I received a question about GSM by email. Here's the question.

---

I have encountered some problems in running GSM with real time initial data and would like to seek your advices.

I downloaded the GSM codes using CVS and followed the on-line documentation to set up the model. I successfully ran the model (gsm6228, MPI version with 4 cpus) with the test data come with the model codes on an IBM p630 machine, which is of POWER4 architecture. (IBM SP is POWER3.)

Then I tried to run the global model on real time analysis data downloaded from NCEP's FTP server. The two initial data files downloaded from NCEP are gfs.t00z.sanl and gfs.t00z.sfcanl. I have to add an extension .ieee to the file names. Jack Ritchie has kindly provided some revised codes to me in order to ingest NCEP's data files.

However, the program scnvrt.x aborted and I got the following message in scnvrt.out:

DIMENSION MISMATCH
IMS= 1152 IDIM= 192
JMS= 576 JDIM= 94

Since Jack mentioned that the NCEP initial data are of higher resolution, I modified the following statements in the script gsm:

/home/gpfs1/user2/gsm/test0/libs/etc/scnvrt gsm $INITIAL_SFC_FILE_TYPE \
382:1152:576:64 \
-1:-1:-1:-1:-1 \
$fmt:bin $INITIAL_SIGMA:$INITIAL_SFC \
cnvrt_initial_sig:cnvrt_initial_sfc || exit 8

Then I got the following messages in fcstout.ft00 and the program fcst.x aborted.

0: spdmx(01:10)= 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.
0: spdmx(11:20)= 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.
0: spdmx(21:30)= 0. 0. 0. 0. 0. 0. 0. 0.
0: run failure. spdmax=0.
0: run failure. spdmax=0.

The leading zero on each row indicated that the message was from MPI task 0.

There is another interesting thing. I wanted to check if the serial version (MARCH=single) worked well. But the compilation of date.f (under the libs directory) failed with a lot of syntax error.

I am posting a question about compiling date.f in a separate thread.

---

There is another interesting thing. I wanted to check if the serial version (MARCH=single) worked well. But the compilation of date.f (under the libs directory) failed with a lot of syntax error.

Is it possbile to use the output from this run as input to the rsim script. What is the format of the sig.ftxx and sfc.ftxx output files (ieee?). I generated 6 h output of sig and sfc output files. I did not try to download the test data (rsim script test) but instead just used the output data from the gsm NCEP run.

The script blew up at rinpg.x. Do I need to do some conversion of the sig and sfc files? Appreciate your advice.

Sample output from rinpg.x

---

vln global coef has max= 3.467909443828019E+043 min= -3.314300416290822E+041
at k= 25
vln global coef has max= 1.085993185725898E+044 min= -1.439076189719652E+042
at k= 26
vln global coef has max= 3.092553898758515E+043 min= -1.089600549367152E+043
at k= 27
vln global coef has max= 3.923198026923298E+043 min= -5.599858670049712E+042
at k= 28
u/m global coef has max= NaN min= 0.000000000000000E+000
at k= 1
u/m global coef has max= 0.000000000000000E+000 min= 0.000000000000000E+000
at k= 2
u/m global coef has max= 0.000000000000000E+000 min= NaN
at k= 3
u/m global coef has max= 0.000000000000000E+000 min= NaN
at k= 4
u/m global coef has max= NaN min= NaN
at k= 5
u/m global coef has max= NaN min= NaN
at k= 6
u/m global coef has max= NaN min= NaN
at k= 7
u/m global coef has max= NaN min= NaN
at k= 8
u/m global coef has max= NaN min= NaN
at k= 9
u/m global coef has max= NaN min= NaN
at k= 10
u/m global coef has max= NaN min= NaN
at k= 11
u/m global coef has max= NaN min= NaN
at k= 12
u/m global coef has max= NaN min= NaN
at k= 13
u/m global coef has max= NaN min= NaN
at k= 14

---

Hi,
I have been running GSM using realtime NCEP GFS native output (surface and sigma files) as initial conditions for more than a year. My experience is that, with such high resolution initial data, GSM is able to forecast development and movement of tropical cyclones over the western North Pacific and the South China Sea. On the contrary, in those GSM runs using re-analysis data as initial conditions, the vortices dissipated too rapidly. As you may be aware, tropical cyclones have a great impact on rainfall forecast.
Apart from tropical cyclones, GSM (initialized by the high resolution initial data) also demonstrated the skill of forecasting cold surges of the northeast monsoon affecting southern China. The timing of temperature drop is quite accurate, though the amplitude or magnitude may need calibration. I will try to upload one or two graphs to illustrate what I have said.
I hope that you can continue to provide support in the initialization of GSM using realtime NCEP GFS native output. Thanks.
Sai Ming