中国民用对地观测卫星在近10 a得到迅速发展，2012年和2013年相继发射的ZY-3和GF-1遥感卫星已成为中国现阶段主要应用的高分影像卫星，但二者对地观测能力是否相同并不清楚。因此，基于2对同日过空的GF-1 PMS1和ZY-3 MUX影像对，利用归一化植被指数NDVI对二者的植被观测能力进行对比。结果表明，GF-1 PMS1和ZY-3 MUX的植被观测能力虽然很接近，但也存在一定的差异。主要表现在ZY-3 MUX植被指数NDVI的信息量和信号总体强于GF-1 PMS1，后者的低估幅度为可达−3%；但随着NDVI的增强，GF-1 PMS1的低估会逐渐减少，在NDVI的高值区甚至可超过ZY-3 MUX。由于二者之间存在差异，因此它们如要应用于同一项目，建议要进行数据转换，以确保结果的准确对比。分析表明，这2种传感器数据之间的差异是二者在光谱响应函数、空间分辨率以及定标精度等方面的差异引起的。%China’s commercial earth observation system has developed rapidly in the last decade. Successively launched respectively in 2012 and 2013, Ziyuan-3 (ZY-3) and Gaofen-1 (GF-1) are currently the main sources of high spatial-resolution imagery in China, which have provided invaluable land-observation information for a variety of applications. Nevertheless, the vegetation mapping ability and the quantitative relationship between the two kinds of sensor data are unclear as these have not yet been explored since the launch of the two satellites. To meet this special requirement, this study compares the vegetation measurement of GF-1 PMS1 and ZY-3 MUX sensor data based on the normalized difference vegetation index (NDVI). Two tandem image pairs of Fuzhou and Jingmen, China, acquired respectively on June 14 and December 12, 2014, were employed for this comparison approach. Because of the difference in spatial resolution between the GF-1 PMS1 (8 m) and ZY-3 MUX (6 m) images, a comparison based on the ROI (region of interest) was carried out, as the pixel-by-pixel comparison method was not suitable. A total of 69 ROIs were collected, which included 44 ROIs from forests and 25 ROIs from paddy lands. The digital numbers of the red and near infrared bands of both sensors were converted to at-satellite reflectance, which were used to calculate the NDVI image pairs of the two sensors. The comparison mainly depended on the statistics of the NDVI data of both sensors and the regression analysis on the sampled ROIs. The results show that the NDVI data of GF-1 PMS1 and ZY-3 MUX sensors are quite similar because they have a very high degree of agreement, with theR2 values of 0.95 for Fuzhou image pair and 0.98 for Jingmen pair. However, the comparison study finds that the differences do occur between both sensors’ NDVI data. The data range and the standard deviation of ZY-3’ NDVI are greater than GF-1’ NDVI. Moreover, the NDVI of the ZY-3 MUX sensor generally has a higher mean value than that of GF-1 PMS1. Compared with ZY-3, the GF-1 PMS1 can underestimate the NDVI by up to -3% in Fuzhou’s paddy observation. Nevertheless, this underestimation can be reduced in the high NDVI data region, where the NDVI values of GF-1 PMS1 are very close to those of ZY-3 MUX and sometimes even higher than ZY-3’ NDVI regardless of vegetation types. Due to the differences between the NDVIs of both sensors revealed in this study, the NDVI derived from GF-1 PMS1 and ZY-3 MUX sensor data should not be compared directly if they are applied in the same project. Data conversion from one sensor to the other is generally required to ensure an accurate comparison between their NDVI-based products. This study shows that such a conversion can improve the agreement between the two sensors’ NDVI data, as the difference in RMSE between the NDVI values of both sensors could be reduced by up to -16.3% based on the conversion model derived from the regression analysis with the data from 69 sampled ROIs. Analysis shows that the differences in two sensors’ NDVI data are caused by the differences in the spectral response function, spatial resolution, and calibration accuracy between the two sensors.
展开▼
