Spatiotemporal High-Resolution Cloud Mapping with a Ground-Based IR Scanner

Abstract

The high spatiotemporal variability of clouds requires automated monitoring systems. This study presents a retrieval algorithm that evaluates observations of a hemispherically scanning thermal infrared radiometer, the NubiScope, to produce georeferenced, spatially explicit cloud maps. The algorithm uses atmospheric temperature and moisture profiles and an atmospheric radiative transfer code to differentiate between cloudy and cloudless measurements. In case of a cloud, it estimates its position by using the temperature profile and viewing geometry. The proposed algorithm was tested with 25 cloud maps generated by the Fmask algorithm from Landsat 7 images. The overall cloud detection rate was ranging from 0.607 for zenith angles of 0 to 10° to 0.298 for 50–60° on a pixel basis. The overall detection of cloudless pixels was 0.987 for zenith angles of 30–40° and much more stable over the whole range of zenith angles compared to cloud detection. This proves the algorithm’s capability in detecting clouds, but even better cloudless areas. Cloud-base height was best estimated up to a height of 4000 m compared to ceilometer base heights but showed large deviation above that level. This study shows the potential of the NubiScope system to produce high spatial and temporal resolution cloud maps. Future development is needed for a more accurate determination of cloud height with thermal infrared measurements.