联合HJ-1/CCD和Landsat8/OLI数据反演黑河中游叶面积指数

目前制约30 m分辨率地表参数遥感提取的主要因素是有限的观测个数,而联合多传感器观测是提高单位时间观测频次的一个有效途径。本文以黑河中游为研究区,利用HJ-1/CCD和Landsat 8/OLI传感器构建多传感器观测数据集。对多传感器观测数据集在观测周期内的有效观测个数、观测角度和双向反射分布函数BRDF分布特征、以及经过预处理后的多传感器数据一致性等问题进行分析。不同传感器观测数据质量差异是多传感器联合反演的主要问题,因此本文首先制定了多传感器数据质量控制方案,然后利用统一模型查找表反演单传感器叶面积指数LAI结果,对10天观测周期内经过质量筛选的单传感器反演结果采用平均方法合成LAI产品。结果表明,LAI有效反演像元占总反演像元比例由单传感器的6.4%—49.7%提高到多传感器的75.9%。利用地面测量数据进行验证分析,LAI反演结果与地面实测数据的均方根误差RMSE均值为0.71。利用30 m分辨率的HJ-1/CCD和Landsat 8/OLI传感器数据可以生产精度可信、时间分辨率连续的LAI产品。

Leaf area index inversion combining with HJ-1/CCD and Landsat 8/OLI data in the middle reach of the Heihe River basin

The main restriction on surface parameter inversion from remote sensing data with 30 m resolution is the limited number of observations. Nevertheless, a network or multiple-sensor method can efficiently increase the number of observations. In this study, a multi-sensor database was generated from HJ-1/CCD and Landsat 8/OLI from June 2013 to August in 2013 in the middle reach of Heihe River Basin. Characteristics, including proportion of valid observations, distribution of observation angles, bidirectional reflectance distribution function, and data consistency among sensors after preprocessing, of the multi-sensor dataset were analyzed. Difference in observation quality from different sensors is a major issue regarding Leaf Area Index (LAI) inversion from a multi-sensor dataset. Therefore, an observation quality control criterion was initially designed. Multi-sensor observations that satisfied the quality control requirements were used to inverse LAI based on a look-up table built by the unified model. The synthesis LAI over 10 days was set as the mean of LAI inversion from each sensor observation because of limited observation number. Analysis and validation were performed based on LAI products produced in the middle reach of the Heihe River Basin. Results show that the percentage of valid LAI inversion significantly increased from 6.4% to 49.7% of the single-sensor inversion to 75.9% of the multi-sensor inversion. Validated results show that the average RMSE between field measurements and LAI inversion was 0.71. The network of HJ-1/CCD and Landsat 8/OLI sensors with 30 m spatial resolution can generate LAI products with reasonable accuracy and continuous temporal resolution.