Reprojection of VIIRS SDR Imagery Using Concurrent Gradient Search

The application of the gradient search method for reprojection of Visible Infrared Imaging Radiometer Suite (VIIRS) satellite data record (SDR) imagery is described. The method is an extension of the scheme developed earlier for reprojection of Moderate Resolution Imaging Spectroradiometer (MODIS) L1B imagery. The new scheme has three important improvements: 1) the interscan and intrascan search steps are combined into a single step to save computational time; 2) one-sided (left-right, up-down) gradients are utilized to improve convergence; and 3) the use of the map projection instead of the latitude-longitude coordinate system to improve performance and robustness. The scheme is computationally very fast, employing only basic arithmetic operations and precalculated matrices of spatial gradients. An average number of iteration steps for the reprojection of mid-latitude quadruple VIIRS SDR granule is less than 1.5, i.e., the scheme usually converges in less than 2 iterations. The ambiguity in the overlapping areas due to the bow-tie effect is resolved by forcing a solution located closer to the scan line center. In addition, the accuracy of VIIRS imagery geo-location was evaluated by comparison against MODIS 250 m images. Absolute geolocation biases of the VIIRS imagery over the 1-year period from June 01, 2016 to May 01, 2017 were found on average to be within 0.004 and -0.003 of the sample size (Δ line) in the along-track direction and 0.055 and 0.035 of the sample size (Δ pixel) in the along-scan direction for bands I2 and M7, respectively. These results demonstrate the excellent geometric performance of the VIIRS Suomi National Polar-orbiting Partnership sensor and are consistent with those reported by the VIIRS geolocation teams.