Moist Total Energy Norm (MTEN) diagnostic
MTEN
shows the sensitivity of the forecast model to different observations withdrawn from the full analysis system. There are two ways of computing the MTEN
diagnostic: A special branch was created in CY40 (see below) where the MTEN diagnostic can be requested. This approach uses Harmonie ensemble system to perform series of observation denial independent runs. This means that the following settings are used in msms/harmonie.pm
'ENSBDMBR' => [ 0 ],
'ENSCTL' => [ '000', '001', '002', '003', '004', '005', '006', '007' ],
'AIRCRAFT_OBS' => [ 0, 1, 1, 1, 1, 1, 1, 1],
'BUOY_OBS' => [ 1, 0, 1, 1, 1, 1, 1, 1],
'AMSUA_OBS' => [ 1, 1, 0, 1, 1, 1, 1, 1],
'AMSUB_OBS' => [ 1, 1, 1, 0, 1, 1, 1, 1],
'POL_OBS' => [ 1, 1, 1, 1, 0, 1, 1, 1],
'HRW_OBS' => [ 1, 1, 1, 1, 1, 0, 1, 1],
'TEMP_OBS' => [ 1, 1, 1, 1, 1, 1, 0, 1],
'IASI_OBS' => [ 1, 1, 1, 1, 1, 1, 1, 0],
In this particular example, we are interested in the impact of aircraft, Buoy, amsu-a, amsu-b/mhs, polar winds, high-resolution geowinds, radiosonde, and iasi observations. This setting is activated in config.exp
with the following choice:
export REFEXP DOMTEN
export SYNOP_OBS=1 # All synop
export AIRCRAFT_OBS=1 # AMDAR, AIREP, ACARS
export BUOY_OBS=1 # Buoy
export POL_OBS=1 # Satob polar winds
export GEO_OBS=0 # Satob geo winds
export HRW_OBS=1 # Satob HRWind
export TEMP_OBS=1 # TEMP, TEMPSHIP
export PILOT_OBS=1 # Pilot, Europrofiler
export SEVIRI_OBS=0 # Seviri radiances
export AMSUA_OBS=1 # AMSU-A
export AMSUB_OBS=1 # AMSU-B, MHS
export IASI_OBS=1 # IASI
export PAOB_OBS=0 # PAOB not defined everywhere
export SCATT_OBS=0 # Scatterometer data not defined everywhere
export LIMB_OBS=0 # LIMB observations, GPS Radio Occultations
export RADAR_OBS=0 # Radar
export GNSS_OBS=0 # GNSS
Where REFEXP
is the reference experiment (see below), and DOMTEN
(yes,no) is activate the MTEN
choice when fetching the First-guess and the VarBC files for the MTEN computation, as follows: in /scr/Fetch_assim_data
:
if [ ${DOMTEN} = "yes" ]; then
HM_REFEXP=/sbt/harmonie/$REFEXP
adir=${ECFSLOC}:${HM_REFEXP}/$YY/$MM/$DD/$HH
else
adir=$( ArchDir $HM_EXP $YY $MM $DD $HH )
fi
in scr/FirstGuess
(be careful this happens twice in the script)
if [ ${DOMTEN} = "yes" ]; then
HM_REFEXP=/sbt/harmonie/$REFEXP
adir=${ECFSLOC}:${HM_REFEXP}/$FGYY/$FGMM/$FGDD/$FGHH
else
adir=$( ArchDir $HM_EXP $FGYY $FGMM $FGDD $FGHH )
fi
- The
MTEN
can be also computed using a deterministic system. In this case, you need to take care of the First-guess and the VarBC files, which should come from the reference experiment. You need to carefully set the choice of the observations to be tested inscr/include.ass
. In this case, you need to adapt the aboveFetch_assim_data
andFirstGuess
scripts accordingly.
The MTEN
diagnostic, similarly to DFS
, is case sensitive, so it's better to male the computation with times and dates enough distant (by 5 days or more).
The MTEN
can be computed the example below:
for EXP in EXP1 EXP2;
for RANGE in 06 12 18 24 30 36 42 48;
do
YY=`echo $DTG | cut -c 1-4`
mm=`echo $DTG | cut -c 5-6`
dd=`echo $DTG | cut -c 7-8`
hh=`echo $DTG | cut -c 9-10`
# -- Get the FA files
# ===================
ecp ec:/$USER/harmonie/$REFEXP/$YY/$mm/$dd/$hh/ICMSHHARM+00$RANGE ./FAREF$RANGE
ecp ec:/$USER/harmonie/${EXP}/$YY/$mm/$dd/$hh/ICMSHHARM+00$RANGE ./${EXP}$RANGE
$MTEN_BIN/MTEN ./FAREF$RANGE ./${EXP}$RANGE
done
done
See (Storto and Randriamampianina, 2010) for more details.