基于最优插值和贝叶斯最大熵的海表温度融合方法研究

Study on the Merging Sea Surface Temperature Data Based on Optimal Interpolation and Bayesian Maximum Entropy Method

海表温度(sea surface temperature,SST)是影响全球气候的重要因素,在海洋科学研究中占有关键位置。论文基于MODIS红外、AMSR-2和HY-2A微波辐射计数据,分别利用最优插值和贝叶斯最大熵方法对SST数据进行融合,并用i Quam实测数据和Argo浮标数据对2015年SST融合数据进行检验。MODIS、AMSR-2、HY-2A辐射计SST的年平均空间覆盖率分别为15.0%,21.6%,22.0%,最优插值和贝叶斯最大熵融合SST产品的年平均空间覆盖率提高到98.6%和99.4%,融合产品空间覆盖率明显提高。与i Quam实测数据对比,最优插值和贝叶斯最大熵融合产品年平均偏差分别为0.07℃,0.04℃,均方根误差皆为0.78℃,其中3-7月最优插值融合产品的精度略优于贝叶斯最大熵融合产品,其它月份则相反;与Argo浮标数据对比,两种融合产品的均值偏差分别为0.06℃,0.01℃,均方根误差分别为0.77℃,0.75℃。整体上,贝叶斯最大熵融合产品的精度略优于最优插值融合产品,但计算成本较高。

Sea surface temperature（SST） is an important factor affecting global climate, and constitutes a key position in oceanographic research. In this study, the SST datasets, derived from MODIS, AMSR-2 and HY-2A radiometer, are merged by using the optimal interpolation（OI） and Bayesian maximum entropy（BME） method,respectively. The merged SSTs over the whole year in 2015 are validated by using the i Quam and Argo SST data.The average daily coverage of MODIS, AMSR-2 and HY-2A radiometer is 15.0%, 21.6% and 22.0%, respectively,while the average daily coverage of the merged SST based on OI and BME is 98.6% and 99.4%, respectively.Compared with single sensor SST, the spatial coverage of the two merged SSTs is significantly improved. In contrast with i Quam SST, the bias of the merged SSTs based on optimal interpolation and Bayesian maximum entropy is 0.07 ℃ and 0.04 ℃, respectively, while the RMSE are both 0.78 ℃, of which from March to July the merged SSTs based on OI show better accuracy than those based on BME, with the other months on the contrary.Compared with the Argo buoy SST data, the bias of the two merged SSTs is 0.06 ℃ and 0.01 ℃, respectively,while the RMSE are 0.77 and 0.75. Overall, the accuracy of the merged SSTs based on BME is slightly better than that based on OI, but meanwhile the computational cost of the BME method is higher.