The Arctic is experiencing rapid warming, a phenomenon known as Arctic Amplification, which is intensifying climate change in the Arctic relative to lower latitudes. This warming is significantly impacting Arctic sea ice, leading to a decline in its volume, extent, and age. Paradoxically, winter sea ice production appears to have been increasing, creating a tension between observed warming and increased ice growth. Previous research suggests that further warming will eventually lead to a decrease in ice production, but a comprehensive explanation of this rise-then-fall pattern has been lacking. This study aims to address this gap by investigating the driving factors behind the observed trends in Arctic sea ice production, focusing on the Kara and Laptev seas, major contributors to Arctic sea ice production. Understanding these factors is crucial because sea ice plays a critical role in the Arctic's radiative budget, ocean-atmosphere heat fluxes, and freshwater redistribution. The study's importance lies in its potential to improve our understanding of the complex interactions driving sea ice change and to provide more accurate predictions of future sea ice conditions.

Declining Arctic sea ice extent and thickness, directly linked to anthropogenic carbon emissions, are well-documented. This decline is both a consequence and a driver of Arctic Amplification, with losses more pronounced in summer than winter. Although summer melting losses are partially offset by increased winter ice production, this compensation is insufficient to prevent the overall decline in sea ice. Studies have shown the importance of winter ice production for restoring the ice pack before the onset of polar day, influencing regional radiative budgets and heat fluxes. Sea-ice growth also impacts ocean hydrography through brine rejection and freshwater redistribution. The apparent contradiction between rising winter ice growth and intensified winter Arctic Amplification is likely due to the interplay of multiple factors. Less summer ice leads to thinner winter ice, promoting higher winter growth rates but also increased ocean-to-atmosphere heat fluxes. Previous modeling studies suggest a weakening of the positive correlation between temperature and winter growth, eventually turning negative, indicating warming overwhelming negative feedbacks. This study aims to provide a physical explanation for the observed rise-then-fall behavior of sea ice production by considering these interacting processes.

This study employs a simple linear model to explain the rise and fall of sea ice production in the Kara and Laptev seas, often referred to as "ice factories." The model, informed by the physics of sea ice growth, is trained on internal variability across 40 ensemble members from the CESM-LE, a state-of-the-art global climate model. The 40 ensemble members consist of a 20th-century run (1920-2005) using historical forcing and a high-emissions RCP8.5 run (2006-2080). The model decomposes total ice production into two components: "usage" (freezing area days) and "efficiency" (mean growth rate). The model considers key climate variables impacting these components. Freezing area days are primarily controlled by September sea surface temperature (SST) and October-December surface air temperature. The growth rate is determined by the balance of heat fluxes at the ice base and the conductive heat loss to the atmosphere, influenced by snow and ice thickness, and surface air temperature (ΔT). To capture the influence of open water, created by summer melting and ice divergence, the model includes terms for September open water area, net winter sea ice area diverged, and compensated winter sea ice area diverged. The model’s equation is: Ice Prod = β₁ΔT/hs + β₂ΔTAsep + β₃ΔTAnet + β₄ΔTAcomp + β₅SSTsep. Multiple linear regression is performed to determine the regression coefficients (β) using the internal variability of the CESM-LE data. The model's skill is assessed by the R² value. The model is then applied to observation-based data (ERA5, NASA Goddard, Polar Pathfinder, NOAA OI SSTV2, HADISST, SnowModel-LG) to estimate historical ice production in the Kara-Laptev region.

Analysis of CESM-LE data reveals that the freezing season in the Kara and Laptev seas is shrinking, while the growth rate is initially increasing. The refreezing process, most productive in early winter, occurs later and is more pinched in shape through the 21st century. Total winter ice production in the ensemble mean rises from about 1970 to 2010, then falls from 2020 onward. Decomposition into freezing area days and growth rate shows that freezing area days decline from the 1990s, while the growth rate rises until about 2030, then declines. The linear model successfully explains a large fraction (76%-81%) of internal variability in ice production across the CESM-LE ensemble. Three regressors are dominant: net divergence × ΔT, September open water × ΔT, and inverse snow depth × ΔT. The model also successfully reconstructs the forced rise-then-fall behavior of ice production in the CESM-LE. The increase in ice production until 2020 is mainly due to the increasing September open water area and decreasing snow thickness. After 2020, atmospheric warming dominates, leading to a decline in ice production. September SST is a significant driver of the forced decline. Applying the model to observational data suggests a similar rise-then-fall pattern, with ice production peaking around 2000 and a slight decline observed thereafter. While the September open water area reached a maximum, future warming will likely lead to negative ice production trends from this term. Snow thinning initially counteracts warming until roughly 2030.

The study's findings demonstrate the utility of a simple linear model in capturing the complex interplay of factors driving sea ice production. The model successfully explains both internal variability and forced trends in ice production, highlighting the contributions of negative feedbacks (increased open water area, divergence, reduced snow depth) and the eventual dominance of warming. The peak in ice production around 2020 in the model aligns with observational data, suggesting that this peak is not simply a model artifact. The results suggest that the continued retreat of the September sea ice edge is a critical factor influencing the timing of peak ice production, and that once this edge approaches zero, atmospheric warming will dominate and ice production will begin to fall. The study highlights the crucial role of upper ocean temperatures in delaying the freezing season and the relatively minor impact of ocean-to-ice heat flux on ice growth in this region. This work improves our understanding of sea ice dynamics and its response to climate change, with implications for regional and global climate patterns.

This study presents a simple yet effective linear model that successfully explains both the internal variability and the forced rise-then-fall trend in Arctic sea ice production in the Kara and Laptev seas. The model highlights the competing influences of several negative feedbacks to sea ice loss and the eventual dominance of atmospheric warming and increases in upper ocean temperature. Observational data supports the model’s prediction of a peak in ice production followed by a decline. Future research could investigate spatial variations in sea ice production and incorporate additional factors to refine the model's predictive capabilities. Furthermore, studying the implications of declining sea ice production on the Arctic halocline and biogeochemical cycles is warranted.

The study focuses on the Kara and Laptev seas, limiting the generalizability of the findings to other Arctic regions. The model's reliance on CESM-LE data introduces potential biases inherent to the model itself. The observational data used have their own uncertainties and limitations, especially concerning sea ice thickness measurements. The assumption of similar sensitivity to regressors between the model and observations may not perfectly hold.