Abstract Dynamical climate models have been extensively used over the last decade to perform seasonal sea ice predictions in the context of ensemble forecasting. To date, the sensitivity to the initial conditions has received the most attention through the evaluation of the theoretical limit of Antarctic sea ice predictability imposed by the chaotic evolution of the climate system. The respective contributions of perturbed ocean–sea ice initial conditions and perturbed atmospheric boundary conditions to this predictability remains unevaluated, though. Using the coupled ocean–sea ice model NEMO3.6–LIM3, we developed a suitable framework for evaluating both influences on seasonal Antarctic sea ice predictions. This study reveals that the uncertainty associated with the evolution of the atmospheric conditions is a major limitation to the realization of skilful sea ice extent (SIE) predictions. However, it has a limited impact on sea ice volume (SIV) predictions. The discrepancies between the SIE and SIV predictabilities have been attributed to the presence of very thin ice, which accounts for much of the SIE variability in winter. We also demonstrated that an incorrect estimate of the ocean–sea ice initial conditions has a weaker, but not negligible influence. The time evolution of different plausible ocean–sea ice initial conditions under perfect knowledge of the atmospheric conditions suggest that the initial SIE errors cannot be totally reduced by the atmospheric forcing. They even increase during the melt season. The high persistence of the SIV anomalies is found to be responsible for this behaviour. These findings imply that a correct initialization of the sea ice thickness (SIT) might be more important than previously thought for seasonal Antarctic sea ice predictions, especially if one wants to predict the SIE during the melt season.

中文翻译：

---

扰动的大气和海-海冰初始条件对南极季节性海冰预测技巧的影响：NEMO3.6-LIM3研究

摘要 动力气候模型在过去十年中被广泛用于在集合预报的背景下进行季节性海冰预测。迄今为止，通过评估气候系统混沌演化所施加的南极海冰可预测性的理论极限，对初始条件的敏感性受到了最多的关注。然而，扰动的海洋-海冰初始条件和扰动的大气边界条件对这种可预测性的各自贡献仍未得到评估。使用耦合的海-海冰模型 NEMO3.6-LIM3，我们开发了一个合适的框架来评估这两种对季节性南极海冰预测的影响。这项研究表明，与大气条件演变相关的不确定性是实现熟练海冰范围 (SIE) 预测的主要限制。然而，它对海冰体积 (SIV) 预测的影响有限。SIE 和 SIV 可预测性之间的差异归因于非常薄的冰的存在，这是冬季 SIE 变化的主要原因。我们还证明了对海洋-海冰初始条件的错误估计具有较弱但不可忽略的影响。在完全了解大气条件的情况下，不同可能的海洋-海冰初始条件的时间演变表明，大气强迫不能完全减少初始 SIE 误差。它们甚至在融化季节会增加。发现 SIV 异常的高度持久性是造成这种行为的原因。这些发现意味着，正确初始化海冰厚度 (SIT) 可能比以前认为的季节性南极海冰预测更重要，特别是如果人们想在融化季节预测 SIE。