Abstract
Abstract. Investigation of the focal plane assembly of the Sentinel-2 satellites show slight delays in the acquisition time of different bands on different CCD lines of about 0.5 to 1 second. This effect was already exploited in the detection of moving objects in very high resolution imagery as from WorldView-2 or -3 and also already for Sentinel-2 imagery. In our study we use the four 10-m-bands 2, 3, 4 and 8 (blue, green, red and near infrared) of Sentinel-2. In the level 1C processing each spectral band gets orthorectified separately on the same digital elevation model. So on the one hand moving objects on the ground experience a shift between the spectral bands. On the other hand objects not on the ground also show a slight shift between the spectral bands depending on the height of the object above ground. In this work we use this second effect. Analysis of cloudy Sentinel-2 scenes show small shifts of only one to two pixels depending on the height of the clouds above ground. So a new method based on algorithms for deriving dense digital elevation models from stereo imagery was developed to derive the cloud heights in Sentinel-2 images from the parallax from the 10-m-bands. After detailed description of the developed method it is applied to different cloudy Sentinel-2 images and the results are cross-checked using the shadows of the clouds together with the position of the sun at acquisition time.
Highlights
Actual high and very-high resolution Satellites like Sentinel2 or WorldView-2/-3 acquire images scanning the ground line by line while traveling along their orbit
In the study we present in this publication we use this second effect
In this paper we presented a new method for deriving cloud heights from spectral shifts in the visual 10-m-bands of the Sentinel-2 MSI instrument
Summary
Actual high and very-high resolution Satellites like Sentinel or WorldView-2/-3 acquire images scanning the ground line by line while traveling along their orbit Such a push-broomscanner consists of different CCD lines for each panchromatic and multispectral band acquiring one line at one time. 3 cause small time gaps between image bands clearly separated in the focal plane assembly – e.g. about one second between the blue and red CCD lines – as listed in table 1 from (ESA, 2021). A resumé of the work is given in the last section Conclusion and Outlook
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have