| 摘要: |
| 海冰密集度数据是开展全球海洋监测和应对气候变化研究的重要数据源,为了研制出分辨率更高,误差更小的北极海冰密集度融合资料,本文使用了多源海冰密集度资料,以OSTIA (Operational Sea Surface Temperature and Ice Analysis)数据集为融合背景场,采用以下方案开展融合研究。首先,对现有5种海冰资料进行质量控制;其次,以OSI SAF (Ocean and Sea Ice Satellite Application Facility)资料为基准,采用概率密度匹配法订正各资料的系统误差;然后,利用小波分解将各资料分解为低频信息和高频信息,对低频信息和高频信息分别计算融合权重和卡尔曼滤波处理;最后,利用小波重构将各资料进行融合,生成0.05°分辨率的北极逐日海冰密集度融合资料。通过对比国际上广泛认可的OISST (Optimum Interpolation Sea Surface Temperature)、OSTIA海冰密集度资料,验证结果显示:融合资料与OISST、OSTIA海冰密集度资料在北极的空间分布上高度一致,相关系数均超过0.967。相对于前人的研究,本融合资料与OISST的偏差由–1.170%减少到–0.108%,与OSTIA的偏差由0.276%减少到–0.156%;与OISST和OSTIA的均方根误差分别由9.835%减少为8.010%以及由7.427%减少为5.140%。本融合资料的偏差以及均方根误差都得到了显著的提升,具有较高的质量。 |
| 关键词: 海冰密集度 小波变换 卡尔曼滤波 概率密度函数匹配法 |
| DOI：10.11759/hykx20230902001 |
| 分类号:P73 |
| 基金项目:国家重点研究与发展计划项目(2021YFC2803301);江苏省研究生研究与实践创新计划项目(Nos.KYCX23_1347) |
|
| A study of data fusion on Arctic sea ice concentration |
|
WANG An1, HE Yijun2,3, YING Qianhui3
|
|
1.School of Electronics and Information Engineering, Nanjing University of Information Science and Technology, Nanjing 210044, China;2.Key Laboratory of Space Ocean Remote Sensing and Applications, National Satellite Ocean Application Service, Beijing 100081, China;3.School of Marine Sciences, Nanjing University of Information Science and Technology, Nanjing 210044, China
|
| Abstract: |
| Sea ice concentration data are important for global ocean monitoring and climate-change-response research. To develop Arctic sea ice concentration fusion data with higher resolution and smaller errors, this study uses multisource sea ice concentration data, with Operational Sea Surface Temperature and Ice Analysis (OSTIA) data as the fusion background field. The following scheme is adopted to perform the fusion study. First, quality control of the existing five sea ice data is performed; second, the systematic error in each data is revised via the probability density matching method using the Ocean and Sea Ice Satellite Application Facility (OSI SAF) data as the benchmark; thereafter, the data are decomposed into low-frequency and high-frequency information by means of wavelet decomposition, and fusion is calculated for both these types of information. Thereafter, wavelet decomposition is used to decompose the data into low-frequency information and high-frequency information, and the low-frequency information and high-frequency information are calculated and processed via Kalman filtering. Finally, wavelet reconstruction is employed to fuse the data to generate the fusion data of day-by-day Arctic sea ice density with a resolution of 0.05°. Upon comparing with the internationally recognized Optimum Interpolation Sea Surface Temperature (OISST) data and OSTIA sea ice density data, the validation results demonstrate that the fusion data, OISST sea surface temperature data, and OSTIA sea ice density data are highly consistent in the spatial distribution of the Arctic, with the correlation coefficients exceeding 0.967. Compared with previous research results, deviation of the fusion data from OISST data is reduced from −1.170% to −0.108% and that from OSTIA data is reduced from 0.276% to −0.156%; in addition, the root mean square error (RMSE) between the fusion data and OISST data is reduced from 9.835% to 8.010% and that between the fusion data and OSTIA data is reduced from 7.427% to 5.140%. The bias as well as RMSE of these fusion data has been significantly improved with high quality. |
| Key words: sea ice concentration wavelet transform Kalman filtering probability density function matching method |
