1. Undetected ice over inland water
Some ice pixels over inland water are found not being detected. This is either due to the failed detection in the upstream snow/ice mask or the failed internal detection in the AOD retrieval algorithm. These undetected ice pixels tend to have unrealistic high AOD retrievals.
2. Stripes in VIIRS AOT images over heavy aerosol region
Sometimes stripes are observed in the VIIRS AOT images over heavy aerosol regions, such as regions with heavy smoke plumes. Figure 1 shows an example. The pixels within the stripes are plotted as missing values. Our investigation shows that the those pixels have low AOT qualities and therefore not plotted, since only pixels with high and medium qualities are plotted in the AOT image. The cause of the problems is further traced back to the heavy aerosol flags from the cloud mask product, which is an upstream product of AOT. The stripe pixels are incorrectly set as "not heavy aerosol" while their surrounding pixels are set as "heavy aerosol". The stripes only occur in some of the bow-tie deletion regions and they are likely to be caused by a code error in the VIIRS cloud mask algorithm.
Figure 1. VIIRS AOT image with stripes over heavy aerosol region on November 14, 2016.
3. Plotting Artifacts over Granule Overlapping Regions
We often observe artifacts in the smoke mask and AOT images over the VIIRS granule overlapping regions. VIIRS is onboard NPP, which is a polar-orbiting satellite. It is possible that the satellite passes some regions twice or more during day time. Since we plot daily VIIRS images of RGB, smoke mask, and AOT on eIDEA, some parts of the images are hidden due to the overlapping of the granules. In preparing the eIDEA images, We treat the overlapping regions as following: (1) plot RGB image from the latest granule; (2) plot smoke mask, AOT image from the latest granule; (3) in areas where smoke mask, AOT data are missing, plot the smoke mask or AOT from the previous granule. By doing this, we can show more retrievals in the images than in those if we only plot the latest retrievals. However, such treatments can also introduce artifacts on smoke mask and AOT images, such as showing up of the underlying granule edge, retrievals over cloud, low AOT in smoke plumes, etc.
We demonstrate this issue with an example of smoke mask plots over RGB in the following figures. In Figure 2, we show RGB images of VIIRS overpasses on August 21,2015 over central United States. There are two overpasses over part of the region as shown in the red circle in the bottom plot in Figure 1: one overpass is at about 1840 UTC (top left) and the other is at about 2020 UTC (top right). On eIDEA, we plot RGB image of the overpass at 2020 UTC on top of that of the overpass at 1840 UTC, as shown in the bottom plot in Figure 1.
Figure 3 shows the smoke mask of the above granules. The smoke mask are not exactly the same for the two granules in the overlapping region. A couple of reasons are the causes of the differences: (1) different Sun-Satellite geometries can cause the algorithm retrieve different results; (2) the distribution of the smoke and the cloud are different between the two granules. On eIDEA, we plot smoke mask of the overpass of 2020 UTC in the overlapping region if it detect smoke, otherwise, we plot smoke mask from the overpass at 1840 UTC if it detects smoke. The image shown on eIDEA is the bottom image in Figure 3. On that image, we can see the artifacts: (1) the edge of the granule 1840 UTC shows up; (2) smoke detected in cloudy region.
Figure 2. VIIRS RGB image on August 21, 2015.
Figure 3. VIIRS smoke mask overlaid on top of RGB image (August 21, 2015).