|
http://www.atmos.washington.edu/academics/classes/2013Q1/380/HW3.html Due Thursday Jan 31 |
|
In this exercise you will learn about the purpose of running multiple ensemble members to investigate error in a forecast associated with errors in the initial conditions. You will also learn about error from model resolution. Our baroclinic wave case study is idealized, so our verification or "truth" is another run but at 0.5 degree resolution. This 0.5 deg run was initialized with the same Gaussian perturbation that you used last week. You will run a 2 deg resolution simulation with the Gaussian perturbation plus some random noise, which represents errors in the observations and or the method of constructing model initial conditions (called "analysis"). Summary: I. Make your very own initial condition files with a random perturbation to the baroclinic wave initial conditions (~5-10 min) II. Build and run CAM (~5 min, plus some wait time) to use your very own initial conditions file. III. Meanwhile analyze CAM output in MATLAB and write up answers to a few questions (a couple of hours max) I. Make your initial conditions Go to your camruns directory and make a subdirectory for this exercise called perturbed2deg and another subdirectory for matlab files for this exercise. qrsh cd /home/disk/p/atms380/$LOGNAME/camruns mkdir perturbed2deg mkdir perturbed2deg/run mkdir perturbed2deg/mfiles cd perturbed2deg/mfiles In this mfiles directory. Copy the matlab script to make the initial conditions and run it in matlab. It is meant to be self explanatory, but ask Cecilia if it is not. Remember, the dot at the end of the cp command puts the file in the current directory (dot = here). The & sign puts matlab in the background so you can continue to use the window to do other things. cp /home/disk/p/atms380/scripts/mk_cam_ICs_perturbed.m . matlab & Run the mk_cam_ICs_perturbed.m script in matlab and be sure to choose "Y" so it makes the output. Check out your initial conditions file back on the unix command line and put it in the run directory. Some of our machines fail when you use ncdump. Don't worry about it. The sys admin is working on fixing it. When done be sure to move your file to the run directory. ls -l *.nc after the dash is the letter l (one key to the right of k) ncdump -h cami_perturbed2deg_$LOGNAME.nc | more the character before more is a vertical bar mv cami_perturbed2deg_$LOGNAME.nc ../run II. Build and run CAM In the "work" directory. Copy the build script from the class script directory here. The dot at the end of the command puts the file in the current directory (dot = here). cd /home/disk/p/atms380/$LOGNAME/camruns/perturbed2deg cp /home/disk/p/atms380/scripts/bld-cam4-perturbed.csh . Execute the script: bld-cam4-perturbed.csh The build script checks to see if you put your perturbed initial conditions files in the run directory (the last step in part I). It will fail if it doesn't find it. If all is well, it will spew some stuff to the screen and take a few minutes before it says it is done. It will make the script run-cam4.csh for you, which you will submit to the queue. Also for sanity check, verify it is the queue. qsub run-cam4.csh qstat qstat -u "*" the latter lets you see all the jobs in the queue. If for some reason you wish to kill your job. qdel xxxxx fill in the x's with the job-ID and this will cancel your job III. Analyze CAM - turn in about a page on a-c below, plus figures. Get the matlab scripts for this exercise cp /home/disk/p/atms380/scripts/ex3* . a) Run ex3_a.m and follow the instructions that matlab displays. Turn in the anomaly correlations for the three pairs of fields. The Anomaly Correlation is a measure of how well the patterns match. We apply this to atmospheric two dimensional fields at a single time. b) Run ex3_b.m in matlab. This script has you examine the role of resolution on the Anomaly Correlation (AC). You will be studying runs I did earlier this week. Run the script and look at the figures. Figure 1 has the AC's for the 1, 2, and 4 deg runs always compared to the 0.5 degree run, which we consider "truth" (our verification). Figure 3 shows you the two dimensional fields that are being used to compute the AC. The AC is computed after the zonal mean has been removed. After looking at a few of days, hit 'S' and the scripts will compute the AC for each day and string them together into a timeseries (see Figure 2). To turn in: Try a range of days and variables and discuss your observations about what controls the Anomaly Correlation. Consider resolution, time, and variable. c) Run ex3_c.m in matlab. This script plots information about the wind component in the eastward direction or U-wind at 226hPa in a series of ensemble members that had initial conditions perturbed with a random number. All runs are 2 deg resolution. Nine ensemble members have been prepared for you in advance. Run the script once and then try to edit the script to include your runs as the 10th, by incrementing the variable ENS_MEMS. You probably won't notice one more line. If you feel ambitious, you can try to edit the script to include all the student's runs, which would increase ENS_MEMS to 18! Be my guest. Figure 1 is the AC for each member of the ensemble where the 0.5 degree run is still considered "truth". Figure 2 is the ensemble mean of U and Figure 3 is like Figure 2, but for each ensemble member. Figures 2 and 3 also have a black line showing the location of the U-wind maximum in the northern hemisphere. This is the jet maximum. To turn in: Try a range of days for Figs 2 and 3 (edit the script where it says "VARY THIS" at line 56) and discuss your observations about how the ensemble members differ. If you want to challenger your matlab skills, try to change the AC from being computed for the whole globe, to just use the northern hemisphere. Skip it if you are not loving matlab yet.
|
| Return to Homepage |