seasonal sensitivity study on cobel-isba local forecast system for fog and low clouds
DESCRIPTION
SEASONAL sensitivity study on COBEL-ISBA LOCAL FORECAST SYSTEM for fog and low clouds at Paris CDG airport. ROQUELAURE Stevie and BERGOT Thierry. Météo-France. WSN05 september 2005. Outlines. + COBEL-ISBA numerical prediction model. - PowerPoint PPT PresentationTRANSCRIPT
SEASONAL sensitivity study on COBEL-ISBA LOCAL FORECAST SYSTEM for fog and low clouds at Paris CDG airport
ROQUELAURE StevieandBERGOT Thierry
Météo-FranceWSN05 september 2005
Outlines
Météo-France
+ COBEL-ISBA numerical prediction model
+ Advection sensitivity on 2002-2003 winter season
+ Cloud (fluxes) sensitivity on 2002-2003 winter season
+ 3h data assimilation vs 1h data assimilation (winter 2003-2004)
+ Conclusions
WSN05 september 2005
Cobel-Isba numerical prediction
Météo-France
3h -Assimilation : 3h -Assimilation : 1D--Var + fog and low Var + fog and low clouds initialisationclouds initialisation
Observations ISBA offline
COBEL/ISBA – 1D
12h local forecasts:Detection of LVP conditions (fog and low clouds) on 30minutesVisibility < 600m & ceiling < 200ft
forecasters
guess
Aladin forecasts (3D)
?
WSN05 september 2005
Advection (T,q) sensitivity
Météo-France
+ Simulations set (2002-2003 winter season): - ref : without advections - adv : temperature & humidity advections - AT + std : advections + temperature std - AT – std : advections - temperature std - Aq + std : advections + humidity std - Aq – std : advections - humidity std
Aq + std
Aq - std
ref (no advections)
Aq
AT
WSN05 september 2005
Temperature advection standard deviation distribution
0
0,1
0,2
0,3
0,4
0,5
0,6
0,7
0,8
-2 -1,5 -1 -0,5 0 0,5 1 1,5 2
Temperature advection mean (K/s)
Sta
nd
ard
de
via
tio
n (
K/s
)
Advections (T,q) STD distributions
Météo-France
+ Advection are calculated from NWP model Aladin - spatial and temporal consistency advection std distributions
Humidity advection standard deviation distribution
0
0,1
0,2
0,3
0,4
0,5
0,6
0,7
0,8
-1 -0,8 -0,6 -0,4 -0,2 0 0,2 0,4 0,6 0,8 1
Humidity advection mean ((g/kg)/s)
Sta
nd
ard
de
via
tio
n (
(g/k
g)/
s)
constant
linear linear
constant
linear linear
WSN05 september 2005
Advections (T,q) effects on LVP forecasts
Météo-France
+ All scores are from:
- 2002-2003 winter season (5 months dec-April) - 3h assimilation scheme (0Z, 3Z …., 21Z) (~1200 simulations)
+ Scores are calculated from all initialisations time
- Hite Rate = ( forecasted&observed LVP / observed LVP ) - False Alarm Rate = ( forecasted&non-observed LVP / non-observed LVP)
WSN05 september 2005
Advections (T,q) effects on LVP forecasts
Météo-France
Advection effect
+ No improvements with advections (HR & FAR)
+ Adv is slightly better than persistence (HR)
WSN05 september 2005
Temperature advection std effects on LVP forecasts
Météo-France
cold effect
+ No symetrical effect (cold/warm advection)
+ AT std impact is important on LVP (colder advection)
+ No major degradation in FAR
WSN05 september 2005
Humidity advection std effects on LVP forecasts
Météo-France
humidification
+ No symetrical effect (moist/dry advection)
+ AQ std impact is important on LVP (moist advection)
+ Degradation in FAR (~10%) with moist advection
WSN05 september 2005
Clouds initialisation
Météo-France
+ Initialisation of fog and low clouds iterative method for minimizing the divergence between observed & cobel radiation fluxes + Simulations set (2002-2003 winter): - ref : with low clouds & fog initialisation - no : no low clouds & fog initialisation - aladin : low clouds & fog initialisation + cloud cover from aladin (IR and VIS) - obs : low clouds & fog initialisation + persistence of observed clouds (IR)
WSN05 september 2005
Clouds (oberved IR vs COBEL IR)
Météo-France
NoBias=-44.2 W/m2
RefBias=-16.6 W/m2
AladinBias=-6.4 W/m2
ObsBias=-2.3 W/m2
WSN05 september 2005
Clouds effects on LVP forecasts
Météo-FranceWSN05 september 2005
LVP: 3h data assimilation vs 1h data assimilation
Météo-France
+ This result is from the work of MARZOUKI Hicham (master student), for winter season 2003-2004
+ The only difference between the 2 simulations is the data assimilation frequency (3h 1h) 8 initialisations/day 24 initialisations/day
WSN05 september 2005
LVP: 3h data assimilation vs 1h data assimilation
Météo-France
+ Improvements can be observed in HR (5%) & FAR by decreasing the data assimilation frequency to 1 hour.
WSN05 september 2005
CONCLUSIONS
Météo-France
+ Advections have a important impact on LVP forecasts - moist/cold advections impact is more significant dry/warm advections - advections impact is cumulative more important for long term simulation (4 -12 hours)
+ Fog and low clouds initialisation is very important - IR radiative impact significant improvement in FAR
+ Improvements can be observed in HR & FAR by decreasing the data assimilation frequency to 1 hour.
WSN05 september 2005
Thank you
Météo-FranceWSN05 september 2005
Advection STD computation
Météo-France
Time UTC
00Z 06Z 18Z12Z 24Z 30Z
+kmx 10
site
R=10 km
1- Advection spatial mean & std
2- Advection temporal mean & std
On all levels:
WSN05 september 2005