论文部分内容阅读
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。
Vegetation remote sensing products for global response to changes in agricultural production management, the application of ecology and other fields of great significance. However, there are still many problems in the existing remote sensing products of phytoplankton, mainly including the use of time series data with different parameters on one hand and the difference of product results caused by different extraction algorithms on the other hand. On the other hand, ground surface observation data and remote sensing inversion Systematic errors in the validation method due to inconsistent physical meaning of the data. In this paper, the Heihe River Basin is used as a research area to compare and verify the remote sensing phenological products of MLCD (MODIS global land cover dynamics product) extracted based on EVI (Enhanced Vegetation Index) time series data and UMPM (Leaf Area Index) product by universal multi-life-cycle phenology monitoring method. At the same time, the differences and characteristics of phenological features extracted based on the EVI and LAI time series are further evaluated by verification analysis, and the deviations of direct validation results due to the inconsistent physical meaning of vegetation phenology extracted by surface observations and remote sensing are discussed. The results showed that UMPM products were more effective than MLCD products in general. However, sparse vegetation areas dominated by grassland and shrubs had more missing data and lower spatio-temporal stability due to LAI accuracy. The ground observation data of corn shows that EVI is more sensitive to LAI than LAI and more suitable for extracting growth starting point. However, the vegetation index is easy to be saturated and the peak starting point is generally advanced. The peak starting point based on LAI extraction is more reasonable. Since the observed phenophases in the ground pay more attention to the fruit growth in the later period, the remote sensing observations only focus on the growth of the leaves. The peak end point and the end point of the growth defined by the remote sensing differ greatly from the mature and mature phases of the corn.