5th Climate Europe Webminar

Model drift analysis to understand the causes of systematic errors in climate

prediction systems

EmiliaSanchez-Gomez

KaterinaGoubanovaandcollaborators

Climate Prediction System

External forcings GHG, Volcanoes, aerosols, solar

Forecast

verifica-on

skill

Observations Ocean, atmosphere, land, sea-ice

Initial conditions

Climate Model

Boundary conditions

Perturbing Ensembles

simulations Drift correction

Forecast

Products

downscaling

Climate Prediction System

External forcings GHG, Volcanoes, aerosols, solar

Forecast

verifica-on

skill

Observations Ocean, atmosphere, land, sea-ice

Initial conditions Boundary conditions

Perturbing Ensembles

simulations Drift correction

Uncertainties affecting forecast performance

Forecast

Products

downscaling

Climate Model

Climate Prediction System

External forcings GHG, Volcanoes, aerosols, solar

Forecast

verifica-on

skill

Observations Ocean, atmosphere, land, sea-ice

Initial conditions Boundary conditions

Perturbing Ensembles

simulations Drift correction

Forecast

Products

downscaling

Climate Model

Evaluation of climate models

ModelEvalua-on

Meanclimate

Variability

Predic-veskill Mechanisms

Drift Analysis to understand model biases

AdaptedfromMagnussonetal.AdaptedfromMagnussonetal.

Observedclimate

Modelclimate

T=0

-me

Unini-alized

simula-on

ini-alized

simula-on

ModeldriAIni-alshock

ThemodeldriAisthesequenceofphysicalprocessesbywhichmodeladjusttoits

equilibriumstateora>ractor

Drift Analysis to understand model biases

AdaptedfromMagnussonetal.AdaptedfromMagnussonetal.

ThedriA(orbiasadjustment)analysisdependsonthespa-al

and-mescalesconsideredinthephysicalproblem:

o  Fast atmospheric processes (convective processes, clouds) hours to days o  Ocean subsurface (mixing processes) days to months

o  Deep ocean circulation and gyres years, decades

Transpose-AMIPprotocol

Seasonalforecast

Decadalforecast

Understanding climate model biases

Coupled models continue to suffer from severe equatorial SST biases over the Tropical Atlantic

SSTbiasintheTropicalOceans(fromToniazzoandWoolnough,2013)

Data:CMIP5

WhatistheoriginofSSTbiasesinthe

SouthEasternTropicalAtlan-c(SETA)?

Drift Analysis to understand model biases

2-monthsaveragedevolu-onoftheSSTbiaswithrespecttoGLORYS2v3

q  Model:

CNRM-CM:ARPEGE(~50km)(T359L31)andNEMO~0.250(ORCA025L75)

q Seasonalforecast:

§  IniQalizaQonfromERAIandGLORYS2v3

§  Startdate:1February

§  10years:2000-2009

§  Three-members,6monthsleadQme

Goubanovaetal.2017(inprep.)

Drift of ocean subsurface temperature

FEB

MAR

APR

MAY

JUN

JUL

Isotherme20°Cinthemodel

Isotherme20°CinGLORYS2v3

⁰C

Goubanovaetal.2017(inprep.)

Talongtheequator(2°S-2°N) Talongthecoast(2°)

Drift in wind stress

Goubanovaetal.2017(inprep.)

Shading:windstressamplitude

Vector:windstressvector

Contour:zonalwindstress

Pa

ERAI CNRM-CMModel

Meanevolu-onoftheequatorialwindstress(2°S-2°N)

Drift in wind stress

Goubanovaetal.2017(inprep.)

Shading:windstressamplitude

Vector:windstressvector

Contour:zonalwindstress

Pa

ERAI CNRM-CMModel

Meanevolu-onoftheequatorialwindstress(2°S-2°N)

Westerlybias

Sensitivity experiments

Goubanovaetal.2017(inprep.)

TAUEQ : Seasonal forecast experiment with wind stress replacement (ERAI) at the equator

TAUEQ

Sensitivity experiments

Thetaalongtheequator(2°S-2°N) Thetaalongthecoast(2°) Thetaalongtheequator(2°S-2°N) Thetaalongthecoast(2°)

FEB

MAR

APR

MAY

JUN

JUL

FEB

MAR

APR

MAY

JUN

JUL

CONTROL TAUEQ

Sensitivity experiments

SETA

SSTbiasdailyevoluQonovertheSETA(⁰C):

--- CTRL-HR

--- TAUEQ

--- CTRL-LR

u Remoteforcingfromtheequatorcontributesto~50%

ofwarmSETASSTbiasinCNRM-CMmodel

Error

reduc-on

Physical analysis SST biases in SETA

CTRL: Spurious particular warm horizontal advection

TAUEQ: warm horizontal advection disappears

CTRL-HR

CTRL-LR

TAUEQ

SSTbiasevoluQon(⁰C)

°C/day

RATE

XY-Adv

Z-Adv

FORC

DIFF

ENTR

CTRL

TAUEQ

SETA

Mixed-layer temperature tendency terms

TemperRATE=XY_Adv+Z_Adv+Atm.FORC+Vert.DIFF+ENTR+Res

Conclusions

q Usefulness of driA analysis in Climate Predic-on Systems tounderstandtheoriginofmodelsystema-cbiases

(Toniazzo and Woolnough 2013; Vanniere et al. 2013,2014;Sanchez-Gomezetal.2016;Goubanovaetal.(inprep))

q Casestudy:TropicalAtlan-cSSTbias

-  SeasonalforecastsCNRM-CMmodel

-  AnalysisofthedriAtounderstandSSTbiasintheSETAregion

ü  50%ofSSTbiascanbeexplainedbyspuriouswarmadvecQonfrom the equator, due to atmospheric biases (equatorialwesterlywindbias).

ü  Transpose-AMIPanalysisshowsthatatmosphericwesterlybiasdevelopveryfast(withinthefirst5days)(Roehrigetal.)