The use of global synthetic data for pre-launch tuning of the VIIRS cloud mask algorithm

Abstract

A methodology is presented to perform pre-launch tuning of thresholds used in the Visible Infrared Imager/Radiometer Suite (VIIRS) cloud mask (VCM) algorithm. The approach relies upon several data sources, including global synthetic data (GSD), Moderate Resolution Imaging Spectroradiometer (MODIS) and VIIRS relative spectral responses (RSRs) and MODIS top-of-atmosphere (TOA) radiance, reflectance and brightness temperature data. The GSD are used first to derive cloud cover distributions, that is, 0%, 50% and 100%, at VIIRS moderate resolution (800 m) for each VCM cloud detection test, based on inputs to the radiative transfer models. These cloud distributions are then used to update the values of the low, mid and high cloud-free confidence thresholds in the VCM algorithm. The approach is demonstrated by using MODIS RSRs with the GSD to set these thresholds and then analysing granules of MODIS data with the updated VCM. Performance is quantified through comparisons with manually generated cloud masks created from the MODIS imagery. The performance of the tuned VCM with MODIS data improved substantially for all major background conditions. The probability of correct typing (PCT) improved nearly 4% over the ocean to 97.5% and nearly 20% over snow-covered surfaces to 95.1%. The PCT values over land improved from 87.1% to 93.4% and over desert from 87.2% to 93.9%. The process was then repeated using VIIRS RSRs and the updated thresholds were forwarded to the National Polar-Orbiting Operational Satellite System (NPOESS) Preparatory Project (NPP) ground segment for incorporation into the operational system. It is concluded that GSD are invaluable for the pre-launch tuning of the VCM algorithm, which is now expected to exceed system requirements soon after the launch of the NPP satellite.