1
The No h A lan ic mean s a e in eddy- esol ing coupled models:
a mul imodel s udy
Amanda F igola1, Eneko Ma in-Ma inez1, Edua do Mo eno-Chama o1,2, Ma ga ida Samsó1, Saskia
Loos el -Tomas1, Pie e-An oine B e onniè e1, Da ia Kuzne so a1, Xia Lin3, Pablo O ega1
1Ba celona Supe compu ing Cen e , Ba celona, Spain 5
2now a Max Planck Ins i u e o Me eo ology, Hambu g, Ge many
3Nanjing Uni e si y o In o ma ion Science and Technology, Nanjing, China
Co espondence o: Amanda F igola (amanda. [email p o ec ed]s)
Abs ac . Ocean mesoscale s uc u es, which a e pa ame e ized in s anda d esolu ion models, play an impo an ole in 10
ea u ing global ocean dynamics. He e we s udy he e ec s o inc easing model ocean esolu ion o eddy- esol ing scales on
he ep esen a ion o he No h A lan ic mean s a e, by compa ing an ensemble o ou HighResMIP coupled his o ical
simula ions wi h nominal ocean esolu ions o a leas 1/10º – co esponding o he models CESM1-CAM5-SE-HR, EC-
Ea h3P-VHR, HadGEM3-GC31-HH, and MPI-ESM1-2-ER – o a baseline o 39 Coupled Model In e compa ison P ojec
phase 6 (CMIP6) simula ions a coa se esolu ion. Despi e a wa m and sal y bias a he su ace, we ind gene ally imp o ed 15
e ical s a i ica ion in he Lab ado (LS) and wes e n I minge Seas (IS) in he high esolu ion ensemble, leading o
s onge deep wa e con ec ion in ha egion, in close ag eemen wi h obse a ions. Bo h he o e u ning and ba o opic
ci cula ions o he No h A lan ic p esen signi ican imp o emen s in he eddy- esol ing models: he A lan ic Me idional
O e u ning Ci cula ion (AMOC) is weake han o lowe esolu ion models and close o RAPID obse a ions; he pa hs,
s eng h, and s uc u e o he Gul S eam (GS), No h A lan ic Cu en (NAC), and subpola gy e (SPG) a e also imp o ed, 20
esul ing in o educed su ace empe a u e and salini y biases no h o Cape Ha e as (NCH) and in he Cen al No h
A lan ic (CNA).
1 In oduc ion
The No h A lan ic is a key egion wi h mul iple impac s on he global clima e sys em. One o i s main oles is he
edis ibu ion o hea om low o high la i udes h ough he A lan ic Me idional O e u ning Ci cula ion (AMOC). A 26.5º 25
N he A lan ic Ocean anspo s ~1.2 PW o hea , which ep esen s ~60–65 % o he combined con ibu ions om he
A lan ic and he Paci ic a hose la i udes (Ganachaud and Wunsch, 2000; Johns e al., 2023; Lumpkin and Spee , 2007;
T enbe h and Fasullo, 2017). Indeed, hea anspo by he AMOC explains he milde empe a u es in he No he n
Hemisphe e compa ed o he Sou he n Hemisphe e (Buckley and Ma shall, 2016). The No h A lan ic is also an impo an
an h opogenic ca bon sink, con ibu ing o educing a mosphe ic global wa ming (B own e al., 2021). This egion exhibi s 30
h ps://doi.o g/10.5194/egusphe e-2025-547
P ep in . Discussion s a ed: 21 Feb ua y 2025
c
Au ho (s) 2025. CC BY 4.0 License.
2
he highes global up ake a e o an h opogenic ca bon pe a ea, which is ela ed o enhanced e ical pene a ion ia he
AMOC uppe cell (G ube e al., 2019).
Changes in he AMOC in he pas ha e been associa ed wi h ab up changes in clima e (Ng e al., 2018), and clima e
p ojec ions indica e consis en AMOC weakening a inc eased CO2 le els (Jackson e al., 2020), wi h impo an e ec s upon
clima e, such as No he n Hemisphe e d ying and cooling, and a sou hwa d shi in he in e opical con e gence zone 35
(ITCZ; Bellomo and Mehling, 2024; Liu e al., 2020). Thus, conside ing i s undamen al ole wi hin he clima e sys em, he
dynamics o he No h A lan ic need o be app op ia ely ep esen ed in clima e models, in o de o us ully e alua e he
u u e impac s o clima e change.
The No h A lan ic ci cula ion is in luenced by a se ies o elemen s and p ocesses ha a e s ongly in e connec ed: he
s eng h, pa hs, and wa e p ope ies o he AMOC no hwa d limb, which de e mine he hea and salini y con en o he 40
wa e s eaching he subpola No h A lan ic (SPNA; Ma zocchi e al., 2015); deep wa e con ec ion a high la i udes, which
is linked o he o ma ion o dense wa e s (Koenigk e al., 2021); su ace wa e mass ans o ma ion in he SPNA, associa ed
wi h cooling h ough a mosphe ic hea luxes (Jackson and Pe i , 2023); densi ica ion o wa e s a he subpola gy e (SPG)
bounda y (Ka sman e al., 2018); o mixing wi h A c ic wa e s en e ing he SPG h ough he G eenland–Sco land Ridge
(Dey e al., 2024). 45
The ho izon al esolu ion o ocean models is c ucial o accu a ely ep esen ing hese p ocesses. A ealis ic ba hyme y
is key in cha ac e izing ocean h ough lows and hei p ope ies, in pa icula hose o he G eenland–Sco land Ridge
(Ka sman e al., 2018; Dey e al., 2024). Addi ionally, ocean ho izon al esolu ion de e mines he ep esen a ion o
mesoscale ea u es, he so-called ocean eddies. These s uc u es impac he ocean dynamics h ough hei undamen al ole in
he anspo o hea and sal (Sun e al., 2019; T eguie e al., 2012). Ocean eddies' a e age ho izon al scale is smalle a 50
high la i udes, con inen al shel es, and a eas o weak s a i ica ion (Hallbe g, 2013). Models wi h ocean esolu ions o a
leas 1/10º a e known as eddy- esol ing models and a e capable o esol ing eddies in ex ensi e a eas o he No h A lan ic –
wi h limi a ions in egions o weak s a i ica ion o shallow ba hyme y. Eddy-pe mi ing models, ins ead, ha e esolu ions in
he o de o 1/4º and a e only able o esol e eddies in he opics. In models wi h ocean esolu ions o 1º o coa se , he
con ibu ion o eddies is pa ame e ized. 55
Associa ed wi h a be e cha ac e iza ion o he ocean mesoscale, inc easing he ocean esolu ion o eddy-pe mi ing
scales has been shown o imp o e he ep esen a ion o bounda y and on al cu en s, such as he Gul S eam (GS) and
No h A lan ic Cu en (NAC), bo h in e ms o loca ion and s uc u e, wi h signi ican u he imp o emen a eddy-
esol ing scales (Hewi e al., 2017; Ma zocchi e al., 2015). This be e cha ac e iza ion o he GS and NAC leads o
educed sea su ace empe a u e (SST) and salini y (SSS) biases no h o Cape Ha e as (NCH) and in he Cen al No h 60
A lan ic (CNA) a eddy- esol ing scales (Ma zocchi e al., 2015). These educed su ace empe a u e biases a e e lec ed in
he a mosphe e mean s a e as well: he win e s o m ack bias gene ally p esen a high la i udes in eddy-pa ame e ized
models is educed a eddy- esol ing scales, associa ed o a weake me idional empe a u e g adien in he No h A lan ic
h ps://doi.o g/10.5194/egusphe e-2025-547
P ep in . Discussion s a ed: 21 Feb ua y 2025
c
Au ho (s) 2025. CC BY 4.0 License.
3
(Mo eno-Chama o e al., 2024); and also he local nega i e bias in p ecipi a ion associa ed o he cold CNA bias is educed
a high esolu ion (Mo eno-Chama o e al., 2022). 65
Inc easing ocean esolu ion o (a leas ) eddy-pe mi ing scales has also been shown o imp o e ai -sea in e ac ions in he
No h A lan ic. Mo e speci ically, eddy-pe mi ing models, when un oge he wi h an a mosphe ic componen o equi alen
esolu ion, exhibi mo e ealis ic nea su ace wind s ess di e gence and cu l ields o e he GS and NAC compa ed o
eddy-pa ame e ized models (Tsa sali e al., 2022). Simila ly, he ep esen a ion o he co a iance be ween SST and hea
luxes is imp o ed in ha egion a eddy-pe mi ing scales (Bellucci e al., 2021). 70
This s udy has i s ocus on he No h A lan ic, and aims a assessing he impac o explici ly esol ing mesoscale ocean
eddies in he ep esen a ion o i s mean s a e in his o ical simula ions, by compa ing an ensemble o ou Coupled Model
In e compa ison P ojec phase 6 (CMIP6) HighResMIP (Sec . 2.1; Haa sma e al., 2016) coupled eddy- esol ing models –
namely CESM1-CAM5-SE-HR (Chang e al., 2020), EC-Ea h3P-VHR (Mo eno-Chama o e al., 2024), HadGEM3-GC31-
HH (Robe s e al., 2019), and MPI-ESM1-2-ER (Gu jah e al., 2019) – wi h a second ensemble o 39 CMIP6 coupled non-75
eddy- esol ing models. This wo k ocuses on desc ibing he dynamics o he No h A lan ic, as well as he p ope ies ha
impac hem, such as he biases in empe a u e and salini y, he s a i ica ion, and he deep ocean con ec ion.
To ou knowledge, only a ew mul imodel compa isons o coupled his o ical expe imen s wi h a ocus on he No h
A lan ic ocean exis , ha include eddy- esol ing simula ions (e.g. Robe s e al., 2020; Koenigk e al., 2021), al hough none
o hem speci ically add esses he impac o esol ing mesoscale ocean eddies. In ha con ex , ou s udy s ands ou o i s 80
pa icula ocus on he added alue o hese eddies, ea u ing he la ges ensemble o coupled eddy- esol ing simula ions
conside ed so a . This ensemble allows us o e alua e mo e consis en ly which aspec s o he mean clima e a e imp o ed a
ha esolu ion.
This manusc ip is s uc u ed as ollows: he da a and me hodological app oach employed a e desc ibed in Sec . 2. The
main esul s o he s udy a e p esen ed in Sec . 3, including a cha ac e iza ion o SST and SSS biases in he No h A lan ic 85
o he high and low esolu ion ensembles (Sec . 3.1), he s a i ica ion (Sec . 3.2) and mixing in he egions o deep wa e
o ma ion (Sec 3.3), he AMOC (Sec 3.4), and he gy e ci cula ions, including he NAC and he SPG (Sec . 3.5). Finally, in
Sec . 4 we del e deep in o he discussion o he main esul s, ela e hem o he cu en li e a u e, and p esen ou
conclusions.
2 Da a and me hods 90
2.1 Model da a and me hodological app oach
In o de o assess he impac o inc eased ho izon al esolu ion on he ep esen a ion o he No h A lan ic mean s a e, we
analyze he ou pu s om ou CMIP6-endo sed HighResMIP (Haa sma e al., 2016) coupled eddy- esol ing his o ical
simula ions (his -1950; HR-HIST he eina e ) – co esponding o he models CESM1-CAM5-SE-HR (Chang e al., 2020),
h ps://doi.o g/10.5194/egusphe e-2025-547
P ep in . Discussion s a ed: 21 Feb ua y 2025
c
Au ho (s) 2025. CC BY 4.0 License.
4
EC-Ea h3P-VHR (Mo eno-Chama o e al., 2024), HadGEM3-GC31-HH (Robe s e al., 2019), and MPI-ESM1-2-ER 95
(Gu jah e al., 2019) – and compa e hem o a baseline ensemble o 39 CMIP6 coupled his o ical uns (Ey ing e al., 2016)
pe o med a coa se esolu ion (LR-HIST he eina e ). We eckon his s anda d esolu ion ensemble as a mo e igo ous
benchma k han he low- esolu ion HighResMIP coun e pa s o he ou eddy- esol ing models, gi en i s much la ge size.
Mo e de ails on he models conside ed a e p o ided in Tables 1 and B1.
100
HR-HIST models
ocean componen
ocean g id
a m. componen
a m. g id
e e ence
CESM1-CAM5-SE-HR
POP2
1/10°; ipola ;
3600x2400 lon/la ; 62 le els;
CAM5.2
25 km; 30 le els;
Chang e al. (2020)
EC-Ea h3P-VHR
NEMO3.6
1/12°; ORCA12 ipola ;
4322 x 3059 lon/la ; 75 le els;
IFS cy36 4
16 km; 91 le els;
Mo eno-Chama o e al.
(2024)
HadGEM3-GC31-HH
NEMO-HadGEM3-
GO6.0
1/12°; eORCA12 ipola ;
4320 x 3604 lon/la ; 75 le els;
Me UM-HadGEM3-
GA7.1
50 km; 85 le els;
Robe s e al. (2019)
MPI-ESM1-2-ER
MPIOM
1/10°; TP6M ipola ;
3602 x 2394 lon/la ; 40 le els;
ECHAM6.3
103 km; 95 le els;
Gu jah e al. (2019)
LR-HIST models
–
25–250 km
(mainly 100 km)
–
100–500 km
(mainly 100–250 km)
–
Table 1: O e iew o models used in he cu en s udy. Ocean g id de ails include: nominal esolu ion; g id ype; size o ho izon al g id;
and numbe o e ical le els. De ails abou he indi idual LR-HIST models can be ound in Table B1 in Appendix B.
All models in he HR-HIST ensemble ha e a nominal ocean esolu ion o a leas 1/10º, allowing hem o ep esen he 105
ocean mesoscale in ex ensi e a eas o he No h A lan ic. By con as , in LR-HIST, he ocean mesoscale is, a bes , only
esol ed in he opics and mo e gene ally pa ame e ized, since ocean esolu ion in ha ensemble anges om 25 o 250 km,
being 100 km he mos common esolu ion ac oss models. A mosphe ic esolu ion is also gene ally highe in he HR-HIST
ensemble compa ed o LR-HIST, anging om 15 o 100 km o HR-HIST, and om 100 o 500 km o LR-HIST (Tables 1
and B1). We no e he he e ogenei y in model componen s employed ac oss he ensembles, which migh a oid a dominan 110
con ibu ion o speci ic indi idual model biases wi hin he ensembles.
Model selec ion c i e ia is based on he a ailabili y o h ee-dimensional empe a u e and salini y, and he o e u ning
mass s eam unc ion (ei he ms mz o ms yz) o he A lan ic Ocean as ou pu a iables. Al hough in mos ocean g ids he
y-g id di ec ion migh di e om he me idional di ec ion a high no he n la i udes, in gene al he o e u ning mass
s eam unc ion calcula ed along lines o cons an y (ms yz) p o ides a good app oxima ion o ha calcula ed along lines o 115
h ps://doi.o g/10.5194/egusphe e-2025-547
P ep in . Discussion s a ed: 21 Feb ua y 2025
c
Au ho (s) 2025. CC BY 4.0 License.
5
cons an la i ude (ms mz) (G i ies e al., 2016). In ou s udy, we selec ms mz o e ms yz when a ailable. The models o
which only ms yz is a ailable as ou pu a iable (ma ked wi h an * in Fig. 9) p esen signi ican g id o a ion only om ≥
40º N no hwa ds (excep o MPI-ESM1-2-ER, which p esen s o a ion al eady a 30º N). Fo cau ion, we es ic ou
analysis o la i udes below 40º N, which ne e heless allows us o ex ac aluable in o ma ion.
Fo cing ields in he HighResMIP coupled his o ical simula ions (HR-HIST ensemble) a e almos iden ical o hose in he 120
CMIP6 his o ical simula ions (LR-HIST ensemble). The only signi ican di e ence conce ns land use, which is ixed in ime
in HighResMIP and ep esen a i e o he p esen -day pe iod (a ound yea 2000) (Haa sma e al., 2016), and ime- a ying in
CMIP6 his o ical simula ions (Ey ing e al., 2016). We do no expec his o cause impo an di e ences in he ocean
a iables conside ed in ou analysis. I is also wo h no ing ha HighResMIP his o ical simula ions a e un wi hou
in e ac i e ae osols, bu his is also he case o se e al CMIP6 his o ical simula ions (e.g. CMCC-CM2-HR4, EC-Ea h3, 125
FGOALS- 3-L, GISS-E2-2-G, IPSL-CM6A-LR, MPI-ESM1-2-HR).
Mo e signi ican di e ences conce n model ini ializa ion and spin-up. Due o he high compu a ional cos s o high-
esolu ion modelling, in HighResMIP, ini ial condi ions o his o ical uns a e aken om a sho spin-up (~30–50 yea s)
wi h ixed 1950’s adia i e o cings and ocean ini ial condi ions (Haa sma e al., 2016), ins ead o om a long p e-indus ial
con ol ep esen a i e o 1850 condi ions. Thus, in he case o he HighResMIP expe imen s, he subs an ially sho e spin-130
up and his o ical pe iod co e ed (1950–2014) can lea e some linge ing d i s. Ne e heless, in some eddy- esol ing
HighResMIP simula ions, he ocean seems o equilib a e as e (Mo eno-Chama o e al., 2024; Robe s e al., 2019)
compa ed o hei lowe esolu ion coun e pa s.
Da a analyses a e ca ied ou using he Ea h Sys em Model E alua ion Tool (ESMValTool 2.10.0), a Py hon package
designed o model in e compa ison pu poses (Andela e al., 2023a, b; Righi e al., 2020). We no e ha he CMIP6 model 135
ICON-ESM-LR was excluded om some o he analyses due o an incompa ibili y o he da a wi h ESMValTool.
Clima ologies a e compu ed o he las 35 yea s o he his o ical uns (1980–2014) o educe he po en ial e ec o model
d i s, which a e expec ed o be la ge in he ea lie yea s o he HighResMIP simula ions.
2.2 Speci ic diagnos ics
Po en ial densi y anomalies wi h espec o a e e ence p essu e o 0 dba (σ0) a e calcula ed om empe a u e and 140
salini y mon hly means, using he polynomial app oxima ion o he TEOS-10 equa ion o s a e o Boussinesq models
(Roque e al., 2015). Mixed laye dep h (MLD) is de ined and calcula ed as he shallowes dep h le el a which mon hly
po en ial densi y σ0 exceeds by a h eshold o 0.03 kg m-3 i s alue a a e e ence dep h o 10 m, as desc ibed in de Boye
Mon égu e al. (2004). This me hod is p e e ed o e employing di ec MLD model ou pu s ha use ins an aneous alues
and a ange o di e en de ini ions, o ensu e a consis en compa ison ac oss models and obse a ions. 145
h ps://doi.o g/10.5194/egusphe e-2025-547
P ep in . Discussion s a ed: 21 Feb ua y 2025
c
Au ho (s) 2025. CC BY 4.0 License.
6
2.3 Obse a ional e e ences
Obse a ional and eanalysis da a a e employed o e alua e model pe o mance. Fo empe a u e and salini y, he Me
O ice Hadley Cen e EN.4.2.2 da ase (Good e al., 2013) wi h he Gou e ski and Reseghe i (2010) expendable
ba hy he mog aph and Gou e ski and Cheng (2020) mechanical ba hy he mog aph co ec ions is used, which has a
esolu ion o 1º. We op o his h ee-dimensional da ase o join ly assess biases a he su ace and dep h. We also use i o 150
de i e an obse a ional e e ence o he MLD ha is physically consis en wi h EN4 salini y and empe a u e ields.
Fo he A lan ic o e u ning mass s eam unc ion and ba o opic s eam unc ion, ORAS5m eanalysis da a (Tie sche e
al., 2020) a e used as ou e e ence (1/4° esolu ion; pe iod 1980–2014). ORAS5m is an imp o ed e sion o he 5 h
ECMWF ocean eanalysis sys em ORAS5 (Zuo e al., 2019), wi h educed SST nudging and inc eased weigh o coas al
obse a ions. This e sion imp o es he ep esen a ion o he AMOC and leads o educed biases in win e e o ecas s o he 155
No h A lan ic.
As a complemen a y e e ence o di ec obse a ional da a, he clima ological e ical p o ile o he RAPID a ay AMOC
s eam unc ion is employed o alida e he simula ed AMOC a 26.5° N. RAPID is a moni o ing p og amme p o iding ime
se ies o AMOC based on empe a u e, salini y and p essu e p o iles om a moo ing a ay c ossing he A lan ic om wes o
eas a 26.5º N (Johns e al., 2023; Moa e al., 2023). The clima ology employed co esponds o he pe iod Ap il 2004 – 160
Feb ua y 2022.
Mon hly a e aged absolu e dynamic opog aphy da a (sea su ace heigh abo e geoid) a e also employed om AVISO
obse a ions a 1/4° esolu ion o he pe iod Feb ua y 1993 – Decembe 2014.
3 Resul s
3.1 Sea su ace biases 165
Tempe a u e and salini y biases, h ough hei impac on he zonal and e ical densi y g adien s, a e impo an o he
ealism o he ocean ci cula ion and deep wa e o ma ion in he No h A lan ic. The mean SST biases o he indi idual LR-
HIST and HR-HIST models a e shown in Fig. 1, and hei espec i e mul i-model means in Fig. 2. In gene al, he HR-HIST
ensemble mean displays wa me su ace wa e s in he SPNA, compa ed o he LR-HIST one. The LR-HIST ensemble shows
wo main SST biases o opposi e sign and simila magni ude. The i s is a wa m bias loca ed along he No h Ame ican 170
coas , a NCH, wi h empe a u es 2–5º C wa me han obse a ions (Fig. 2). This bias has p e iously been associa ed wi h a
mis ep esen a ion o he posi ion o he GS sepa a ion om he coas (Ma zocchi e al., 2015). The o he is a cold bias in he
CNA (2–5º C), which ea lie s udies ha e linked o an o e ly weak NAC and an unde es ima ion o he ho izon al hea
anspo in o he CNA domain (Lin e al., in e iew). The NCH bias has been shown o ha e an impo an impac on he
global a mosphe ic ci cula ion, h ough a Rossby wa e esponse o local changes in e ical mo ion in he oposphe e (Lee 175
h ps://doi.o g/10.5194/egusphe e-2025-547
P ep in . Discussion s a ed: 21 Feb ua y 2025
c
Au ho (s) 2025. CC BY 4.0 License.
7
Figu e 1. Sea su ace empe a u e (SST) bias (shading; in ºC) o he indi idual LR-HIST ( ows 1–8) and HR-HIST (las ow) models
wi h espec o EN4 o he pe iod 1980–2014. EN4 clima ology shown in con ou lines (in ºC). Values in pa en hesis in each sub igu e
heade show he spa ially a e aged absolu e mean bias o each indi idual model.
h ps://doi.o g/10.5194/egusphe e-2025-547
P ep in . Discussion s a ed: 21 Feb ua y 2025
c
Au ho (s) 2025. CC BY 4.0 License.
8
180
e al., 2018); he CNA bias has an e ec on local p ecipi a ion (Mo eno-Chama o e al., 2022). In he HR-HIST mean, he
NCH and CNA biases a e signi ican ly educed compa ed o LR-HIST. By con as , he HR-HIST mean shows a signi ican
posi i e bias o 1–3º C in he Lab ado Sea (LS), which is weake in he LR-HIST mean (Fig. 2). We no e, howe e , ha he
ange o biases is la ge in he LR-HIST ensemble compa ed o he HR-HIST one, wi h some LR-HIST models showing
la ge posi i e biases han he HR-HIST ones (Fig. 1; e.g. CAS-ESM2-0 and CESM2-WACCM-FV2), while o he LR-HIST 185
models p esen subs an ial biases o opposi e sign (e.g. CanESM5-1 and E3SM-1-1).
Figu e 2. SST bias o he mul i-model mean o he (a) LR-HIST and (b) HR-HIST ensembles. Plo ing de ails as in Fig. 1. Colou ed 190
polygons delinea e he main bias egions add essed in he pape : no h o Cape Ha e as (NCH) in ed [edges: (78º W, 34º N), (61º W, 41º
N), (61º W, 46º N), (71º W, 44º N)], Cen al No h A lan ic (CNA) in g een (30º–45º W, 42º–52º N), and Lab ado Sea (LS) in blue (44º–
60º W, 52.5º–65º N).
Analogously o he SSTs, SSS biases om he indi idual models and he co esponding ensemble means a e desc ibed in 195
Figs. 3 and 4. The mul i-model mean SSS biases show a simila pa e n o he empe a u e ones (Fig. 4). LR-HIST p esen s a
posi i e salini y bias o 1–3 a NCH, and a nega i e bias o 0.5–1.5 in he CNA. No e ha salini ies a e p esen ed on he
p ac ical salini y scale h oughou he manusc ip , wi h no associa ed uni s. In con as o he SST biases, he SSS CNA
nega i e bias is no a common ea u e in all LR models, al hough i is indeed dominan ac oss hem (Fig. 3). Fo HR-HIST,
he NCH and CNA biases a e signi ican ly educed wi h espec o LR-HIST, al hough a posi i e bias o 0.5–1 appea s in he 200
LS ha is no p esen in he LR-HIST ensemble mean, p obably because biases o di e en models compensa e wi h each
o he . We no e also ha in he LS, models end o show SST and SSS biases o he same sign, wi h CIESM, GISS-E2-2-G,
INM-CM4-8, and INM-CM5-0 as excep ions. This migh lead o a compensa ing con ibu ion o he su ace densi y biases.
Despi e he appa en LS deg ada ion o HR-HIST, he spa ially-a e aged absolu e SSS biases in he No h A lan ic a e
h ps://doi.o g/10.5194/egusphe e-2025-547
P ep in . Discussion s a ed: 21 Feb ua y 2025
c
Au ho (s) 2025. CC BY 4.0 License.
9
205
Figu e 3. As in Fig. 1 bu o he sea su ace salini y (SSS) biases.
h ps://doi.o g/10.5194/egusphe e-2025-547
P ep in . Discussion s a ed: 21 Feb ua y 2025
c
Au ho (s) 2025. CC BY 4.0 License.
16
300
Figu e 10. AMOC (in S ) by g oups, o (a) LR-HIST, (b) HR-HIST and (c) ORAS5m eanalysis. The black e ical line indica es he
la i ude om which g id o a ion could ha e an e ec (Sec 2.1).
p o ile shows a pa icula ly good i wi h he RAPID a ay one abo e ~1000 m, al hough, in gene al, he AMOC is oo
shallow bo h o LR-HIST and HR-HIST. This is in pa due o di e ences in he me hodological app oach (Danabasoglu e 305
al., 2021). HR-HIST models emain ela i ely close o he RAPID da a, and he main ou lie s bo h in e ms o unde - and
o e es ima ion o he AMOC a e LR-HIST models (Fig. 11a).
To pe o m a quan i a i e compa ison be ween he wo ensembles, we ex end he analysis o AMOC p o iles by
compu ing wo me ics ha measu e he deg ee o ag eemen o he di e en models wi h obse a ions, as diagnosed by he
Pea son co ela ion (x-axis in Fig. 11b) and he RMSE (y-axis in Fig. 11b) ac oss he e ical dimension. Figu e 11b 310
con i ms ha , al hough none o he HR-HIST models is sys ema ically be e han all he LR-HIST ones, HR-HIST models
lie wi hin he ange o bes pe o ming models, bo h in e ms o e ical co ela ion and RMSE agains RAPID, wi h HR-
HIST models concen a ed close o he bo om- igh co ne o he igu e. To complemen his analysis o he impac o
esolu ion on he o e u ning ci cula ion, nex sec ion looks a he impac on he gy e ci cula ions in he No h A lan ic.
3.5 Gy e ci cula ions 315
In his sec ion, he main gy e ci cula ions o he No h A lan ic a e examined, as desc ibed by he ba o opic
s eam unc ion (BSF), a measu e o he e ically in eg a ed olume anspo . The gy e ci cula ions play a key ole in
clima e in e ms o no hwa d ocean hea and eshwa e anspo and deep wa e o ma ion. In o de o alida e he posi ion
o he GS and NAC in models, we also plo he ze o con ou line o absolu e dynamic opog aphy om AVISO obse a ions,
which delimi s he in e gy e bounda y (dashed lines in Figs. 12 and 13). Ideally, his ze o line in obse a ions would o e lap 320
he ze o line o he model BSF, as o ORAS5m in Fig. 13c.
h ps://doi.o g/10.5194/egusphe e-2025-547
P ep in . Discussion s a ed: 21 Feb ua y 2025
c
Au ho (s) 2025. CC BY 4.0 License.
17
Figu e 11. (a) Clima ological AMOC p o ile a 26.5º N (in S ). Model and eanalysis (ORAS5m) da a co espond o he in e al 1980–
2014 (mon hly da a). RAPID obse a ions a e a e aged o e he pe iod Ap il 2004 – Feb ua y 2022. (b) Pea son co ela ion coe icien
(ho izon al axis) and Roo Mean Squa e E o (RMSE; in S ) ( e ical axis) o AMOC p o iles a 26.5º N agains RAPID, bo h es ima ed 325
ac oss he e ical dimension.
In he mul i-model mean o LR-HIST, he GS sepa a es oo a no h om he Ame ican coas compa ed o AVISO,
which implies ha i s NCH bias egion (Fig. 13a, ed polygon) is only in luenced by wa me , mo e saline wa e s o sou he n
o igin, in con as wi h ORAS5m, whe e sligh en ainmen o colde , eshe wa e s om he no h occu s. This can 330
he e o e explain he posi i e empe a u e and salini y biases desc ibed in Sec . 3.1.
In he mul i-model mean o HR-HIST, he GS sepa a es om he coas u he sou h compa ed o LR-HIST, bu also
sligh ly compa ed o AVISO, and he NCH egion is pa ially in luenced by wa e s o no he n o igin, as in ORAS5m (Fig.
13b, c). These esul s can explain why he HR-HIST models show compa ably educed SST and SSS biases wi h espec o
LR-HIST (Figs. 2 and 4). Fu he mo e, HR-HIST displays an imp o ed GS s uc u e (e.g. in he Flo ida Cu en ) wi h a 335
h ps://doi.o g/10.5194/egusphe e-2025-547
P ep in . Discussion s a ed: 21 Feb ua y 2025
c
Au ho (s) 2025. CC BY 4.0 License.
18
Figu e 12. Ba o opic s eam unc ion (BSF; in S ) o LR-HIST ( ows 1–4) and HR-HIST models and eanalysis (ORAS5m) (las ow).
Ze o con ou o absolu e dynamic opog aphy om AVISO obse a ions (dashed black line) co esponding o he pe iod 1993–2014 is
also shown. Boxes as in Figs. 2 and 4: NCH in ed, CNA in g een, LS in blue. No e: some models did no p esen a BSF ou pu (e.g. 340
CESM1-CAM5-SE-HR) o his p esen ed un ealis ic alues, hence he eason ewe models a e shown in his igu e.
na owe and locally s onge cu en han LR-HIST (Fig. 13b), in close ag eemen wi h ORAS5m, al hough his could be
pa ly explained by he ac ha ORAS5m was p oduced wi h an eddy-pe mi ing ocean.
The NAC is oo zonal in mos LR-HIST models (Fig. 12). In he LR-HIST ensemble mean, he CNA egion is only 345
ouched a i s sou he n edge by he wa me /sal ie NAC wa e s, emaining p edominan ly exposed o he in luence o he
SPG (Fig. 13a). By con as o ORAS5m and HR-HIST, he SPG has a mo e es ained in luence on ha egion in a o o a
less zonal NAC (Fig. 13b, c), which could explain he educed biases a high esolu ion in he CNA o HR-HIST (Sec .
3.1.). This is a common ea u e in all a ailable indi idual HR-HIST models o which he BSF is a ailable: EC-Ea h3P-
VHR, HadGEM3-GC31-HH, and MPI-ESM1-2-ER. 350
The SPG is gene ally s onge in HR-HIST compa ed o LR-HIST, and close in s eng h and s uc u e o ORAS5m
(Figs. 12 and 13), al hough some indi idual LR-HIST models like No ESM2-LM and SAM0-UNICON ha e gy es o
h ps://doi.o g/10.5194/egusphe e-2025-547
P ep in . Discussion s a ed: 21 Feb ua y 2025
c
Au ho (s) 2025. CC BY 4.0 License.
19
compa able in ensi y o HadGEM3-GC31-HH, which has he s onges SPG ac oss HR-HIST models. We no e ha a
s onge SPG is consis en wi h he s onge con ec ion iden i ied in Sec . 3.3, as his la e enhances he p esence o dense
wa e s in he deep ocean, subsequen ly s eng hening he ba oclinic p essu e g adien ha d i es he gy e (Yashayae and 355
Lode , 2009).
Figu e 13. BSF (in S ) by g oups o (a) LR-HIST, (b) HR-HIST and (c) eanalysis (ORAS5m). Plo ing de ails as in Fig. 12.
4 Main conclusions and discussion 360
In his s udy, we analyse he impac o ho izon al esolu ion on he ep esen a ion o he No h A lan ic mean s a e, by
compa ing wo ensembles o coupled his o ical simula ions: ou HighResMIP expe imen s a eddy- esol ing scales (HR-
HIST ensemble; a leas 1/10º nominal esolu ion) and 39 CMIP6 expe imen s wi h eddy-pa ame e ized o eddy-pe mi ing
ocean esolu ions (LR-HIST ensemble).
The main biases o key he modynamic and dynamical a iables o he No h A lan ic a e analysed o he wo 365
ensembles. In pa icula we examine i) he main su ace empe a u e and salini y biases; ii) s a i ica ion and iii) deep wa e
con ec ion in he No h A lan ic; i ) he ep esen a ion o he AMOC; and ) he gy e ci cula ions o he No h
A lan ic, including he GS, NAC, and he SPG. In he ollowing, he main indings o he pape a e desc ibed and hei
implica ions discussed in ligh o he p e ious li e a u e.
Th ee main SST and SSS bias egions a e ound in he simula ions, loca ed a NCH, he CNA, and he LS, which show 370
signi ican di e ences ac oss he wo ensembles. In he NCH egion, we ind educed posi i e empe a u e and salini y
su ace biases o he mul i-model HR-HIST mean wi h espec o LR-HIST, associa ed wi h a mo e sou hwa d posi ion o
he GS sepa a ion, in ag eemen wi h p e ious indi idual model s udies (Robe s e al., 2019; Gu jah e al., 2019; Ma zocchi
e al., 2015).
Then, he CNA cold and esh biases in he mul i-model LR-HIST mean a e also educed in HR-HIST – as in Gu jah e 375
al. (2019) and Ma zocchi e al. (2015) – which desc ibes a less zonal NAC and a mo e es ic ed in luence o SPG wa e s in
h ps://doi.o g/10.5194/egusphe e-2025-547
P ep in . Discussion s a ed: 21 Feb ua y 2025
c
Au ho (s) 2025. CC BY 4.0 License.
20
ha egion, in close ag eemen wi h obse a ions and eanalysis. Sein e al. (2018) ound simila esul s when compa ing
HighResMIP coupled simula ions ob ained wi h he AWI-CM model, showing ha ul ima ely i was an inc ease in ocean
esolu ion ha shi ed he NAC pa h owa ds he no h, wi h no signi ican in luence o a mosphe ic esolu ion. Su ace
biases in he NCH and CNA egions migh be connec ed wi h each o he due o hei ul ima e link wi h he GS/NAC 380
dynamics. As expec ed, he co ela ion be ween he su ace biases o hese wo egions inc eases when he HR-HIST models
a e excluded om he compu a ion (Fig. A1), which e lec s ha he NCH egion is ou side he GS domain in HR-HIST
models.
The LS egion s ands ou o a wa m and sal y bias in he HR-HIST mul i-model mean. Fo empe a u e, i is also p esen
in he LR-HIST ensemble mean, al hough much weake , whe eas o salini y, i does no show in he LR-HIST ensemble 385
mean, e en i i is p esen in some o he indi idual models. Indeed, some indi idual LR-HIST models show LS biases o
compa able s eng h o he HR-HIST ones, bu hei signal is compensa ed in he mul i-model mean by models wi h biases o
opposi e sign. The o igin and dynamical impac s o he LS biases a e a cu en ma e o deba e (Mena y e al., 2015; Robe s
e al., 2020). These su ace biases migh ha e an e ec on LS deep wa e con ec ion h ough hei decisi e in luence on
e ical s a i ica ion and, he e o e, co ec ing hem migh help ob aining mo e ealis ic p esen -day AMOC es ima es and 390
eliable u u e p ojec ions. Fu he e o s should hus aim a iden i ying he sou ce/o igin o hese biases. Ou s udy hin s
ha CNA and LS biases migh be ac ually ela ed. Sca e plo s o SSS biases be ween bo h egions (Fig. A2) indica e a
s ong co ela ion be ween hem ( = 0.86, p < 0.001), which sugges s a po en ial link be ween LS salini y biases and he
NAC, h ough he e ec o he NAC on he no hwa d salini y anspo . We no e, howe e , ha he LS and CNA a e also
connec ed h ough he SPG ci cula ion, which could also pa ly explain why hei SSS biases a e ela ed. The co ela ion 395
be ween he SST biases o he LS and CNA egions is also signi ican , al hough weake compa ed o he SSS biases ( =
0.54, p < 0.001; Fig. A2), which migh indica e addi ional di e ences be ween he mechanisms exe ing con ol o e he
SSTs in bo h egions, like he local a mosphe ic o cing. S udies such as Chang e al. (2020) and Robe s e al. (2019) epo
inc eased hea anspo by he AMOC in eddy- esol ing models, u he suppo ing he idea o inc eased no hwa d
anspo as a po en ial o igin o he LS biases. 400
In e es ingly, we ind imp o ed e ical p o iles o empe a u e and salini y in he LIS box o HR-HIST: despi e he
la ge biases ound a he su ace wi h espec o LR-HIST, he subsu ace is colde and eshe in HR-HIST. This migh be
ela ed o inc eased e ical (upwa ds) hea and sal anspo s by ocean mesoscale eddies (Hewi e al., 2017), p o iding a
po en ial explana ion o he wa m and sal y su ace biases in he LS. The mean densi y p o ile, which is he key p ope y
con olling he e ical mixing, is also imp o ed o he LIS in HR-HIST wi h espec o LR-HIST, showing compa a i ely 405
educed s a i ica ion. This explains why HR-HIST p esen s deepe mixed laye s in he LS and along he eas G eenland
coas han LR-HIST, in close ag eemen wi h EN4-de i ed alues, an imp o emen ha is also ound in he No dic Seas.
Since EN4 migh p esen some unce ain ies in he ocean subsu ace, in pa icula be o e he A go pe iod (1999 o p esen ), a
compa ison wi h o he mo e ecen MLD es ima es is wa an ed o assess he consis ency o ou esul s. A MLD clima ology
o he 2000–2016 pe iod based on A go densi y p o iles by Hol e e al. (2017) shows mixed laye s down o a maximum o 410
h ps://doi.o g/10.5194/egusphe e-2025-547
P ep in . Discussion s a ed: 21 Feb ua y 2025
c
Au ho (s) 2025. CC BY 4.0 License.
21
1800 m in he LS. Time- a ying es ima es o win e maximum MLDs in he LS ob ained om A go loa s, he AR7W line,
and moo ed measu emen s, sugges alues a ound 1200–1700 m in he 2002–2015 in e al (Yashayae and Lode , 2016),
showing an in ensi ica ion in ecen yea s, wi h a eco d alue o 2100 m in 2016 (Yashayae and Lode , 2017). O e all, he
maximum alues in he HR-HIST ensemble mean o he LS (1800–2000 m) a e sligh ly la ge han he di ec obse a ional
es ima es, excluding he eco d alues in Yashayae and Lode (2017). Howe e , i is impo an o no e ha some 415
di e ences a e expec ed as ou HR-HIST alues a e compu ed om 1980 o 2014, and use gene ally smoo he p o iles
associa ed wi h he coa se empo al esolu ion o he model da a compa ed o he indi idual p o iles om obse a ional
s udies.
The AMOC is weake in he mul i-model HR-HIST mean compa ed o LR-HIST and i s s eng h and s uc u e exhibi a
be e i wi h RAPID obse a ions and he ocean eanalysis ORAS5m. Addi ionally, he HR-HIST AMOC p esen s sha pe 420
ea u es, as shown e.g. in Sein e al. (2018), be e esembling eanalysis da a. Ne e heless, as poin ed ou in p e ious
s udies (Robe s e al., 2020; Hi schi e al., 2020), he AMOC emains oo shallow compa ed o RAPID bo h in he LR-HIST
and HR-HIST ensemble means.
The ole o esolu ion in AMOC s eng h is a cu en ma e o deba e in he li e a u e, wi h di e en indi idual model
s udies poin ing a di e en esul s. Win on e al. (2014) epo AMOC s eng hening wi h inc eased ocean esolu ion o he 425
GFDL CM2.6 and CM2.5FLOR models a 0.1º and 1º esolu ion, espec i ely. They also ind AMOC s eng h is sensi i e o
ho izon al ic ion and mesoscale eddy pa ame e iza ions. Hewi e al. (2016) show s eng hening in he mean AMOC a a
concomi an inc ease in ocean ( om 1/4º o 1/12º) and a mosphe ic esolu ion ( om 60 o 25 km) in he GC2.1 model.
Simila esul s a e ound by Mo eno-Chama o e al. (2024) o he EC-Ea h3P model when inc easing ocean and
a mosphe ic esolu ion om 0.25º o 0.08º and om ~54 km o ~12 km, espec i ely. On he o he hand, a s udy assessing 430
he sepa a e e ec s o enhanced a mosphe ic and ocean esolu ion on AMOC beha iou wi h he AWI-CM model, desc ibes
a weakening a inc eased a mosphe ic esolu ion ( om 1.9º o 0.9º) associa ed wi h educed winds, bu bo h a weakening a
~45º N and a s eng hening a ~20º N ela ed o ocean g id e inemen ( om 1º o 1/4° nominal esolu ions, and he la e wi h
g id e inemen s in eddy- ich a eas; Sein e al., 2018). Fu he mo e, Gu jah e al. (2019) show li le di e ence in AMOC
s eng h be ween MPI-ESM1-2-HR and MPI-ESM1-2-ER, which use he same a mosphe e and e ical mixing 435
pa ame e iza ion ye di e en ocean esolu ion (0.4º s 0.1º, espec i ely). Tha s udy also shows AMOC can be e y
sensi i e o he e ical mixing scheme.
Mul imodel s udies on his opic ha e also been conduc ed (Robe s e al., 2020; Hi schi e al., 2020). Hi schi e al.
(2020) analyze 28 models (22 ocean-only and six coupled models) wi h ocean esolu ions anging om 2º o 0.05º and ind
inc eased AMOC s eng h a eddy- esol ing scales ( hei Fig. 2). Robe s e al. (2020) compa e he AMOC in HighResMIP 440
simula ions wi h se en di e en coupled models, no inding a consis en e ec o enhancing ocean and/o a mosphe ic
esolu ion on he AMOC s eng h in he dep h space. This also applies o he wo simula ions a eddy- esol ing scales in
ha s udy, pe o med wi h HadGEM3-GC31 and CESM1.3 (ou CESM1-CAM5-SE), he i s showing a s onge AMOC in
h ps://doi.o g/10.5194/egusphe e-2025-547
P ep in . Discussion s a ed: 21 Feb ua y 2025
c
Au ho (s) 2025. CC BY 4.0 License.
22
dep h-space han i s low esolu ion coun e pa , and he second showing a weake AMOC ins ead. In e es ingly, in ha s udy
esul s con e ge owa ds a s onge AMOC in he eddy- esol ing simula ions when densi y coo dina es a e used ins ead. 445
Meanwhile, ou esul s show a weake AMOC a eddy- esol ing scales. Ne e heless, we no e ha ou AMOC p o iles
o eddy- esol ing models a 26.5º N display simila alues o hose in Robe s e al. (2020) and Hi schi e al. (2020). The
di e ences in he esul s lie a he in he cha ac e is ics o he low esolu ion model ensembles, which in hose s udies show
conside able lowe AMOC alues compa ed o ou s, p obably in ela ion wi h he high sensi i i y o AMOC o model
schemes and pa ame e iza ions (see e.g. Win on e al., 2014; Gu jah e al., 2019 abo e). Ne e heless, all h ee mul imodel 450
s udies – Hi schi e al. (2020), Robe s e al. (2020), and ou s udy – poin a an imp o ed AMOC mean-s a e ep esen a ion
a enhanced esolu ion.
Al hough a link exis s be ween AMOC s eng h and SPNA densi ies/mixed laye s (O ega e al., 2021; Mena y e al.,
2020; Ma in-Ma inez e al., in e iew), in ou s udy, he deepe mixed laye s in he mul i-model HR-HIST mean wi h
espec o LR-HIST despi e a weake AMOC, e lec a di e en ep esen a ion o deep wa e sinking mechanisms in high 455
esolu ion models, as desc ibed in Ka sman e al. (2018). Tha s udy shows ha deep wa e sinking in eddy-pe mi ing
models occu s only a he con inen al slopes – a he bounda y cu en o he SPG – and no also in he open ocean whe e
MLDs each hei maximum dep hs, as in 1º ocean models. The sinking mechanism desc ibed o eddy-pe mi ing models
can be explained by buoyancy loss along he bounda y cu en pa h, igge ing a c oss-sho e ba oclinic low and subsequen
sinking o ced by mass conse a ion (Ka sman e al., 2018; S aneo, 2006; Spall and Picka , 2001). The mo e ealis ic SPG 460
s uc u e and MLDs nea he con inen al bounda ies in HR-HIST, e.g along he G eenland Cu en , migh hus be key ac o s
o he imp o emen in modeled AMOC s eng h in HR-HIST. Resul s by Ma in-Ma inez e al. (in e iew) show a as e
sou hwa d p opaga ion o he MLD signal in o he AMOC in eddy- esol ing models, indica ing ha ocean esolu ion a ec s
also he imescales o he dynamics o he No h A lan ic. Finally, a s onge SPG is a obus ea u e in HR-HIST wi h
espec o LR-HIST, as al eady poin ed ou by Hi schi e al. (2020) when compa ing a ange o ocean esolu ions om 1º o 465
0.08º, and has been shown o be highly co ela ed wi h a deepe mixed laye (Koenigk e al., 2021).
To summa ize, compa ed o a la ge low- esolu ion CMIP6 ensemble, e ical s a i ica ion in he LS and wes e n IS is
imp o ed in eddy- esol ing models, leading o s onge deep wa e con ec ion close o obse a ions. Su ace empe a u e
and salini y biases a e educed in he NCH and CNA egions associa ed wi h imp o ed GS and NAC pa hs. The AMOC is
weake and close o RAPID and he SPG is s onge and p esen s an imp o ed s uc u e. 470
475
h ps://doi.o g/10.5194/egusphe e-2025-547
P ep in . Discussion s a ed: 21 Feb ua y 2025
c
Au ho (s) 2025. CC BY 4.0 License.
23
Appendix A: Sca e plo s o su ace biases
Figu e A1. Sca e plo s o (a) SST (in ºC) and (b) SSS biases be ween he Cen al No h A lan ic (CNA) and No h Cape Ha e as (NCH) 480
egions de ined in Figs. 1 and 2. Biases a e calcula ed as spa ially a e aged empo al means in he model minus he co esponding EN4
alues in each o he selec ed boxes. The co esponding co ela ion coe icien s and hei p- alues a e shown nex o he i lines. Dashed
lines a e eg ession lines ob ained a e emo al o HR-HIST models. No e: all eg ession lines and alues in (b) a e calcula ed excluding
he wo ou lie s a he bo om le o he igu e.
485
h ps://doi.o g/10.5194/egusphe e-2025-547
P ep in . Discussion s a ed: 21 Feb ua y 2025
c
Au ho (s) 2025. CC BY 4.0 License.
24
Figu e A2. As in Fig. A1 bu be ween he Cen al No h A lan ic (CNA) and he Lab ado Sea (LS).
490
495
500
505
h ps://doi.o g/10.5194/egusphe e-2025-547
P ep in . Discussion s a ed: 21 Feb ua y 2025
c
Au ho (s) 2025. CC BY 4.0 License.
25
Appendix B: Addi ional model de ails
Table B1. O e iew o indi idual models used in he cu en s udy. Ocean g id de ails include: nominal esolu ion; g id ype; size o
ho izon al g id; and numbe o e ical le els. Expanded e sion o Table 1 (pa 1).
ocean componen
ocean g id
a m. componen
a m. g id
HR-HIST
CESM1-CAM5-SE-HR
POP2
1/10°; ipola ;
3600x2400 lon/la ; 62 le els;
CAM5.2
25 km; 30 le els;
EC-Ea h3P-VHR NEMO3.6 1/12°; ORCA12 ipola ;
4322 x 3059 lon/la ; 75 le els;
IFS cy36 4 16 km; 91 le els;
HadGEM3-GC31-HH NEMO-HadGEM3-GO6.0 1/12°; eORCA12 ipola ;
4320 x 3604 lon/la ; 75 le els;
Me UM-HadGEM3-GA7.1 50 km; 85 le els;
MPI-ESM1-2-ER MPIOM 1/10°; TP6M ipola ;
3602 x 2394 lon/la ; 40 le els;
ECHAM6.3 103 km; 95 le els;
LR-HIST
ACCESS-CM2
ACCESS-OM2
100 km; GFDL-MOM5 ipola ;
360 x 300 lon/la ; 50 le els;
Me UM-HadGEM3-GA7.1
250 km; 85 le els;
ACCESS-ESM1-5 ACCESS-OM2 100 km; MOM5 ipola ;
360 x 300 lon/la ; 50 le els;
HadGAM2 250 km; 38 le els;
CAS-ESM2-0 LICOM2.0 100 km;
362 x 196 lon/la ; 30 le els;
IAP AGCM 5.0 100 km; 35 le els;
CESM2 POP2 100 km; gx1 7 displaced pole;
320x384 lon/la ; 60 le els;
CAM6 100 km; 32 le els;
CESM2-FV2 POP2 100 km; gx1 7, displaced pole;
320 x 384 lon/la ; 60 le els;
CAM6 250 km; 32 le els;
CESM2-WACCM POP2 100 km; gx1 7 displaced pole;
320 x 384 lon/la ; 60 le els;
WACCM6 100 km; 70 le els;
CESM2-WACCM-FV2 POP2 100 km; gx1 7 displaced pole;
320 x 384 lon/la ; 60 le els;
WACCM6 250 km; 70 le els;
CIESM CIESM-OM
100 km; mod. POP2 displ. pole;
320 x 384 lon/la ; 60 le els;
CIESM-AM
(modi ied CAM5)
100 km; 30 le els;
CMCC-CM2-HR4 NEMO3.6 25 km; ORCA0.25;
1442 x 1051 lon/la ; 50 le els;
CAM4 100 km; 26 le els;
CMCC-CM2-SR5 NEMO3.6 100 km; ORCA1 ipola ;
362 x 292 lon/la ; 50 le els;
CAM5.3 100 km; 30 le els;
CMCC-ESM2 NEMO3.6 100 km; ORCA1 ipola ;
362 x 292 lon/la ; 50 le els;
CAM5.3 100 km; 30 le els;
CanESM5 NEMO3.4.1 100 km; ORCA1 ipola ;
361 x 290 lon/la ; 45 le els;
CanAM5 500 km; 49 le els;
CanESM5-1 NEMO3.4.1 100 km; ORCA1 ipola ;
361 x 290 lon/la ; 45 le els;
CanAM5.1 500 km; 49 le els;
E3SM-1-1 MPAS-Ocean ( 6.0)
30-60 km; oEC60 o30
uns uc u ed; 60 le els;
EAM ( 1.1) 100 km; 72 le els;
h ps://doi.o g/10.5194/egusphe e-2025-547
P ep in . Discussion s a ed: 21 Feb ua y 2025
c
Au ho (s) 2025. CC BY 4.0 License.
32
Johns, W. E., Elipo , S., Smeed, D. A., Moa , B., King, B., Volko , D. L., and Smi h, R. H.: Towa ds wo decades o 640
A lan ic Ocean mass and hea anspo s a 26.5° N, Philos. T ans. R. Soc. A, 381, 20220188,
h ps://doi.o g/10.1098/ s a.2022.0188, 2023.
Ka sman, C. A., D ij hou , S. S., Dijks a, H. A., and Spall, M. A.: Sinking o Dense No h A lan ic Wa e s in a Global
Ocean Model: Loca ion and Con ols, J. Geophys. Res. Oceans, 123, 3563–3576, h ps://doi.o g/10.1029/2017JC013329,
2018. 645
Khos a i, N., Wang, Q., Kolduno , N., Hin ichs, C., Semmle , T., Danilo , S., and Jung, T.: The A c ic Ocean in CMIP6
Models: Biases and P ojec ed Changes in Tempe a u e and Salini y, Ea h’s Fu u e, 10, e2021EF002282,
h ps://doi.o g/10.1029/2021EF002282, 2022.
Koenigk, T., Fuen es-F anco, R., Meccia, V. L., Gu jah , O., Jackson, L. C., New, A. L., O ega, P., Robe s, C. D.,
Robe s, M. J., A souze, T., Io ino, D., Moine, M.-P., and Sein, D. V.: Deep mixed ocean olume in he Lab ado Sea in 650
HighResMIP models, Clim. Dyn., 57, 1895–1918, h ps://doi.o g/10.1007/s00382-021-05785-x, 2021.
Lee, R. W., Woollings, T. J., Hoskins, B. J., Williams, K. D., O’Reilly, C. H., and Masa o, G.: Impac o Gul S eam
SST biases on he global a mosphe ic ci cula ion, Clim. Dyn., 51, 3369–3387, h ps://doi.o g/10.1007/s00382-018-4083-9,
2018.
Liu, W., Fedo o , A. V., Xie, S.-P., and Hu, S.: Clima e impac s o a weakened A lan ic Me idional O e u ning 655
Ci cula ion in a wa ming clima e, Sci. Ad ., 6, eaaz4876, h ps://doi.o g/10.1126/sciad .aaz4876, 2020.
Lumpkin, R. and Spee , K.: Global Ocean Me idional O e u ning, J. Phys. Oceanog ., 37, 2550–2562,
h ps://doi.o g/10.1175/JPO3130.1, 2007.
Ma zocchi, A., Hi schi, J. J.-M., Holliday, N. P., Cunningham, S. A., Blake , A. T., and Cowa d, A. C.: The No h
A lan ic subpola ci cula ion in an eddy- esol ing global ocean model, J. Ma . Sys ., 142, 126–143, 660
h ps://doi.o g/10.1016/j.jma sys.2014.10.007, 2015.
Mena y, M. B., Hodson, D. L. R., Robson, J. I., Su on, R. T., Wood, R. A., and Hun , J. A.: Explo ing he impac o
CMIP5 model biases on he simula ion o No h A lan ic decadal a iabili y, Geophys. Res. Le ., 42, 5926–5934,
h ps://doi.o g/10.1002/2015GL064360, 2015.
Mena y, M. B., Jackson, L. C., and Lozie , M. S.: Reconciling he Rela ionship Be ween he AMOC and Lab ado Sea in 665
OSNAP Obse a ions and Clima e Models, Geophys. Res. Le ., 47, e2020GL089793,
h ps://doi.o g/10.1029/2020GL089793, 2020.
Moa , B. I., Smeed, D., Rayne , D., Johns, W. E., Smi h, R. H., Volko , D. L., Ba inge , M. O., Collins, J.: A lan ic
me idional o e u ning ci cula ion obse ed by he RAPID-MOCHA-WBTS (RAPID-Me idional O e u ning Ci cula ion
and Hea lux A ay-Wes e n Bounda y Time Se ies) a ay a 26N om 2004 o 2022 ( 2022.1), NERC EDS B i ish 670
Oceanog aphic Da a Cen e NOC [da a se ],
h ps://www.bodc.ac.uk/da a/published_da a_lib a y/ca alogue/10.5285/04c79ece-3186-349a-e063-6c86abc0158c, 2023.
h ps://doi.o g/10.5194/egusphe e-2025-547
P ep in . Discussion s a ed: 21 Feb ua y 2025
c
Au ho (s) 2025. CC BY 4.0 License.
33
Mo eno-Chama o, E., Ca on, L.-P., Loos eld Tomas, S., Vegas-Regido , J., Gu jah , O., Moine, M.-P., Pu asahan, D.,
Robe s, C. D., Robe s, M. J., Senan, R., Te ay, L., Tou igny, E., and Vidale, P. L.: Impac o inc eased esolu ion on long-
s anding biases in HighResMIP-PRIMAVERA clima e models, Geosci. Model De ., 15, 269–289, 675
h ps://doi.o g/10.5194/gmd-15-269-2022, 2022.
Mo eno-Chama o, E., A souze, T., Acos a, M., B e onniè e, P.-A., Cas illo, M., Fe e , E., F igola, A., Kuzne so a, D.,
Ma in-Ma inez, E., O ega, P., and Palomas, S.: The e y-high esolu ion con igu a ion o he EC-Ea h global model o
HighResMIP, Geosci. Model De . Discuss. [p ep in ], h ps://doi.o g/10.5194/gmd-2024-119, 2024.
Ng, H. C., Robinson, L. F., McManus, J. F., Mohamed, K. J., Jacobel, A. W., I ano ic, R. F., G egoi e, L. J., and Chen, 680
T.: Cohe en deglacial changes in wes e n A lan ic Ocean ci cula ion, Na . Commun., 9, 2947,
h ps://doi.o g/10.1038/s41467-018-05312-3, 2018.
O ega, P., Robson, J. I., Mena y, M., Su on, R. T., Blake , A., Ge me, A., Hi schi, J. J.-M., Sinha, B., He manson, L.,
and Yeage , S.: Lab ado Sea subsu ace densi y as a p ecu so o mul idecadal a iabili y in he No h A lan ic: a mul i-
model s udy, Ea h Sys . Dyn., 12, 419–438, h ps://doi.o g/10.5194/esd-12-419-2021, 2021. 685
Righi, M., Andela, B., Ey ing, V., Laue , A., P edoi, V., Schlund, M., Vegas-Regido , J., Bock, L., B ö z, B., de Mo a,
L., Diblen, F., D eye , L., D os , N., Ea nshaw, P., Hassle , B., Kolduno , N., Li le, B., Loos eld Tomas, S., and
Zimme mann, K.: Ea h Sys em Model E alua ion Tool (ESMValTool) 2.0 – echnical o e iew, Geosci. Model De ., 13,
1179–1199, h ps://doi.o g/10.5194/gmd-13-1179-2020, 2020.
Robe s, M. J., Bake , A., Blockley, E. W., Cal e , D., Cowa d, A., Hewi , H. T., Jackson, L. C., Kuhlb od , T., 690
Ma hio , P., Robe s, C. D., Schiemann, R., Seddon, J., Vanniè e, B., and Vidale, P. L.: Desc ip ion o he esolu ion
hie a chy o he global coupled HadGEM3-GC3.1 model as used in CMIP6 HighResMIP expe imen s, Geosci. Model De .,
12, 4999–5028, h ps://doi.o g/10.5194/gmd-12-4999-2019, 2019.
Robe s, M. J., Jackson, L. C., Robe s, C. D., Meccia, V., Docquie , D., Koenigk, T., O ega, P., Mo eno-Chama o, E.,
Bellucci, A., Cowa d, A., D ij hou , S., Exa chou, E., Gu jah , O., Hewi , H., Io ino, D., Lohmann, K., Pu asahan, D., 695
Schiemann, R., Seddon, J., Te ay, L., Xu, X., Zhang, Q., Chang, P., Yeage , S. G., Cas uccio, F. S., Zhang, S., and Wu, L.:
Sensi i i y o he A lan ic Me idional O e u ning Ci cula ion o Model Resolu ion in CMIP6 HighResMIP Simula ions and
Implica ions o Fu u e Changes, J. Ad . Model. Ea h Sys ., 12, e2019MS002014, h ps://doi.o g/10.1029/2019MS002014,
2020.
Roque , F., Madec, G., McDougall, T. J., and Ba ke , P. M.: Accu a e polynomial exp essions o he densi y and 700
speci ic olume o seawa e using he TEOS-10 s anda d, Ocean Model., 90, 29–43,
h ps://doi.o g/10.1016/j.ocemod.2015.04.002, 2015.
Sein, D. V., Kolduno , N. V., Danilo , S., Sido enko, D., Weke le, C., Cabos, W., Rackow, T., Scholz, P., Semmle , T.,
Wang, Q., and Jung, T.: The Rela i e In luence o A mosphe ic and Oceanic Model Resolu ion on he Ci cula ion o he
No h A lan ic Ocean in a Coupled Clima e Model, J. Ad . Model. Ea h Sys ., 10, 2026–2041, 705
h ps://doi.o g/10.1029/2018MS001327, 2018.
h ps://doi.o g/10.5194/egusphe e-2025-547
P ep in . Discussion s a ed: 21 Feb ua y 2025
c
Au ho (s) 2025. CC BY 4.0 License.
34
Spall, M. A. and Picka , R. S.: Whe e Does Dense Wa e Sink? A Subpola Gy e Example, J. Phys. Oceanog ., 31, 810–
826, h ps://doi.o g/10.1175/1520-0485(2001)031<0810:WDDWSA>2.0.CO;2, 2001.
S aneo, F.: On he Connec ion be ween Dense Wa e Fo ma ion, O e u ning, and Polewa d Hea T anspo in a
Con ec i e Basin, J. Phys. Oceanog ., 36, 1822–1840, h ps://doi.o g/10.1175/JPO2932.1, 2006. 710
Sun, B., Liu, C., and Wang, F.: Global me idional eddy hea anspo in e ed om A go and al ime y obse a ions, Sci.
Rep., 9, 1345, h ps://doi.o g/10.1038/s41598-018-38069-2, 2019.
Tie sche, S., Balmaseda, M., Zuo, H., Robe s, C., Maye , M., and Fe an i, L.: The impo ance o No h A lan ic Ocean
anspo s o seasonal o ecas s, Clim. Dyn., 55, 1995–2011, h ps://doi.o g/10.1007/s00382-020-05364-6, 2020.
T eguie , A. M., Deshayes, J., Lique, C., Dussin, R., and Molines, J. M.: Eddy con ibu ions o he me idional anspo 715
o sal in he No h A lan ic, J. Geophys. Res. Oceans, 117, C05010, h ps://doi.o g/10.1029/2012JC007927, 2012.
T enbe h, K. E. and Fasullo, J. T.: A lan ic me idional hea anspo s compu ed om balancing Ea h’s ene gy locally,
Geophys. Res. Le ., 44, 1919–1927, h ps://doi.o g/10.1002/2016GL072475, 2017.
Tsa sali, E. E., Haa sma, R. J., A hanasiadis, P. J., Bellucci, A., de V ies, H., D ij hou , S., de V ies, I. E., Pu asahan,
D., Robe s, M. J., Sanchez–Gomez, E., and Robe s, C. D.: Impac o esolu ion on he a mosphe e–ocean coupling along 720
he Gul S eam in global high esolu ion models, Clim. Dyn., 58, 3317–3333, h ps://doi.o g/10.1007/s00382-021-06098-9,
2022.
Win on, M., Ande son, W. G., Delwo h, T. L., G i ies, S. M., Hu lin, W. J., and Rosa i, A.: Has coa se ocean esolu ion
biased simula ions o ansien clima e sensi i i y?, Geophys. Res. Le ., 41, 8522–8529,
h ps://doi.o g/10.1002/2014GL061523, 2014. 725
Yashayae , I. and Lode , J. W.: Enhanced p oduc ion o Lab ado Sea Wa e in 2008, Geophys. Res. Le ., 36, L01606,
h ps://doi.o g/10.1029/2008GL036162, 2009.
Yashayae , I. and Lode , J. W.: Recu en eplenishmen o Lab ado Sea Wa e and associa ed decadal-scale a iabili y,
J. Geophys. Res. Oceans, 121, 8095–8114, h ps://doi.o g/10.1002/2016JC012046, 2016.
Yashayae , I. and Lode , J. W.: Fu he in ensi ica ion o deep con ec ion in he Lab ado Sea in 2016, Geophys. Res. 730
Le ., 44, 1429–1438, h ps://doi.o g/10.1002/2016GL071668, 2017.
Zuo, H., Balmaseda, M. A., Tie sche, S., Mogensen, K., and Maye , M.: The ECMWF ope a ional ensemble eanalysis–
analysis sys em o ocean and sea ice: a desc ip ion o he sys em and assessmen , Ocean Sci., 15, 779–808,
h ps://doi.o g/10.5194/os-15-779-2019, 2019.
735
h ps://doi.o g/10.5194/egusphe e-2025-547
P ep in . Discussion s a ed: 21 Feb ua y 2025
c
Au ho (s) 2025. CC BY 4.0 License.
