However, with 3.2.1 beta release, things changed considerably. Among all news announced in this new release, the possibility to start reanalysis with no effort is, IMHO, the greatest one enhancement to WRF EMS 3.2.1.
You can still apply same procedure and end up with the same result, but now, Robert did step forward and enhanced it's scripts so they can be used directly to engage history reanalyse, straight from the shell command line! No need to find history grib download server, no need to register anywhere, no need to download by hand anything or adjust any naming convention... let's see how things work!
Running history reanalysis with WRF EMS 3.2 is very easy. You just have to install WRF EMS (you probably have already?) and create your domain (with or without nests) for simulation. You do that same as you would for any realtime forecast, like explained in official WRF EMS users guide, using domain wizard (dwiz), or, like I always do, using ems_domain.pl script from Robert. That did not changed much from previous version.
Now, when you prepared your domain, all you need is to decide which dataset you want to use. For that purporse, look at this file:
Code: Select all
wrfems/docs/wrfems/strc_datasets.txt
I will show you an example, how to start reanalysis based on ECMWF Interim Global Reanalysis data(*). This is also one of great news, Robert did included not only the scripts for ECMWF Interim gribs, but also he put whole grib dataset on his download server! Data is available from 1989 up to 2008 year. It is in very good 0.7 degree grid resolution, much better than NCEP/FNL reanalysis (2.5 degrees). That means you can start your simulation with pretty fine mother domain grid, and still achieve very high resolutions with only two or three nest levels (of course, always use ARW core for very fine grids).
After creating domain and setting configuration details like you want, starting reanalysis is a piece of cake. Let say, you want to run simulation on your domain with 4 nest levels, starting at date 21/12/1998, with 00z ECMWF Interim data, and simulation should be run up to 00z, 22/12/1998 (that is, 24 hours). You just change directory into your domain:
Code: Select all
cd $EMS/runs/<domainname>
Code: Select all
ems_prep --dset ecmwf --analysis --date 19981221 --cycle 00 --domain 1,2,3,4 --length 24
Code: Select all
ems_run --domain 1,2,3,4
Code: Select all
ems_post --domain 1
ems_post --domain 3
ems_post --domain 4
(*) If you use ECMWF data, and want to publish your work, you must follow some legal requirements. Please read more about that in strc_datasets.txt file.