Unlike Optically thick clouds, Cirrus Clouds are thin, high altitude clouds formed in the upper troposphere layer of the Earth's Atmosphere. These Cirrus Clouds are not easily identifiable in the satellite images acquired with Passive Remote Sensing Sensors such as Landsat MSS, TM, ETM+, ASTER, SPOT, etc. Although there are different kinds of Cirrus Clouds, the sub-visible Cirrus Clouds are particularly of interest because they can be hiding in plain sight and affect the measurements. However, they can be detected within the Short-wave infrared (SWIR) portion of the electromagnetic spectrum, specifically at ~1.38 µm bouncing off of the ice-crystals in these clouds but are absorbed by water vapor in the lower part the atmosphere. Due to the benefits of this wavelength at 1.38 µm, MODIS (1999 onwards), VIIRS (2011 onwards), Landsat 8 (2013 onwards) and Sentinel 2 (2015 onwards) were introduced with their respective Cirrus Cloud detection bands.
However, in the absence of Cirrus detection bands in passive satellite sensors operating before 1999, is there anyway to pin-point the presence of Cirrus Clouds in historical satellite images? It may be possible to identify Cirrus Clouds in satellite images acquired without cirrus band by comparing it with contemporary/concurrent satellite images acquired with sensors having cirrus band. But otherwise, is there any other alternative way? Is anybody aware of any operational tool/algorithm/products that can identify cirrus clouds in past satellite imagery and provide means for their masking/correction?
This topic may be of particular relevance in time-series studies where historical satellite images are frequently compared with the present. For example, if cirrus scattering affects are not corrected, they can lead to incorrect interpretation in Vegetation Indices such as NDVI.
Sources:
- https://earthobservatory.nasa.gov/images/81210/new-landsat-finds-clouds-hiding-in-plain-sight
- https://earthobservatory.nasa.gov/images/2106/invisible-cirrus-clouds
- https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2008JD009972
- (https://www.mdpi.com/2072-4292/9/8/834)
- https://www.metoffice.gov.uk/weather/learn-about/weather/types-of-weather/clouds/high-clouds/cirrus
In such cases hyperspactral satellite images analysis followed by modelling technology play a lead role to identify the thin clouds and other similar phenomena. ..
An old and still valid question, indeed.
It might, in fact, still be difficult to catch these thin cirrus with passive instruments. Even test based on information from 1.38 micron channel may fail in case of low integrated water vapour content (e.g. in polar regions).
However, there are some more information even in thermal infrared channels. The split-window channels (around 11 and 12 microns) on older AVHRR instruments allow for detection of thin high cirrus to some extent. If you go further back in time it gets harder of course. Then success depends strongly on independent auxiliary information. For example accurate data of surface temperature together with satellite observations in thermal infrared might give some hints.
There are some long time series of space-born cloud observations, e.g.:
Article CLARA-A2: the second edition of the CM SAF cloud and radiati...
It really depends on the type of imaging. If, as mentioned before, there is spectrally resolved information, one can make use of the fact that the contribution functions of some channels peak at different altitudes and apply some kind of minimization technique (e.g. see our algorithm for IR sounders https://bit.ly/2VAG2NZ).
However, if the image was obtained only at one wavelength or the contribution functions are too broad or they do not "scan" the whole altitude range then it's more difficult. One can train neural network using this kind of images and a combination of active and passive sounders (e.g. AIRS + CALIOP) serving as a reference, but it's easier to say than to make :)
- Badges
- Science topic
- Similar topics
- Geoscience
- Remote Sensing