Evaluation of a Cloud Detection Technique Using Spatial and Radiometric Thresholds for Near-Infrared Satellite Imagery

Knowledge of cloud location in near infrared (NIR) imagery is of interest to the meteorological community given its greater spatial resolution compared to longwave infrared and its potential nighttime applications. The NIR sensor bands of the Sentinel-2 and VIIRS satellite instruments are used to evaluate the performance of a monochromatic change detection technique designed to locate areas of cloud cover. The VIIRS imagery is examined for its higher temporal resolution compared to Sentinel-2; whereas, Sentinel-2 imagery is examined for its higher spatial resolution. This method consists of an algorithm that analyzes changes between satellite image radiances and a seasonal synthetic background radiance image. These images are differenced based on user-defined radiance, size, and shape thresholds. Pixels that meet these thresholds are flagged as cloud cover. Initial findings indicate that the developed algorithm can identify cloud cover above a specified size threshold, but optically thin clouds and small cloud clusters still present limitations.