GRAPES MESO模式学习笔记05——脚本运行之模式预报

模式积分原理完全不理解，只做脚本运行。

输入

grapesinput
grapesbdy
*.TBL
RRTM_DATA
namelist.input

输出

postvar  积分输出

modelvar_static

脚本

[shell]
#!/bin/ksh
# @ comment = GRAPES
# @ job_type = parallel
# @ input = /dev/null
# @ output = ./out/fcst_$(jobid).out
# @ error = ./out/fcst_$(jobid).err
# @ initialdir =./
# @ notification = error
# @ checkpoint = no
# @ node = 16
# @ tasks_per_node = 8
## @ wall_clock_limit = 4800
## @ class = benchmark
## @ class = normald
## @ class = smalljob
# @ class = normal
## @ class = middlejob
## @ class = largemem
# @ network.MPI = sn_single,,US
# @ queue
export MP_EAGER_LIMIT=32000
export MP_INFOLEVEL=2
export MP_BUFFER_MEM=64M
export XLSMPOPTS=”parthds=1:spins=0:yields=0:schedule=affinity:stack=50000000″
export OMP_NUM_THREADS=1
export AIXTHREAD_MNRATIO=1:1
export SPINLOOPTIME=500
export YIELDLOOPTIME=500
export OMP_DYNAMIC=FALSE,AIX_THREAD_SCOPE=S,MALLOCMULTIHEAP=TRUE
set -x
set -u
set -e
date
######################
# directory setting
######################
BASE_DIR=./
BASE_RUN_DIR=${BASE_DIR}/grapes_meso/my_grapes_meso/run
RUN_DIR=${BASE_RUN_DIR}/fcst
AN_RUN_DIR=${BASE_RUN_DIR}/an
SI_RUN_DIR=${BASE_RUN_DIR}/si
CONDAT_DIR=${BASE_DIR}/grapes_meso/my_grapes_meso/condat
GRAPES_BIN=${BASE_DIR}/grapes_meso/GRAPES_MESO3.3.2.4/fcst/grapes_model/run/grapes.exe
#####################
# enter work space
#####################
test -d ${RUN_DIR} || mkdir -p ${RUN_DIR}
cd ${RUN_DIR}
#####################
# prepare data
#####################
# tbl data
test -f GENPARM.TBL || cp ${CONDAT_DIR}/fcst/GENPARM.TBL .
test -f GEOGRID.TBL || cp ${CONDAT_DIR}/fcst/GEOGRID.TBL .
test -f LANDUSE.TBL || cp ${CONDAT_DIR}/fcst/LANDUSE.TBL .
test -f SOILPARM.TBL || cp ${CONDAT_DIR}/fcst/SOILPARM.TBL .
test -f VEGPARM.TBL || cp ${CONDAT_DIR}/fcst/VEGPARM.TBL .
test -f RRTM_DATA || cp ${CONDAT_DIR}/fcst/RRTM_DATA .
test -f grapesbdy || exit 1
test -f grapesinput || exit 1
###########
# namelist.input
###########
test -f namelist.input || rm -f namelist.input
cp ${SI_RUN_DIR}/namelist.input .
#####################
# run program
#####################
${GRAPES_BIN}
######################
# END
######################
date
[/shell]

后记

2014年7月1日
最近将测试数据换成2014年T639的26层输出，结果运行fcst模块一直积分溢出，不做同化依旧溢出，很可能是namelist中的参数设置有问题或者是其他什么原因，导致后续工作一直卡在这里。接下来我应该换另外一个时间段的数据试试，或者做好SMS监控系统便于设置参数。