MSG/SEVIRI -- Frequently Asked Question
What is MSG/SEVIRI-HDF data?
It is MSG/SEVIRI data in HDF format.
Each file contains radiance count for 11 bands of MSG/SEVIRI and calibration parameters for 12 bands. HDF was used because of its compressing feature (For the case of MSG/SEVIRI,65% of disk space will be saved.)
Recently, we updated this dataset by fixing a problem of geolocation shift in the previous version. (We will discuss this below in detail.) Only the version 2 is available online: "SEVIRI_HDF_v2"
What is the source of MSG/SEVIRI-HDF data?
Data were downloaded from the following McIDAS servers:
- SATEPSDIST6E.NESDIS.NOAA.GOV (realtime MSG data server)
- GEOARC02.SSEC.WISC.EDU (archived MSG data server)
- GP47.SSD.NESDIS.NOAA.GOV (retired in 2008)
Does the geolocation of pixels on Full Disk images remain the same all the time?
Yes. When MSG/SEVIRI Full disk data are released from EUMESAT, they already adjusted so that the center of image always is located at [0W,0N].
Do we do re-mapping or navigation when MSG/SEVIRI-HDF data are created?
No, we simply download the data using the McIDAS command below and convert them to HDF file.
IMGCOPY AMET08/FD GS.179 DAY=2006213 TIME=00:00:00 00:00:00 BAND= 1 SIZE=SAME
Is the SEVIRI_HDF data the same as the Full Disk data distributed by EUMESAT?
Yes for the current version: SEVIRI_HDF_v2.
For the first version of SEVIRI_HDF (which is not available now. The file have been converted to SEVIRI_HDF_v2), the answer is "No".
The difference between these two datasets is caused by the commands used:
- IMGCOPY AMET08/FD GS.179 DAY=2006213 TIME=00:00:00 00:00:00 BAND= 1 SIZE=3712 3712 LATLON=0 0 ( SEVIRI_HDF)
- IMGCOPY AMET08/FD GS.179 DAY=2006213 TIME=00:00:00 00:00:00 BAND= 1 SIZE=SAME (SEVIRI_HDF_v2)
Since the center of SEVIRI full disk is at 0N,0W, literally, it looks like the above commands should return the same array. But in fact, the returned images are not exactly the same: there are 2 pixels shift in both x and y directions. In order to get the same full disk image as EUMESAT, we have to use the second command.
Here is explanation of McIDAS help desk:
Although I agree that it seems like copying an image that IMGLIST says is 3712 x 3712 and centered at 0N 0W should give identical results whether you use IMGCOPY with LATLON=0 0 SIZE=3712 3712 or IMGCOPY with SIZE=SAME, that's not the case.
If you want to copy an entire band, using SIZE=SAME is always the best method. It literally copies the pixel-by-pixel data the way it is in the source image, whereas your example with LATLON=0 0 SIZE=3712 3712 has to do computations such that there will likely be a shift. This might help make the reason clearer: there is no true center pixel in the image because it has an even number of lines and elements, so if you specify an IMGCOPY with something like LATLON= (rather than SIZE=SAME) it has to choose a center pixel for the destination image. So in most (or maybe all?) cases, there will be an offset.
Fortunately, the data values and their lat/lon and image line/element will be the same whether you IMGCOPY with SIZE=SAME or with LATLON= and SIZE=. For example, if you run the following commands, you'll see that the lat/lon, image line/element, raw, radiance, reflectance and brightness temperature values are all the same. And if you flip back and forth between the images you'll see that their placement is the same within the frames (no navigation shift).
The calibration in HDF was received for McIDAS system.
---- "gain" and "offset" are for convert count to radiance. They come with SEVIRI data from EUMESAT.
---- "c1w3, c2w, alpha, beta" are provided by McIDAS system. It may change with the McIDAS software version.
Where does the missing blocks on data image come from?
When data are transmitted from EUMESAT to McIDAS server, each image is cut into many blocks. Full disk image will be put together when data pieces arrive the server. It is possible to loss some data block during data transmission.