SMHI meeting March 3, 2010

Draft Agenda

Meeting starting at 12:00
Ending around 17:00

1) Lunch
2) mpop – Status and plans (Martin)
3) mipp – Status and plans (Lars)
4) pyresample – Status and plans (Esben)
5) Information model (Adam)
6) Future plans
6.1 Database oriented processing (Esben)
6.2 System design ideas (Martin)
6.3 Discussion

7) Actions and next meeting
8) Prepare presentation for Nordsat meeting

Of course each point is open for discussion as needed.

Summary minutes

Esben & Lars arrived around 12:30 (train delayed)
Meeting participants: Esben S Nielsen and Lars Ørum Rasmussen (DMI
Anna Geidne, Martin Raspaud and Adam Dybbroe (SMHI)
1) Lunch was good.
Adam writes the minutes.
2) Martin presented the mpop status and plans
A little discussion on overlays (coastlines and political borders). mpop currently adopts the PPS overlay interface, but this will be decoupled. We will still need to be able to overlay contours for internal use (not for external product dissemination) but it is low priority. Check GMT contours. DMI use GMT for contour overlays. PPS reads one static file (which is not even global) directly in Python and use PIL. Go for a simple solution.

3) Lars presented the mipp software – status and plans
We discussed that hdf5dmi.py should be placed outside the package.
Also mipp should take care of missing segments (make sure mipp does not break down if a segment is missing which happens occasionally).
Slicing is needed: Should make it possible to extract only a subset of the whole (geostationary) satellite disc
Simplify the interface.
Check for possible (vicarious) calibration information for Met-7. Currently the data handled by mipp is not calibrated – no physical units (8 bit counts – no Tbs/Rad)
We need to extend and define a common configuration file format. At the moment mpop and mipp has different configuration file formats.

Coffee

4) pyresample overview by Esben.
Esben briefly described the re-projection method implemented in pyresample.
The trick is to map both the input array and the output (target) area onto the same 3d Cartesian grid.
kd-tree algorithm (scipy) is used for fast nearest neighbor search.