MODIS-VIIRS Continuity: The Impact of Spatial Sampling on Global Land (Level-2) Products

Abstract:

The Visible Infrared Imaging Radiometer Suite (VIIRS) onboard Suomi-NPP (S-NPP) has been providing daily global observations of the Earth surface since early 2012. With the decade-long observations made by MODIS onboard Terra and Aqua, one of the goals of the S-NPP mission is to provide continuity in producing land products that have been generated using heritage MODIS observations. The Land Data Operational Products Evaluation (LDOPE) team uses MODIS-derived products to evaluate land products obtained from VIIRS top-of-atmosphere (TOA) measurements generated through the Land Product Evaluation and Analysis Tool Element (LPEATE). However, due to inherent differences in their observation methods and the corresponding algorithms and post-processing techniques, the products generated from MODIS and VIIRS retain some discrepancies. Amongst all the differences between the two wide-swath radiometers, this study aims at analyzing the impact of differences in the corresponding spatial sampling. In particular, the VIIRS unique sampling scheme can introduce relative biases when comparing products (or observations) obtained from the two sensors. We use Landsat-8’s Operational Land Imager (Level-1T data) scenes acquired within a set of 10 x 10 degree, i.e., “Golden tiles” (used for evaluation purposes by LDOPE) to examine how the discrepancies in the spatial responses manifest in measured radiances on a daily basis (for 16 days). The (band-detector averaged) prelaunch Line Spread Functions (LSFs) were used to represent spatial responses for each sensor. Although the impact of differences in sensors’ spatial responses depends heavily on the spatial heterogeneity of a region-of-interest, the initial results, on average, indicate up to 0.8% and 5% difference (at the swath level) in the TOA radiances and TOA-based NDVI, respectively. The disparity (calculated for three sample scenes collected over the Golden sites) differs for different days (orbital configurations) and for different MODIS sampling sizes (500 m versus 1000 m). In other words, the plots of difference-in-scan-angle versus spatial-response show different trends for the MODIS land and ocean channels. For this ongoing study, the impact of gridding and differences in the spectral response functions will also be investigated.