引用本文:邰文飞,张新胜,蔡明勇,申振,申文明,史雪威,陈绪慧.两种应用场景下NDVI时间序列数据拟合方法研究[J].环境监控与预警,2022,14(3):19-26
TAI Wen-fei, ZHANG Xin-sheng, CAI Ming-yong, SHEN Zhen, SHEN Wen-ming,SHI Xue-wei, CHEN Xu-hui.Study on NDVI Time Series Data Fitting Method Based on Two Application Scenarios[J].Environmental Monitoring and Forewarning,2022,14(3):19-26
【打印本页】   【HTML】   【下载PDF全文】   查看/发表评论  【EndNote】   【RefMan】   【BibTex】
←前一篇|后一篇→ 过刊浏览    高级检索
本文已被:浏览 291次   下载 50 本文二维码信息
码上扫一扫!
分享到: 微信 更多
两种应用场景下NDVI时间序列数据拟合方法研究
邰文飞,张新胜,蔡明勇,申振,申文明,史雪威,陈绪慧
生态环境部卫星环境应用中心,北京 100094
摘要:
植被覆盖指数(NDVI)时间序列数据集包含地表植被的长势、生长周期、时空变化等信息,其拟合重建结果可应用于物候信息提取、生态质量评价、人类活动扰动识别、覆被变化动态监测等方面。基于TIMESAT软件,选取物候参数提取和扰动识别2个应用场景,结合地面站点数据和Jacknife法模拟数据,对比分析非对称高斯函数拟合法(AG法)、双Logistic函数拟合法(D-L法)和Savitzky-Golay滤波法(S-G法)3种方法的拟合效果。结果表明:(1)3种方法拟合重建后提取的生长开始时间(SOS)、生长结束时间(EOS)、生长周期(LOS)等物候参数接近站点数据,AG法和D-L法保持NDVI时序曲线整体变化特征的能力较强,提取的SOS和EOS更接近站点数据;(2)人类活动扰动识别应用场景中,S-G法在滤波时能够最大限度地保留时序曲线细节变化,恢复速率相关系数达到0.618,回归估计标准差低于AG法和D-L法,因此识别精度最优。
关键词:  植被覆盖指数  时间序列  遥感  TIMESAT软件  拟合重建  物候参数  扰动识别
DOI:
分类号:X87;TP79
基金项目:国家重点研发计划基金资助项目(2021YFB3901103);地质灾害防治与地质环境保护国家重点实验室开放基金资助项目(SKLGP2020K005)
Study on NDVI Time Series Data Fitting Method Based on Two Application Scenarios
TAI Wen-fei, ZHANG Xin-sheng, CAI Ming-yong, SHEN Zhen, SHEN Wen-ming,SHI Xue-wei, CHEN Xu-hui
Ministry of Ecology and Environment Center for Satellite Application on Ecology and Environment, Beijing 100094, China
Abstract:
The NDVI time series data contains information about the growth, growth period, spacial temproral changes of surface vegetation, and the fitting reconstruction results are applied to phenological information extraction, ecological quality evaluation, human activity disturbance recognition, dynamic monitoring of mulching change, etc. Based on TIMESAT software, we select two application scenarios, combining ground station data and Jacknife method, simulation data, and analyze the fitting effect of asymmetric Gaussian function (AG), double Logistic function (D-L) and Savitzky Gray method (S-G). The results showed that the extracted growth start time (SOS), growth end time (EOS) and growth cycle (LOS) and other terminal parameters are close to site data, AG and D-L maintain the overall change characteristics of NDVI timing curve, and the extracted growth start time and growth end time are closer to site data. In human activity disturbance recognition application scenarios, S-G method can maximize the timing curve detail changes during filtering, so the recognition accuracy is better than AG and D-L methods, with recovery rate reaching at 0.618 and estimated standard deviation of regression lower than AG and D-L method.
Key words:  NDVI  Time series  Remote sensing  TIMESAT software  Fitting reconstruction  Phenological parameters  Disturbance