An experimental sky-image-derived cloud validation dataset for Sentinel-2 and Landsat 8 satellites over NASA GSFC

Availability of a reliable cloud mask for optical satellite imagery is prerequisite, when generating high-quality high-level geoinformation products. Creation reference (ground truth) moderate spatial resolution sensors, such as Operational Land Imager (OLI) aboard Landsat 8 and Multispectral Instrument (MSI) Sentinel-2A/B satellites, challenging time-consuming task. Existing datasets were mainly produced through photointerpretation images by an analyst, which can introduce subjectivity in detecting clouds. Therefore, other methods data shall be explored evaluated that complement existing datasets. In this paper, we document generation provide the description new dataset, named GSFC-Cloud, based on extensive use ground-based sky. The dataset collected over same area, covers various conditions, available six twenty-eight Sentinel-2 scenes spanning period September 2017 to November 2018. vector format, so masks at resolutions validated. We also describe system automate process collection using low-cost off-the-shelf parts with long-term objective replicate set-up multiple locations around world. proposed validate improve Surface Reflectance Code (LaSRC) detection imagery. show adding parallax feature estimate subpixel shift between red green bands phase correlation method reduce overdetection clouds performance LaSRC.