Data & Products
Missions & Sensors
Cloud physical parameters over Europe and Africa from MSG/SEVIRI
Cloud physical parameters from SEVIRI at WDC-RSAT
The lower image in Fig. 1 is the typical quicklook of an extract of the APOLLO/SEV cloud-mask. The original APOLLO/SEV cloud-mask contains the information of all applied cloud detection algorithms. In this quicklook, the colud-mask content is reduced to an overall cloud/snow/ice-information without details on the algorithms combined with a land/sea/sunglint-information from another APOLLO/SEV mask.
In APOLLO/SEV the values of the cloud optical depth can range from 0.5 to 500. Optical depths lower 0.5 are more typical for aerosols than for clouds and therefore not displayed in the lower image (Fig. 2). The highest values are associated with the highest water path values and the lowest cloud top temperatures over Central Africa.
The four images below (Fig. 3) show a subset in higher resolution around the Canary Islands. Besides the composite (upper-left) the cloud coverage (upper-right), cloud top temperature (lower-left) and cloud optical depth (lower-right) are displayed.
The AVHRR Processing scheme Over cLouds Land and Ocean (APOLLO) was the
first AVHRR data processing scheme to make use of all five spectral channels
during daytime. It was adapted to Meteosat Second Generation (MSG) SEVIRI
May 2004 has been chosen as epoch to compare SEVIRI-based with AVHRR-based monthly mean cloud fractions. The SEVIRI-based monthly mean cloud fractions have been computed from a collection of 245 full disk individual scenes in the time interval 9:00 – 11:00 UT. The time interval has been chosen to fit the morning overpass time range of NOAA-17. The AVHRR-based monthly mean cloud fractions have been computed from 78 individual NOAA-17 AVHRR scenes for covering the region of the "European Cloud Climatology" ECC, a DLR long-term data set of cloud parameters from AVHRR using APOLLO.
A comparison between cloud fractions from SEVIRI and from AVHRR both computed by the same methodology, APOLLO, can be done in various ways. Here, the most suitable comparison to find out the major differences using different data sources should be to look at mean values instead of individual scenes. For the region of the European Cloud Climatology ECC (34 N to 72 N, 11 W to 32 E) (Meerkötter et al, 2004) the comparison has been carried out on the basis of the mean cloud cover for May 2004. Since the AVHRR mean cloud cover has been composed of NOAA-17 morning data between 9:00 and 11:00 UT, the SEVIRI scenes from this time interval have been used to create the SEVIRI mean cloud cover. Actually the monthly mean cloud cover from AVHRR data is provided in the ECC map which is a rectangular projection with equidistant latitude-longitude grid. It has been remapped to the MSG projection for the comparison. To display the main results the mean cloud fractions are presented in the following for the full disk and for a European subset, together with the differences of the mean cloud fractions and some 2-dimensional histograms.
Figure 4 shows the full MSG disk monthly mean cloud fractions derived with APOLLO from SEVIRI scenes. The cloud cover is only displayed where it is composed of at least 100 values from 245 available individual scenes. The large scale global distribution of the cloud cover is represented clearly with low values along the Subtropics, the lowest values over the desert areas in Africa and around the Arabian Peninsula and the higher values in the Tropics and Mid-latitudes. Also some medium and small scale features caused by the Trade Winds can be seen, e.g. the lee-side effects south-east off the Canary Islands, the low cloud cover west off the African coast where dry air from the Sahara permanently suppresses the development of clouds, and the windward side orographic cloud cover over the south-east of Madagascar. The increased cloud cover over mountainous regions, e.g. the Alps, is partly caused by convective clouds which frequently appear over mountains in summer. On the other hand it can also be caused by snow pixels erroneously detected as clouds. Also this image shows that the cloud cover tends to increase systematically to the edge of the MSG disk due to slant-viewing effects. One should take into account that the satellite zenith angles for SEVIRI can reach more than 80º but for AVHRR they range only up to 69º. One outcome of this comparison is that cloud fractions from SEVIRI should be treated the more carefully the closer the edge of the disk is or even should be restricted to satellite zenith angles up to 69º at the maximum. This corresponds to about 61º latitude (at 0º longitude). Nevertheless in this study all pixels have been taken into account for the statistics.
Figure 5 and 6 show the monthly mean cloud cover from SEVIRI for the MSG European subset together with the region of the Canary Islands and from AVHRR for the ECC area mapped to the MSG European subset respectively.
The differences between SEVIRI and AVHRR monthly mean cloud fractions are shown in figure 7. The following conclusions can be made looking at the difference image:
For a quantitative analysis a statistical comparison of the SEVIRI-based monthly mean cloud fractions with the AVHRR-based has been done using 2-dimensional histograms. 2-d-histograms have been calculated for the total ECC area, for the northern and the southern half, and for sea respectively land only. This means that either all pixels of the ECC area count for the histogram or only the pixels of the northern/southern part or only sea/land pixels respectively. Such histograms put all value combinations into suitably predefined classes and provide statistical quantities of the resulting distribution. Since we deal with cloud fractions and for a later comparison with ground-based observations so-called Octa-classes have been chosen. This means that cloud fraction combinations have been sampled to boxes with the size of 12.5% by 12.5%. The figures 8-12 show the 2-d-histograms where the SEVIRI-based fractions are assigned to the y-axis and the AVHRR-based to the x-axis for five different regions or surface types. Boxes with no values are plotted in white. Boxes with less than 1% population, compared to the total number of pixels counting for the histogram, are plotted in grey. Boxes with at least 1% population are plotted in colours from blue over green, yellow and orange to red. The red colour indicates the boxes with the maximum population which varies depending on the distribution. The maximum occurring population is given together with the colour scale. Additionally the total populations of all Octa-boxes along the main as well as along the +1-Octa and -1-Octa diagonals are given. The total population of all these three diagonals (percentage above the percentages of the single diagonals) varies between about 70% and 80%, i.e. 70% - 80% of all pixels show differences with ±1 Octa. This result is comparable to the result of the comparison of APOLLO/AVHRR cloud fractions with ground observations (synoptic data) (Kriebel et al, 2003).
A major reprocessing is currently ongoing starting from February 2004 to today. Data is available on request.
Monthly animated gif formatted quicklooks (in reduced 928x928 pixel resolution) based on 1200 UTC images
of color composites (comp), cloud masks (mask) and cloud optical depth (odp) can
be found here.
For questions and more information please contact the SEVIRI team at DLR
Gesell, G., 1989: An Algorithm for Snow and Ice Detection Using AVHRR Data: An Extension to the APOLLO Software Package. International Journal of Remote Sensing, Vol. 10, Nos. 4 and 5, pp. 897-905
Kriebel, K.T., R.W. Saunders and G. Gesell, 1989: Optical Properties of Clouds Derived from Fully Cloudy AVHRR Pixels. Beiträge zur Physik der Atmosphäre, Vol. 62, No. 3, pp. 165-171, August 1989
Kriebel K. T., Gesell G., Kästner M., Mannstein H., The cloud analysis tool APOLLO: Improvements and Validation, Int. J. Rem. Sens., 24, 2389-2408, 2003
Meerkötter R., C. König, P. Bissolli, G. Gesell and H. Mannstein, 2004, A 14-year European Cloud Climatology from NOAA/AVHRR data in comparison to surface observations, Geophysical Research Letters, Vol. 31, L15103, doi:10.1029/2004GL020098
Saunders, R.W. and K.T. Kriebel, 1988: An improved method for detecting clear sky and cloudy radiances from AVHRR data. International Journal of Remote Sensing, 9, 123-150
Saunders, R.W., 1988: Cloud top temperature/height: A high resolution imagery product from AVHRR data, Meteorological Magazine, Vol 117, pp 211-221