面向地表参数演变过程的时序遥感变化检测技术暨三北工程评估

Faced with the gaps between remote sensing information extraction and land surface evolution process analysis, it is of great scientific significance and applications prospects to advance the change detection methodologies oriented on land surface biophysical parameters evolution process based on long-term time series images. In order to handle the challenges of insufficient support of change detection techniques to research on land surface processes, this project will investigate the basic theories, key technologies and typical applications of remote sensing time series processing. Two key techniques will be proposed, including the automatic change extraction of specific land surface cover transformation and automatic change identification of the evolution processes of land surface biophysical parameters. The established methodology can succeed in automatically extraction of four basic types (bare land, water body, impervious cover) oriented land surface cover transformation and vegetation cover conversion, temporal pattern recognition, information on dynamic process such as the magnitudes, speed, start & end time. The Three-North Project has built the crucial ecological barrier in North China. The central and local governments have invested Hundreds of billions of YMB in the Three-North Project since it started in the late 1970s. Multiple ecological projects aimed to afforestation have been launched after that, including the Beijing-Tianjing Sand Source Control Program, the Nature Forest Conservation Program and the Grain to Green Program. The benefits of the Three-North Project are always debatable due to the complex spatiotemporal heterogeneity of vegetation and ecological dynamic processes and limitations of time series analysis methods. Therefore, the proposed time series change detection approach will be applied to the long-term monitoring and evaluation of the Three North project. The evolution processes of four kinds of biophysical parameters, the vegetation cover, soil moisture, conservation of soil and water, and land surface radiation, will be investigated. The contributions of climate changes and projects inputs will be taken into considerations in the long-term monitoring and combined evaluations of the Three-North Project. The outcomes are expected to promote the developments of time series change detection methods of remote sensing images and also the scientific evaluations of the Three-North Project.

当前遥感地学研究面临遥感信息提取和地表演变过程分离的挑战。基于长时间序列遥感影像实现面向地表参数演变过程的自动遥感变化检测，具有重要科学意义和应用前景。本项目迎接当前遥感变化检测技术难以满足地表过程演变研究的挑战，研究面向地表参数演变过程的时序遥感变化检测理论与技术方法。力图攻克目标导向型地表覆盖变化自动检测、地表参数演变过程识别等关键技术。实现地表覆盖四大基本类型、植被覆盖类型变化自动提取，突变型、扰动型、渐变型、混合型等时序模式识别，获取变化类型、幅度、速率、起始时间等地表变化过程信息。三北工程筑牢我国北方重要生态屏障，但是对于三北工程效果存在不同看法。因此利用面向地表参数演变过程的时序遥感变化检测技术，综合植被覆盖、土壤湿度、水土保持、地表辐射等多种地表参数演变过程，纳入气候变化与工程投入因素，开展三北地区长时间序列时空连续监测与综合评估。拓展遥感变化检测技术，促进三北工程科学评价。

本项目立足于国家粮食与生态安全等重大战略需求，开展了时序遥感变化检测技术创新及其应用研究，取得了以下主要研究成果：(1)创建了植被变化时空连续建模框架，设计了植被流失去向自动识别方法，能精确获取变化区域、变化时间、演变模式等演变过程信息；利用该方法获得了全国2001-2017逐年新增不透水面、裸土和水体数据集，进而揭示出城市分级背景下耕地流失的非均衡性演变规律，以及耕地抛荒与城市化时空耦合关系；(2)建立了涵盖退耕还林、耕地开垦、复种变化等耕地变化自动监测方法，其具有无需依赖训练数据的优点；采用该方法基于MODIS时序数据集，揭示了我国中东部地区耕地开垦发生在相比退耕还林自然条件更差的区域，值得高度关注；(3)提出了若干基于知识的农作物时序遥感制图方法，具有在小样本或无样本情况下实现大尺度高精度自动制图能力；利用所设计的基于Sentinel-2色素指数的大尺度花生自动制图方法，绘制了首张覆盖我国东北三省的20米高精度花生空间分布图；(4)开展了上世纪八十年代以来全国多个大型生态工程的综合评估，揭示了我国生态工程建设成效显著从而推动中国引领全球绿化进程，工程效益与投资金额密切相关但存在显著的时空异质性，半干旱区生态工程效益增值空间很大；进一步阐明了遥感指标或数据集选取对全球植被绿化程度评估的影响机制，结果表明基于NDVI和MODIS GPP估算的全球绿化程度远高于LAI等其它指标，全球变暖情景下不同指标评估结果的不确定性加剧。项目成果，将促进时序遥感分析技术的发展，提高植被演变大尺度时空连续监测水平，为针对性地精准实施生态工程措施、保障国家粮食安全与可持续发展提供时空大数据技术支撑。.在该项目的资助下，项目负责人以第一作者/通讯作者在ISPRS Journal of Photogrammetry and Remote Sensing、Land Use Policy、Land Degradation & Development、Computers and Electronics in Agriculture和《地理科学进展》等地理学和农学主流期刊上发表研究成果，授权专利6项，获2020年福建省科学技术进步二等奖（排名1）。项目研发的时序遥感监测技术成果及其相关数据产品，已公开发布共享。