short range ensemble forecasting at météo-france status and plans j. nicolau, météo-france
DESCRIPTION
Short range ensemble forecasting at Météo-France status and plans J. Nicolau, Météo-France. Outline. Operational configuration Post-processing LAMEPS experiment Combination of different SVs sets (Hungary) Forecasting Extreme Water Levels in Estuaries Forecasting uncertainty on wind power - PowerPoint PPT PresentationTRANSCRIPT
SRNWP workshop - Bologne
Short range ensemble forecasting at Météo-Francestatus and plans
J. Nicolau, Météo-France
SRNWP workshop - Bologne
Outline• Operational configuration• Post-processing• LAMEPS experiment• Combination of different SVs sets (Hungary)• Forecasting Extreme Water Levels in Estuaries• Forecasting uncertainty on wind power• Future plans
SRNWP workshop - Bologne
• Ensemble based on Météo-France operational global model ARPEGE (variable mesh) PEACE (Prévision d’Ensemble A Courte Échéance)
• T358c2.4 version (operational ARPEGE resolution ~20km over France)
• Initial state uncertainties with SVs • 11 members (ARPEGE operational + 10 perturbed
members)• 1 run/day (18H)• Up to 60h
Operational configuration
SRNWP workshop - Bologne
Singular Vectors computation
• Targeting over Atlantic Ocean and Western Europe
• Optimization time window : 0-12h• Total Energy norm (initial and final time)• Selection of the 16 first SVs• Linear combination according to the statistical
errors of analysis• No physics (except diffusion)• Low resolution : TL95 regular truncation
SRNWP workshop - Bologne
Svs targetting area
SRNWP workshop - Bologne
SRNWP workshop - Bologne
SRNWP workshop - Bologne
SRNWP workshop - Bologne
Post-processing : statistical adaptation
• Multi-linear regression (Pseudo-PP)• 2m Temperature, 10m Wind speed• Training (3 years) on the control run • Coefficients computed over the first 24h forecast• Then applied to the corresponding steps (ie coeffs
6h applied at 30h and 54h) + more spread - less unbiaised• ~500 sites over France
SRNWP workshop - Bologne
Calibration• Talagrand diagrams• Dynamical (last 40 days) and global (over all of the sites)
training– + sample size– - no local effects
• ~500 sites over France• 2m temperatures, 10m Wind speed, 24hTP
SRNWP workshop - Bologne
• PEACE coupled with ALADIN• 48H forecasts• 10km mesh grid• 20 cases of heavy precipitations• Compared to observations data (~1100 rain
gauges)
LAMEPS experiment (1)
SRNWP workshop - Bologne
HRHR
LAMEPS experiment (2)
SRNWP workshop - Bologne
Observations
Probabilities from PEACE/ALADIN
Probabilities from PEACE
Probabilities for 12hRR >5,10 and 20mm - Step 24h
24 Sept. 1999
LAMEPS experiment (3)
SRNWP workshop - Bologne
Combination of SVs– Collaboration MF/HMS (E.Hägel) – Experiments on different targetting areas– Computation of a set of SVs on Central Europe – Selection of independent SVs from the first set
and combination– Then integration with the global ARPEGE
based ensemble system for a ten day period
SRNWP workshop - Bologne
MSLP
T850 z500
WS10
Average standard deviation over Hungary for the period 10/07/2004-19/07/2004 for MSLP, ws10, T850, z500 for different experiments with the model ARPEGE
results with combination of SVs
results with the former PEACE target domain
SRNWP workshop - Bologne
Combination of SVs
• The improvement we achieved is comperable with the case when we changed only the target domain
• It would also be possible to use different target times for computing the two sets of SVs, maybe further improvement could be achieved
• Possibility to implement this new computation in the operational PEACE configuration.
• See Edit’s poster for further informations !
SRNWP workshop - Bologne
Uncertainty on wind power
• Project with :– EDF (wind power production data)– EMP (wind power production model)– MF (wind forecast distributions from PEACE)
• Evaluation over 18 months• Selection of wind power sites over France
SRNWP workshop - Bologne
SRNWP workshop - Bologne
Over levelObserved level
Astronomic tide
Under level
Nuclear power plant cooling
Forecasting Extreme Water Levels (1)
Storm surges
SRNWP workshop - Bologne
Storm Surges ModelInitial State
Bathymetry
Storm surges
Atmospheric Forcing (WS, WD, MSLP)
Forecasting Extreme Water Levels (2)
SRNWP workshop - Bologne
Wind speed
MSLP
Wind direction
Forecasting Extreme Water Levels (3)
SRNWP workshop - Bologne
Future plans
• Work on SVs (stretched, physics, area, time, combination)
• Model perturbation• Evolved perturbation• Downscaling
– Coupling with the limited area model ALADIN heavy precipitations predictions
• Post-processing
SRNWP workshop - Bologne
• Great scale forecast : – deterministic model – multi-model approach – different lagged forecast– Ensemble : stamps, spread– Multi-analysis ensemble (initial state sensitivity)
• Sensitive weather– Probabilistic forecat– Multi-ensemble approach
SRNWP workshop - Bologne
Discussion• How to deal with different
models/ensembles ?• How to rely uncertainty from ensembles to
the deterministic forecast from high resolution models (eg epsgrams vs det. Atmograms) ?
• How to rely different distributions (eg « mixed epsgrams ») ?