Global- -M Mso so et son Impact sur la Prvision et son Impact sur - - PowerPoint PPT Presentation

▶
global m m so so et son impact sur la pr vision et son
SMART_READER_LITE
LIVE PREVIEW

Global- -M Mso so et son Impact sur la Prvision et son Impact sur - - PowerPoint PPT Presentation

Sep 19, 2022 587 likes •1.14k views

Global- -M Mso so et son Impact sur la Prvision et son Impact sur la Prvision Global Moyenne- -chance chance au Canada au Canada Moyenne Stphane Blair (MRB) Stphane Laroche ( MRB ) Michel Roch (MRB) Anne-Marie Leduc

slide-1
SLIDE 1

Global Global-

  • M

Méso éso et son Impact sur la Prévision et son Impact sur la Prévision Moyenne Moyenne-

  • Échéance

Échéance au Canada au Canada

Environnement Canada

Stéphane Bélair (MRB) Stéphane Laroche (MRB) Michel Roch (MRB) Anne-Marie Leduc (CMC) François Lemay (CMC) Jean-Marc Bélanger (CMC) Paul Vaillancourt (MRB)

slide-2
SLIDE 2

Environment Canada

Objectif du Projet Objectif du Projet Fournir à Fournir à CMC CMC-

  • Opérations

Opérations une nouvelle une nouvelle version de GEM qui permet d’améliorer version de GEM qui permet d’améliorer la prévision météorologique à la prévision météorologique à moyenne moyenne-

  • échéance

échéance au Canada au Canada

slide-3
SLIDE 3

Environment Canada

Current Current Operational Operational Configuration Configuration

  • Hydrostatic primitive equations
  • Implicit two time-level semi-Lagrangian

scheme; time step = 2700 sec

  • 3D finite elements
  • periodicity in the horizontal
  • No motion across top and bottom

surfaces

  • Global uniform lat-lon grid with

computational poles rotated with respect to geographic poles.

  • 0.9° horizontal resolution
  • 400 X 200 grid points in the horizontal
  • 28 η levels, top at 10 hPa
  • ∇6 horizontal diffusion on momentum

variables.

  • Stratospheric sponge
  • Solar and infrared radiation interactive

with water vapor, carbon dioxide, ozone and clouds (Garand + Fouquart and Bonnel)

  • Surface processes with simple force-

restore system

  • TKE scheme for PBL vertical diffusion,

diffusion coefficients based on stability and kinetic energy

  • Simple diagnostic clouds in PBL with

impact on stability (and vertical diffusion)

  • Surface layer based on Monin-Obukhov

similarity theory

  • Kuo-type deep convection scheme
  • Sundqvist condensation scheme for

stratiform precipitation

  • Gravity wave drag (MacFarlane 1987)
  • Low-level blocking (Lott and Miller)

DYNAMICS PHYSICS

slide-4
SLIDE 4

Environment Canada

Dicton Populaire Dicton Populaire

Tu prends le global Tu prends le global T’augmentes la résolution T’augmentes la résolution Tu mets plus de niveaux Tu mets plus de niveaux Tu prends la physique du régional Tu prends la physique du régional … et that’s it !!! … et that’s it !!!

slide-5
SLIDE 5

Main Features of Global Main Features of Global-

  • Meso

Meso

Environment Canada

  • Increased horizontal and vertical resolution
  • 800x600x58L (33 km) compared to 400x200x28L (100 km)
  • Numerical poles at geographic locations
  • Representation of clouds and precipitation
  • Shallow convection with Kuo Transient
  • Deep convection with Kain-Fritsch
  • Modified Sundqvist scheme for grid-scale condensation
  • Bougeault-Lacarrère for the turbulent mixing length
  • Constant thermodynamic roughness length over water in the

Tropics

  • ISBA land surface scheme with sequential assimilation of soil

moisture (based on OI)

slide-6
SLIDE 6

Meteorological Meteorological Performance Performance

(If it looks good, it must be good…) (If it looks good, it must be good…)

Environment Canada

60-h forecasts of instantaneous precipitation rates valid at 1200 UTC 25 Feb. 2001 using the current operational model and the proposed configuration. The black and white image shows the radar reflectivities valid at the same time

OP MESO RADAR

slide-7
SLIDE 7

Environment Canada

MESO OP (Michel Roch)

slide-8
SLIDE 8

Vers une Implémentation Méso Vers une Implémentation Méso-

  • 4D

4D

Environment Canada

Deux cycles d’assimilation Été: du 8 juillet au 10 septembre 2004 Hiver: du 8 décembre 2004 au 3 février 2005

ΔX1 ΔX0 b X X

X

Modifications à l’analyse:

  • Analyses faites sur les 58 niveaux du Global-Méso
  • Nouvelles statistiques d’erreurs (méthode NMC, 58 niveaux)
  • Optimisation du code du modèle et du variationnel
  • Moins d’itérations de la boucle interne (55 au lieu de 70) rendu

possible par l’utilisation d’une matrice Hessienne enrichie

slide-9
SLIDE 9

Northern Hemisphere

96-h forecasts RMSE and BIAS 101 WINTER cases Blue: Operational model Red: Global Meso

slide-10
SLIDE 10

Southern Hemisphere

96-h forecasts RMSE and BIAS 101 WINTER cases Blue: Operational model Red: Global Meso

slide-11
SLIDE 11

Tropics

96-h forecasts RMSE and BIAS 101 WINTER cases Blue: Operational model Red: Global Meso

slide-12
SLIDE 12

Northern Hemisphere

96-h forecasts RMSE and BIAS 110 SUMMER cases Blue: Operational model Red: Global Meso

slide-13
SLIDE 13

Southern Hemisphere

96-h forecasts RMSE and BIAS 110 SUMMER cases Blue: Operational model Red: Global Meso

slide-14
SLIDE 14

Tropics

96-h forecasts RMSE and BIAS 110 SUMMER cases Blue: Operational model Red: Global Meso

slide-15
SLIDE 15

Environment Canada

Objective Evaluation Against Analyses Objective Evaluation Against Analyses Anomaly Correlations Anomaly Correlations

Geopotential Height – 500 hPa Northern Hemisphere Southern Hemisphere 101 Winter (DJF 2005) Cases

24h 120h 48h 72h 96h 24h 120h 48h 72h 96h

slide-16
SLIDE 16

Error Spectra Error Spectra

OP MESO

slide-17
SLIDE 17

Northern Hemisphere

96-h forecasts RMSE and BIAS 101 WINTER cases Filtered (red) vs non-filtered (blue) Filtered Meso (red) vs Filtered OP (blue)

slide-18
SLIDE 18

Meso Meso-

  • 4D will be Born During the

4D will be Born During the Year of the Dog Year of the Dog

Environment Canada

Loyal, faithful, and honest Trouble trusting others Trustworthy, except for occasional little « white lies » Excellent listener Narrow-minded and stubborn Possibly temperamental Happy when physically active

slide-19
SLIDE 19

Environment Canada

Objective Evaluation of Precipitation Objective Evaluation of Precipitation

SHEF US surface stations

Bias ETS

NEW OP NEW OP

55 summer 2004 cases

Precip 48-72h

slide-20
SLIDE 20

Global Evaluation of Clouds with SSM Global Evaluation of Clouds with SSM/I /I

WINTER (Paul Vaillancourt)

slide-21
SLIDE 21

Radar Echoes for Hurricane Rita Radar Echoes for Hurricane Rita

Environment Canada

(Michel Roch)

slide-22
SLIDE 22

Hurricane Rita: MESO vs OP Hurricane Rita: MESO vs OP

Environment Canada

Observed trajectory MESO OP Other models (MR)

slide-23
SLIDE 23

Systematic Evaluation of Systematic Evaluation of Hurricanes and Typhoons Hurricanes and Typhoons

Environment Canada

Trajectory EXAMPLE: KATRINA Central Pressure OP MESO OBS OP MESO OBS (Anne-Marie Leduc)

slide-24
SLIDE 24

800 x 600 x 58* x 45 400 x 200 x 28 x 15 2 x 3 x 2-3 x 3 = 30-60 NEW OP

The Cost of Global The Cost of Global-

  • Meso

Meso

ni nj nk Δt

Estimation: around 380 CPUs will be required to produce 10-day forecasts in about 100 min (wallclock)

Environment Canada

Size of output files: at least 40 Gb for a 10-day forecast (compared to 2 Gb for the current

  • perational system)

Impact on development and future implementations

slide-25
SLIDE 25

Pendant ce temps … Ailleurs … Pendant ce temps … Ailleurs …

Environment Canada

NCEP ECMWF UK Met Office Japan Met Agency T382 (35 km) L64 Summer 2005 3 Feb. 2006 T799 (25 km) L91 432 x 325 (60 km) x L38 640 x 320 (60 km) x L40 2001

slide-26
SLIDE 26

Nous et les Autres Nous et les Autres

Environment Canada

~ 12 m De combien allons-nous réduire le gap?

slide-27
SLIDE 27

Calendar Calendar

Environment Canada

Late Jan. Final assimilation cycles and 10-day integrations are on the way Mid to End of March Proposition to CPOP Spring Parallel run Early Summer Operational Implementation

slide-28
SLIDE 28

Future Developments Future Developments

Environment Canada

  • Stratospheric version of GEM (model + assimilation) (CHARRON, DUGAS)
  • New radiation code with improved optical properties of clouds (VAILLANCOURT)
  • Improvement in the representation of grid-scale and subgrid-scale orography,

together with improved representation of subgrid-scale roughness and low-level blocking effect of mountains (ZADRA, BELAIR)

  • First version of the Canadian Land Data Assimilation System (MAHFOUF, BELAIR,

DEBLONDE)

  • Off-line modeling of sea-ice (PELLERIN, BEAUDOIN)
  • Further improvements to clouds and precipitation (BELAIR, MAILHOT, LEDUC)
  • Second version of CaLDAS (assimilation of L-band data + possibility of CLASS)

(MAHFOUF, BELAIR, DEBLONDE)

slide-29
SLIDE 29

Backup Charts Backup Charts

Environment Canada

slide-30
SLIDE 30

Environment Canada

Ceux qui ont Contribué Ceux qui ont Contribué

(liste non exhaustive) (liste non exhaustive)

Maryse Beauchemin Stéphane Bélair Jean-Marc Bélanger Bernard Bilodeau Bruce Brasnett Gilbert Brunet Mark Buehner Corinne Burlaud Cécilien Charrette Martin Charron Yves Chartier Luc Corbeil Jean Côté Michel Desgagné Bernard Dugas Pierre Gauthier Claude Girard Jacques Hallé Carol Hopkins Pierre Koklas Manon Lajoie Ervig Lapalme Stéphane Laroche Anne-Marie Leduc Vivian Lee Louis Lefaivre Francois Lemay Jean-François Mahfouf Jocelyn Mailhot André Méthot Josée Morneau Alain Patoine Pierre Pellerin Paul Pestieau André Plante Abdessama Qaddouri Lise Rivard Michel Roch Donald Talbot Monique Tanguay André Tremblay Paul Vaillancourt Michel Valin Gilles Verner Ayrton Zadra

MERCI a TOUS

slide-31
SLIDE 31

NCEP

03/1991 T126 L18 06/1998 T170 L42 10/1998 T126 L18 01/2000 T170 L42 06/2000

  • Ens. T126

10/2002 T254 L64 /2005 T382 L64?

ECMWF

05/1985 T106 L16 05/1986 T106 L19 09/1991 T213 L31 04/1998 T319 L31 03/1999 T319 L50

  • Ens. T159 L31

10/1999 T319 L60

  • Ens. T159 L40

11/2000 T511 L60

  • Ens. T255 L40

2003 T511 L90 Future T719-T799 L90

METEO-FRANCE: T358 L41 UK Met Office: 432 x 325 x 38L Japan Met Agency: 630 x 320 (T213) x 40L

Environment Canada

Other Major Centers Other Major Centers

slide-32
SLIDE 32

Si on part de:

R E E

  • f
  • f

σ σ σ σ 2

2 2 2 2

− + + =

Et si on suppose que le biais et la variance du modèle ne changent pas durant les intégrations du modèle, on peut exprimer la croissance d’erreur MSE de la manière suivante:

t R t E

  • f

∂ ∂ − = ∂ ∂ σ σ 2

2

t R t E

  • f

∂ ∂ = ∂ ∂ σ σ 2

2

⇔

Si on suppose encore que le facteur de décorrélation est le même pour les deux modèles, alors on obtient que la croissance d’erreur MSE dépend linéairement de la variance du modèle. Étant donné que deux signaux se “décorrèlent” plus rapidement quand la variance est plus grande (ce qui n’est pas montré ici), on peut s’attendre à ce que le lien entre la croissance d’erreur et la variance soit plus sévère que cette relation linéaire.

Le Le probl problème de la variance ème de la variance

Environment Canada

slide-33
SLIDE 33

Filtrage spectrale Filtrage spectrale de la solution de la solution Global Global-

  • Méso

Méso

Pour une comparaison “fair”, les deux modèles doivent être évalués au même niveau de variance. C’est-à- dire, on doit filtrer la solution du modèle à plus haute résolution pour

  • btenir un spectre de variance

semblable à celui du modèle à plus basse résolution. Si les deux variances sont les même, alors on obtient:

( )

( )

OP GMfilt

  • filt

OP GMfilt OP GMfilt

R R E E E E − − ⎟ ⎠ ⎞ ⎜ ⎝ ⎛ − = − σ σ 2

2 2 2 2

T199 Solution filtrée T70

Environment Canada

slide-34
SLIDE 34

Northern Hem.

24-h forecasts Standard Dev. and BIAS 118 WINTER cases Blue: Global Meso (control) Red: Global Meso (filtered)

Environment Canada

slide-35
SLIDE 35

T30 T200 Due to its increase in resolution, the variance spectrum produced by GLOBAL MESO for waves between truncatures 30 and 200 appears to be more realistic.

Variance Spectra of OP and MESO Variance Spectra of OP and MESO

slide-36
SLIDE 36

Objective Evaluation of Precipitation Objective Evaluation of Precipitation Over Other Parts of the World Over Other Parts of the World

World Asia Day 1 MESO OP

slide-37
SLIDE 37

MESO SSM/ I

Frequency Maps of Clouds with SSM/I Frequency Maps of Clouds with SSM/I (Stogryn (Stogryn’s ’s Criteria) Criteria)

(Gracieuseté de Corrine Burlaud)

slide-38
SLIDE 38

Hurricane Rita with Other Models Hurricane Rita with Other Models

Environment Canada

slide-39
SLIDE 39

Operational

10hPa 28 η levels

ΔZ (km)

New

10hPa 58 η levels ΔZ (km)

Computational Grid Computational Grid

Environment Canada

slide-40
SLIDE 40

GPCP MESO OP JUL 2002

Contours: (mm/day) 0-1 1-5 5-10 10-20 >20

Global Evaluation of Precipitation Global Evaluation of Precipitation

Environment Canada

slide-41
SLIDE 41

Global Evaluation of Clouds with SSM Global Evaluation of Clouds with SSM/I /I

WINTER SUMMER

slide-42
SLIDE 42

A Brief History of CMC’s A Brief History of CMC’s Global Forecasting System Global Forecasting System

Environment Canada

March 1991 SEF T79 L21 June 1993 SEF T119 L21 Significant changes to the physics (Kuo, shallow convection, radiation) June 1995 SEF T199 L21 June 1997 SEF T199 L21 Global variational assimilation system

  • Oct. 1998

GEM 400x200 L28 Sundqvist scheme

  • Sept. 2000

GEM 400x200 L28 TOVS + ACARS/AMDARS

  • Dec. 2001

GEM 400x200 L28 Eta analysis + Low-level blocking March 2005 GEM 400x200 L28 4DVAR

slide-43
SLIDE 43

Operational New

Computational Grid Computational Grid

Environment Canada

slide-44
SLIDE 44

Toutes les Obs. 7.5h 4.5h 0h 6h 3h 9h Fenêtre d’assimilation Intégration du modèle opérationnel pour générer le champ d’essai ou l’analyse finale. Intégration du modèle opérationnel (ou trajectoire pleine résolution) pour le calcul des innovations (O-P) au temps approprié. Analyse Champ d’essai ATOVS Autres Observations ΔX1 ΔX0 b X X

X

4DVAR 4DVAR

Environment Canada

slide-45
SLIDE 45

06 UTC 12 UTC 18 UTC 00 UTC

OBS OBS T + 6h OBS T + 9h OBS T + 9h OBS T + 9h OBS T + 3h OBS T + 3h OBS T + 3h

00 UTC

240h global fst 6-h global

first guess

6-h global

first guess

6-h global

first guess

4D-Var Analysis 4D-Var Analysis 4D-Var Analysis 4D-Var Analysis

6-h global

first guess

6-h global

first guess

4D-Var Analysis 4D-Var Analysis 4D-Var Analysis 4D-Var Analysis

T + 6h OBS T + 5h30

6-h regional

first guess

3D-Var Analysis

6-h regional

first guess T + 1h40 OBS

3D-Var Analysis

48h regional fst

T + 1h40 OBS OBS T + 5h30

3D-Var Analysis

6-h regional

first guess

6-h regional

first guess

3D-Var Analysis

144h global fst 6-h regional

first guess T + 1h40 OBS

3D-Var Analysis

48h regional fst

G1 G2 R1

Assimilation “Flowchart” Assimilation “Flowchart”

CONTINUOUS GLOBAL ASSIMILATION CYCLE

slide-46
SLIDE 46

Global Evaluation of Precipitation Global Evaluation of Precipitation

Environment Canada

slide-47
SLIDE 47

Winter (DJF) 2001-2002

Reduction of Precipitation Bias Over Oceans Reduction of Precipitation Bias Over Oceans

Environment Canada

Global Land Oceans

slide-48
SLIDE 48

Objective Evaluation of Precipitation Objective Evaluation of Precipitation

  • ver North America
  • ver North America

Environment Canada

0.2 mm 25 mm Day 3 Bias Threat score Threat score MESO OP

slide-49
SLIDE 49

Typhoon and Hurricanes Prediction: Typhoon and Hurricanes Prediction: Model or Assimilation? Model or Assimilation?

MESO OP

Environment Canada

Japan

slide-50
SLIDE 50

20K 10 K

Distribution Functions of Distribution Functions of Optically Thick Clouds Optically Thick Clouds

(Gracieuseté de Corrine Burlaud)

Environment Canada

Feb 2002, F13, F14 et F15 (801 510 cases)

slide-51
SLIDE 51

Soil Moisture over North America Soil Moisture over North America (2004 (2004-

  • 2003)

2003)

June July August

0.01 0.02 0.05 0.10

  • 0.01
  • 0.02
  • 0.05
  • 0.10

m3/m-3

Soil Moisture Precipitation

  • 10
  • 2
  • 1

1 2

  • 5

10 5 25 (mm/day)

August

slide-52
SLIDE 52

Coordinated Enhanced Observing Period Coordinated Enhanced Observing Period (CEOP) (CEOP)

Environment Canada

MSC’s CEOP Experiment: Based on the new mesoscale version of the Global Enrvironmental Multiscale (GEM) model that is currently being developed at MSC for medium-range weather forecasts

1 May 2002 1 Oct 2002 1 Oct 2003 31 Dec 2004 SPIN-UP EOP3 EOP4

PERIOD of INTEGRATION: CYCLING and ASSIMILATION STRATEGY:

6h 6h 6h 6h Upper-air component of the analyses is directly

  • btained from CMC’s archive, i.e., no 4DVAR is

performed for atmospheric observations Surface component of the analyses is cycled from the previous 6-h forecast, with sequential assimilation of soil moisture, surface temperature, and snow

CONTINUOUS EVOLUTION of ATMOSPHERE and SURFACE:

36-h run 36-h run

12 UTC 12 UTC 00 UTC 00 UTC 00 UTC

X X

slide-53
SLIDE 53

Soil Moisture at CEOP Reference Sites Soil Moisture at CEOP Reference Sites Spatial and Temporal Variability Spatial and Temporal Variability

2002 2003 2004 2002 2003 2004 Soil moisture Soil moisture

J F M A M J J A S O N D O N D F M A M J J A S J

Evolution of soil moisture at Lindenberg, Germany

(m3m-3)

Time averages over 10 days for the model closest point, with (min,max) during that period Time and spatial averages

  • ver 10 days for a 5x5 box

(25 points, ~160 km), with (min,max) during that period

slide-54
SLIDE 54

JAN 2002

Contours: (mm/day) 0-1 1-5 5-10 10-20 >20

Global Evaluation of Precipitation Global Evaluation of Precipitation

Environment Canada

GPCP MESO OP