GEM Model - Status Nov 2006

Subject: [Wxo_grib] Changes to CMC's global forecast system --
Changements au système global de prévision du CMC

La version française de la note suit ci-dessous.

NOCN03 CWAO 241910

GENOT TLTP. NO.001

ON TUESDAY OCTOBER 31TH 2006, AT 12 UTC, THE CANADIAN
METEOROLOGICAL CENTRE WILL IMPLEMENT MAJOR CHANGES TO THE
OPERATIONAL GLOBAL FORECAST SYSTEM INCREASING SIGNIFICANTLY THE
RESOLUTION OF THE GLOBAL ENVIRONMENTAL MULTISCALE (GEM) MODEL AND
IMPROVING ITS PHYSICAL PARAMETERIZATION.

SUMMARY OF THE CHANGES TO THE GLOBAL CONFIGURATION OF THE GEM MODEL

DYNAMICAL CONFIGURATION

THE FORECAST MODEL HORIZONTAL RESOLUTION WILL SIGNIFICANTLY
INCREASE FROM ABOUT 100 KM (400 X 200 GRID POINTS) TO NEARLY 33 KM
AT MID LATITUDES (800 X 600 GRID POINTS). THE NUMBER OF VERTICAL
LEVELS WILL INCREASE FROM 28 TO 58, THE TOP OF THE MODEL REMAINING
AT 10 HPA. THE TIME STEP WILL CONSEQUENTLY BE REDUCED FROM 2700 TO
900 SECONDS.
THE SPONGE LAYER NEAR THE MODEL LID WILL INCLUDE 4 LEVELS INSTEAD
OF 1. IN THIS LAYER, THE HORIZONTAL DIFFUSION IS INCREASED IN ORDER
TO MINIMIZE THE NEGATIVE IMPACT OF SPURIOUS WAVES REFLECTED AT THE
LID.
PHYSICAL PARAMETERIZATIONS
THE PHYSICAL CONFIGURATION OF THE FORECAST MODEL WILL BE
SUBSTANTIALLY MODIFIED COMPARED TO THE OLD OPERATIONAL VERSION. THE
CONDENSATION AND PRECIPITATION PACKAGE, IN PARTICULAR, WAS CHANGED
QUITE DRASTICALLY.
THE KAIN-FRITSCH SCHEME WILL REPLACE THE KUO SCHEME FOR THE DEEP
CONVECTION. THIS ALLOWS FOR A LARGER CONTRIBUTION OF THE GRID-SCALE
(RESOLVED) CONDENSATION PROCESSES WHICH IS CONSISTENT WITH THE
INCREASE IN RESOLUTION. THE SUNDQUIST GRID-SCALE CONDENSATION
SCHEME WAS MODIFIED MOSTLY WITH RESPECT TO THE PRECIPITATION
EVAPORATION BELOW CLOUD BASE, WHICH IS DONE OVER SEVERAL LEVELS IN
THE NEW VERSION. ANOTHER CHANGE TO THE CONDENSATION SUITE IS THE
INCLUSION OF A SHALLOW CONVECTION SCHEME, BASED ON A KUO SCHEME
CLOSURE, CALLED KUO TRANSIENT.
THE SURFACE MODELING SCHEME KNOWN AS ISBA (INTERACTIONS, SURFACE,
BIOSPHERE, AND ATMOSPHERE) WILL REPLACE THE SO-CALLED
'FORCE-RESTORE' MODULE. ISBA IS MORE SOPHISTICATED IN ITS TREATMENT
OF SOIL, VEGETATION, AND SNOW. TOGETHER WITH ISBA, A LAND-SURFACE
6-HOURS DATA ASSIMILATION SYSTEM WILL BE IMPLEMENTED IN ORDER TO
PROVIDE INITIAL CONDITIONS OF SURFACE TEMPERATURES AND MOISTURE.
THE INCREASE IN RESOLUTION OF THE MODEL AND THE USE OF ISBA MAKES
THE SNOW ANALYSIS MORE PRECISE.
ANOTHER MODEL PHYSICS MODIFICATION IS THE USE OF THE BOUGEAULT AND
LACARRERE MIXING LENGTH FOR VERTICAL DIFFUSION DUE TO ATMOSPHERIC
TURBULENCE, PROVIDING A CLEAR IMPROVEMENT OVER THE ONE PREVIOUSLY
USED, ESPECIALLY FOR CONVECTIVE, WELL-MIXED, BOUNDARY LAYERS.

SUMMARY OF THE CHANGES TO THE GLOBAL DATA ASSIMILATION SYSTEM

A NEW SET OF BACKGROUND ERROR STATISTICS ON THE 58 MODEL LEVELS HAS
BEEN COMPUTED USING THE SO-CALLED NMC METHOD. AS IN THE PREVIOUS
IMPLEMENTATION OF 4D-VAR, A LOW-RESOLUTION MODEL IS USED TO
PROPAGATE THE ANALYSIS INCREMENTS (T108) OVER THE 6 HOURS DATA
ASSIMILATION WINDOW. THE SET OF PHYSICAL PARAMETERIZATIONS FOR THIS
MODEL HAS BEEN CHANGED TO THOSE NOW USED IN THE NEW FORECAST MODEL
DESCRIBED ABOVE. HOWEVER, THE CORRESPONDING SIMPLIFIED PHYSICAL
PARAMETERIZATIONS USED IN THE TANGENT LINEAR MODEL AND ITS ADJOINT
MODEL REMAIN THE SAME AS BEFORE.
THE COMPUTATIONAL EFFICIENCY OF THE 4D-VAR DATA ASSIMILATION SYSTEM
HAS BEEN IMPROVED BY 40 OVERALL. THE CUT-OFF TIMES FOR AVAILABILITY
OF OBSERVATIONS HAVE BEEN SLIGHTLY MODIFIED TO DELIVER THE
OPERATIONAL ANALYSES AT ABOUT THE SAME TIME AS IN THE PREVIOUS
GLOBAL FORECAST SYSTEM.

IMPACTS ON PRODUCTS AND POST-PROCESSING

ALL PRODUCTS BASED ON OUTPUTS OF THE NEW GLOBAL MODEL HAVE BEEN
GENERATED AND VALIDATED DURING THE PARALLEL RUN. SOME CHANGES WERE
DONE TO THE AVIATION WEATHER PACKAGE. THE THRESHOLD VALUES USED TO
ASSESS MODERATE AND SEVERE TURBULENCE WERE ADJUSTED TO BE
CONSISTENT WITH THOSE USED ON THE REGIONAL MODEL. AESTHETIC
MODIFICATIONS HAVE ALSO BEEN DONE ON THE CONTOURING OF THE FREEZING
LEVEL AND TROPOPAUSE HEIGHT CHART. IN THE CLASSICAL 4-PANEL CHARTS,
THE MEAN SEA LEVEL PRESSURE IS NOW FILTERED. THE FILTERING OF
VORTICITY AND VERTICAL MOTION HAS BEEN INCREASED. THE 10 AND 30
PERCENT CONTOURS OF RELATIVE HUMIDITY HAVE BEEN ELIMINATED. ALL
OTHER PRODUCTS WERE VALIDATED WITHOUT PROBLEMS.

THE UMOS SYSTEM HAS BEEN RUNNING IN THE NEW GLOBAL MODEL PARALLEL
RUN SINCE THE VERY BEGINNING. THIS ALLOWED UMOS TO BEGIN ITS
ADAPTATION PERIOD EARLY IN THE MODEL TRANSITION PROCESS.
VERIFICATIONS OF THE UMOS TEMPERATURES BASED ON OUTPUTS OF THE NEW
GLOBAL MODEL SHOW THAT THE BIASES ARE VERY SIMILAR IN AMPLITUDE TO
WHAT IS OBSERVED IN THE UMOS TEMPERATURES BASED ON THE OLD GLOBAL
MODEL, BUT SHOW A MORE REALISTIC DIURNAL CYCLE. HOWEVER, THE ROOT
MEAN SQUARE ERROR OF THE UMOS TEMPERATURES BASED ON THE NEW GLOBAL
MODEL IS ABOUT ONE DEGREE CELSIUS LOWER AT ALL PROJECTION TIMES
THAN DIRECT MODEL OUTPUT.

THE SCRIBE FORECASTS GENERATED IN AUTOMATED MODE FROM THE OUTPUTS
OF THE NEW GLOBAL MODEL WERE VERIFIED. IT IS NOTED THAT THE SCRIBE
FORECASTS ARE IMPROVED FOR ALL WEATHER ELEMENTS AND AT ALL
PROJECTION TIME FROM DAY ONE TO DAY FIVE INCLUSIVE, EXCEPT PERHAPS
FOR THE WIND SPEED. SCRIBE HAS A TENDENCY TO FORECAST SOMEWHAT
STRONGER WIND SPEEDS BASED ON THE NEW GLOBAL MODEL COMPARED TO THE
OLD MODEL AND TO UNDERESTIMATE SOMEWHAT THE FREQUENCY OF LIGHT
WINDS. THIS DECREASES SOMEWHAT THE PERCENTAGE OF THE WIND SPEED
FORECASTS THAT ARE CORRECT CALCULATED FROM CONTINGENCY TABLES.
HOWEVER, THE WIND DIRECTION IS SIGNIFICANTLY IMPROVED.

PROFILES AT SPECIFIC POINTS OF THE MODEL OUTPUT SENT TO THE
REGIONAL OFFICES IN BUFR FORMAT WILL CONTAIN THE FULL VERTICAL
RESOLUTION (58 LEVELS) OF THE NEW MODEL. HOWEVER, THE GRIDDED DATA
IN GRIB FORMAT WILL NOT CHANGE IN RESOLUTION FOLLOWING THE
IMPLEMENTATION. OF COURSE EVEN AT THEIR CURRENT RESOLUTION, THE
GRIB PRODUCTS WILL BENEFIT FROM THE METEOROLOGICAL GAINS OF THE NEW
MODEL. THE FEASIBILITY OF INCREASING THE GRIB DATASET RESOLUTION TO
33 KM WILL BE EVALUATED AGAINST USER REQUIREMENTS AND AVAILABLE
BANDWIDTH RESOURCES BEFORE A DECISION IS MADE.

USERS WILL SEE CHANGES IN THE RECEPTION TIME OF THE GLOBAL
OPERATIONAL PRODUCTS WITH THIS IMPLEMENTATION. PRODUCTS FROM THE
GLOBAL SUITE WILL BE RECEIVED ABOUT 40 TO 55 MINUTES LATER WITH
RESPECT TO THE CURRENT RECEPTION TIME. SPECIFICALLY THE GLOBAL
SCRIBE MATRICES WILL BE DELIVERED 55 MINUTES LATER. WE WILL BE IN A
POSITION TO CORRECT THIS SITUATION WITH AN INCREASE IN COMPUTING
POWER SCHEDULE FOR THE FIRST MONTHS OF 2007.

SUMMARY OF THE OBJECTIVE AND SUBJECTIVE EVALUATIONS

EXTENSIVE PRELIMINARY TESTING IN BOTH SUMMER AND WINTER CONDITIONS
WAS NECESSARY TO ACHIEVE THE FINAL CONFIGURATION. THE MODEL AND THE
UPGRADED DATA ASSIMILATION SYSTEM RAN IN PARALLEL FOR 5 MONTHS,
DURING WHICH TIME IT WAS COMPARED AND EVALUATED BOTH OBJECTIVELY
AND SUBJECTIVELY.
OVER THE COURSE OF THE PARALLEL RUN PERIOD, THE MESO-GLOBAL MODEL
GENERALLY OUTPERFORMED THE OPERATIONAL MODEL QUITE SIGNIFICANTLY AS
MEASURED AGAINST THE GLOBAL RADIOSONDE OBSERVATION NETWORK. OVER
NORTH AMERICA, ROOT MEAN SQUARE ERRORS OF THE MASS FIELDS WERE
REDUCED SIGNIFICANTLY, ESPECIALLY FOR GEOPOTENTIAL HEIGHT AND
TEMPERATURES IN THE 1-5 DAY PERIOD. THE SIGNAL FOR THE WIND
FORECASTS WAS POSITIVE AT JET LEVEL, BUT NEUTRAL IN THE MID LEVELS
OF THE ATMOSPHERE AND SLIGHTLY POORER IN THE LOW LEVELS. THE
STRONGEST IMPROVEMENTS ARE NOTED OVER THE TROPICS, WHILE THE
MESO-GLOBAL ALSO PERFORMED WELL OVER THE SOUTHERN HEMISPHERE AND
OVER ASIA.

FOR THE PRECIPITATION FORECASTS, THE MESO-GLOBAL SCORES AGAINST
BOTH THE NORTH AMERICAN SYNOPTIC NETWORK OBSERVATIONS AND THE
HIGH-RESOLUTION U.S. SHEF NETWORK WERE MUCH IMPROVED IN THE 0-0.2
MM CLASS, INDICATIVE OF STRONGER DISCRIMINATION BETWEEN
PRECIPITATION AND NO PRECIPITATION CASES. MEANWHILE THE MODEL ALSO
PERFORMED MUCH BETTER IN THE HIGHER AMOUNTS (25 MM). MODEL BIASES
WERE STRONGLY IMPROVED FOR ALL PRECIPITATION CLASSES. OVERALL THE
SCORES INDICATE MUCH IMPROVED UTILITY OF THE MESO-GLOBAL MODEL
PRECIPITATION FORECASTS.

THE NEW SYSTEM WAS SUBJECTIVELY EVALUATED IN COMPARISON WITH THE
OLD SYSTEM. DURING THE PARALLEL RUN, EVALUATION IN OPERATIONS OF
THE NEW SYSTEM WAS GENERALLY PREFERRED TO THE OLD GLOBAL FORECAST
SYSTEM FOR BOTH THE MASS FIELDS AND THE QUANTITATIVE PRECIPITATION
FORECAST (QPF). THE NEW MODEL WAS MORE ACTIVE AND DEVELOPED
STRONGER SYSTEMS WHICH WERE GENERALLY MORE REALISTIC INCLUDING FOR
TROPICAL SYSTEMS. HURRICANE TRACKS AND TRANSITION TO EXTRATROPICAL
STORMS WERE BETTER HANDLED. PRECIPITATION AMOUNTS AND PATTERNS WERE
NOTED TO BE MUCH IMPROVED WITH THE NEW FORECAST SYSTEM.

IMPACT ON REGIONAL MODEL:

SINCE THE REGIONAL MODEL FEEDS FROM THE GLOBAL ASSIMILATION CYCLE
OBJECTIVE ANALYSIS EVERY 12 HOURS, THE IMPLEMENTATION OF A NEW
GLOBAL HAS AN IMPACT ON THE REGIONAL CYCLE ITSELF. THE MASS FIELD
SCORES FOR THE GEM REGIONAL MODEL SPUN UP FROM THE MESO-GLOBAL WERE
GENERALLY NEUTRAL IN COMPARISON WITH THE OPERATIONAL MODEL FOR DAY
1 AND SLIGHTLY, BUT CONSISTENTLY, IMPROVED FOR DAY 2. PRECIPITATION
SCORES WERE GENERALLY NEUTRAL. THE SUBJECTIVE VERIFICATION HAS
SHOWN SMALL BENEFITS IN THE FORECASTS OF THE MASS FIELDS BUT ALSO
IN THE PRECIPITATION VERIFICATION.

ADDITIONAL DETAILS ABOUT THIS IMPLEMENTATION WILL BE AVAILABLE ON
OCTOBER 30TH 2006 AT THE FOLLOWING URL:
HTTP://WWW.SMC.EC.GC.CA/CMC/OP_SYSTEMS/RECENT_E.HTML

FOR QUESTIONS ABOUT THIS IMPLEMENTATION, YOU CAN WRITE AT:
PRODUCTION-INFO(AT)CMC.EC.GC.CA

MARDI LE 31 OCTOBRE 2006, A 12 UTC, LE CENTRE METEOROLOGIQUE
CANADIEN IMPLANTERA UN NOUVEAU SYSTEME GLOBAL DE PREVISION AUX
OPERATIONS. LES CHANGEMENTS SONT MAJEURS AUTANT DANS LA RESOLUTION
DU MODELE GEM (GLOBAL ENVIRONNEMENTAL MULTI-ECHELLE) QUE DANS LA
PARAMETRISATION PHYSIQUE UTILISEE.

RESUME DES CHANGEMENTS A LA CONFIGURATION GLOBALE DU MODELE GEM

CONFIGURATION DYNAMIQUE

LA RESOLUTION HORIZONTALE DU MODELE DE PREVISION AUGMENTERA DE
FACON SIGNIFICATIVE. ELLE PASSERA D'ENVIRON 100 KM (400 X 200
POINTS DE GRILLE) A PRES DE 33 KM AUX LATITUDES MOYENNES (800 X 600
POINTS DE GRILLE). LE NOMBRE DE NIVEAUX DANS LA VERTICALE
AUGMENTERA DE 28 A 58, LE TOIT DU MODELE DEMEURANT A 10 HPA. LE PAS
DE TEMPS SERA REDUIT EN CONSEQUENCE, PASSANT DE 2700 A 900 SECONDES.

LA COUCHE EPONGE PRES DU TOIT DU MODELE COMPRENDRA 4 NIVEAUX AU
LIEU D'UN SEUL NIVEAU. DANS CETTE COUCHE, LA DIFFUSION HORIZONTALE
EST AUGMENTEE DANS LE BUT DE MINIMISER L'IMPACT NEGATIF D'ONDES
REFLECHIES DE FACON IRREALISTE PAR LE TOIT.

PARAMETRAGES PHYSIQUES

LA CONFIGURATION PHYSIQUE DU MODELE DE PREVISION SERA EGALEMENT
SUBSTANTIELLEMENT MODIFIEE COMPARATIVEMENT AU MODELE OPERATIONEL
PRECEDENT. EN PARTICULIER, L'ENSEMBLE CONDENSATION-PRECIPITATION
CHANGERA CONSIDERABLEMENT.

LE SCHEMA DE CONVECTION PROFONDE DE KAIN-FRITSCH REMPLACERA LE
SCHEMA DE KUO. CE CHANGEMENT PERMET UNE PLUS GRANDE CONTRIBUTION DU
SCHEMA DE CONDENSATION QUI OPERE A L'ECHELLE DE LA MAILLE, CE QUI
EST CONSISTANT AVEC L'AUGMENTATION DE LA RESOLUTION. IL S'AGIT DU
SCHEMA DE SUNDQUIST QUI A ETE REVISE AFIN DE MIEUX REPRESENTER
L'EVAPORATION DES PRECIPITATIONS SOUS LA BASE DES NUAGES, PHENOMENE
QUI SERA DESORMAIS PRIS EN COMPTE SUR PLUSIEURS NIVEAUX. ENFIN,
L'ENSEMBLE CONDENSATION-PRECIPITATION A ETE COMPLETE PAR
L'INTRODUCTION D'UN SCHEMA DE CONVECTION RESTREINTE APPELE KUO
TRANSIENT.

LE SCHEMA DE SURFACE APPELE ISBA (INTERACTION SURFACE BIOSPHERE ET
ATMOSPHERE) REMPLACE LE MODULE APPELE FORCE ET RESTORE. ISBA TRAITE
LE SOL, LA VEGETATION, ET LA COUVERTURE NIVALE AVEC PLUS DE
SOPHISTICATION. UN SYSTEME D'ASSIMILATION SEQUENTIEL DE DONNEES DE
SURFACE TERRESTRE SERA IMPLANTE AFIN DE FOURNIR LES CONDITIONS
INITIALES DE TEMPERATURE ET D'HUMIDITE DE SURFACE, LORS DU
LANCEMENT DU MODELE. LA RESOLUTION ACCRUE DU MODELE ET ISBA RENDENT
L'ANALYSE DE NEIGE PLUS PRECISE.

NOTONS QUE L'ON UTILISERA LA FORMULATION DE BOUGEAULT ET LACARRERE
POUR LA LONGUEUR DE MELANGE POUR LA DIFFUSION VERTICALE CAUSEE PAR
LA TURBULENCE ATMOSPHERIQUE, CE QUI REPRESENTE UNE AMELIORATION
CERTAINE EN PARTICULIER LORSQUE LA COUCHE LIMITE EST CONVECTIVEMENT
ACTIVE ET BIEN MELANGEE.

L'EFFICACITE COMPUTATIONNELLE DU SYSTEME D'ASSIMILATION DE DONNEES
4D-VAR A ETE AMELIOREE DE 40 AU TOTAL ET LES TEMPS DE COUPURE POUR
L'INCLUSION DES OBSERVATIONS ONT ETE LEGEREMENT MODIFIES POUR
LIVRER LES ANALYSES OPERATIONNELLES AVEC APPROXIMATIVEMENT LES
MEMES TEMPS QU'AVANT.

RESUME DES CHANGEMENTS AU SYSTEME GLOBAL D'ASSIMILATION DE DONNEES

LES STATISTIQUES D'ERREUR DE PREVISION ONT ETE RECALCULEES SUR LES
58 NIVEAUX DU MODELE EN UTILISANT LA METHODE DITE DU NMC. DE MEME
QUE DANS L'IMPLANTATION PRECEDENTE DU 4D-VAR, UN MODELE A BASSE
RESOLUTION EST UTILISE AFIN DE PROPAGER LES INCREMENTS D'ANALYSE
(T108) SUR LA FENETRE D'ASSIMILATION DE 6 HEURES. L'ENSEMBLE DE
PARAMETRAGES PHYSIQUES DE CE MODELE A ETE MODIFIE DE LA MEME FACON
QUE LE MODELE DE PREVISION DECRIT CI-HAUT. NOTONS TOUTEFOIS QUE
L'ENSEMBLE DES PARAMETRAGES UTILISE AU SEIN DU LINEAIRE MODELE
TANGENT AINSI QUE SON MODELE ADJOINT DEMEURENT LES MEMES.

IMPACTS SUR LES SORTIES ET PRODUITS DERIVES

TOUS LES PRODUITS ONT ETE GENERES ET VALIDES A PARTIR DES SORTIES
DU NOUVEAU MODELE GLOBAL PENDANT LA PASSE PARALLELE. CERTAINS
CHANGEMENTS ONT ETE APPORTES DANS L'ENSEMBLE DES CARTES POUR LA
METEOROLOGIE POUR L'AVIATION. LES SEUILS DE TURBULENCE MODEREE ET
SEVERE ONT ETE AUGMENTES POUR ETRE CONFORMES A CEUX UTILISES AVEC
LE MODELE REGIONAL. DES MODIFICATIONS ESTHETIQUES ONT AUSSI ETE
APPORTEES AUX CONTOURS DU NIVEAU DE CONGELATION ET DE LA HAUTEUR DE
LA TROPOPAUSE. CONCERNANT LES CARTES 4-PANNEAUX CLASSIQUES, LA
PRESSION AU NIVEAU MOYEN DE LA MER EST MAINTENANT FILTREE. LE
FILTRAGE DU TOURBILLON ET DU MOUVEMENT VERTICAL A ETE ACCENTUE. LES
CONTOURS DE 10 ET 30 POUR CENT D'HUMIDITE RELATIVE ONT ETE
ELIMINES. TOUS LES AUTRES PRODUITS DE POST-TRAITEMENT ONT ETE
VALIDES ET AUCUN PROBLEME N'A ETE NOTE.

LE SYSTEME UMOS A ETE INCLUS AU TOUT DEBUT DE LA PASSE PARALLELE DU
NOUVEAU MODELE GLOBAL, CE QUI A PERMIS A UMOS DE COMMENCER SA
PERIODE D'ADAPTATION TOT DANS LE PROCESSUS DE TRANSITION DE MODELE.
LES VERIFICATIONS DES TEMPERATURES UMOS BASEES SUR LE NOUVEAU
MODELE GLOBAL INDIQUENT QUE LES ERREURS MOYENNES SONT SEMBLABLES EN
AMPLITUDE A CE QUI EST NOTE DANS LES TEMPERATURES UMOS BASEES SUR
L'ANCIEN MODELE GLOBAL, MAIS AVEC UN CYCLE DIURNE PLUS REALISTE. DE
PLUS, L'ERREUR QUADRATIQUE MOYENNE DES TEMPERATURES UMOS BASEES SUR
LE NOUVEAU MODELE GLOBAL EST DIMINUEE D'ENVIRON UN DEGRE CELSIUS A
TOUS LES TEMPS DE PROJECTION PAR RAPPORT AUX SORTIES DIRECTES DU
MODELE.

LES PREVISIONS SCRIBE GENEREES EN MODE AUTOMATISE A PARTIR DES
SORTIES DU NOUVEAU MODELE GLOBAL ONT ETE VERIFIEES EN PROFONDEUR.
ON NOTE UNE AMELIORATION DES PREVISIONS SCRIBE POUR TOUS LES
ELEMENTS DU TEMPS A TOUS LES TEMPS DE PROJECTION DU JOUR UN AU JOUR
CINQ INCLUSIVEMENT, SAUF PEUT-ETRE POUR LA VITESSE DU VENT. IL EST
A NOTER QUE SCRIBE A TENDANCE A PREVOIR DES VENTS UN PEU PLUS FORTS
A PARTIR DU NOUVEAU MODELE GLOBAL QU'A PARTIR DE L'ANCIEN MODELE ET
A SOUS-ESTIMER QUELQUE PEU LA FREQUENCE DE VENTS FAIBLES. CECI
DIMINUE QUELQUE PEU LE POURCENTAGE DE PREVISIONS CORRECTES CALCULE
A PARTIR DE TABLES DE CONTINGENCES. CEPENDANT, LA DIRECTION DES
VENTS EST AMELIOREE DE FACON SIGNIFICATIVE.

LES PROFILS A DES POINTS SPECIFIQUES TRANSMIS AUX BUREAUX REGIONAUX
DANS LE FORMAT BUFR AURONT LA PLEINE RESOLUTION VERTICALE (58
NIVEAUX) DU NOUVEAU MODELE. CEPENDANT LES DONNEES AUX POINTS DE
GRILLE EN FORMAT GRIB NE VERRONT PAS LEUR RESOLUTION AUGMENTEES
DANS UN PREMIER TEMPS. MEME EN RESTANT A LEUR RESOLUTION ACTUELLE
LES PRODUITS GRIB PROFITERONT DES GAINS DU NOUVEAU MODELE. LA
FAISABILITE D'AUGMENTER LA RESOLUTION DES DONNEES GRIB A 33 KM SERA
EVALUEE SELON LES BESOINS DES USAGERS ET LES RESSOURCES DE
TRANSMISSION DISPONIBLES.

LES PRODUITS OPERATIONNELS DU SYSTEME GLOBAL SERONT DISPONIBLES UN
PEU PLUS TARD AVEC L'IMPLANTATION DU NOUVEAU MODELE GLOBAL SOIT
APPROXIMATIVEMENT DE 40 A 55 MINUTES PLUS TARD QUE PRESENTEMENT. EN
PARTICULIER LES MATRICES SCRIBE GLOBALES SERONT LIVREES 55 MINUTES
PLUS TARD. NOUS SERONS EN MESURE DE CORRIGER CETTE SITUATION AVEC
L'INSTALLATION D'UNE PLUS GRANDE PUISSANCE DE CALCUL AU DEBUT 2007.

VERIFICATIONS OBJECTIVES ET SUBJECTIVES

DES TESTS PRELIMINAIRES COMPLETS ONT ETE EFFECTUES SUR DES
SITUATIONS D'HIVER ET D'ETE POUR EN ARRIVER A UNE CONFIGURATION
FINALE. PAR LA SUITE LE MODELE ET LE CYCLE D'ASSIMILATION ONT
TOURNE EN PARALLELE PENDANT 5 MOIS PENDANT LESQUELS IL A ETE
COMPARE ET EVALUE OBJECTIVEMENT ET SUBJECTIVEMENT.

DURANT LA PASSE PARALLELE, LE MODELE GLOBAL PARALLELE A
GENERALEMENT BATTU LE MODELE OPERATIONNEL DE FACON SIGNIFICATIVE
LORSQUE VERIFIE CONTRE LE RESEAU GLOBAL DES STATIONS DE
RADIOSONDAGE. SUR L'AMERIQUE DU NORD, LES ERREURS QUADRATIQUES
MOYENNES DES CHAMPS DE MASSE ONT ETE REDUITES DE FACON
SIGNIFICATIVE, PARTICULIEREMENT POUR LES HAUTEURS GEOPOTENTIELLES
ET LES TEMPERATURES POUR LES JOURS 1 A 5. LE SIGNAL POUR LES VENTS
ETAIT EN FAVEUR DU MODELE PARALLELE AU NIVEAU DU COURANT-JET MAIS
NEUTRE AUX NIVEAUX MOYENS ET ENFIN MOINS BON AUX BAS NIVEAUX. ON
NOTE LES GAINS LES PLUS FORTS AUX TROPIQUES, MAIS LE NOUVEAU MODELE
A BIEN FAIT DANS L'HEMISPHERE SUD ET EN ASIE.

POUR CE QUI EST DES PRECIPITATIONS, LES MODELES ONT ETE EVALUES EN
UTILISANT LE RESEAU SYNOPTIQUE ET LE RESEAU AMERICAIN A HAUTE
DENSITE SHEF. LE MODELE GLOBAL PARALELE A DEMONTRE UNE NETTE
AMELIORATION DANS LA CATEGORIE 0-0,2 MM, MONTRANT UNE PLUS GRANDE
HABILITE A DISCRIMINER ENTRE LES SITUATIONS DE PRECIPITATIONS OU DE
NON-PRECIPITATIONS. LE MODELE PARALLELE A AUSSI MIEUX REUSSI DANS
LES CATEGORIES PLUS ELEVEES (25 MM). LE BIAIS DU NOUVEAU MODELE EST
MEILLEUR POUR TOUTES LES CATEGORIES. SOMME TOUTE, LES PREVISIONS DE
PRECIPITATIONS SONT BEAUCOUP AMELIOREES AVEC LE NOUVEAU MODELE.

LE NOUVEAU SYSTEME A ETE EVALUE SUBJECTIVEMENT EN COMPARAISON DE
L'ANCIEN SYSTEME. DURANT LA PASSE D'EVALUATION PARALLELE, LES
SORTIES DU NOUVEAU SYSTEME GLOBAL ONT ETE PREFEREES AUTANT POUR LES
CHAMPS DE MASSE QUE LES PREVISIONS DE QUANTITES DE PRECIPITATIONS.
LE NOUVEAU MODELE EST PLUS ACTIF ET PREVOIT DES SYSTEMES PLUS CREUX
ET INTENSES QUI SONT GENERALEMENT PLUS REALISTES ET CECI EST AUSSI
VRAI POUR LES SYSTEMES TROPICAUX. LES TRAJECTOIRES DES OURAGANS ET
LEURS TRANSITIONS AUX LATITUDES MOYENNES ONT ETE MIEUX PREVUES. LES
QUANTITES ET LES PATRONS DE PRECIPITATIONS SONT NETTEMENT AMELIORES.

IMPACT SUR LE MODELE REGIONAL

L'EVALUATION OBJECTIVE DU MODELE REGIONAL PARALLELE EST NECESSAIRE
PUISQU'IL SE NOURRIT DES ANALYSES DU CYCLE GLOBAL. LA VERIFICATION
OBJECTIVE DES CHAMPS DE MASSE DU MODELE REGIONAL PARALLELE EN
COMPARAISON DU MODELE REGIONAL OPERATIONNEL A MONTRE UN
COMPORTEMENT NEUTRE POUR LE JOUR 1 MAIS UN GAIN FAIBLE MAIS
CONSTANT POUR LE JOUR 2. LA VERIFICATION OBJECTIVE DES
PRECIPITATIONS (QPF) NE MONTRE AUCUNE DIFFERENCE. LA VERIFICATION
SUBJECTIVE MONTRE UN LEGER GAIN DU MODELE REGIONAL PARALLELE POUR
LES CHAMPS DE MASSE DE MEME QUE POUR LES PRECIPITATIONS.

LES DETAILS DE CETTE IMPLEMENTATION SERONT DISPONIBLE A PARTIR DU
30 OCTOBRE 2006 A L'URL SUIVANT:
HTTP://WWW.SMC.EC.GC.CA/CMC/OP_SYSTEMS/RECENT_F.HTML

POUR DES QUESTIONS SPECIFIQUES, S'ADRESSER A:
PRODUCTION-INFO(AT)CMC.EC.GC.CA

EVERELL / ADMA / SMA-SMC TORONTO