On the Compatibility between Morning and Afternoon Overpass Data Products: A Case Study Using Terra and Aqua MODIS Vegetation Indices

Radiometric compatibility between morning and afternoon overpass data were assessed for two vegetation indices (VIs), the Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI), using Terra- and Aqua-Moderate Resolution Imaging Spectroradiometer (MODIS) standard 1 km 16-day composite VI products (MOD13A2 and MYD13A2). Terra and Aqua MODIS products provide good opportunities to examine the potential effects of morning vs. afternoon overpasses because the data are acquired by sensors with the same characteristics and the products are generated using the same algorithms, and thus any differences between their observations can be attributed to diurnal changes in atmospheric and surface conditions (e.g., cloud covers and soil moisture contents). 210 sites were located throughout the North and South American continents, including the EOS core validation and fluxnet sites, where data were extracted over a 3 km-by-3 km window for the overlapping period of three years starting in June 2002. Mean differences of the NDVI and EVI between Terra and Aqua (i.e., Aqua minus Terra) were computed and used as a measure of compatibility (discrepancy) with the criterion that the mean differences should be equal to zero or less than the VI uncertainty levels due to radiometric calibration and atmospheric correction. Mean differences of both the NDVI and EVI showed latitudinal, land cover, and seasonal dependencies. Aqua MODIS NDVI/EVI were statistically larger at around 50 N and smaller at around 23.5 N and 23.5 S than the Terra counterparts. Aqua NDVI/EVI values were larger for water and forest land cover types and smaller for the other land cover types, including woodland, grassland, shrubland, and cropland, than the Terra counterparts. Aqua MODIS VIs were also statistically smaller than the Terra counterparts in the July-October time period, growing seasons in the northern hemisphere. These dependencies may be related with diurnal cycles of cloud formations/dissipations. Regardless of the statistical significance, however, the maximum, mean difference estimates (0.02 and 0.01 units for the NDVI and EVI, respectively) were all comparable or less than the VI uncertainties due to radiometric calibration (0.01 and 0.02 for the NDVI and EVI) and atmospheric correction (0.02 and 0.01 for the NDVI and EVI). Continental-scale, mean difference estimates of the indices computed using all of the 210 sites were found to be statistically "zero" at the 95 % confidence level (-0.002 for the NDVI and 0.001 for the EVI). Overall, these results indicate that morning and afternoon overpass data sets can be interchangeably used for the geographic coverage examined here.