基于自适应时空张量补全的甲醛浓度时间序列重建  

Formaldehyde concentration time series reconstruction based on adaptive spatio temporal tensor completion

作　　者：刘多多 袁强强[1] 汪源 LIU Duoduo;YUAN Qiangqiang;WANG Yuan(School of Geodesy and Geomatics,Wuhan University,Wuhan 430079,China)

机构地区：[1]武汉大学测绘学院,武汉430079

出　　处：《测绘工程》2023年第4期13-20,共8页Engineering of Surveying and Mapping

摘　　要：甲醛是大气中重要的微量气体,可作为挥发性有机化合物氧化的敏感性指标,同时也是一种有刺激性气味的有毒气体,易对人体健康产生不利影响。然而,遥感卫星获得的HCHO数据不可避免地受到云层、气溶胶和其他不利大气条件的影响而普遍存在缺失,阻碍了后续的分析和应用。本研究采用时空自适应低秩张量补全方法来解决这一问题。该方法能充分利用空间邻域、时间邻域和周期特征的相关信息来构建高度相关的张量,并通过ST-Tensor模型来重建缺失的信息。本研究利用该方法对2019—2021年在亚洲中南部获得的TROPOMI近实时L_3级HCHO数据进行重建。定性和定量研究结果表明,ST-Tensor方法比最邻近时序内插法具有更好的稳定性、准确性和空间连续性,均方根误差、平均绝对偏差、相关系数分别提高了1.78×10^(-5)、1.7×10^(-6)和0.19。利用重建结果获得的四季平均时空分布图,表明HCHO浓度的高低与人口密度、温度、气候特征等有关。Formaldehyde(HCHO)is an important trace gas in the atmosphere.It can be used as a sensitive indicator of oxidation of Volatile Organic Compounds(VOCs).It is also a toxic gas with irritating odor,which is prone to adverse effects on human health.However,the HCHO data obtained by remote sensing satellites are inevitably affected by clouds,aerosols and other adverse atmospheric conditions,and are generally missing,which hinders the subsequent analysis and application.This study adopts an Spatio-Temporal Adaptive Low-Rank Tensor Completion(termed ST-Tensor)method to solve this problem.This method can make full use of the relevant information of spatial neighborhood,temporal neighborhood and periodic features to construct highly correlated tensors,and reconstruct the missing information through the ST-Tensor model.In this study,this method was used to reconstruct the near real-time L_3 level HCHO data of TROPOMI acquired in south-central Asia during 2019—2021.Qualitative and quantitative results show that the ST-Tensor method has better stability,accuracy and spatial continuity than the nearest neighbor time series interpolation method,and the root mean square error,mean absolute deviation and correlation coefficient have been improved by 1.78×10^(-5),1.7×10^(-6) and 0.19,respectively.The average spatial-temporal distribution map of the four-season average obtained from the reconstruction results shows that the concentration of HCHO is related to population density,temperature,and climate characteristics.

关 键 词：甲醛 缺失值重建 自适应 低秩张量补全 时空信息 TROPOMI 

分 类 号：P407[天文地球—大气科学及气象学]