ems ljubljana, 2006 mathias d. müller 1, c. schmutz 2, e. parlow 3 an ensemble assimilation and...
DESCRIPTION
Initial conditions Initialization: - observations of temperature & humidity - 3D model data: aLMo, NMM-22, NMM-4, NMM-2 Data assimilationTRANSCRIPT
EMS LJUBLJANA, 2006
Mathias D. Müller1, C. Schmutz2, E. Parlow3
An ensemble assimilation and forecast system for 1D fog prediction
1,3) Institute of Meteorology, Climatology & Remote SensingUniversity of Basel, [email protected]
www.meteoblue.ch
2) MeteoSwiss
1D fog modeling (COBEL-NOAH and PAFOG)
Radiation land surface model
Turbulence microphysics
+ initial (IC) and boundary conditions (BC)
Initial conditions
Initialization:
- observations of temperature & humidity
- 3D model data: aLMo, NMM-22, NMM-4, NMM-2D
ata
assi
mila
tion
Boundary conditionsBoundary conditions:
From 3D models: aLMo, NMM-22, NMM-4, NMM-2
- Clouds
- Advection of temperature & humidity
Valley fog
3D
t
Initialization – Data assimilation
15 16 17 18 19 20 21 22 23 24 25 26 27 28
Temperature (°C)
analysis (x)
observation (y)background (xb)
error:
„the magic“
Temperatur20 2221.5
observationbackground analysis
B and R determine the relative importance
NMM-4 1400 UTC
large model and time dependence
Assimilation - B for 3 different 3D models (Winter)NMM-22 00 UTC
NMM-4 00 UTC
aLMo 00 UTC
Initialization – Data assimilation (example)
28 Nov 2004Zürich Airport
21 hour forecastof NMM-2
The ensemble forecast system
var
iatio
nal a
ssim
ilatio
n
B-m
atric
es
CO
BE
L-N
OA
H P
AFO
G
Obser -vations
3D-Model runs
post
-pro
cess
ing
Fog
fore
cast
per
iod
NM
M-4
NM
M-2
NM
M-2
2aL
Mo
3D - Forecast time
www.meteoblue.ch
1D-models
Different IC and BC
Ensemble Forecast - Example
fogHEI
GH
T (m
)
2 m Temperature (°C) 2 m rel. Hum. (%)
INITIALIZED:14 OCTOBER 2005 1500 UTC
100
90
80
70
60
50
14
12
16
10
8
6
4
Verification of the 1D ensemble forecast - ROC
FALSE ALARM RATE
HIT
RA
TE
no sk
ill
0
1
1
1040
60
Fog (observation) = visibility < 1000 m
Fog (model) = liquid water content > threshold has probability x
ROC
fog: 106060
Fog – yes/no?
Importance of Advection Sensitivity to humidity assimilation
Verification of the 1D ensemble forecast - ROC
03-11 UTC from 1 November 2004 until 30 April 2005
advection of cooler and drier air
cool warm dry humid
Hourly advection estimates (different 3D models)
03-11 UTC from 1 November 2004 until 30 April 2005
- Initialisierungszeitpunkt
- Multimodel
PAFOG
MODEL-ENSEMBLECOBEL-NOAH
15:00 UTC 18:00 UTC
21:00 UTC 00:00 UTC
Verification of the 1D ensemble forecast - ROC
• 1D ensemble forecast has the potential to improve fog prediction at Zürich airport:
• Advection (of cooler and drier air) is very important
• Humidity assimilation with large uncertainty → more observations, humidity ensemble
• COST-722
• MeteoSwiss
Conclusions
Ensemble Hit Rate False Alarm rate
COBEL-NOAHPAFOG
60 %80 %
30 %45 %1D
Than
ks
3D simulations even more promising
Model
satellite
Assimilation – R für Radiosonde in Payerne
Write in incremental Form
Introduce T and U transform to eliminate B from the cost function
(physical space)
(Control variable space)
Assimilation – inkrementelle cost function
NMC-Method (use 3D models):
Assimilation – Error covariance Matrix