Before you begin
What computer should I get?
Basically any up-to-date machine should work fine. Of course, the faster it is the quicker you will get your result. Watch out that the main memory supports your expected domain size (see page 14 of the documentation). METRAS-PCL runs in shared memory parallelisation, so more cores in one node make your runs faster. If you want to use your compute cores in task parallelisation (i.e. run several similar jobs at the same time) that would be even more efficient, but note that in this case your memory needs to contain all your runs at once. See page 14 of the documentation for some ressource estimates.
Linux?! Do I need a dedicated machine for this?
That would certainly be simpler in the long run. But if you want to start easy, you can also install a dual boot machine. Or if you want a bit more comfort run a virtual Linux machine. And maybe later you run your machine on Linux and install another operating system in a virtual machine.
And what Linux distribution is the best?
We use several of the big distributions in different versions and notice no difference. METRAS-PCL uses some system libraries that should be available on any recent distribution. We have not had any user who had problems due to the distribution they chose. Use whatever looks fairly easy to install and maintain. Remember to keep your system updated (all of them)!
AP output differs from model documentation (page 58):
- Structure number 1000 in the AP-file contains only the four variables (zeit, jn, dt, dtold) there are not two dummy variables which are described in the documentation. This will be corrected in the next model version.
- In the BP-file structure number 1000 is written as described in the documentation.
System issues before the program starts
METRAS-PCL_5.0.0: Command not found (also applies to gritop-l)
The system cannot automatically find the program executable. Check in which directory METRAS-PCL is stored. If it is in the current directory use '
./METRAS-PCL_5.0.0' to start the program. Otherwise include the full path in the command, e.g. '
Experts: For a permanent solution include the installation directory in your search path variable (
PATH). If you keep the model in the working directory include '
./' in the search path.
./METRAS-PCL_5.0.0: Permission denied (also applies to gritop-l)
When the program was installed locally the file permissions were not set to executable as required. Use '
chmod u+x METRAS-PCL_5.0.0' to make it executable for the current user or '
chmod ug+x METRAS-PCL_5.0.0' to make it available to other local users as well.
System issues at runtime
Segmentation fault (some users may not see the actual message, but runs just disappear without trace)
The memory requirements exceed the system settings. Increase allowed memory use according to the documentation section 3.3.6.
GRITOP-L seems to be taking a long time to run. Is it stuck?
The program basically compares every grid cell with every available data point. The supplied data covers Germany in a 10x10m2 resolution. That is many points. Allow some time for this. We have not experienced any case where the program is stuck in an infinte loop.
Hint: If the extent of the domain is given in geographic coordinates
(domain_geo = .TRUE.) the program should run substantially faster.
Geographic coordinates in METRAS-PCL:
- All geographic coordinates in METRAS-PCL use the format
- The only exception is the input for GRITOP-L, to simplify the input from common map applications you can set the model domain in decimal degrees
DD.ddddas well. However, the output will still be in