不同时间尺度多源时序数据的FEEMD分解比较研究  

Comparative study of FEEMD decomposition of multisource time series remote sensing data at different time scales

作　　者：王正 邱士可[1,2] 曾群 吕言利[4] 王超 张起萍[4] 李双权 WANG Zheng;QIU Shike;ZENG Qun;LYU Yanli;WANG Chao;ZHANG Qiping;LI Shuangquan(Institute of Geographical Science,Henan Academy of Science,Zhengzhou 450052,China;Key Laboratory of Remote Sensing and GIS of Henan Province,Zhengzhou 450052,China;College of Urban and Environmental Sciences Research Institute of Sustainable Development,Central China Normal University,Wuhan 430079,China;Earth View Image Co.,Ltd.,Beijing 100083,China)

机构地区：[1]河南省科学院地理研究所,郑州450052 [2]河南省遥感与GIS重点实验室,郑州450052 [3]华中师范大学城市与环境科学学院,武汉430079 [4]北京国遥新天地信息技术股份有限公司,北京100083

出　　处：《华中师范大学学报（自然科学版）》2023年第6期821-836,共16页Journal of Central China Normal University：Natural Sciences

基　　金：河南省重点研发与推广专项(科技攻关)项目(232102321100,222102320467);河南省科学院中央引导地方科技发展专项项目(211201004);河南省科学院重大聚焦项目(210101007);河南省科学院特聘研究员项目(230501008);河南省软科学研究计划项目(232400411139)。

摘　　要：中国南海东北部海区的叶绿素a浓度及相关环境因子受多尺度物理强迫影响,具有非线性非平稳态特征,对该区域的数据进行分解存在一定困难.该文利用一种自适应、非线性、非平稳态的FEEMD方法对研究区8 d尺度和月尺度长时序叶绿素a浓度及相关环境因子数据进行分解,结果发现:1)FEEMD有效避免了EMD和EEMD的高频模态混叠问题;2)FEEMD的运行速度比EMD和EEMD快10倍以上;3)基于8天和月尺度数据分解出的21年数据总趋势一致;4)相较于月尺度数据,8天尺度数据能分解出更多具有实际物理意义的高频模态,计算这些高频模态的周期发现基于8天尺度数据能分解出短至约2个月、4个月(季节)、6个月的周期;5)8天尺度叶绿素a浓度数据能分解出长达5年左右周期,其他相关环境因子可分解出10~14年超长周期,而月尺度数据一般只能分解出年尺度周期.该文研究结果表明,FEEMD方法可在环境复杂、动态度高、因子多变的区域进行长时间序列数据分解,并能取得理想效果,能为复杂环境条件下多因子间驱动关系研究提供借鉴.The long time series chlorophyll a concentration and related environmental factors in the northeastern South China Sea are affected by multi-scale physical forcing,which have nonlinear and non-stationary characteristics.Therefore,it is difficult to decompose the data in this region.In this study,an adaptive,non-linear,non-stationary FEEMD method is utilized to decompose the 8-day-scale and monthly-scale datasets of chlorophyll a concentration and associated environmental factors.The results are shown as follows.1)FEEMD can effectively overcome the high-frequency mode mixing problem of EMD and EEMD;2)FEEMD is 10 times faster than EMD and EEMD;3)The overall trends of the 21-year data decomposed based on 8-day and monthly scale data are consistent;4)The 8-day scale datas can be decomposed into more physically significant high-frequency modes than monthly data.The calculation of these high-frequency modes reveals that the 8-day scale datas can be decomposed into periods as short as about 2 months,4 months(seasons),and 6 months(half year);5)The 8-day chlorophyll a concentration data can be decomposed into cycles of up to about 5 years,and other related environmental factors can be decomposed into very long cycles of 10-14 years,while the monthly scale data can only be decomposed into annual scale cycles.The analysis results of this paper demonstrated that the FEEMD method can effective decompose long time series data in the study area with complex environment,high dynamic factors.The optimal results achieved by FEEMD in data decomposition in complex regions can provide implications for subsequent studies of multifactor-driven relationships in this area.

关 键 词：FEEMD 数据分解 叶绿素A浓度 环境因子 南海东北部 

分 类 号：R407.8[医药卫生—临床医学]