地球信息科学学报 ›› 2011, Vol. 13 ›› Issue (4): 439-446.doi: 10.3724/SP.J.1047.2011.00439

• 地球信息综合分析 • 上一篇    下一篇

地表环境参数时间序列重构的方法与应用分析

江东, 付晶莹, 黄耀欢, 庄大方   

  1. 中国科学院地理科学与资源研究所,资源环境科学数据中心,资源与环境信息系统国家重点实验室,北京 100101
  • 收稿日期:2011-02-16 修回日期:2011-06-26 出版日期:2011-08-25 发布日期:2011-08-23
  • 作者简介:江东(1972-),男,安徽寿县人,博士,副研究员。研究方向为资源环境遥感应用。 E-mail:jiangd@igsnrr.ac.cn
  • 基金资助:

    国家科技支撑计划课题"中国重大自然灾害孕险环境分析技术"(2008BAK50B01); 国家科技支撑计划课题"南水北调水资源综合配置技术研究"(2006BAB04A16);国家航天局航天遥感论证中心环境星应用推广工程课题"北部湾环境综合评价与监测"(2008A02A09)。

Reconstruction of Time Series Data of Environmental Parameters: Methods and Application

JIANG Dong, FU Jingying, HUANG Yaohuan, ZHUANG Dafang   

  1. Data Center for Resources and Environmental Sciences, Institute of Geographic Sciences and Natural Resources Research, CAS, Beijing 100101, China
  • Received:2011-02-16 Revised:2011-06-26 Online:2011-08-25 Published:2011-08-23

摘要: 卫星传感器数据提取的地表环境参数是全球变化和区域资源环境研究的重要信息源。时间序列数据的重构旨在利用多种统计和数值分析方法,模拟参数的季节/年度变化规律,从而插补缺失观测值,优化时间序列数据,为相关研究提供更加完备的基础数据。目前,数据重构主要应用在植被指数、叶面积指数、地表能量平衡参数等,重构的方法包括阈值法、滤波方法和非线性拟合等。随着卫星传感器时空、波谱分辨率的提高和研究的深入,耦合了遥感反演模型和陆面过程模型的数据同化重构方法,是今后研究和应用的重要方向。

关键词: 遥感参数, 时间序列重构, 数据同化

Abstract: The satellite-derived environmental parameters play important roles in global change and regional resources and environment researches. Atmosphere effects and sensor limitations often lead to data products of inherently variable quality. The main goals of time series data reconstruction are to remove cloud affected observations and create gapless dataset at a prescribed time with multiple spatio-temporal interpolation and statistical methods. The conventional reconstructing algorithm include threshold method, temporal filtering method, and nonlinear simulation, etc., and mainly applied on satellite derived vegetation index, leaf area index and surface energy balance parameters. Although methods mentioned above have been applied in related researches and received good results, they suffer from their own drawbacks which limit their use. The algorithm of MVC and MVI ignores the remarkable anisotropy of reflectivity and VI, which may conceal details of vegetation with too large interval and lead to the result of reconstructed VI data from MVC and MVI remain a lot of noises. The BISE and other threshold-based methods may make the extracted temporal information unreliable. The Fourier analysis algorithm is sensitive to spurious peaks of the VI trend line, so it may depart from the truth a lot. Much attention should be paid to data assimilation based approaches which couple remote sensing retrieval model with surface processing model.

Key words: remotely sensed parameters, time series data reconstruction, data assimilation