Le sytème régional de prévisions des glaces (RIPS)Le sytème régional de prévisions des glaces (RIPS)
Optimisation du modèle et vérification des prévisions Optimisation du modèle et vérification des prévisions
Jean-François Lemieux, Christiane Beaudoin Jean-François Lemieux, Christiane Beaudoin
Collaborateurs Collaborateurs
FranFrançoisçois Roy Roy (CMC)(CMC), Gregory Smith , Gregory Smith (RPNE)(RPNE), Frédéric Dupont , Frédéric Dupont (CMC)(CMC),, Mark BuehnerMark Buehner (ARMA) (ARMA), Alain Caya , Alain Caya (ARMA),(ARMA), Patricia DeRepentigny Patricia DeRepentigny (CMC),(CMC), André Plante André Plante (CMC),(CMC), Paul Pestieau Paul Pestieau (CIS),(CIS), Tom Carrières Tom Carrières (CIS(CIS),), Pierre Pellerin (RPNE),Pierre Pellerin (RPNE), Gilles Garric Gilles Garric (Mercator(Mercator), Nicolas Ferry ), Nicolas Ferry (Mercator)(Mercator)
CMC 15 mars 2013CMC 15 mars 2013
• Plan
• 1) Brève descriptionde RIPS• 2) Optimization du modèle avec les bouées dérivantes• 3) Méthode de vérification• 4) Expériences de sensibilité du modèle :• - profondeur de la couche de mélange (MLD)• - taille des floes• 5) Évaluation de RIPS• 6) Résumé et futur
Responsabilités du Canada dans le projet METAREA- Émettre et disséminer les prévisions atmosphériques et maritimes (incluant la couverture de
glace) pour les régions Metarea 17 et 18- Phase 1 : RIPS (modèle de glace offline) utilisé pour produire les prévisions de glace- Phase 2 : Modèle de glace (CICE 4.1) couplé au modèle océanique (NEMO) CREG12 (F. Dupont)
- Phase 3 : Modèle pleinement couplé Atmosphère/Glace/Océan
Description of the ice model
The model used is the CICE Los Alamos sea ice model CICE version 4.1 (E.Hunke, W.Lipscomb - Documentation Nov 2008 )
It has several components :
- a thermodynamic model that computes local growth rates of snow and ice - a ice dynamics model that predicts the velocity field of the ice pack - a transport model that describes the advection of the ice concentration, ice volumes and others state variables - a ridging parameterization that transfers ice among thickness categories The number of ice categories used : ncat = 8 WMO standard ice thickness categories + 1 category :
10 - 15 – 30 – 50 – 70 – 120 – 200 >200 cm
Atmospheric forcing fields CMC RDPS forecast - Wind components - Temperature - Humidity - SW and LW Fluxes - Precipitation rates
Sea ice model
Ice forecastIce concentration (A)Ice velocity (u)Ice pressure (P)
OFF LINE ICE FORECAST Initial time
- 3d-var ice analysis Ice concentration (A) - Glorys1v1 climatology - Ice thickness (h) - Mixed layer depth (mld) - CMC SST analysis Sea surface temperature
- Previous ice forecast Ice velocity (u0) Mixed-layer ocean
Verification packageIce concentration
Ocean forcing field - Glorys1v1 climatology Ocean current (Uw)
Climatologie - Glorys1v1 - réanalyses océaniques globales - période 7ans 2002-2008 - résolution .25 deg - forçages atmosphériques dérivés des analyses
opérationnelles ECMWF - modèle océanique :Nemo - modèle de glace : Lim2 (2 catégories de glace)
Ice concentration 3d-var NA Analysis Valid 06 May 2010
- Ice model is run on
3d-var North American
ice analysis grid
- 5 km resolution (1640x1080)
- Forced by Gem regional
forecasts at 10km resolution
- Time step = 1200s
- Outputs every 3 hours
- Issued 4 times a day
00z, 06z, 12z, 18z
in experimental mode
R&D since july 2012
We optimize using the ice strength parameter P*We optimize using the ice strength parameter P*The resistance of ice to deformation P is proportional to P*.The resistance of ice to deformation P is proportional to P*.
P = P* h exp [-C (1-A)]P = P* h exp [-C (1-A)]P = ice strengthP = ice strengthh = ice thicknessh = ice thicknessC = empirical constant = 20C = empirical constant = 20A = total ice concentrationA = total ice concentration
Dansereau and Tremblay (in prep)Dansereau and Tremblay (in prep)Kreysher et al. 2000Kreysher et al. 2000
RIPS and drifting buoys optimizationRIPS and drifting buoys optimization
RIPS and drifting buoys optimization
Averaging over one yearAveraging over one yearAbout 20 buoys per dayAbout 20 buoys per day
RIPS and drifting buoys optimization
We calculate the following:We calculate the following:
RMSE and bias calculationsRMSE and bias calculations
where DSLO < 0.5 day and where DSLO < 0.5 day and
Verification mask against 3Dvar analysis - dslo (days since last obs) < 0.5 and- dslo (days since last obs) < 0.5 and - change in Aice (ice concentration) > 10%- change in Aice (ice concentration) > 10%
VerificationsVerifications
AA0h0h
AA48h48h
FF48h48h
Sensitivity to mixed layer depthSensitivity to mixed layer depth
48h forecast48h forecastNA regionNA region
- The mixed layer depth - The mixed layer depth (MLD)(MLD) best constant value was found for each month best constant value was found for each month - Climatological values - Climatological values (2-d fields)(2-d fields) give results as good as best value for each month give results as good as best value for each month
Sensitivity to ice floes diameterSensitivity to ice floes diameter(affecting the lateral melt)(affecting the lateral melt)
48h forecast48h forecastNA regionNA region
- - The value of 30m was found optimalThe value of 30m was found optimal
Prévisions faites pour toute l’année 2010 aves les paramètres optimaux :
- Epaisseur de couche de mélange climatologique- Diamètre des floes de glace = 30m- P* = 12,5kN/m2
Vérification des prévisions RIPS
region=NAregion=NAlead=48lead=48
region=Beringregion=Beringlead=48lead=48
region=NAregion=NAlead =24hlead =24h
region=NAregion=NAlead =48hlead =48h
Error field for 48h forecast starting 8 march Error field for 48h forecast starting 8 march 2010 18z2010 18z bb : dynamics + thermodynamics : dynamics + thermodynamics cc : dynamics only : dynamics only
Monthly verificationsMonthly verifications
- better than persistence for all months of 2010 but- better than persistence for all months of 2010 but- not statistically significant in January and March (bootstrap method 95%)- not statistically significant in January and March (bootstrap method 95%)
monthly RMSEmonthly RMSE
March persistenceMarch persistence
October persistenceOctober persistence October forecastOctober forecast
March forecastMarch forecast
Monthly verificationsMonthly verifications - using the latest and improved RIPS2 3D-Var analyses- using the latest and improved RIPS2 3D-Var analyses- note that RMS and bias values of persistence and forecast are reduced - note that RMS and bias values of persistence and forecast are reduced
Merci!!!
RIPS ouputs available everydayRIPS ouputs available everydayhttp://whxlab3.dart.ns.ec.gc.ca/~murthaj/rips/rips.phphttp://whxlab3.dart.ns.ec.gc.ca/~murthaj/rips/rips.php
SummarySummary
• RIPS is in mode R&D (4-48h forecasts / day) since july 2012. RIPS is in mode R&D (4-48h forecasts / day) since july 2012.
• Objective tuning of RIPS against drifting buoys.Objective tuning of RIPS against drifting buoys.
• Mixed layer depth (Mixed layer depth (MLDMLD) climatology improves skill during growing ) climatology improves skill during growing season. season.
• RIPS beats persistence almost all year RIPS beats persistence almost all year (more difficult in january, february, (more difficult in january, february, march).march).
• The two thickness category climatology is a weakness.The two thickness category climatology is a weakness.
•Publication soumise Q. J. R. Meteorol. Soc. (fév 2013) :Publication soumise Q. J. R. Meteorol. Soc. (fév 2013) :• The Regional Ice Prediction System (RIPS) : model optimization The Regional Ice Prediction System (RIPS) : model optimization and forecasts verificationand forecasts verification
FuturFutur
- Présentation CPOP 19 mars 2013 :- Présentation CPOP 19 mars 2013 :
Proposition de passe expérimentale pour le système régional des prévisions Proposition de passe expérimentale pour le système régional des prévisions des glaces (RIPS)des glaces (RIPS)
- RIPS- RIPS livrélivré aux opérations du CMC printemps 2013aux opérations du CMC printemps 2013
- Migration à la grille CREG12- Migration à la grille CREG12
De vieux proverbes nous donnent enfin des réponses !!!
Processus thermodynamique simplifié : << Il n’y a ni neige ni glace que le soleil ne fonde >>
Pour des prévisions de glace à long terme simples et précises : << A la Saint-Mathias se fond et se brise la glace (14mai) >>