Embed Size (px)
Citation preview
Takuya KAWABATA ([email protected]), Tohru KURODA, Hiromu SEKO, Kazuo SAITO (Meteorological Research Institute/JMA)
NHM-4DVAR i s a cloud-resolvingnonhydrostatic 4D-Var data assimilation system
NHM-4DVAR v1 with dry dynamics and water vaporNerima heavy rainfall event on 21 July 1999
NHM-4DVAR v2 with warm rainHeavy rainfall event around Tokyo on 4 5 Sep 2005o yd ostat c a data ass at o syste
based on the JMA Nonhydrostatic Model (NHM), toinvestigate the mechanism of heavy rainfall eventsinduced by mesoscale convective systems (MCSs).The horizontal resolution of the NHM-4DVAR is set to2 km to resolve MCSs. .
NHM-4DVAR v1: The forward model is thefull nonlinear NHM, while the adjoint model onlyconsiders perturbations to the dry dynamics and theadvection of water vapor. Doppler radar’s radial winddata (RW), GPS precipitable water vapor data (GPS-PWV) and surface temperature and wind data areassimilated. The system was applied to a heavyrainfall event in Tokyo in July 1999 called the
Model specificationsCategory Forward model
(nonlinear NHM)adjoint model
Basic equations
Fully compressiblewith a map factor
O
Prognostic variables
Momentums, potential temperature, pressure, mixing ratios of water
Momentums, potential temperature, pressure, mixing ratio of
Nerima heavy rainfall event on 21 July 1999 Heavy rainfall event around Tokyo on 4-5 Sep. 2005
adjoint model
O
Momentums,potential temperature, pressure, mixing ratio of
Tokyo
A heavy rainfall event occurred around Tokyo on 4-5 Sep. 2005. Maximum amount of total precipitation reached 264 mm at Suginami in Tokyo.
rainfall event in Tokyo in July 1999, called theNerima heavy rainfall.. Time sequence of observedrainfall intensity is well reproduced in the assimilationwindow and the subsequent forecast. .
NHM-4DVAR v2: The warm rain cloudmicrophysical process has been implemented to theadjoint model. New control variables relevant to thewater substances are introduced, and the observationoperator of the radar reflectivity is developed. Thissystem was applied to a heavy rain fall event aroundTokyo occurred on 4-5 Sep. 2005 . With theassimilation of the radar reflectivity and mesoscaledata, location, horizontal size, and rainfall intensity ofthe observed rain band is well reproduced
pCR
p⎟⎟⎞
⎜⎜⎛
=π
∫+=z
surfsurfBdzzϑ
ππ )( Control variables
gvapor, cloud water, cloud ice, rain, snow and graupel, andkinetic turbulent energy.
gwater vapor
Cloud microphysics
3 ice bulk microphysics
Advection of water vapor
Variables Formula
gwater vapor,cloud water,rain water
Warm rain process
Variables Formula Assimilation area
Assimilation period peak of heavy rainfall
the observed rain band is well reproduced. .p ⎟⎟⎠
⎜⎜⎝
=0
π Wind u, v, w
Non-Hydrostatic pressure πU = π - πB
Potential temperature θ
Surface pressure Ps
Pseudo relative humidity qu = qv / qsb
Wind u, v, w
Non-Hydrostatic pressure πU = π - πB
Potential temperature θ
Surface pressure Ps
Total water qv+qc
Rain water qr / qsb
Pseudo relative humidity(for LBCs)
qu = qv / qsb
Assimilation resultLeft : PWVcircles : observationscontour : analysis
Right: surface windred arrows : observationsblack arrows : analysiscolor contour : topography
Assimilation area
( )( )75.110log10 qrdnscqrZ qr ×××=
cqr : constant , dns : density, qr : mixing ratio of rain waterAssumption: Marshall-Palmer drop size distributionObservational error:15dBZ
051015202530
1510 1520 1530 1540 1550 1600 1610 1620 1630 1640 1650
mm
JST
Observation ForecastForecast results
Observation
Ob i
Radar reflectivity (dBZ)
1448JST 1507JST 1525JST
10-minute rainfall amount at Nerima
1545JST
Hourly accumulated rainfall amount
Observation operator for reflectivity (Z-Qr relation)
HANEDA airport radar elevation angle = 0.7
Formation and development of the convective cells, which caused the Nerima heavy rainfall, are well reproduced.
10-minute rainfall amount of the forecast are quantitatively in good agreement with the observation.
Assimilation resultsAnalysis reproduces the location, intensity and horizontal size of themain rain band, furthermore, convective cells located on the east of themain rain band.
0.1 1.0 10.0 20.0 30.0 50.0 70.0(mm/h)
1999. 7.21.15-16JST
0 200km
(Valid:21.1600JST) 2hour 0min
Forecast
Observation
Forecast
1450JST 1510JST 1530JST 1550JST Nerimaheavy rainfall
A
AA A A
A A A
BBB
B
CC
Radar reflectivity (dBZ)Observation
2059JST2039JST2019JSTHANEDA airport radar elevation angle = 0.7
1999.07.2114JST
15JST 16JST 17JST
Assimilation window (1 hour)
3-hour forecast
Time evolution of Nerima heavy rainfall Formation Develop Mature
Design of the assimilation experiment 0.1 1.0 10.0 20.0 30.0 50.0 70.0
max
=10
3.90
0 50km
min
=-.
0001
9
(Valid:21.1600JST) 5hour 0min
Fist guess
Noheavy rainfall
C C C
Analysis2059JST2039JST2019JST
GPS-PWV5 minute interval
Radial Wind1 minute interval
Surface wind & temperature
10 minute interval
2005.09.04Assimilation window (10 min)
GPS-PWV5 minute interval
Radial Wind1 minute interval
Surface wind & temperature
10 minute interval
Reflectivity1 minute interval
0.1 1.0 10.0 20.0 30.0 50.0 70.0
max
=14
.643
mi
0 50km Rainfall intensity(mm/h)Observation2100JST
Observation
nothing
If reflectivity isover 10 dBZ in the first
guess field andunder 10 dBZ in the
observations, the grid is regarded as the false precipitation to be
li i t d d “0 dBZ”
Assimilate “0 dBZ”
nothing
Assimilation of “0 dBZ” observation
2010 2020 2030 2040 2050 210020JST
Design of the assimilation experimentAnalysis
AcknowledgmentThe GPS precipitable water vapor data are reanalyzed and provided by Mr. Shoji of MRI, the Doppler radar data are provided by Dr. Akaeda of JMA, and the UMIHOTARU surface observation data are provided by the Ministry of the Environment of Japan. The authors express our gratitude to them.
ReferenceKawabata et al, 2007: An Assimilation and Forecasting Experiment of the Nerima Heavy Rainfall with a Cloud-Resolving Nonhydrostatic 4-Dimensiojnal Variational Data Assimilation System, Journal of the Meteorological Society of Japan, Vol. 85, No. 3, 255-276
First guess
failure
eliminated, and “0 dBZ” information is used.
Observational error:normal error x 4
“0 dBZ” observations are NOT real observations.
Without “0 dBZ”
failure
0.1 1.0 10.0 20.0 30.0 40.0 50.0
max
=23
4.16
Prec mm/h (z*= 20m) (Valid:04.2100JST) 0hour 10min
0 50km
Initial : 2005.09.04.1150UTC2100JST
