Assimilation of Accumulated Precipitation Data

References:

• Precipitation Data Assimilation in the WRFDA Users' Guide
• Brief presentation on advanced exercises

Source code

Get the pre-compiled code, if you have not done so.

Rainfall can only be assimilated with 4DVAR, so be sure that is the code you are using for this exercise.

WRFDA/var/build/da_wrfvar.exe is the executable that will be used in this practice session.

Working directory

We recommend you create a new working directory for this exercise:

mkdir /kumquat/users/$USER/DA/rainfall
cd /kumquat/users/$USER/DA/rainfall

Input data

The test data for this rainfall tutorial case can be found in /kumquat/wrfhelp/DATA/WRFDA/Rainfall

ln -fs /kumquat/wrfhelp/DATA/WRFDA/Rainfall/ob/ob.rain.2008020518.06h ob07.rain
ln -fs /kumquat/wrfhelp/DATA/WRFDA/Rainfall/rc/wrfinput_d01 ./fg
ln -fs /kumquat/wrfhelp/DATA/WRFDA/Rainfall/rc/wrfinput_d01 .
ln -fs /kumquat/wrfhelp/DATA/WRFDA/Rainfall/rc/wrfbdy_d01 .
ln -fs /kumquat/wrfhelp/DATA/WRFDA/Rainfall/be/be.dat .

These are the observation, first guess, boundary, and background error files, respectively. It is necessary to link a few more files for a 4D-VAR precipitation run

ln -fs /kumquat/users/${USER}/DA/WRFDA/run/RRTM_DATA_DBL ./RRTM_DATA
ln -fs /kumquat/users/${USER}/DA/WRFDA/run/*.TBL .

Edit namelists

A very basic namelist.input file for running this tutorial case is available at /kumquat/wrfhelp/DATA/WRFDA/Rainfall/namelist.input.

cp /kumquat/wrfhelp/DATA/WRFDA/Rainfall/namelist.input .
vi namelist.input

Pay special attention to the following namelist variables, making sure they are set to these values:

&wrfvar1
var4d=true,
var4d_lbc=false,
var4d_bin=3600,
var4d_bin_rain=21600,
...
/
...
&wrfvar4
use_rainobs=true,
thin_rainobs=true,
thin_mesh_conv=20.,
/

Run 4D-VAR with Accumulated Rainfall Observations

mpirun -np 6 /kumquat/users/${USER}/DA/WRFDA/var/build/da_wrfvar.exe

Note! In the provided namelist, we haven't set a small limit for ntmax (maximum number of iterations of the cost function minimization) like we did for the original practice 4DVAR case. Therefore, on the classroom computers this case will take some time, probably more than 10 minutes. If you would rather move on quickly and do not care about reaching the full minimization, set ntmax=5; this will still not be instantaneous, but will only take a few minutes.