Evaluating the consistency of multi-source wideband remote sensing images: A band simulation approach using Hyperion data

Inconsistencies often occur among same band images from multi-source remote sensing platforms because of differences in imaging environments and satellite sensor design parameters. Differences caused by imaging environments have been explored in numerous quantitative studies. However, differences caused by the design parameters of sensors are given scant attention, thus leading to several inaccuracies and misinterpretations of study results. Therefore, based on the combination of Earth Observing-1 (EO-1) Hyperion hyperspectral data and two key functions of a sensor imaging system, i.e., Spectral Response Function (SRF) and Point Spread Function (PSF), Near-Infrared (NIR) band images of three typical sensors, namely, Landsat Enhanced Thematic Mapper Plus (ETM+), Terra Moderate-Resolution Imaging Spectroradiometer (MODIS), and National Oceanic and Atmospheric Administration Advanced Very High Resolution Radiometer (AVHRR 3), were simulated, compared, and consistently evaluated. The study obtained the following results. (1) The SRF can cause several discrepancies among the three simulated NIR bands, namely S_MODIS, S_ETM+, and S_AVHRR3. The difference between S_MODIS and S_ETM+ was relatively small (approximately -0.940% with a mean difference of -0.675), whereas the difference between S_ETM+ and S_AVHRR3 was relatively large (approximately 10.129% with a mean difference of 3.448). Thus, a thorough evaluation of the consistency of same band images from multi-source sensors should be conducted before they could be applied to further research. (2) The accuracy of simulating moderate or low spatial resolution imagery by aggregating high spatial resolution imagery can be improved by incorporating the PSF, particularly in heterogeneous areas. (3) The simulation method adopted in this study, i.e., EO-1 Hyperion imagery combined with the PSF and SRF of one sensor, can simulate wideband moderate-or low-resolution images of that sensor effetely with a relative mean difference less than -2.371%. This study does not only provide an effective means of evaluating the consistency of same band images from multi-source wideband remote sensing platforms, but also offers a better method of selecting more accurate multi-source remotely sensed data.