Hi Aleksandar,
>> I was wondering if it is possible to describe the pixel geometry of
>> satellite measurements in CF. In atmospheric trace gas remote sensing,
>> an individual measurement's location is often described by the center
>> point's lat/lon coordinates plus the four pixel corner points' lat/lon
>> coordinates. One measurement's location is thus described by a total of
>> five lat/lon points.
>
> I am not clear in what form is your data. Is it still in the sensor's
> field-of-view projection (swath) or already massaged into a grid. You
> mention "pixel corner points", that sounds like a grid. Typical sensor
> field-of-view projection onto the Earth geoid is an ellipse so no real
> corners there.
My data are still in swath form, there's no gridding involved. In all
cases I'm working with, the satellite data *do* have "corner points".
See for example the SCIAMACHY SciaL1c Software User Manual (SUM),
available at
http://earth.esa.int/services/SciaL1C_tool/SUM_SciaL1c_2B.pdf. Therein,
on page 38, A9, fields 8-10, it says
8 Sub-satellite point at the middle of the integration time
9 4 corner co-ordinates of the nadir ground pixel(the first
co-ordinate is the one which is the first in time and flight
direction, the second the first in time and last in flight
direction, the third the last in time and first in flight direction
and the fourth the last in time and flight direction)
10 Centre co-ordinate of the nadir ground pixel
For the GOME-2 instruments, the definition is similar (see p. 7 in
http://www.eumetsat.int/groups/ops/documents/document/pdf_ten_97232-eps-gome-pfs.pdf)
I'm interested in saving fields 9 and 10 in the files I intend to produce.
> The convention does not yet
> have a standard representation for satellite swath data.
Is this something which might change? I could try to make an according
proposal, if you guys think this would be within the scope of the CF
convention.
> The
> trajectory feature type is one way of improvising some types of swath
> data if each observation is done at a different time.
>
> Not in a standard way. Your new variables holding lat and lon corner
> points must share the dimension used for the trajectory points: the
> "observation" dimension.
I will probably go for one single dimension, "observation", and then add
variables as I need, adhering to the conventions where they exist. I
would be interested in getting a standard representation for satellite
swath data into CF, to have my files as standards-conforming as possible.
Cheers, Andreas.
Received on Thu Mar 21 2013 - 03:26:20 GMT
