Impact of the ice thickness distribution discretization on the sea ice concentration variability in the NEMO3.6-LIM3 global ocean–sea ice model

Model output and observational data and scripts corresponding to the manuscript "Impact of the ice thickness distribution discretization on the sea ice concentration variability in the NEMO3.6-LIM3 global ocean–sea ice model" Abstract. This study assesses the impact of different sea ice thickness distribution (ITD) configurations on the sea ice concentration (SIC) variability in ocean-standalone NEMO3.6-LIM3 simulations. Three ITD configurations with different numbers of sea ice thickness categories and boundaries are evaluated against three different satellite products (hereafter referred to as “data”). Typical model and data interannual SIC variability is characterized by k-means clustering both in the Arctic and Antarctica between 1979 and 2014 in two seasons, January–March and August–October, which show the largest coherence across clusters in individual months. Analysis in the Arctic is done before and after detrending the series with a 2nd degree polynomial to separate interannual from longer-term variability. Before detrending, winter clusters capture SIC response to atmospheric variability at both poles and summer cluster a positive and negative trend in the Arctic and Antarctic SIC respectively. After detrending, Arctic clusters reflect SIC response to interannual atmospheric variability predominantly. Model–data cluster comparison suggests that no specific ITD configuration or category number increases realism of the simulated Arctic and Antarctic SIC variability in winter. In the Arctic summer, more thin-ice categories decrease model–data agreement without detrending but increase agreement after detrending. Overall, a single-category configuration agrees the worst with data. Direct model–data comparison of SIC anomaly fields shows that more thick-ice categories improve winter SIC variability realism in Central Arctic regions with very thick ice. By contrast, more thin-ice categories reduce model–data agreement in the Central Arctic in summer, due to an overly large simulated sea ice volume. In summary, whereas better resolving thin ice in NEMO3.6-LIM3 can hamper model realism in the Arctic but improve it in Antarctica, more thick-ice categories increase realism in the Arctic winter. And although the single-category configuration performs the worst overall, no optimal configuration is identified. Our results suggest that no clear benefit is obtained from increasing the number of sea ice thickness categories beyond the current usual standard of 5 categories in NEMO3.6-LIM3.本数据集对应论文《海冰厚度分布离散化对NEMO3.6-LIM3全球海洋-海冰模式中海冰浓度变率的影响》的摘要内容，包含模型输出、观测数据与配套脚本。本研究评估了不同海冰厚度分布（ice thickness distribution, ITD）配置对离线运行的NEMO3.6-LIM3海洋-海冰模式模拟中海冰浓度（sea ice concentration, SIC）变率的影响。 研究针对三种具有不同海冰厚度分类数量与边界的ITD配置，结合三种不同卫星反演产品（下文简称“观测数据”）开展评估。本研究选取1979年至2014年的1-3月与8-10月两个季节为研究时段，通过k-means聚类（k-means clustering）分析北极与南极区域的典型模式与观测数据年际海冰浓度变率，该时段内各聚类在单个月份中呈现出最高的簇间一致性。 北极区域的分析会先基于二阶多项式对序列进行去趋势处理前后分别开展，以此将年际变率与长期变率区分开来。去趋势处理前，冬季聚类可捕捉两极海冰浓度对大气变率的响应；而夏季聚类则分别对应北极与南极海冰浓度的正、负长期趋势。经去趋势处理后，北极区域的聚类则主要反映海冰浓度对年际大气变率的响应。 模式与观测数据的聚类对比结果显示，并无特定的ITD配置或海冰厚度分类数量，能够提升北极与南极冬季模拟海冰浓度变率的真实性。在北极夏季，未进行去趋势处理时，增加薄冰分类数量会降低模式与观测数据的一致性；但经去趋势处理后，二者的契合度会得到提升。总体而言，单分类配置与观测数据的契合度最差。 直接对比海冰浓度异常场的模式与观测数据结果显示，在北极中央区域的厚冰区中，增加厚冰分类数量能够提升冬季海冰浓度变率的模拟真实性。与之相反，北极夏季中央区域的模式与观测数据契合度会因模拟海冰体积过大，而随着薄冰分类数量的增加出现下降。 综上，尽管在NEMO3.6-LIM3中优化薄冰分辨率会损害北极区域的模式真实性，但可提升南极区域的表现；而增加厚冰分类数量则能提升北极冬季的模拟真实性。尽管单分类配置的整体表现最差，但本研究并未识别出最优配置。结果表明，在NEMO3.6-LIM3中，将海冰厚度分类数量提升至当前通用标准（5类）以上，并不会带来明确的收益。