Regional Climate Simulation With RCMs
Regional Climate Simulation With RCMs
Regional Climate Simulation With RCMs
Michel Dqu
Mto-France/CNRM, CNRS/GAME 42 avenue Coriolis 310! "oulouse, France
#e$ue%&eteo'(r
"o )e $uote# as*
+$u, M', 2010* Re,ional cli&ate si&ulation -it. a &osaic o( RCMs' Meteorolo,isc.e
/eitsc.ri(t, 10 132, 20-233, #oi*10'112!/0041-2044/2010/04'
1
A)stract
".e ai& o( t.e 5resent stu#6 is to e75lore t.e 5ossi)ilit6 o( ,ettin, .i,. .ori8ontal resolution over
Euro5e )6 usin, a cluster o( lo- co&5utational cost at&os5.eric ,eneral circulation &o#els
1AGCM2' ".e a55roac. is )ase# on ,lo)al &o#els -it. stretc.e# ,ri#, #riven )6 a &e#iu&-
resolution &o#el servin, as a 9con#uctor: to &aintain ti&e consistenc6 o( t.e s6no5tic events an#
avoi# cli&ate #e,ra#ation in t.e lo- resolution 5art o( t.e ,ri#s' ".us t.e ,lo)al &o#els )e.ave as
re,ional cli&ate &o#els 1RCM2' A 5er(ect &o#el a55roac. is use# -it. a .i,. resolution AGCM
servin, as a re(erence' ".e results s.o- t.at -.en t.e #i((erent &o#els are &er,e# on a .i,.
resolution ,ri#, t.e &ean cli&ate is si&ilar to -.at is o)taine# -it. an e$uivalent .i,. resolution
AGCM -it. an unstretc.e# ,ri#' A cli&ate c.an,e nu&erical e75eri&ent s.o-s t.at t.is is also
vali# (or t.e cli&ate res5onse to increase# ,reen.ouse ,as concentration'
2
1. Introduction
".e e75ectation in t.e (ort.co&in, #eca#es o( a -ar&in, o( t.e 5lanet )6 a cou5le o( #e,rees an#
t.e (ear t.at t.e (re$uenc6 o( e7tre&e events -ill increase are co&&onl6 &entione# )6 t.e &ass
&e#ia, )ut are not su((icient (or #ecision &a;ers at a re,ional level' <n t.e 4t. <=CC Assess&ent
Re5ort 1<=CC, 200!2, so&e results at re,ional-scale -ere ,iven t.rou,. statistical #o-nscalin, o(
lar,e-scale 1200 to 400 ;&2 ,eneral circulation &o#els 1GCM2, an# so&e re,ional cli&ate &o#els
1RCM2 si&ulations at .i,.er resolution 130 to 100 ;&2' At Euro5ean level t.e F=-=R>+ENCE
coor#inate# e75eri&ent 1C?R<S"ENSEN et al' 2002, @ACAB et al' 200!2 5ro5ose# an ense&)le o( RCM
si&ulations at 0 ;& resolution'
".e ne7t <=CC e7ercise 1AR2 -ill inclu#e contri)utions o( re,ional &o#els on areas o( t.e ,lo)e
t.at .ave not )e covere# recentl6 )6 international &ulti&o#el 5roCects li;e NARCCA= over Nort.
A&erica 1.tt5*//---'narcca5'ucar'e#u/2 or ENSEMBDES over Euro5e 1?EE<"" AN+ GR<GGS, 20042'
".is 5roCect, en#orse# )6 t.e Eorl# Cli&ate Researc. =ro,ra& 1ECR=2, is na&e# CAR+EF
1CAor#inate# Re,ional +o-nscalin, EF5eri&ent2' ".e tar,et resolution -ill )e 0 ;&' <#eall6, t.e
)est e7ercise -oul# )e to run ,lo)al &o#els at 0 ;& resolution 1or .i,.er, e',' MA"S>E+A et al',
2000 use# a 20 ;& resolution2 #riven )6 sea sur(ace te&5eratures 1SS"2 (ro& lo-er resolution
cou5le# &o#el 1)et-een 100 an# 200 ;& in AR2' But suc. an e7ercise is not a((or#a)le (or &an6
&o#elin, ,rou5s, an# ver6 .eav6 (or ot.er ,rou5s' Since one ;e6 issue o( re,ional res5onse to
cli&ate c.an,e is t.e evaluation o( uncertainties 1+GH>G et al', 200!2, t.e nu&)er o( availa)le
in#e5en#ent nu&erical e75eri&ents is as i&5ortant as t.e .ori8ontal resolution o( t.e &o#el
out5uts'
Anot.er a55roac. is to use several &o#els (ocusin, on various re,ions' ".e tra#itional a55roac. is
)ase# on li&ite# area &o#elin, 1G<ARG<, 10002' But ,lo)al &o#elin, is also (easi)le, as lon, as
.ori8ontal resolution is not constant 1FAF-RAB<NAE<"/ et al', 20042' ".e total co&5utation cost o(
&an6 re,ional &o#els over t.e ,lo)e is lar,er t.an t.e cost -it. a sin,le .i,. resolution &o#el, )ut*
one is not o)li,e#, in 5ractice, to cover t.e -.ole ,lo)e, )ecause t.ere are areas o( interest
t.e 5resent ,eneration o( co&5uters (avors t.e #istri)ution o( co&5utations over several
no#es, an# t.e ela5se# ti&e -it. 5arallel in#e5en#ent &o#els &a6 )e less t.an runnin, a
sin,le .i,. resolution &o#el
t.e co&5utations can )e s.are# )et-een #i((erent centers .avin, #i((erent areas o( interest
1t.is is t.e 5rinci5le o( CAR+EF2
Recent results 5resente# in t.e Dun# 2000 -or;s.o5 an# availa)le (ro& t.e aut.or u5on re$uest as
a tec.nical re5ort 1+GH>G, 20002 s.o- t.at coverin, t.e ,lo)e -it. 20 varia)le resolution &o#els
5ro#uce a sea&less ,lo)al &ean cli&ate co&5ara)le to t.e cli&ate ,enerate# )6 a sin,le .i,.
resolution version o( t.e sa&e &o#el' <ntro#ucin, lateral )oun#ar6 con#itions to &aintain a ti&e
consistenc6 )et-een t.e 20 varia)le resolution AGCMs intro#uces a &inor #e,ra#ation in )oreal
-inter te&5erature'
".e ai& o( t.e 5resent stu#6 is to e75lore t.e a)ove a55roac. -it. a (ocus on Euro5e' <n section 2
t.e &o#el c.aracteristics are 5resente# an# in section 3 t.e &et.o# to ,et .i,. resolution over a 5art
o( t.e ,lo)e is #evelo5e#' ".e cost an# )ene(its o( t.e a55roac., in ter&s o( co&5utation are
e7a&ine# in section 4' <n section , t.e &ean cli&ate o( a cluster o( ,lo)al varia)le resolution
AGCMs is co&5are# -it. t.e &ean cli&ate o( a sin,le .i,. resolution AGCM' <n section 3, t.e
res5onse o( t.is cluster to a cli&ate c.an,e scenario is co&5are# -it. t.e res5onse o( t.e .i,.
resolution AGCM' A conclusion an# t-o 5ers5ectives are ,iven in section !'
3
2. ARPEGE-Climate AGCM
AR=EGE-<FS is t.e nu&erical &o#el use# (or o5erational (orecastin, at Mto-France an#
Euro5ean Centre (or Me#iu&-ran,e Eeat.er Forecasts 1ECMEF2 since t.e 1000s' <t is also use# as
a cli&ate &o#el 1+GH>G et al', 10042' ".is &o#el .as )een nativel6 )uilt to allo- .avin, t.e 5ole o(
lon,itu#e-latitu#e coor#inates ever6-.ere on t.e ,lo)e 1not necessaril6 in t.e Arctic Acean2' <n
a##ition, t.e latitu#es circles can )e #istri)ute# in a non-uni(or& -a6, -it. a .o&ot.etic trans(or&
on t.e 5olar stereo,ra5.ic 5roCection 5lan' ".is is t.e onl6 trans(or&ation (a&il6 -.ic. ena)les
isotro56 ever6-.ere over t.e s5.ere 1CA>R"<ER an# GEDEIN, 10442' ".e 5rinci5le o( t.e
trans(or&ation is*
1'5roCect t.e ,lo)e onto a 5lan )6 a stereo,ra5.ic 5roCection centere# on t.e 5ole o( stretc.in,
2'&ulti5l6 all #istances to t.is 5ole )6 a constant (actor c, na&e# t.e stretc.in, (actor
3',o )ac; to t.e s5.ere )6 an inverse stereo,ra5.ic 5roCection
At a ,iven location, t.e &es. si8e in an6 #irection is &ulti5lie# )6 a &a5 (actor m -.ic. #e5en#s
onl6 on t.e an,ular #istance to t.e 5ole o( stretc.in,*
m=
2c
c
2
1 c
2
1 sin
112
<sotro56 o( t.e .ori8ontal #iscreti8ation 1&a#e 5ossi)le -it. a s5ectral #eco&5osition )ase# on a
trian,ular truncation2 is a use(ul 5ro5ert6 )ecause t.e at&os5.ere e$uations 1e7ce5t t.e Coriolis
ter&2 are in#e5en#ent o( t.e #irection o( t.e a7es' So, intro#ucin, anisotro56 in t.e nu&erics
intro#uces arti(acts in t.e solutions, e',' )6 (ilterin, &ore e((icientl6 -aves co&in, (ro& t.e Eest
t.an -aves co&in, (ro& t.e Nort.' ".e #ra-)ac; o( t.e AR=EGE-<FS tec.ni$ue, na&e# 5olar
stretc.in,, is t.at t.ere is less (le7i)ilit6 in #esi,nin, t.e area o( .i,. resolution t.an -.en 5la6in,
in#e5en#entl6 -it. latitu#e an# lon,itu#e #iscreti8ations 1S"AN<FAR"? an# M<"C?EDD, 10!4, D<, 20032'
".is a55roac., 5ri&aril6 #esi,ne# (or Nu&erical Eeat.er =re#iction, .as )een vali#ate# in cli&ate
&o#e )6 +GH>G an# =<E+ED<EJRE 11002'
".e cli&ate &o#el use# .ere is version '1 o( AR=EGE cli&ate' ".is version is an a#a5tation o( t.e
version #escri)e# in G<BED<N an# +GH>G 120032 to a &ore recent c6cle o( AR=EGE-<FS 1c6cle 32
instea# o( c6cle 142' ".e c.aracteristics o( its #6na&ical core as -ell as t.e 5.6sical
5ara&eteri8ations are ver6 si&ilar' Ane can (in# a #etaile# #escri5tion at
.tt5*//---'cnr&'&eteo'(r/,&,ec/ar5e,e-cli&at/AR=CD<-J'1/in#e7'.t&l ' ".is s5ectral &o#el .as
31 levels on t.e vertical an# a ti&e ste5 o( 20 &in -.atever t.e s5ectral truncation use# in t.e
5resent stu#6, t.an;s to its se&i-i&5licit an# se&i-Da,ran,ian #6na&ics'
As (ar as t.e 5.6sical 5ara&eteri8ations are concerne#, t.e convection sc.e&e is a &ass-(lu7
sc.e&e -it. conver,ence o( .u&i#it6 closure #escri)e# )6 BA>GEA>D" 11042 in its ori,inal
(or&ulation' ".e clou#-#i((usion-5reci5itation sc.e&e is 5resente# in R<CAR+ an# RAIER 110032' ".e
ra#iative sc.e&e is #erive# (ro& MARCRE""E 110002' ".e oro,ra5.ic ,ravit6 -ave sc.e&e is )ase#
on DA"" 110002 #evelo5&ents' ".e soil-ve,etation-sno- sc.e&e is <SBA 1+A>J<DDE et al', 20002'
".e Sea Sur(ace "e&5eratures 1SS"2 are 5rescri)e# -it. &ont.l6 &eans co&in, (ro& F=3-
ENSEMBDES strea& 2 cou5le# si&ulation 1100-21002' ".is scenario -as run -it. a "D33
version o( AR=EGE cou5le# -it. A=A'4 ocean &o#el 1MA+EC et al', 10042' Be6on# 6ear 2000 1see
section 32, t.e A1B scenario -as use# (or t.e ra#iative (orcin,' ".e SS" are correcte# )6
su)tractin, t.e calen#ar &ont.l6 &ean )ias calculate# -it. t.e 1031-2000 5erio#' <n t.e 5resent
stu#6, onl6 t-o ti&e slices 11031-1000 an# 20!1-21002 .ave )een consi#ere#'
4
3. An AGCM used as an RCM
Eit. t.e 5olar stretc.in, tec.ni$ue, t.e area o( &a7i&u& resolution is a s5.erical ca5' ".e .i,.er
t.e stretc.in, (actor, t.e s&aller t.e area o( .i,. resolution 1e',' t.e area -it. &ore t.an 40K o( t.e
&a7i&u& resolution2' An alternative to t.e use o( a lo- stretc.in, (actor to cover a lar,e area is t.e
use o( several &o#els' <( one -ants to cover t.e ,lo)e -it. a uni(or& #istri)ution o( 5oles, t.e
&a7i&u& nu&)er is 20, #ue to t.e li&ite# series o( 5latonic soli#s' ?o-ever, one can c.eat a little,
)6 consi#erin, t.e 20 (aces o( an icosa.e#ron* t.e6 are e$uilateral trian,les an# can )e easil6
#ivi#e# into e$ual-area e$uilateral trian,les' For e7a&5le, i( eac. trian,le is #ivi#e# into 0 trian,les,
t.e total nu&)er o( trian,les is 140' ".e 02 trian,le vertices are al&ost uni(or&l6 #istri)ute# on t.e
s5.ere' Gettin, &ore vertices is eas6 to o)tain )6 usin, a (iner #iscreti8ation o( t.e 5ri&ar6
trian,les' ?ere 0 5oles, locate# in an# aroun# Euro5e, are consi#ere# a&on,st t.ose 02 vertices'
".eir coor#inates are ,iven in "a)le 1' For eac. &o#el a "D33 s5ectral truncation 1312 ;&
resolution2 .as )een use#' ".is resolution .as )een t65icall6 use# in <=CC-AR4 an# can )e run at a
s5ee# o( a (e- si&ulate# 6ears 5er real #a6 on a stan#ar# 5ersonal co&5uter' ".is is a c.ea5 )ut
reasona)le truncation in 2000 co&5uter stan#ar#s' Eit. a stretc.in, (actor c o( , t.e &a7i&u&
resolution at t.e 5ole is 32 ;& an# t.e &ini&u& resolution, at t.e anti5o#es is 130 ;&' <( -e
&er,e t.e 0 stretc.e# ,ri#s, retainin, at eac. location t.e &o#el -.ic. .as t.e .i,.est resolution,
-e ,et a co&5osite ,ri#' Fi,ure 1 s.o-s t.e resolution o( t.is ,ri# over Euro5e' <ts &ini&u&
resolution )et-een t-o 5oles is 43 ;& an# t.e &ean resolution over t.e #is5la6 area is !0 ;&'
Contrar6 to +GH>G 120002 suc. &o#els cannot )e run in stan#alone &o#e 1or GCM &o#e2 )ecause
t.eir too coarse resolution outsi#e Euro5e is #etri&ental to t.e si&ulation ever6-.ere' ".ese
&o#els nee# a #river, li;e stan#ar# li&ite# area &o#els 1DAM2' ?ere t.e #river is a non-stretc.e#
version o( t.e sa&e &o#el -it. a "D12! truncation 113 ;&2 -.ic. is enou,. to re5resent t.e
s6no5tic circulation' ".e stretc.e# &o#els are rela7e# eac. ti&e ste5 over t.e ,lo)e to-ar#s t.e
"D12! si&ulation, -.ic. acts also as a 9con#uctor: )6 ena)lin, t.e nine si&ulations to re&ain in
a,ree&ent at t.eir )oun#aries' At.er-ise, eac. &o#el -oul# .ave an in#e5en#ent c.ronolo,6' ".e
lo-est vertical level is let (ree, as -ell as t.e soil varia)les, -it. a linear transition o( t.e rela7ation
coe((icient in t.e lo-est levels' ".e e-(ol#in, ti&e o( t.e rela7ation is one ti&e ste5 120 &in2 (or
-in#, te&5erature an# sur(ace 5ressure, -.ereas it is 3 .ours (or &oisture' <n#ee# t.e #rivin,
1"D12!2 an# #riven 1stretc.e#2 &o#els .ave t.e sa&e 5.6sics an# t.e sa&e vertical #iscreti8ation*
i( t.e6 use t.e sa&e 5ro,nostic varia)les as an in5ut, t.e6 calculate t.e sa&e (lu7es' But t.e #rivin,
an# #riven &o#els .ave a #i((erent sur(ace elevation, -.ic. intro#uces #i((erences in t.e 5lanetar6
)oun#ar6 la6er' <t is t.ere(ore tra#itional to let so&e (ree#o& at t.e ti&e ste5 scale to t.e #ia)atic
5rocesses in t.e #riven &o#el (or nu&erical a#Cust&ents 1e',' G>D+BERG et al', 2002' Suc.
#i((erences are .ar&less )ecause t.e6 are little a#vecte# (ro& (ar areas insi#e t.e .i,. resolution
5art o( t.e stretc.e# ,ri#'
".ese nine stretc.e# &o#els are RCMs, so t.e6 &ust .ave a rela7ation-(ree area' "o t.is 5ur5ose, a
s5atial &as; is a55lie# to t.e rela7ation coe((icients' For &o#el i 1iL1 ''' 02 at location x, t.is &as;
is *
M
i
x =Z
d
i
xd
0
,d
i
xd
1
x , ' '' d
i
xd
0
x
T
d
i
x / d
0
122
-.ere d
i
(x) is local &es. si8e o( &o#el i, calculate# -it. E$' 112M d
0
is t.e constant &es. si8e o( t.e
"D12! #river' Z is 8ero -.en t.e con#ition o( its ar,u&ent is &et, i'e' -.en &o#el i .as t.e
&a7i&u& resolution, an# is one ot.er-ise' Z cannot )e 8ero (or t-o &o#els at t.e sa&e location' T
is a 5iece-ise linear (unction )uilt as*
TL0 )elo- 0'4
TL1/24 (or d
i
(x)/d
0
L0'3
TL1/4 (or d
i
(x)/d
0
L0'4
TL1 (or d
i
(x)/d
0
L1
TL10 )e6on# 2
".e )ase i#ea is t.at t.e lo-er t.e RCM resolution, t.e .i,.er t.e constraint' Accor#in, to E$' 122,
an RCM is let (ree 1M
i
(x)L02 at an6 ,ri# 5oint -.ere its resolution is t.e &a7i&u& out o( all
stretc.e# &o#els an# #river' As a conse$uence, t.e co&5osite ,ri# 1Fi,ure 12 is )uilt )6 collectin,
out o( eac. stretc.e# &o#el t.e ,ri# 5oints (or -.ic. ZL0' Fi,ure 2 s.o-s t.is &as; (or t.e &o#el
-it. 5ole 4'
<n or#er to c.ec; t.e relia)ilit6 o( t.e co&5osite ,ri#, a sin,le situation is #is5la6e# as an e7a&5le*
@anuar6 1
st
, 12 >"C, 6ear 30 o( t.e control si&ulation 16ear 1000 in t.e c.ronolo,6 o( t.e SS"
(ro& t.e cou5le# scenario2' ".is &a5 is not t.e result o( a selection* no ot.er #ail6 &a5 .as )een
5lotte#' Fi,ure 3 s.o-s &ean sea level 5ressure (or t.e #river 1"D12!2 an# t.e #riven 1co&5osite2
&o#els' E.en loo;in, care(ull6 at t.e lo-er 5anel, one can see so&e sea&s sout. o( <celan#, nort.
o( Gotlan# ''' )ut (ro& a s6no5tic 5oint o( vie-, t.e t-o &a5s are i#entical' ?o-ever insi#e eac.
contri)utin, su)-#o&ain a #i((erent &o#el .as )een run -it. a 5ro,ressive constraint outsi#e its
)oun#ar6'
A static situation is not enou,. to convince t.at transient 5.eno&ena translate nicel6 t.rou,. t.e
#o&ain' Fi,ure 4 s.o-s a ?ov&oeller #ia,ra& alon, t.e 40NN 5arallel #urin, t.e last -inter o( t.e
si&ulation' ".is traCector6 #oes not corres5on# to a stor& trac;, )ut ena)les to cross e$uivalentl6 4
su)-#o&ains 1see ca5tion2' <n t.e -estern .al(, trou,. are &ovin, ra5i#l6 east-ar#s, -.ereas in t.e
eastern .al(, $uasi-stationar6 .i,. 5ressure 1a)ove 1030 .=a2 a55ear (ro& ti&e to ti&e' ".e
co&5osite is close to t.e #river, an# no sea& a55ear at 32NE, 0NE an# 32NE, -.ereas #i((erent
si&ulations are 5ut si#e-)6-si#e' <n or#er to ,et a &ore s6ste&atic evaluation o( t.e #a6-to-#a6
varia)ilit6, Fi,ure s.o-s t.e #ail6 stan#ar# #eviation o( @anuar6 &ean sea level 5ressure' ?ere
a,ain, t.e 0 #i((erent &o#els &atc. in a satis(actor6 -a6' <n 5articular, t.e un#ulations over Atlas
&ountain or Sout. o( <celan# are not a lateral )oun#ar6 e((ect, )ecause t.e6 are also 5resent in t.e
#river si&ulation'
. Com!utation costs and "ene#its
<( t.e ai& -ere to run 02 &o#els in or#er to cover t.e -.ole ,lo)e, t.e a55roac. -oul# )e &uc.
&ore e75ensive t.an runnin, a sin,le ,lo)al &o#el at e$uivalent .i,. resolution' An t.e ot.er .an#,
i( t.e ai& is to (ocus on a s&all re,ion, a sin,le stretc.e# &o#el is su((icient, an# t.e a55roac. .as
)een alrea#6 #iscusse# in FAF-RAB<NAE<"/ et al' 120042' ?ere t.e re,ion -e -ant to #o-nscale
covers Nort. Atlantic an# Euro5e at 30 ;& resolution' A ,lo)al version o( t.e &o#el at suc.
resolution corres5on#s to a "D310 s5ectral truncation an# involves 144004 ,ri# 5oints' <n t.e &ulti-
&o#el a55roac. use# .ere, -e nee# a #river &o#el at "D12! truncation, involvin, 24!2 ,ri#
5oints, an# 0 stretc.e# &o#els involvin, 3232 ,ri# 5oints' Ee use t.us 40330 ,ri# 5oints, so 3K
o( t.e cost o( a ,lo)al &o#el'
?o-ever, countin, t.e ,ri# 5oints ,ives onl6 a 5artial vie- o( t.e real cost, )ecause t.ere are
over.ea#s #ue to in5ut/out5ut, &e&or6 #istri)ution, etc' Anot.er evaluation, closer to t.e actual
nee# o( a scientist 5er(or&in, a si&ulation, is t.e ela5se# ti&e (or a 30-6ear si&ulation' ".is ti&e is
o( course &ac.ine #e5en#ent' ".e e75eri&ents 5resente# .ere .ave )een run on a 32-core no#e o(
t.e &ain(ra&e <BM co&5uter availa)le at ECMEF in 2000' ".e "D310 run laste# 2 #a6s, t.e
"D12! run laste# ! #a6s, an# t.e 0 stretc.e# &o#el runs laste# 4 #a6s' ".e ratio is 30K, close to
-.at -e ,et -.en countin, t.e ,ri# 5oints'
".e c.oice o( nine &o#els .ere .as )een #ictate# )6 t.e nee# to .ave enou,. sea&s to 5rove t.at
3
t.e O&o8aicO is not a O5atc.-or;O' <( -e -ant Cust a #o-nscalin, over Euro5e, (our stretc.e#
&o#els are enou,. an# t.e )ene(it is .i,.er 1ratio 34K in ,ri# 5oint an# 44K in ela5se ti&e2'
Anot.er a#vanta,e o( t.e &ulti-stretc.e# a55roac. is t.at a "D310 version o( t.e &o#els re$uires a
)i, co&5uter, )ecause o( t.e &e&or6' A "D12! version can )e run on a )i, 5ersonal co&5uter -it.
12 GB &e&or6, an# a stretc.e# "D33 nee#s 1 #a6s (or a 30-6ear run on a stan#ar# #es;to5
5ersonal co&5uter' E.en t.e constraint is not onl6 real ti&e (or e75eri&ent, )ut access to a
&ain(ra&e co&5uter, -e .ave .ere anot.er )ene(it' But -e &ust ;ee5 in &in# t.at t.is 5rovi#es a
#o-nscalin, a55roac., -.ic. is not as relia)le as an ori,inal ,lo)al .i,.-resolution (ree si&ulation'
<n t.e (ollo-in,, -e -ill e7a&ine .o- close t.e Oco56O is (ro& t.e Oori,inalO'
$. Per#ect model %alidation
At t.is sta,e, -e s.all e7a&ine t.e cli&ate o( t.e co&5osite &o#el' Ee -ill not co&5are &o#el
#ata -it. o)servations, )ut use a 5er(ect &o#el a55roac.' <n#ee#, t.e sco5e o( t.e tec.ni$ue use#
.ere is not to co&5ensate so&e errors o( AR=EGE-<FS cli&ate &o#el )6 ot.er errors co&in, (ro&
t.is o## co&5osite a55roac.' ".e ai& o( t.e co&5osite is to &i&ic at a lo-er co&5utation cost t.e
)e.avior o( a .i,. resolution &o#el over a ,iven re,ion' ".e re(erence .ere is t.ere(ore a 30 6ear
si&ulation o( t.e sa&e &o#el, use# at "D310 truncation, -.ic. corres5on#s to 32;&, t.e &a7i&u&
resolution o( t.e co&5osite &a5'
Fi,ure 3 s.o-s t.e 2& te&5erature #i((erences' <n -inter, t.e onl6 re,ion -.ere t.e RCM )e.aves
in a #i((erent -a6 (ro& t.e .i,. resolution AGCM is Greenlan# an# Nort. Cana#a' <n su&&er,
Euro5e is 0' P -ar&er -it. t.e RCM, an# t.is -ar&in, reac.es 1 P in sout.-eastern Euro5e' All
t.ese (eatures are si,ni(icant at 0K level 1si&5le Gaussian test on t.e avera,e o( 30 in#e5en#ent
values2' A t-test is not a55lica)le .ere )ecause t.e si&ulations are #riven )6 t.e sa&e SS" an# are
not co&5letel6 in#e5en#ent' ".ese (eatures are a res5onse o( t.e &o#el to t.e lateral con#itions,
-.ic. &eans t.e6 are not a result o( t.e sa&5lin, varia)ilit6 1co&in, (ro& t.e #rivin, run2 or (ro&
t.e &o#el internal varia)ilit6' ".is )e.avior .as )een o)serve# in 5er(ect &o#el e75eri&ents -it. a
DAM 1RA+> et al', 20042' ".e -ar& s5ot near t.e Re# Sea in su&&er is outsi#e t.e #o&ain an# is
an e7tra5olation arti(act'
Fi,ure ! s.o- t.e 5reci5itation s6ste&atic error' ".e a&5litu#e is )et-een -0'3 an# 0'3 &&/#a6 an#
t.e error 5attern #oes not e7.i)it an6 stron, structure, e7ce5t a 0'3 &&/#a6 #r6 )ias over Euro5e in
su&&er, consistent -it. t.e -ar& )ias an# statisticall6 si,ni(icant'
&. Res!onse to a climate chan'e scenario
Ane o( t.e &aCor a55lications o( GCM/RCM is t.e res5onse to ,reen.ouse ,as increase' ?ere -e
-ill consi#er t.e A1B scenario (or t.e 5erio# 20!1-2100' ".e re(erence 5erio# is 1031-1000,
e7a&ine# in section '
<n -inter, t.e te&5erature res5onse o( t.e .i,. resolution &o#el 1not s.o-n2 is a)out 2 P in
sout.ern an# -estern Euro5e an# 2' P in nort.-eastern Euro5e 1u5 to 4 P in nort.ern Finlan#2'
".e #i((erence )et-een t.e co&5osite an# t.e .i,. resolution res5onse 1Fi,ure 4a2 s.o- t.at t.e
co&5osite is 0' P -ar&er in nort.-eastern Euro5e' <n su&&er, t.e -ar&in, reac.es P in
sout.ern Euro5e an# is )et-een 2 an# 4 P in t.e nort.ern 5art 1not s.o-n2 ' ".e co&5osite is
col#er )6 0' P over t.e nort.ern 5art 1Fi,ure 4)2' Ane can re&ar; t.at in )ot. seasons, t.e RCM
a55roac. increases t.e s5atial contrasts'
As (ar as t.e 5reci5itation res5onse is concerne#, -e ,et a #ecrease in -inter 1-0'3 to -0'3 &&/#a62
over t.e Me#iterranean area, -it. an e7tension to t.e A8ores area' Nort. o( t.e 4N 5arallel,
5reci5itation increases -it. a &a7i&u& o( 1 &&/#a6 Sout. o( <celan# an# alon, t.e Nor-e,ian
coast' ".e co&5osite &o#el 1Fi,ure 0a2 #oes not &o#i(6 t.is 5attern )ut re#uces t.e #r6 res5onse in
!
t.e -estern 5art o( t.e Me#iterranean an# a&5li(ies it in t.e -estern 5art' <n su&&er, t.e
5reci5itation #ecrease #e5icte# )6 t.e .i,. resolution AGCM 1not s.o-n2 is ,enerali8e# over
Euro5e, -it. a &a7i&u& o( -0'0 &&/#a6 over Ger&an6' ?ere, t.e co&5osite re#uces t.e res5onse
-it. a 5ositive #i((erence o( 0'3 &&/#a6 over nort.ern Euro5e'
Given t.e stron, uncertainties -.ic. are attac.e# to cli&ate c.an,e over Euro5e, one can consi#er
t.at t.e #i((erences -e o)serve in Fi,ures 4 an# 0 #o not e75ress a real #isa,ree&ent )et-een t.e
.i,. resolution AGCM an# t.e co&5osite RCM'
(. Conclusion and !ers!ecti%es
<t .as )een s.o-n t.at t.e ,ri# stretc.in, tec.ni$ue -it. a .i,. (actor, associate# -it. a 5ro,ressive
#rivin, in t.e lo- resolution 5art o( t.e ,ri# o((ers an interestin, alternative to t.e tra#itional
li&ite# area &o#elin, constraine# )6 +avies lateral rela7ation' Ane can easil6 )uil# non-
rectan,ular #o&ains 1even non-connecte# #o&ains li;e Euro5e an# Nort. A&erica2' ".e (act t.at
t.e constraint is rat.er loose at t.e li&it )et-een t-o nei,.)ors &a;es t.e sea&s $uasi-invisi)le
an# avoi#s t.e a#Cust&ents 5ro)le&s at t.e lateral )oun#aries' Eit. an a55ro5riate a#a5tation, t.is
tec.ni$ue coul# also )e a55lie# to a &osaic o( li&ite# area &o#els'
".is series o( nine si&ulations is 5reli&inar6, )ecause t.e e$uivalent resolution is not #i((erent (ro&
a sin,le 120-;&-,ri# &o#el run -it. a stretc.in, (actor o( 2' Ee ,ot 5ositive results in ter&s o(
ti&e consistenc6 )et-een nei,.)or AGCMs, as -ell as in ter&s o( si&ilarit6 o( t.e &ean cli&ates
an# &ean cli&ate res5onses -it. a ,lo)al 30-;&-,ri# &o#el' ".is su,,ests t.at t.e resolution o( t.e
stretc.e# AGCMs coul# )e increase# 1e',' ,oin, to 100 ;&2' Ane coul# also increase t.e stretc.in,
(actor c -.ic. -as in t.is e75eri&ent' A (actor cL4 see&s (easi)le (ro& 5reli&inar6 tests' A (actor
cL12 6iel#s to nu&erical insta)ilities -it. AR=EGE-<FS, #ue to t.e #iver,ence ter& in t.e se&i-
i&5licit sc.e&e' Eit. an increase# stretc.in, (actor t.e nu&)er o( 5oles s.oul# increase, in or#er to
avoi# lar,e s5atial variations in resolution' ".is is not a tec.nical 5ro)le&* an e$uilateral trian,le is
eas6 to #ivi#e into e$ual areas' ".e a55roac. use# .ere can )e e7ten#e# to reac. a &es. si8e o( 10-
1 ;& over a (le7i)le #o&ain'
A secon# 5ers5ective is to 5ass t.e .i,. resolution in(or&ation 1e',' a #ee5 c6clone2 (ro& one
#o&ain to its nei,.)or )6 a t-o -a6 nestin, tec.ni$ue 1Doren8 an# @aco), 2002' ".is -ill &i&ic
-.at occurs in a true .i,. resolution &o#el' ".e counter5art o( t.is is t.at all si&ulations 1t.e
#river an# t.e #riven ones2 &ust )e s6nc.ronous, -.ic. &a;es t.e co&5utation less (le7i)le' But
on a #istri)ute# co&5uter, t.e e7c.an,e o( in(or&ation )et-een t.e nei,.)ors -ill occur ever6 2-3
.ours instea# o( ever6 ti&e ste5 in a .i,. resolution &o#el 120 &in2, -.ic. &a;es t.e a55roac.
5otentiall6 (ruit(ul'
".e vali#ation a55roac. use# in t.is stu#6 ,ives a (ra&e-or; to e75eri&ent t.e (easi)ilit6 o( t.e
a)ove &entione# t-o 5ers5ectives'
4
Ac)no*led'ments
".is -or; -as 5artl6 su55orte# )6 t.e Euro5ean Co&&ission =ro,ra&&e F=3 un#er contract
GACE-C"-2003-03!00 1CEC<D<A2 an# )6 Frenc. national 5ro,ra& ANR 1JMCS04-SCAM=E<2'
".an;s are also #ue to t.e &an6 #evelo5ers o( AR=EGE-<FS &o#el at Mto-France an# ECMEF'
0
References
C?R<S"ENSEN @'?', "'R' CAR"ER, F' G<ARG<, 2002' =R>+ENCE e&5lo6s ne- &et.o#s to assess
Euro5ean cli&ate c.an,e'-- E'A'S' 43, 14!'
CA>R"<ER =', @'F' GEDEIN, 1044* A ,lo)al nu&erical -eat.er 5re#iction &o#el -it. varia)le
resolution* A55lication to t.e s.allo- -ater e$uations'-- Huart' @' Ro6' Meteor' Soc' 114, 1321--
1343'
BA>GEA>D" =', 104* A si&5le 5ara&eteri8ation o( t.e lar,e-scale e((ects o( cu&ulus convection'--
Mon' Eeat.er Rev' 113, 2104--2121'
+GH>G M', C' +REJE"AN, A' BRA>N, +' CAR<ADDE, 1004* ".e AR=EGE/<FS at&os5.ere &o#el* a
contri)ution to t.e Frenc. co&&unit6 cli&ate &o#ellin,'-- Cli&' +6n' 10, 240--233'
+GH>G M', @'=' =<E+ED<EJRE, 100* ?i,. resolution cli&ate si&ulation over Euro5e'-- Cli&' +6n' 11,
321--330'
+GH>G M', +'=' RAEED, +' DQ"?<, F' G<ARG<, @'?' C?R<S"ENSEN, B' RACPED, +' @ACAB, E' P@EDDS"RRM, M'
+E CAS"RA, B' JAN +EN ?>RP, 200!* An interco&5arison o( re,ional cli&ate si&ulations (or Euro5e*
assessin, uncertainties in &o#el 5roCections'-- Cli&atic C.an,e 41, 3--!0'
+GH>G M', 2000* Re,ional cli&ate si&ulation -it. &osaic GCMs* (ree versus #riven si&ulations'--
CNRM "ec.nical Re5ort 3 1availa)le u5on re$uest (ro& t.e aut.or2, 1! 55'
+A>J<DDE ?', S' =DAN"AN, @'F' RAIER, +'B' S"E=?ENSAN, S' "I"ECA, D' PERGAA", S' DAFAN", R'A' BE""S,
2000* ".e i&5ortance o( ve,etation (ee#)ac;s in #ou)le#-CA2 ti&e-slice e75eri&ents'-- @'
Geo5.6s' Res' 10, 14441--14431'
FAF-RAB<NAE<"/ M', @' CA"G, B' +>GAS, M' +GH>G, @'D' MC GREGAR, A' BEDAC?<"SP<, 2004* Stretc.e#-
,ri# Mo#el <nterco&5arison =roCect* #eca#al re,ional cli&ate si&ulations -it. en.ance# varia)le
an# uni(or&-resolution GCMs'-- Meteorol' At&os' =.6s' 100, 10--1!!'
G<BED<N A'D', M' +GH>G, 2003* Ant.ro5o,enic cli&ate c.an,e over t.e Me#iterranean re,ion
si&ulate# )6 a ,lo)al varia)le resolution &o#el'-- Cli&' +6n' 20, 32!S330'
G<ARG< F', 1000* Si&ulation o( re,ional cli&ate usin, a li&ite# area &o#el neste# in a ,eneral
circulation &o#el'-- @' Cli&ate 3, 041--033'
G>D+BERG A', E' PAAS, M' +GH>G, S' IANG, S' JES"ER "?ARSEN, 200* Re#uction o( s6ste&atic errors
)6 e&5irical &o#el correction* i&5act on seasonal 5re#iction s;ill'-- "ellus !A, !--44'
?EE<"" C'+', +'@' GR<GGS, 2004* Ense&)les-)ase# 5re#ictions o( cli&ate c.an,es an# t.eir
i&5acts'-- E'A'S' 4, 33'
<=CC, 200!* Cli&ate C.an,e 200!* ".e =.6sical Science Basis' Contri)ution o( Eor;in, Grou5 <
to t.e Fourt. Assess&ent Re5ort o( t.e <nter,overn&ental =anel on Cli&ate C.an,e'o( t.e
inter,overna&ental 5anel o( cli&ate c.an,e'-- Solo&on S', +' Hin, M' Mannin,, /' C.en, M',
Mar$uis, P'B' Aver6t, M' "i,nor, ?'D' Miller 1E#s'2' Ca&)ri#,e >niversit6 =ress, >P, 003 55'
@ACAB +',D' BTRR<NG, A'B' C?R<S"ENSEN, @'?' C?R<S"ENSEN, M' +E CAS"RA, M' +GH>G, F' G<ARG<, S'
?AGEMANN, M' ?<RSC?<, R' @ANES, E' P@EDDS"RRM, G' DEN+ER<NP, B' RACPED, E'S' SUNC?E/, C' SC?TR,
S'<' SENEJ<RA"NE, S' SAMA", A' JAN >D+EN, B' JAN +EN ?>RP, 200!* An inter-co&5arison o( re,ional
cli&ate &o#els (or Euro5e* +esi,n o( t.e e75eri&ents an# &o#el 5er(or&ance'-- Cli&atic C.an,e
41, 31--2'
D< /'F', 2003* At&os5.eric GCM res5onse to an i#eali8e# ano&al6 o( t.e Me#iterranean sea
sur(ace te&5erature'-- Cli&' +6n' 2!, !4!--413'
LORENZ P., D. JACOB, 200*<n(luence o( re,ional scale in(or&ation on t.e ,lo)al circulation*A t-o-
10
-a6 nestin, cli&ate si&ulation'-- Geo5.6s' Res' Detters 32, D14!03, #oi*10'1020/200GD02331'
DA"" F', 1000* Alleviation o( stationar6 )iases in a GCM t.rou,. a &ountain #ra, 5ara&eteri8ation
sc.e&e an# a si&5le re5resentation o( &ountain li(t (orces'-- Mon' Eeat.er Rev' 12, !44--401'
MA+EC G', =' +EDECD>SE, M' <MBAR+, M' DGJI, 1004* A=A 4'1 ocean ,eneral circulation &o#el
re(erence &anual'-- Notes #u =Vle #e Mo#lisation 11, "ec.' re5', <=SD, =aris, 01 55'
MA"S>E+A M', R' M</>"A, S' P>S>NAP<, 2000* Future c.an,e in -interti&e at&os5.eric )loc;in,
si&ulate# usin, a 20-;&-&es. at&os5.eric ,lo)al circulation &o#el' @' Geo5.6s' Res' 114,
+12114, #oi*10'1020/2000@+011010'
MARCRE""E @'@' 1000* <&5act o( c.an,es to t.e ra#iation trans(er 5ara&eteri8ations 5lus clou#
o5tical 5ro5erties in t.e ECMEF &o#el'-- Mon' Eeat.er Rev' 114, 44!--4!3'
RA+> R', M' +GH>G, S' SAMA", 2004* S5ectral nu#,in, in a s5ectral re,ional cli&ate &o#el'-- "ellus
30A12, 44--40!'
R<CAR+ @'D', @'F' RAIER 1003* A statistical clou# sc.e&e (or use in an AGCM'-- Ann' Geo5.6sicae
11, 100--111'
S"AN<FAR"? A'N', ?'D' M<"C?EDD, 10!4* A varia)le-resolution (inite-ele&ent tec.ni$ue (or re,ional
(orecastin, -it. t.e 5ri&itive e$uations'-- Mon' Eea' Rev' 103, 430--44!'
11
=ole nu&)er 1 2 3 4 3 ! 4 0
lon,itu#e 0' 4' -4' 0' 0' -34'0! 34'0! -1'4 1'4
latitu#e 20'01 3'23 3'23 4'24 !4'33 4'!3 4'!3 40'!3 40'!3
"a)le 1* Don,itu#es an# latitu#es o( t.e nine 5oles use# in t.is stu#6'
12
Fi,ure 1* ?ori8ontal resolution o( t.e co&5osite ,ri# )ase# on t.e 0 stretc.e# ,ri#s' Contour
interval 10 ;&'
13
Fi,ure 2* Mas; a55lie# to t.e rela7ation coe((icient (or t.e stretc.e# &o#el -it. 5ole 4'Contours
0'01, 0'1, 1 an# '
14
Fi,ure 3* Mean sea level 5ressure o( @anuar6 1, 6ear 30 o( t.e control si&ulation (or #rivin, &o#el
1to52 an# co&5osite &o#el 1)otto&2' Contour interval 10 .=a
1
Fi,ure 4* ?ov&oeller #ia,ra& o( &ean sea level 5ressure o( t.e last -inter 1+@F2 alon, 40NN
5arallel (or #rivin, &o#el 1to52 an# co&5osite 1)otto&2' ".e co&5osite corres5on#s to 5ole 2
1E- 32E2, 5ole 3 132E-0E2, 5ole 4 10E-32E2 an 5ole 1 132E-E2' >nit is #i((erence to 1000.=a,
contour interval is 10 .=a, #ata a)ove 30 or )elo- 0 are s.a#e#'
13
Fi,ure * +ail6 stan#ar# #eviation o( @anuar6 &ean sea level 5ressure (or #rivin, &o#el 1to52 an#
co&5osite 1)otto&2M contour interval 2 .=a'
1!
Fi,ure 3* 2& te&5erature s6ste&atic error in t.e control run in +@F 1to52 an# @@A 1)otto&2' Contour
interval 0' P, 0-contour o&itte#, s.a#in, a)ove 1 P an# )elo- 1 P
14
Fi,ure !* As (i,ure 3 (or 5reci5itation' Contour interval 0'3 &&/#a6, s.a#in, a)ove 0'3 an# )elo-
-0'3
10
Fi,ure 4* +i((erence in 2& te&5erature cli&ate c.an,e res5onse (or +@F 1to52 an# @@A 1)otto&2'
Contour interval 0' P, 0-contour o&itte#, s.a#in, a)ove 1 P an# )elo- 1 P
20
Fi,ure 0* As (i,ure 4 (or 5reci5itation' Contour interval 0'3 &&/#a6, s.a#in, a)ove 0'3 an# )elo-
-0'3
21