南大洋长时间序列海冰密集度遥感数据产品比较评估

Intercomparison of Long-Term Sea Ice Concentration Remote Sensing Products in the Southern Ocean

摘要: 目前在气候研究中使用的时间序列连续且长度最长的被动微波遥感海冰密集度数据产品主要有美国雪冰数据中心(NSIDC)发布的NASA Team算法数据产品(NSIDC NASA Team数据产品)和Bootstrap算法数据产品(NSIDC Bootstrap数据产品)，以及欧洲气象卫星应用组织(EUMETSAT)海洋海冰卫星应用中心(OSI SAF)发布的气候数据记录产品(OSI SAF数据产品)，这三套长时间序列海冰密集度数据产品都是基于SMMR、SSM/I、SSMIS被动微波传感器，但源数据、反演算法和质量订正方法等方面的差异，可能引起数据产品之间的精度差异。为选取和应用数据产品开展南大洋海冰有关气候问题的研究，本文对上述三套数据产品在南大洋地区进行较长时间的比较评估。以基于AMSR-E、AMSR2被动微波传感器的ASI算法数据产品作为参照数据，分类分析按照“最近距离”原则匹配的格点上的海冰密集度。结果显示，长时间序列数据产品相对参照数据产品表现为漏估和空估的格点主要出现在海冰边缘区，海冰融化期漏估和空估的比例高于海冰冻结期。长时间序列数据产品整体上低估了海冰密集度。NSIDC Bootstrap数据产品相对参照数据产品漏估和空估的总比例低于另外两套评估数据产品，平均偏差的数值和均方根偏差小于另外两套评估数据产品。整体上，NSIDC Bootstrap数据产品与参照数据产品更为接近。

Abstract: At present, the most consistent and longest satellite passive microwave (PM) sea ice concentration (SIC) products for climate studies include the NASA Team algorithm product and the Bootstrap algorithm product distributed by the National Snow and Ice Data Center (NSIDC), as well as the Climate Data Record (CDR) product developed at the EUMETSAT’s Ocean and Sea Ice Satellite Application Facility (OSI SAF). The three long-term PM SIC products are all derived from measured brightness temperatures from the Scanning Multichannel Microwave Radiometer (SMMR), the Special Sensor Microwave/Imager (SSM/I) and the Special Sensor Microwave Imager Sounder (SSMIS), while the precision of the three products may be inconsistent due to the differences in input data, algorithms and quality control methods. In order to select and apply long-term PM SIC products for climate studies related to the sea ice in the Southern Ocean, this paper conducted a comparative assessment of the three PM SIC products on a relatively long time series length in the Southern Ocean. For this purpose, the PM SIC product based on the Advanced Microwave Scanning Radiome-ter for the Earth Observing System (AMSR-E) and the Advanced Microwave Scanning Radiometer 2 (AMSR2) with the ARTIST Sea Ice (ASI) algorithm was used as reference data, and the cell-by-cell comparisons were performed on three categories of co-located data pairs using the near-est-neighbour scheme, including nonsingular sea ice samples, singular open water samples (omis-sive estimation samples) and singular sea ice samples (redundant estimation samples). The results show that omissive estimation samples and redundant estimation samples mainly appear in the marginal ice zone, and the proportion of omissive estimation samples and redundant estimation samples during the melting period is higher than that during the freezing period. For nonsingular sea ice samples, the three long-term PM SIC products overall underestimate SIC relative to the reference product. The proportion of omissive estimation samples and redundant estimation samples for each month in the NSIDC Bootstrap algorithm product is lower than that in the NSIDC NASA Team algorithm product and the OSI SAF CDR product. The NSIDC Bootstrap algorithm product shows the smallest values of mean difference (bias) and root mean square error (RMSE) compared to the other two long-term PM SIC products. The results indicate that the NSIDC Bootstrap algorithm product is closer to the reference product than the other two long-term PM SIC products.