Leveraging Data Assimilation for Accurate Sea Ice State Prediction

The Arctic is warming at a rate four times faster than the global average, as highlighted by Rantanen et al. (2022). This accelerated warming is evident in the rapid decline of sea ice, which not only serves as a critical indicator of climate change but also drives further warming through the ice albedo feedback loop. As sea ice diminishes, the Earth's reflectivity decreases, leading to greater absorption of solar energy by the ocean, which in turn accelerates ice melt and amplifies Arctic warming. The decline in Arctic sea ice has far-reaching implications, including impacts on national security, industrial development, and biodiversity. Accurately predicting future Arctic sea ice states, especially during the September sea ice minimum, is therefore a critical objective within the sea ice modeling community. A perfect model experiment within Icepack, a one-dimensional single-column ice model derived from the comprehensive CICE sea ice model, has been conducted to explore the potential of assimilating sea ice albedo to improve estimation of the mean sea ice state. In this study, one ensemble member of a provisional ensemble without assimilation designated as a hypothetical ‘truth’, and the albedo from this truth member is assimilated into a second ensemble to better align the second ensemble mean sea ice state with that of the truth. Data assimilation was performed using the Data Assimilation Research Testbed (DART), which employs the Ensemble Adjustment Kalman Filter (EAKF) to directly solve the forward problem, retrieving mean and analysis perturbations. The model then propagates background error forward in time, producing multiple realizations of analysis states. This approach was used to evaluate the significance of remote sensing for future ice albedo observations. Sea ice albedo can be categorized into three main types: snow-covered (high albedo), bare ice (medium albedo), and melt ponds (low albedo due to the pooling of liquid water). The study found that albedo critically enhances predictions of future sea ice states when compared to commonly assimilated variables such as sea ice concentration and thickness. As a result, increasing the frequency and spatial distribution of albedo observations in the Arctic is essential for improving predictions of mean sea ice state under scenarios of increased greenhouse gas forcing. Enhanced observation networks will provide more accurate data, leading to better forecasting and a deeper understanding of sea ice dynamics in a changing climate.