An evaluation of the potential to combine MODIS-Himawari-8 vegetation indices after normalising geometrical differences across Australian landscapes

The Himawari-8 Advanced Himawari Imager (H8/AHI) is a novel geostationary (GEO) satellite that provides the potential to gap-fill low-earth orbiting (LEO) data, such as MODIS (MODerate-resolution Imaging Spectroradiometer), with its very high, 10minute temporal resolution. However, different sun-sensor geometries between GEO and LEO observations challenge this goal. In this study, we investigated the similarity between vegetation indices derived from Himawari-8 AHI and Terra and Aqua MODIS in capturing the seasonal greenness profiles of different Australian vegetation types, including grasslands, croplands, woodlands and forests. We analysed diurnal and seasonal patterns of AHI reflectances and vegetation index (VI) values to optimise daily compositing methodologies over our sites. From the 10 minute AHI surface reflectances, we generated two-years of AHI daily-composited time series reflectances and VIs, and compared these with equivalent MODIS time series using the MOD/MYD13 products (BRDF-uncorrected), and MODIS geometrical adjusted data corrected to local sun and view angles of selected sites (BRDF-corrected). Our results show (1) the large diurnal variations in VI and reflectance values with maximum NDVI values at the most extreme sun angles, minimum NDVI values around solar noon, and mixed EVI2 diurnal changes depending on landscapes; (2) seasonal geometrical changes due to sun angle variations resulted in strong hotspot phenomena around the equinox periods, amplifying the diurnal variations in surface reflectances but with mixed diurnal VI results ; (3) daily H8/AHI-NDVI values were consistently lower, and daily H8/AHI-EVI2 was consistently higher than their equivalent MODIS values at all selected sites; (4) while the VI seasonal profiles of MODIS and daily extracted H8/AHI were consistently similar in shape but different in amplitudes at grassland and cropland sites, their similarities in more densely vegetated forest sites were lower than in the other sites; (5) adjustments for sun-view angle differences between the two sensors resulted in very high correlations and consistent values between MODIS and H8/AHI VI time-series for all landscape types, including forests. The disparities between MODIS and H8/AHI VI over forest areas were attributed to both Bidirectional Reflectance Distribution Function (BRDF) effects and to the much lower seasonal VI variations. This highlights the very strong BRDF influences in the cross-sensor datasets, particularly associated with the H8/AHI sensor at a fixed view angle. Our outcomes also provide concrete evidence that data harmonisation from these two sensor sources with varying geometrical orientations, can only be accomplished with BRDF correction to a common sun and view angle.