Calibration of global MODIS cloud amount using CALIOP cloud profiles

Abstract. The Moderate Resolution Imaging Spectroradiometer (MODIS) cloud detection procedure classifies instantaneous fields of view (IFOV) as either confident cloudy, probably cloudy, probably clear, or confident clear. The cloud amount calculation requires quantitative cloud fractions to be assigned to these classes. The operational procedure used by NASA assumes that confident clear and probably clear IFOV are cloud-free (cloud fraction 0 %), while the remaining categories are completely filled with clouds (cloud fraction 100 %). This study demonstrates that this best guess approach is unreliable, especially on a regional/ local scale. We use data from the Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) instrument flown on the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) mission, collocated with MODIS/ Aqua IFOV. Based on 33,793,648 paired observations acquired in January and July 2015, we conclude that actual cloud fractions to be associated with MODIS cloud mask categories are 21.5 %, 27.7 %, 66.6 %, and 94.7 %. Spatial variability is significant, even within a single MODIS algorithm path, and the operational approach introduces uncertainties of up to 30 % of cloud amount, notably in the polar regions at night, and in selected locations over the northern hemisphere. Applications of MODIS data at ~10 degrees resolution (or finer) should first assess the extent of the error. Uncertainties were related to the efficiency of the cloud masking algorithm. Until the algorithm can be significantly modified, our method is a robust way to calibrate (correct) MODIS estimates. It can be also used for MODIS/ Terra data, and other missions where the footprint is collocated with CALIPSO.