Cloud mask and cloud fraction derived from solar reflectivity measurements of the imaging spectrometer specMACS on HALO during the EUREC4A field campaign

The data of the shortwave-infrared (SWIR) line camera of the spectrometer of the Munich Aerosol Cloud Scanner (specMACS) of the LMU was used to provide a cloud mask. During EUREC4A specMACS was mounted onboard the research aircraft HALO in downward looking perspective. The SWIR camera measures in the spectral range between 1015 nm and 2496 nm with a spectral bandwidth between 7.5–12 nm at a constant frame rate of 30 Hz. The instrument has a field-of-view of 35.5° which corresponds to a ground swath width of about 6.4 km at a typical flight altitude of 10 km.
The cloud mask is based on two criteria: the overall brightness of the observed pixels and the strength of absorption due to water vapor. From the downward looking perspective clouds are brighter than the dark sea surface most of the time. Nonetheless, sometimes the sunlight reflected by the sea surface – the sunglint – can lead to the opposite. Then the observed water vapor absorption is used as a secondary signal. In clear sky situations light which has been reflected on the ocean surface is absorbed more strongly due to water vapor, compared to light which is reflected at cloud tops before reaching the atmosphere layers with the largest amount of water vapor close to the ocean surface.
For evaluating the scene’s brightness and water vapor absorption strength reference spectra (1015 - 1300 nm) are fitted to the data. For the brightness signal two different thresholds are used to discriminate between “most likely cloudy” (if the brightness of a pixel is higher than the upper threshold), “probably cloudy“ (if the brightness is between both thresholds) and “cloud free” pixels (if the brightness is smaller than the lower threshold). This brightness criterion is not sufficient for ocean areas influenced by sunglint. Sunglint situations are identified by theoretical considerations depending on solar illumination and viewing geometry. In case of sunglint the method additionally evaluates the water vapor absorption for the classes “probably” and “most likely cloudy” pixels. Pixels with a strong vapor absorption signal are set to “cloud free”. A pixel is classified as “unknown”, if the measurement is not suitable for deriving a cloud mask due to, e.g., window icing.