黑河流域遥感物候产品验证与分析

Validation and analysis of remote sensing phenology products in the Heihe River Basin

植被物候遥感产品对全球变化响应、农业生产管理、生态学的应用等多领域研究具有重要意义。但现有植被物候遥感产品还有较多问题,主要包括一方面使用不同参数的时间序列数据以及不同提取算法导致的产品结果差异较大,另一方面在地面验证中地面观测数据与遥感反演数据的物理含义不一致导致的验证方法的系统性误差。本文以黑河流域为研究区,对比验证基于EVI(Enhanced Vegetation Index)时间序列数据提取的MLCD(MODIS global land cover dynamics product)植被遥感物候产品和基于LAI(Leaf Area Index)时间序列数据提取的UMPM(product by universal multi-life-cycle phenology monitoring method)植被遥感物候产品的有效性及精度等。同时,通过验证分析进一步评估基于EVI和LAI时间序列提取的物候特征的差异及特点,探讨由于地面观测植被物候与遥感提取植被物候的物理意义的不一致问题导致的直接验证结果偏差。结果表明:UMPM产品有效性整体高于MLCD产品,但在以草地和灌木为主的稀疏植被区,由于LAI取值精度的原因,UMPM产品存在较多缺失数据,且时空稳定性较低;基于玉米地面观测数据表明,EVI对植被开始生长的信号比LAI更加敏感,更适合提取生长起点,但植被指数易饱和,峰值起点普遍提前,基于LAI提取的峰值起点更加合理。由于地面观测的物候期在后期更加关注果实生长,遥感观测仅关注叶片的生长,遥感定义的峰值终点和生长终点与玉米的乳熟期和成熟期差异较大。

Land surface phenology is of great significance in the fields of global change response, agricultural management, and ecological applications. However, compared with the strong demand for vegetation phenology mapping for global and regional research and applications, the development of remote sensing vegetation phenology products is slower than that of other remote sensing parameters, such as Leaf Area Index （LAI） and Vegetation Index （VI）. Although Enhanced Vegetation Index （EVI） and LAI are the most widely used vegetation parameters for remote sensing phenology extraction, this study aims to compare the remote sensing phenology products derived from MOD- erate-resolution Imaging Speetroradiometer （MODIS） EVI and global land surface satellite product LAI in Heihe River Basin. Moreover, this study aims to assess the difference inphenology information that was extracted from EVI and LAI time series. The validity and accuracy between the MODIS global Land Cover Dynamics product （MLCD） and Universal Multi-life-cycle Pheno- logy Monitoring Method （UMPM） product in the Heihe River Basin were compared. Validity contains the missing product rate and stability. Accuracy was assessed using representative observation sites. The sites were the basis of the two proposed indications （the mean bias and mean absolute bias） for the evaluation of remote sensing phonological metrics. Furthermore, phenology information that was extracted from EVI and LAI products was compared. Then, the variances between ground observations and phenology products were discussed in detail. The validity of UMPM is better than that of MLCD as a whole. However, in sparse vegetation that is mainly composed of shrubs and grasslands, UMPM has more missing data and lower spatial-temporal stability because its precision value is lower than that of EVI. For maize, EVI is more suitable for extracting the start of growing season, whereas LAI performs better for extracting the peak of growing season. However, field observations focus on fruit development during the later period of maize growth, whereas remote sensing phenology detection is specific to leaf development. Both MLCD and UMPM are inconsistent with ground observations after the peak of the growing season. Either MLCD or UMPM has advantages in different vegetation types and various phenological developmental stages. Uniting multisource data can improve the accuracy and validity of remote sensing phenology products. In addition, owing to the coarse spatial resolution of current remote sensing phenology products, it inevitably includes other plants within one square kilometers, which may lead to variability in phenological developmental stages and a weak relationship between remote sensing data and ground observations. Improving the spatial resolution of remote sensing phenology products is significant in the future promotion of their application and development.