基于等纬度DINEOF的遥感SST产品缺失数据重构算法及精度验证分析

Data reconstruction algorithm of remote sensing SST products and accuracy analysis based on the same latitude DINEOF

由于云覆盖、算法失效等原因造成SST遥感数据产品大量缺失，严重影响其在海洋环境监测中的应用．对缺失数据进行重构是提高SST数据产品质量及应用效果的一项重要工作．在已发展的DINEOF（Data Interpolating Empirical Orthogonal Function）基础上提出了一种适用于海量数据、大空间尺度的数据重构改进算法——等纬度正交经验函数重构方法（Same Latitude-DINEOF，SL-DINEOF）．利用此方法对中国近海的MODISSST数据产品进行重构，并与原始的非等纬度方法进行重构精度对比分析．结果表明，基于SL-DINEOF的重构精度呈季节性变化，受SST量值、时间变异及原始数据缺失率3方面影响，其中在南海表现最优（RMSE为0．73l～0．957）；对比可知，在各个海区SL-DINEOF均优于DINEOF，RMSE的最大降幅达21．28％，且纬度越高优势越明显；此外，SL-DINEOF最优通常出现在春秋季．

Due to the cloud cover or invalid algorithm, a large number of missing data of remote sensing SST products seriously impact on the remote sensing SST data application in the marine environmental monitoring. Reconstruction is an important task to improve the quality of SST data products and its application effects. This paper presents an improved reconstruction algorithms same latitude empirical orthogonal function reconstruction method （SL-DINEOF） based on the de veloped Data Interpolating Empirical Orthogonal Function （DINEOF）, which is suitable for massive amounts of data in large spatial scales. The new method can be used for the reconstruction of MODIS SST data products in Chinese offshore, and then the precision comparative analysis between SL-DINEOF and DINEOF is carried out. The results show that there are different seasonal variations on SL-DINEOF reconstruction accuracy, which are affected by the magnitude of SST, SST temporal variability and missing data rate. The performance of SL-DINEOF is better than DINEOF method in all sea areas with the maximum reduce extent of RMSE to 21.28%. The optimal performance of SL-DINEOF accuracy is appeared in South China Sea （RMSE： 0. 731-0. 957）. The advantages are more obvious in the higher latitudes; additionally, the optimal performance of SL-DINEOF method generally appears in spring and autumn.