X img





作者投稿 专家审稿 编辑办公 编委办公 主编办公

冰川冻土 ›› 2020, Vol. 42 ›› Issue (3): 734-744.doi: 10.7522/j.issn.1000-0240.2019.1049

• 冰冻圈与全球变化 • 上一篇    下一篇


尹鹏1,2(), 王常颖1, 杨俊钢2()   

  1. 1.青岛大学 数据科学与软件工程学院,山东 青岛 266071
    2.自然资源部 第一海洋研究所,山东 青岛 266061
  • 收稿日期:2019-03-20 修回日期:2019-09-09 出版日期:2020-10-31 发布日期:2020-12-08
  • 通讯作者: 杨俊钢 E-mail:yinpengcom@163.com;yangjg@fio.org.cn
  • 作者简介:尹鹏(1996 - ), 男, 山东泰安人, 2017年在青岛大学获学士学位, 现为青岛大学在读硕士研究生, 从事机器学习与大数据分析研究. E-mail: yinpengcom@163.com
  • 基金资助:

Comparison and assessment of long-time series sea ice concentration remote sensing datasets in the Arctic

Peng YIN1,2(), Changying WANG1, Jungang YANG2()   

  1. 1.Data Science and Software Engineering College,Qingdao University,Qingdao 266071,Shandong,China
    2.The First Institute of Oceanography,Ministry of Natural Resources,Qingdao 266061,Shandong,China
  • Received:2019-03-20 Revised:2019-09-09 Online:2020-10-31 Published:2020-12-08
  • Contact: Jungang YANG E-mail:yinpengcom@163.com;yangjg@fio.org.cn


基于空间分辨率和精度更高的BLM海冰密集度数据集, 评估了两种时间跨度较长、 应用广泛的25 km分辨率海冰密集度遥感数据集——NSIDC数据集和SICCI数据集。两种数据集与BLM数据集的海冰面积变化趋势相同, 但均低于基于BLM数据集得到的海冰面积, 其中基于SICCI数据集得到的海冰面积更接近BLM数据集。相比于NSIDC数据集, SICCI数据集的年、 月平均和日海冰面积偏差分别低81.88%、 80.90%、 81.44%, 且其海冰密集度平均偏差为-3.28%, 低于NSIDC数据的4.36%, 因此在进行北极地区整体海冰面积及海冰密集度分析时应选用SICCI数据集。按纬度、 海冰密集度值分情况对两种数据进行比较, 发现NSIDC数据集对开阔水域和浮冰区的区分效果较差, 其在低纬度和低密集度区域的平均偏差分别为10.11%和13.13%, 而SICCI数据集的平均偏差达到0.05%和0.44%, 是研究低纬度和中低海冰密集度区域的首选数据。与之相对, NSIDC数据集对中高纬度高海冰密集区域, 特别是近北极点区域的反映能力优于SICCI数据集, 平均偏差为1.08%, 均方根偏差为7.76%, 因此进行中高纬度高海冰密集度区域分析时首选NSIDC数据集。对两类数据集在北极东北航道上的分段评估结果发现, 低纬度海冰边缘地带或中低海冰密集度区域占比较高的航段区, SICCI数据集更接近BLM数据集, 这些航段应使用SICCI数据集进行分析; 而在中高纬度高海冰密集度区域占比较高的航段区, NSIDC数据集更加贴合, 应为首选数据集。

关键词: 海冰密集度, 长时间序列, 北极海冰, 数据评估


Arctic sea ice plays a very important role in global climate change and sea-ice concentration (SIC) is a crucial parameter for sea ice monitoring. The accuracy of SIC data is an essential basis for Arctic research. Therefore, comparing and assessing the products of sea ice data retrieved from different satellite observations are necessary. The results of existing research indicate that the AMSR2/ASI is a best dataset for SIC quantity studies and real-time shipping guide. However, the temporal coverage of these datasets is relatively short. There are some limits in the study of long-time series. In this paper, the SICCI dataset released by ESA SICCI and the NSIDC dataset released by NSIDC were compared and assessed using the BLM dataset released by Bremen University. The results showed that sea ice area obtained by the three datasets all had significant trends of reduction during 2003 - 2010 and 2013 - 2015. Sea ice concentration and sea ice area of SICCI are lower than BLM and those of NSIDC are higher than BLM. Compared with the NSIDC dataset, the bias of annual and monthly mean and daily sea ice area of SICCI are lower about 81.88%, 80.90%, 81.44%, respectively. The bias of SIC between SICCI and BLM is -3.28%, which is lower than 4.36% of NSIDC. Hence, the SICCI dataset is the best dataset for SIC quantity studies and Arctic sea ice area study. Latitude mean comparisons demonstrate that the SICCI dataset is able to successfully detect the small ice floe area near the continent. The bias of SIC between SICCI and BLM are 0.05% and 0.44% in the region of low latitude and low SIC, respectively, so the SICCI dataset is the best dataset to study sea ice in the regions of low-mid latitude or low SIC. In contrast, NSIDC has the lowest bias of 1.08% and root-mean-squared error (RMSE) of 7.76% in the regions near North Polar. Therefore, it is the most suitable for study in Arctic Polar and should be used for further study in the regions of middle and high latitudes. In addition, quantitative evaluation via BLM indicates that the SICCI dataset is suitable for study sea ice in the shipping route of low latitude or low SIC areas, and the NSIDC dataset performs well in the shipping route of high latitude or high SIC area.

Key words: sea ice concentration (SIC), long-time series, Arctic sea ice, data assessment


  • P731.32