the remaining ~30%  attributable to 2023 atmospheric circulation. Finally, an ensemble forecast suggests that Antarctic SIA is likely to remain significantly below climatology in austral winter 2024 due to continuing warm Southern Ocean conditions.

Photosynthetic eukaryotic algae survived the Neoproterozoic Snowball Earth events, indicating that liquid-water refugia existed somewhere on the surface. We examine the potential for refugia at the coldest time of a snowball event, before CO2 had risen and with high-albedo ice on the frozen ocean, before it became darkened by dust deposition. We use the Community Earth System Model to simulate a “modern” Snowball Earth (i.e., with continents in their current configuration), in which the ocean surface has frozen to the equator as “sea glaciers”, hundreds of meters thick, flowing like ice shelves. Despite global mean surface temperatures below -60°C, some areas of the land surface reach above-freezing temperatures because they are darker than the ice-covered ocean. With low CO2 (10 ppm) and land-surface albedo 0.4 (characteristic of bright sand-deserts), 0.1 percent of the land surface could host liquid water seasonally; this increases to 12 percent for darker land of albedo 0.2, characteristic of polar deserts. Narrow bays intruding from the ocean to these locations (such as the modern Red Sea) could provide a water source protected from sea-glacier invasion, where photosynthetic life could survive. The abundance of potential refugia increases more strongly in response to reducing the land albedo than to increasing the CO2, for the same global radiative forcing.

The lapse-rate feedback is the dominant driver of stronger warming in the Arctic than the Antarctic in simulations with increased CO2. While Antarctic surface elevation has been implicated in promoting a weaker Antarctic lapse-rate feedback, the mechanisms in which elevation impacts the lapse-rate feedback are still unclear. Here we suggest that weaker Antarctic warming under CO2 forcing stems from shallower, less intense climatological inversions due to limited atmospheric heat transport above the ice sheet elevation and elevation-induced katabatic winds. In slab ocean model experiments with flattened Antarctic topography, stronger climatological inversions support a stronger lapse-rate feedback and annual-mean Antarctic warming comparable to the Arctic under CO2 doubling. Unlike the Arctic, seasonality in warming over flat Antarctica is mainly driven by a negative shortwave cloud feedback which exclusively dampens summer warming, with a smaller contribution from the winter-enhanced lapse-rate feedback.

The retreat of Arctic sea ice is enabling increased ocean surface wave activity at the sea ice edge, yet the physical processes governing interactions between waves and sea ice are not fully understood. Here, we use a collection of in situ observations of waves in ice to evaluate a recent global climate model experiment that includes coupled interactions between ocean waves and the sea ice floe size distribution. Observations come from subsurface moorings and free-drifting buoys spanning 2012-2019 in the Beaufort Sea, and we group the data based on distance inside the ice edge for comparison with model results. Locally generated wind waves are relatively prevalent in observations beyond 100 km inside the ice but are absent in the model. Low-frequency swell, however, is present in the model, while subsurface moorings located more than 100 km inside the ice do not report any swell with significant wave height exceeding the instruments' detection limits. These results motivate further model development and future observing campaigns, suggesting that local wave generation inside the ice edge may play a significant role for floe fracture while demonstrating a need for more robust constraints on wave attenuation by sea ice.

The retreat of Arctic sea ice coincides with increased ocean surface wave activity, and wave-ice interactions are consequently poised to have a growing influence on the Arctic climate system. Recent field campaigns have focused on rectifying the scarcity of wave measurements inside the marginal ice zone, and work is now underway to incorporate wave-ice interactions in global climate models. Here, we apply a collection of in situ wave observations spanning multiple years in the Beaufort Sea and including wave activity beyond 100 kilometers inside the sea ice edge. To better understand waves in the presence of sea ice, we connect the in situ data with satellite-derived ice concentrations across the Arctic and compare the observations with a recent global climate model experiment that includes coupled interactions between waves and a sea ice floe size distribution. We present a series of comparisons focused on wave energy and wind-wave relationships in partial ice cover. These analyses provide a framework for assessing the impact of uncertainty in wave-ice physics on the marginal ice zone in new experiments in the coupled wave-ice model. Our work guides further model development and future observational campaigns.

The lateral size of sea ice floes is receiving increasing attention as an important variable for the polar climate system. We have developed a model for prognostic evolution of the floe size distribution, which emerges due to five key physical processes: new ice formation, welding of floes in freezing conditions, lateral growth and melt, and fracture of floes by ocean surface waves. As a result of the model’s foundation in the governing physics, free parameters occurring in the equations can be directly constrained by observations. Initial model experiments provided insight into the relative importance of various processes, showing floe freezing processes were particularly important for simulation of the floe size distribution. Previously, physical descriptions of lateral growth and welding together of floes had not been constrained by observations. This motivated an analysis of images obtained by drifting wave buoys (SWIFTs) deployed in the autumn Arctic Ocean to quantify these processes in-situ for the first time. We separated floe area growth due to welding from that due to lateral expansion, and compared observations to our physical descriptions of the individual processes. We also found a strong limitation on floe sizes imposed by the wave field. These results have been used to inform new physical descriptions of processes important for the sea ice floe size distribution.

Sea ice motions play an important role in the polar climate by transporting pollutants, heat, water and salt as well as changing the ice cover. Numerous physics-based models have been constructed to represent the sea ice dynamical interaction with the atmosphere and ocean. Here we propose a new data-driven deep-learning approach which utilizes a convolutional neural network (CNN) to model how Arctic sea ice moves in response to surface winds given its initial ice velocities and concentration a day earlier. Results show that CNN computes the sea ice response with a correlation of 0.82 on average with respect to reality, which surpasses a set of pixel-based predictions, such as persistence (PS), linear regression (LR), random forest (RF), multiple layer perceptrons (MLP) and CICE, a leading physics-based model. The superior predictive skill of CNN suggests the important role played by the connective patterns of the predictors of the sea ice motion.