Development of a daytime polar cloud mask using theoretical models of near-infrared bidirectional reflectance for ARM and CERES

Scene identification from most satellites over polar regions is difficult because clouds are often similar to the underlying surface in terms of temperature and visible reflectance. During the day, however, the brightness temperature difference (BTD) between the 3.7-11.0 µm channels on many satellites is greater for clouds than for clear snow allowing for discrimination between clouds and snow. This difference is in the reflected solar component of the solar infrared (SIR) 3.7-µm channel. Similarly, clouds are usually more reflective than snow at 1.6-µm, a near-infrared (NIR) channel on several new satellites. Typically, these channels are used for detecting snow based on either reflectance or temperature difference thresholds that are empirically established. While more accurate than using visible or infrared thresholds alone, these empirical NIR or SIR thresholds are less than satisfactory because of their wide variability, especially with viewing and illumination angles. To minimize the need for empirically adjusting the thresholds for a given set of conditions, to reduce the error accrued from using single-value thresholds, and to facilitate more accurate automated scene identification over snow-bound regions, a better characterization of the bidirectional reflectance patterns of snow at 3.7-µm is needed. To improve the cloud masks used by the Atmospheric Radiation Measurement (ARM) satellite program and the Clouds and Earth’s Radiant Energy System (CERES) project, bidirectional reflectance models of snow at the NIR and SIR wavelengths are constructed from the results of radiative transfer calculations using several different large ice crystal shapes to simulate snow grains. The SIR models are used to compute BTDs at different angles and compared to results from manual analyses of Advanced Very High Resolution Radiometer (AVHRR) NOAA-12 and 14 1998 daytime images taken during the Surface Heat Budget of the Arctic Ocean (SHEBA) and First International Satellite Cloud Climatology Project (ISCCP) Regional Experiment Arctic Cloud Experiment (FIRE-ACE). The same models plus the NIR models are also used to classify polar scenes observed with the Moderate-Resolution Imaging Spectroradiometer (MODIS) onboard the Terra satellite. The resulting scene identification imagery are evaluated manually to estimate large-scale errors and using data taken over the ARM Site at Barrow, Alaska. Recommendations for tuning or selecting specific theoretical reflectance models are developed from an analysis of the results. By developing theoretically based masks for these two wavelengths, it should be possible to consistently detect clouds over the Arctic and other snow-covered areas using a variety of different satellites.