Passive microwave remote sensing provides unique pan-Arctic light- and cloud-independent daily coverage of lead fraction (LF) for Arctic winter and spring. In this study, we conducted a quantitative assessment of various sea ice concentration (SIC) data products and LF retrieval algorithms to evaluate their accuracy in deriving lead fractions at both overall and pixel-wise levels. Our results indicate that SIC data products are not sensitive to refrozen leads in winter but tend to display clear lead structures in spring. However, the absolute SIC values differ significantly from LF and cannot be directly used as a proxy. As for the LF retrieval algorithms, we proved that the overall accuracy can be largely improved by adjusting upper tie-points. To further minimize errors, we developed an Artificial Neural Network model that outperformed conventional approaches at the pixel-wise level, offering a more reliable estimation method for absolute fraction values.

中文翻译：

---

评估来自被动微波图像的铅分数并改进像素级估计


被动微波遥感为北极冬季和春季提供了独特的泛北极光和云的铅分数 （LF） 日覆盖率。在这项研究中，我们对各种海冰浓度 （SIC） 数据产品和 LF 检索算法进行了定量评估，以评估它们在整体和像素水平上推导铅分数的准确性。我们的结果表明，SIC 数据产品在冬季对再冻结的铅不敏感，但在春季往往显示出清晰的铅结构。但是，绝对 SIC 值与 LF 值有很大不同，不能直接用作代理。至于 LF 检索算法，我们证明通过调整上连接点可以大大提高整体准确性。为了进一步减少误差，我们开发了一个人工神经网络模型，该模型在像素级别上优于传统方法，为绝对分数值提供了更可靠的估计方法。