emo e sensing
A icle
Sensi i i y o Modeled CO2Ai –Sea Flux in a Coas al
En i onmen o Su ace Tempe a u e G adien s,
Su ac an s, and Sa elli e Da a Assimila ion
Rica do To es 1,* , Yu i A ioli 1, Vassilis Ki idis 1, S e ano Cia a a 1,
Manuel Ruiz-Villa eal 2, Jamie Shu le 3, Luca Polimene 1, Vic o Ma inez 1,
Clai e Widdicombe 1, E. Malcolm S. Woodwa d 1, Timo hy Smy h 1, James Fishwick 1and
Ga in H. Tils one 1
1Plymou h Ma ine Labo a o y, Plymou h PL1 3DH, UK; [email p o ec ed] (Y.A.); [email p o ec ed] (V.K.);
[email p o ec ed] (S.C.); [email p o ec ed] (L.P.); [email p o ec ed] (V.M.); [email p o ec ed] (C.W.);
[email p o ec ed] (E.M.S.W.); [email p o ec ed] (T.S.); [email p o ec ed] (J.F.); [email p o ec ed] (G.H.T.)
2Ins i u o Español de Oceanog a ía (IEO), Cen o oceanog á ico da Co uña, Paseo Ma í imo Alcalde
F ancisco Vázquez, nº10, 15001 A Co uña, Spain; [email p o ec ed]
3Exe e , Uni e si y, Pen yn Campus, Geog aphy depa men , T elie e Road, Pen yn,
Co nwall TR10 9FE, UK; j.d.shu le @exe e .ac.uk
*Co espondence: [email p o ec ed]
Recei ed: 31 May 2020; Accep ed: 22 June 2020; Published: 25 June 2020


Abs ac :
This wo k e alua es he sensi i i y o CO
2
ai –sea gas exchange in a coas al si e o
ou di e en model sys em con igu a ions o he 1D coupled hyd odynamic–ecosys em model
GOTM–ERSEM, owa ds iden i ying c i ical dynamics o ele ance when speci ically add essing
quan i ica ion o ai –sea CO
2
exchange. The Eu opean Sea Regional Ecosys em Model (ERSEM) is a
biomass and unc ional g oup-based biogeochemical model ha includes a comp ehensi e ca bona e
sys em and explici ly simula es he p oduc ion o dissol ed o ganic ca bon, dissol ed ino ganic
ca bon and o ganic ma e . The model was implemen ed a he coas al s a ion L4 (4 nm sou h o
Plymou h, 50
°
15.00’N, 4
°
13.02’W, dep h o 51 m). The model pe o mance was e alua ed using
mo e han 1500 hyd ological and biochemical obse a ions ou inely collec ed a L4 h ough he
Wes e n Coas al Obse a o y ac i i ies o 2008—2009. In addi ion o a e e ence simula ion (A),
we an h ee dis inc expe imen s o in es iga e he sensi i i y o he ca bona e sys em and modeled
ai –sea luxes o (B) he sea-su ace empe a u e (SST) diu nal cycle and hus also he nea -su ace
e ical g adien s, (C) biological supp ession o gas exchange and (D) da a assimila ion using sa elli e
Ea h obse a ion da a. The e e ence simula ion cap u es well he physical en i onmen (simula ed
SST has a co ela ion wi h obse a ions equal o 0.94 wi h a p
>
0.95). O e all, he model cap u es
he seasonal signal in mos biogeochemical a iables including he ai –sea lux o CO
2
and p ima y
p oduc ion and can cap u e some o he in a-seasonal a iabili y and sho -li ed blooms. The model
co ec ly ep oduces he seasonali y o nu ien s (co ela ion
>
0.80 o silica e, ni a e and phospha e),
su ace chlo ophyll-a (co ela ion >0.43) and o al biomass (co ela ion >0.7) in a wo yea un o
2008–2009. The model simula es well he concen a ion o DIC, pH and in-wa e pa ial p essu e
o CO
2
(pCO
2
) wi h co ela ions be ween 0.4–0.5. The model esul sugges ha L4 is a weak ne
sou ce o CO
2
(0.3–1.8 molCm
−2
yea
−1
). The esul s o he h ee sensi i i y expe imen s indica e
ha bo h esol ing he empe a u e p o ile nea he su ace and assimila ion o su ace chlo ophyll-a
signi ican ly impac he skill o simula ing he biogeochemis y a L4 and all o he ca bona e chemis y
ela ed a iables. These esul s indica e ha ou o ecas ing abili y o CO
2
ai –sea lux in shel seas
en i onmen s and hei impac in clima e modeling should conside bo h model e inemen s as
means o educing unce ain ies and e o s in any u u e clima e p ojec ions.
Remo e Sens. 2020,12, 2038; doi:10.3390/ s12122038 www.mdpi.com/jou nal/ emo esensing
Remo e Sens. 2020,12, 2038 2 o 29
Keywo ds: da a assimila ion; 1D ecosys em model; CO2; ai –sea gas exchange
1. In oduc ion
Recen esea ch e o s in ma ine and clima e disciplines highligh he e ec s ha inc easing
a mosphe ic CO
2
concen a ions a e ha ing on he Ea h’s clima e and in i s con ol o e he oceanic
ca bona e sys em and pH [
1
]. As one o he g eenhouse gases, i s ai –sea lux is a c i ical pa o he
clima e sys em.
A mosphe ic ca bon dioxide (CO
2
) has been inc easing since he indus ial e olu ion due o
he bu ning o ossil uels, cemen p oduc ion and land use change. While i is gene ally accep ed
ha he ocean sys em annually abso bs up o 30% o he CO
2
eleased by bu ning ossil uels [
2
], i is
no clea whe he his ocean ca bon ‘sink’ is inc easing o dec easing, and conside able empo al and
spa ial in e -annual a ia ion appea s o occu [
3
]. The lux o CO
2
be ween he a mosphe e and he
ocean (ai –sea) is p ima ily con olled by condi ions o he physical en i onmen like wind speed,
sea s a e and sea-su ace empe a u e bu also by biological p ocesses aking place in he eupho ic zone.
Adding CO
2
o seawa e dis up s he ca bona e sys em, leading o an inc ease in ee CO
2
/ca bonic
acid and he bica bona e ion concen a ion while dec easing he ca bona e ion, pH and ca bona e
sa u a ion s a es [
4
]. All o hese ha e implica ions o a ange o biological and chemical p ocesses
like calci ica ion, p ima y p oduc ion and ep oduc ion [5,6].
In shel seas, CO
2
dynamics a e highly a iable in space and ime [
7
,
8
] pa ly d i en by he high
p ima y p oduc i i y, leading o a la ge d aw-down o dissol ed ino ganic ca bon (DIC, he sum o
CO
2
, bica bona e and ca bona e ions) and consequen ise in pH [
9
]. This is u he complica ed by
he close link be ween ben hic p ocesses and he pelagic ca bon cycle and alkalini y [
10
] and i e ine
inpu s o DIC and o al alkalini y (TA) in o coas al sys ems [11].
In addi ion o hese p ocesses, diapycnal mixing ac oss he he mocline d i es a con inuous
change in su ace CO
2
as a esul o e ical luxes o DIC and nu ien s ha s imula e p ima y
p oduc ion [
12
]. The exis ence o nea -su ace empe a u e a ia ions was iden i ied by
Wa d e al. [13]
and hei signi icance o ai –sea CO2 luxes has been ecen ly highligh ed by Wool e al. [14].
Clea ly, u he unde s anding he CO
2
pa hways, sou ces, sinks, hei espec i e budge s
and hei impac s on Ea h’s clima e sys em is essen ial o moni o ing and p ojec ing u u e
clima e scena ios.
In his wo k we e alua e he sensi i i y o model de i ed CO
2
luxes o physical and biological
d i e s in a shallow shel sea loca ion (L4) using he GOTM–ERSEM model [
15
]. The use o a 1D model
he e is app op ia e because in shallow shel seas, he e ical mixing is mainly d i en by he balance
be ween idal mixing and a mosphe ic luxes, especially sola hea ing.
L4 (4 nm sou h o Plymou h, 50
°
15.00
0
N, 4
°
13.02
0
W, dep h o 51 m) is a coas al s a ion
in he wes e n English Channel whe e sus ained obse a ions ha e been ca ied ou since he
la e 1980s and whe e a iabili y in plank onic communi ies has been desc ibed ex ensi ely [
16
].
In ecen yea s, physical obse a ions ha e in ensi ied in he amewo k o he Wes e n Channel
Obse a o y [
17
] un by he Plymou h Ma ine Labo a o y (PML) in coope a ion wi h he Ma ine
Biological Associa ion (MBA). The en i onmen al and wa e condi ions a L4 ha e been sampled
weekly and physico–chemical and biological obse a ions a e eely a ailable. Recen ly, a mid-ocean
buoy has been ins alled a L4 [
18
]. As in o he shel seas, physics a L4 a e mainly he esul o a balance
be ween idal mixing and a mosphe ic o cing and a seasonal empe a u e s a i ica ion appea s when
summe hea ing o e comes he idal induced mixing. Nu ien dynamics and ecosys em dynamics
a e s ongly dependen on his a iabili y [
17
]. The e a e nume ous epo s o s ong in a-seasonal
and in e -annual a iabili y in he plank on p oduc ion and communi y composi ion in he wes e n
English Channel and in o he shel seas e.g., [
16
,
19
,
20
], all o which help o modula e and al e he
Remo e Sens. 2020,12, 2038 3 o 29
ai –sea lux o CO
2
. Collec i ely, hese da a p o ide a subs an ial da ase o pa ame e ize and e alua e
he pe o mance o coupled hyd odynamic–ecosys em models.
2. Me hods
In his wo k we e alua e he sensi i i y o modeled CO
2
luxes o h ee me hodological app oaches
which can a ec he simula ion o he ca bona e sys em wi hin ecosys em models. In doing so, we ha e
de ined h ee nume ical expe imen s and one e e ence simula ion ha cons i u es ou baseline da a o
compa e he changes associa ed wi h each expe imen .
2.1. Physics Model Desc ip ion: Go m
The Gene al Ocean Tu bulence Model, (GOTM [
21
]) is a 1D physical model ha is designed as
a gene ic sys em o ma ine modeling. GOTM dynamically simula es he e olu ion o empe a u e,
densi y and e ical mixing when o ced wi h me eo ological da a (e.g., [
15
]). F om all he u bulence
model op ions a ailable wi hin GOTM we ha e chosen he second o de u bulence model
κ
–
e
wi h he
coe icien s sugges ed in Canu o e al.
[22]
( e sion A) as Umlau and Bu cha d
[23]
demons a ed ha
i can simula e u bulence in s a i ied shea lows in a physically sound and nume ically obus way.
The sensible and la en hea luxes a e calcula ed wi hin GOTM using s anda d o mulae [
24
].
The calcula ions o he la en and sensible hea a e dependen upon he calcula ion o hei coe icien s,
which a e a unc ion o ai –sea empe a u e di e ence, wind speed and a s abili y c i e ion o he
su ace wa e s [
25
]. The long wa e (back) adia ion is calcula ed using he May o mula ion [
26
].
The su ace s ess is calcula ed om he wind s ess. These o mulae we e chosen based on a sensi i i y
s udy o su ace hea lux o mula ion on he model esponse [27].
The physical model supplies he empe a u e o he ecosys em p ocess desc ip ions and
de e mines he luxes o ecosys em componen s be ween adjacen boxes and o laye s. The coupling o
ERSEM and GOTM is desc ibed in [28].
2.2. Biological Model Desc ip ion: E sem
ERSEM is a gene ic ecosys em model based on he “ unc ional g oup app oach” [
29
,
30
];
he bio a in he ecosys em a e di ided in o h ee unc ional ypes: p ima y p oduce s, consume s,
and decompose s, which a e subdi ided on he basis o ophic links and/o size. ERSEM can p oduce
a ange o dynamics and communi y s uc u es wi h a consis en pa ame e iza ion, when applied o a
wide a ie y o physical scena ios. The consis en pa ame e iza ion is only possible because o he
ange o p ocess desc ip ions now included in he model.
The model is capable o accu a ely simula ing he spa ial pa e n o ecological luxes h oughou
he seasonal cycle and he main phy oplank on succession (cg. [
28
,
31
]). I includes ep esen a ions o
he ben hic sys em, which a e i al o he co ec ea men o shel seas. Bo h physiological (inges ion,
espi a ion, exc e ion and eges ion) and popula ion (g ow h and mo ali y) p ocesses a e desc ibed
by luxes o ca bon and nu ien s be ween unc ional g oups. The associa ed decoupled ca bon and
nu ien dynamics gi es a a be e app oxima ion o how nu ien limi a ion ac s on cells han ixed
quo a models [
30
]. I can simula e bo h he classical, la ge cell dynamics and he small cell mic obial
loop, he eby ep esen ing he con inuum o ophic pa hways. ERSEM s anda d pelagic compa men s
include ou phy oplank on unc ional g oups; picoplank on (0.2–2
µ
m), nano lagella es (2–20
µ
m),
dino lagella es (20–200
µ
m) and dia oms (20–200
µ
m) h ee zooplank on g oups; he e o ophic
nano lagella es (<20
µ
m), mic ozooplank on (20–200
µ
m) and mesozooplank on (>200
µ
m) and a
bac e ia loop. F ac iona ion o dissol ed and pa icula e ma e and ins an aneous ligh a e also
key aspec s o he cu en model e sion. A mo e in-dep h desc ip ion o he model can be ound
in
Allen e al. [28]
. The model has been ex ensi ely alida ed wi h espec o he ca bon sys em [
8
],
chl-a [
32
,
33
], and nu ien s and biomass [
34
,
35
]. The ull ERSEM model open sou ce code and manual
can be ound a h p://www.pml.ac.uk/Modelling/Models/ERSEM.
Remo e Sens. 2020,12, 2038 4 o 29
2.3. Obse a ions
Bo h in si u and sa elli e de i ed biological measu emen s we e used o model skill assessmen
and/o assimila ion. These obse a ions a e desc ibed in he ollowing sec ions.
2.3.1. In Si u Biological Measu emen s
The obse a ions we e ob ained unde he weekly sampling s a egy o he Wes e n Channel
Obse a o y. The wa e samples we e collec ed using Niskin bo les moun ed on he CTD ose e
ame. Su ace nu ien measu emen s o silica e, phospha e, ni a e, ni i e and ammonium whe e
measu ed colo ime ically using a B an and Luebbe AAIII segmen ed low au oanalyze [
36
]. Fo chl-a
measu emen s, 1 o 2 L o seawa e om ou dep hs (0, 10, 20 and 50 m) was il e ed on o a
Wha man
®
GF/F glass mic o ibe il e and he il e s o ed in liquid ni ogen un il analysis. Fo he
analysis s age, pigmen s we e ex ac ed om he hawed GF/F il e in o 2 mL me hanol [
37
]
and sonica ed o 35 s. These ex ac s we e hen cen i uged o emo e il e and cell deb is
(5 min a 4000 pm) and analyzed using e e sed-phase HPLC [
38
]. Pigmen s, including chl-a,
we e iden i ied using e en ion ime and spec ally ma ched using pho o-diode a ay spec oscopy [
39
].
Wa e samples om 10 m dep h we e used o quan i ying he phy oplank on and mic ozooplank on
communi y composi ion and abundance om mic oscopic analysis o samples p ese ed wi h Lugol’s
iodine. Cell we e iden i ied o species-le el we e possible and assigned o h ee unc ional g oups
(dia oms (cen ic and penna e), dino lagella es and lagella es). The con e sion om cell numbe s o
biomass was based on olumes acco ding o geome ic shapes and o mulae o Olenina e al.
[40]
and
Menden-Deue and Lessa d [41].
Pho osyn hesis–i adiance pa ame e s (P–E) o he es ima ion o in eg a ed p ima y p oduc ion
ollowed he me hodology p esen ed in Tils one e al.
[42]
and Pla e al.
[43]
. The samples we e
incuba ed o 2–3 h a e which he suspended ma e ial was il e ed h ough 25 mm Wha man GF/F
il e s. The il e s we e exposed o concen a ed HCl umes o 12 h and adioca bon ac i i y on he
il e s was de e mined by a Packa d T i-Ca b 2500 TR liquid scin illa ion analyze using he ex e nal
s anda d and he channel a io me hods o co ec o quenching. The b oadband P–E pa ame e s we e
es ima ed by i ing he da a o he model o Pla e al.
[43]
. The daily in eg a ed p oduc i i y was
es ima ed as in Tils one e al.
[42]
a 1 m in e als om he sea su ace down o 0.1% su ace i adiance.
Time se ies o in si u p ima y p oduc ion a L4 is gi en in Ba nes e al. [44].
The ca bona e sys em was sampled a wo dep hs; 2 m and bo om, using 250 mL glass bo les and
p ese ed wi h 50
µ
L o sa u a ed HgCl
2
solu ion. The samples we e analyzed wi hin 12 mon hs by
i a ion ( o o al alkalini y) and coulome ic analysis ( o DIC). Full de ails o he analy ical p ocedu e
and a ho ough discussion o he ca bona e sys em da a can be ound in Ki idis e al. [45].
2.3.2. Sa elli e Obse a ions
Le el 1a (calib a ed and geoloca ed adiance) 1 km spa ial esolu ion sa elli e emo e sensing
Ea h obse a ion (EO) da a collec ed by he Mode a e-Resolu ion Imaging Spec ome e (MODIS,
onboa d he Aqua pla o m) we e downloaded om he NASA Ocean Colo da a dis ibu ion websi e
(h p://oceanda a.sci.gs c.nasa.go /) o he 2009–2010 pe iod. These we e p ocessed using he
SeaWiFS Da a Analysis Sys em (SeaDAS, e sion 6.4) o gene a e 1 km chl-a da a using he h ee-band
algo i hm, OC3 [
46
]. All p oduc s we e quali y masked using he de aul SeaDAS masks o land,
cloud and high/sa u a ed adiance [47].
2.4. Expe imen s
2.4.1. Expe imen A
The e e ence simula ion o Expe imen A is desc ibed in Sec ions 2.1 and 2.2. In his wo k,
he GOTM model was con igu ed wi h 50 e ical laye s wi h a ime s ep o 10 s. The e ical esolu ion
is inc eased owa ds he bo om laye o be e cap u e he dynamics o he bo om bounda y laye .
Remo e Sens. 2020,12, 2038 5 o 29
This esul s in a laye hickness o 30 cm a he bo om inc easing loga i hmically o 1.6 m om 15 m
abo e he bo om o he su ace.
Fo me eo ological o cing, we use 6 hou ly da a om he Na ional Cen e o En i onmen al
P edic ion/Na ional Cen e o A mosphe ic Resea ch (NCEP/NCAR) eanalysis [
48
], which has a
spa ial esolu ion o 1.9
◦×
1.9
◦
. The model is o ced wi h ai empe a u e, a mosphe ic p essu e,
10 m winds and dew empe a u e. The sho wa e adia ion is imposed sepa a ely a 5 min in e als
using locally obse ed alues measu ed wi h a Li-Co py anome e (LI-200SZ) on a coas al s a ion
main ained by PML (h p://www.wes e nchannelobse a o y.o g.uk/pml_wea he _s a ion).
Tidal o cing was included as ime se ies o su ace ele a ion g adien s gene a ed using 15 idal
ha monics ex ac ed om he esul s o a high esolu ion (1.8
×
1.8 km g id spacing) 3D ba o opic
model simula ion o he UK shel [
49
]. The ha monics used a e all he ele an semi-diu nal and
diu nal idal cons i uen s such as M2, Q1, O1, P1, S1, K1, 2N
2
,
µ2
, N2, u
2
, L2, T2, S2, K2 and M4.
Su ace ele a ion g adien s we e ob ained by a i icially gene a ing hou ly su ace ele a ion ime
se ies on adjacen g id poin s o he L4 s a ion posi ion using he - ide MATLAB oolbox [50].
The model is ini ialized in win e when he wa e column is well-mixed wi h obse ed
empe a u e and salini y p o iles (see Sec ion 2.3). The wa e column e olu ion is u he cons ained
by nudging he model esul s o he obse ed empe a u e and salini y p o iles wi h a elaxa ion ime
scale o wo weeks [
15
,
21
]. In his way, GOTM can accoun o missing 3-dimensional dynamics no
esol ed wi hin 1D models like he long- e m ho izon al ad ec ion o empe a u e and salini y ha is
impo an o esol ing he e ical wa e column s uc u e and hence gas luxes in coas al egions [
51
].
An al e na i e app oach o he in oduc ion o wa e masses wi h di e en empe a u e and salini y
signa u es in o he 1D wa e column is h ough he speci ica ion o ime- a ying, e ically esol ed
ho izon al empe a u e and salini y g adien s. This op ion is no employed he e o his pu pose due
o he absence o such in o ma ion, bu he ea u e is used in Expe imen D as a way o a i icially
in oduce e o s in he mixing dynamics as desc ibed in Sec ion 2.4.4.2. The model is ini ialized whe e
possible om obse a ions aken in Decembe 2005 (plank on biomass and nu ien s) and un o
wo yea s un il Decembe 2007 o c ea e a common s a ile o each o he expe imen s. The yea s
2006–2007 a e aken as a model spin-up pe iod and he model is e alua ed agains obse a ions o he
pe iod 6 Janua y 2008 un il 20 Decembe 2009. The model pa ame e s used ollow Black o d e al.
[31]
as modi ied by A ioli e al. [8].
2.4.2. Expe imen B
Expe imen B looks in o he e ec s ha esol ing he diu nal SST cycle has on he ai –sea luxes
o CO
2
. By he diu nal cycle o SST we conside he changes o he empe a u e e ical s uc u e in
he op ew me e s o he ocean caused by he di ec day ime sola hea ing and nigh ime cooling.
The e o e, o model hese diu nal changes, he model esolu ion is inc eased loga i hmically om 2 cm
a he su ace o 1 m a 10 m deep.
2.4.3. Expe imen C
Expe imen C e alua es he po en ial impac s ha biologically gene a ed su ac an s ha e on
ai –sea luxes. Sea-su ace su ac an s o o ganics can in luence gas exchange in wo ways, as a
monolaye physical ba ie and h ough modi ica ion o sea-su ace hyd odynamics and hence
u bulen ene gy ans e [
52
,
53
], mani es ing as a educed gas ans e eloci y (
K
). Fo Expe imen
C, we ha e chosen o implemen a p ima y p oduc i i y-based educ ion on he gas ans e eloci y.
Se e al s udies ha e ied o quan i y his supp ession o
K
(e.g., and e e ences he ein [
53
]). Howe e ,
he exac magni ude and na u e o he supp ession is unclea as he alues ound among s udies
(in bo h he na u al en i onmen and wi hin labo a o ies) a y conside ably. So he e o e we ha e
chosen o in es iga e he sensi i i y o he luxes o his ype o phenomenon a he han ying o
quan i y i .

Remo e Sens. 2020,12, 2038 6 o 29
2.4.4. Expe imen D
The hi d Expe imen (D) e alua es he impac s o using da a assimila ion o emo ely sensed
chlo ophyll-a (chl-a) on he abili y o ERSEM o simula e he ca bona e sys em and i s associa ed
con ol on ai –sea luxes o CO2.
2.4.4.1. The Ensemble Kalman il e (EnKF)
Expe imen D e alua es he sensi i i y o modeled CO
2
o da a assimila ion o su ace chl-a as
es ima ed om sa elli e. Following p e ious wo k we ha e chosen he Ensemble Kalman Fil e [
54
]
as implemen ed in he 1D GOTM–ERSEM model by To es e al. [15] and used by [34] o his egion.
The Ensemble Kalman Fil e me hod was o mula ed o wo k wi h non-linea models and in pa icula
o handle he e o co a iance e olu ion [
54
]. He e he analysis scheme desc ibed in E ensen
[55]
is
implemen ed. Fi s he model is in eg a ed o wa d in ime wi h he addi ion o a s ochas ic e m),
Ψ
k+1= (Ψa
k) + qk, (1)
q
ep esen ing he model e o s and
ep esen s a non-linea ope a o .
Ψ
is a ma ix holding he model
s a e a iables. The
k
subsc ip indica es he ime index, while he supe sc ip s
and
a
co espond
o he o ecas and analysis. The model e o s a e ep esen ed h ough he co a iances o he s a e
a iable e o s. These a e de ined a p io i and a e e ol ed o wa d in ime by he Kalman il e .
The in o ma ion con ained by a ull p obabili y densi y unc ion can be ep esen ed by an
ensemble o model s a es. Co a iances can be de ined wi h espec o a known ue s a e o as
expec a ions, ensemble co a iance ma ices a ound he ensemble mean,
ψ
. Hence he ensemble o
model s a es will p o ide in o ma ion on s a is ical momen s o in e es .
P ≃(ψ −ψ )(ψ −ψ )T(2)
Pa≃(ψa−ψa)(ψa−ψa)T. (3)
The s anda d EnKF analysis equa ion [
56
] is applied he e. The analysis s ep in ol es upda ing
each o he ensemble membe s aking in o conside a ion he di e ence be ween he obse a ions and
he model s a e as well as he e o s associa ed wi h he model s a e a iables and he obse a ions.
As each ensemble is upda ed indi idually he e is no need o egene a e hem. The ensemble mean will
be he upda ed s a e ha minimizes he model a iance. A mo e in-dep h desc ip ion o he algo i hm
can be ound in Cia a a e al. [34].
The a iables included in he analysis a e log ans o med so ha hei a iabili y is educed,
hei Gaussiani y is imp o ed and posi i i y o he solu ion is gua an eed. The a iables and he assimila ed
obse a ions ha e been ans o med by a Gaussian anamo phosis unc ion na u al log as sugges ed in
Be ino e al.
[57]
. O he applica ions ha e been epo ed in To es e al.
[15]
, Cia a a e al.
[34]
, Ne ge and
G egg [58].
Compa ed o he o he model imp o emen s om expe imen s B and C, he EnKF equi es
addi ional compu ing esou ces which scale wi h he numbe o ensembles equi ed o co ec ly
app oxima e he e olu ion o he p obabili y densi y unc ion o he model s a e. In ou implemen a ion,
he se up equi ed unning 100 di e en model simula ions and was he e o e 100 imes mo e
compu a ionally expensi e han ei he Expe imen B o C. Fo his eason, Expe imen D ocusses on a
single yea (2009).
2.4.4.2. Model E o s
The ensemble o model s a es de ined in Equa ion (1) is ob ained by adding model e o s o he
assimila ion scheme. They a e applied o all he pelagic a iables included in he co a iance ma ix o
s a e a iables bu no o he ben hic s a e a iables. The la e a e no included in he analysis scheme
and hence a e no upda ed.
Remo e Sens. 2020,12, 2038 7 o 29
Two ypes o model e o s a e included in he simula ions: hose ha modi y he s a e a iables
and hose ha pe u b he o cing unc ions. E o s assigned o mos ERSEM s a e a iables we e 20%
Gaussian e o ; hey we e educed o 5% in he de i us ela ed a iables o limi la ge co ec ions o
he la ges pool in he model and in he nu ien a iables as hey cons i u e he bes esol ed a iables
in ERSEM. These e o s a e in oduced once as ini ial condi ions o gene a e he ini ial ensemble o
model s a es, and a each analysis s ep be o e sol ing he Kalman il e equa ion.
Addi ional e o s a e in oduced a un- ime by pe u bing he o cing unc ions ha d i e he
p ima y p oduc ion and he e ical anspo o he biological s a e a iables. A each ime s ep
o each ensemble simula ion, e o s a e added o he su ace i adiance alues. Addi ional e o s
associa ed wi h he in e nal p essu e g adien s and backg ound sedimen concen a ion cons an a e
andomly ini ialized and kep ixed du ing each assimila ion cycle so ha hey a e di e en o each
ensemble membe . The sea-su ace i adiance and he backg ound suspended sedimen concen a ion
a e wo o cing a iables which ha e a la ge ecological impo ance as he ligh egime de e mines in
g ea measu e he sho - e m a iabili y o phy oplank on. Bo h we e assigned a 25% Gaussian e o .
The in e nal p essu e g adien s e o s ac o pe u b he wa e column s uc u e and consequen ly he
e ical anspo associa ed wi h mixing. The in e nal p essu e g adien s e o s we e calcula ed as
pseudo- andom ields as sugges ed by E ensen
[56]
wi h a co ela ion leng h equal o he o al wa e
dep h (51 m).
2.4.4.3. Obse a ional E o s
The Gaussian measu emen e o s o he MODIS de i ed su ace chl-a we e se o 30% o he
mon hs Janua y o Ma ch and Oc obe o Decembe when he EO chl-a e ie als a e expec ed o
be a ec ed by high suspended sedimen s concen a ions in he wa e and high sola zeni h angle
(Figu e 1) [
59
]. Chl-a alues we e gene ally less han 1 mgCm
−3
in hose mon hs. Du ing he
sp ing–summe pe iod, EO e o s we e se a 10% o e lec he dominance o chl-a o e he suspended
sedimen concen a ions. We jus i y his by he good ag eemen ound by G oom e al.
[60]
in
compa ing HPLC in si u chl-a wi h MODIS and SeaWiFS da a who ound no bias due o coas al
adjacency in he ocean colou da a a L4. They concluded ha L4 could be conside ed case 2 wa e s in
la e au umn and win e and case 1 wa e s o he es o he yea .
Figu e 1.
Su ace chl a L4 du ing 2009 as obse ed om sa elli e (blue wi h assigned e o ba s) and
in si u HPLC analysis ( ed).
Remo e Sens. 2020,12, 2038 8 o 29
3. Resul s
The in si u obse a ions used in his wo k ha e been ex ensi ely desc ibed in nume ous
pape s (e.g., [
16
,
17
,
37
,
45
]) bu we p o ide a b ie o e iew he e. The seasonali y o nu ien s show
ypical maximum alues in Janua y (
8µM m−3 o Ni a e
,
0.5 µM m−3Phospha e
,
5µM m−3Silica e
)
and minima a e he sp ing bloom ( om May) un il i s a s inc easing again in Sep embe
ollowing he weakening o he he mal s a i ica ion. Chl-a also shows a ma ked seasonali y a
L4 (e.g., Figu e 1). Two peaks a e iden i ied yea on yea in he da a, one coinciden al wi h he
sp ing bloom (Ap il–May) and one in Sep embe –Oc obe . The sp ing bloom is gene ally domina ed
by dia oms bu wi h signi ican con ibu ions om coccoli opho es and phaeocys is. The analysis
o 15 yea s o phy oplank on abundance a L4 [
16
] has shown signi ican seasonal and in e -annual
changes in abundance and composi ion al hough seasonali y is he la ges de e minan o hese changes.
The au ho s ound ha he win e communi y seems ela i ely s able o e he yea s. The sp ing pe iod
ep esen s he la ges change e en in bo h composi ion and abundance while he summe pe iod is
he mos suscep ible o in e -annual changes in composi ion [
16
]. O e all, lagella es a e he dominan
g oup ep esen ing 87% o he o al phy oplank on cell abundance on a e age wi h he smalle cells
(2–4
µ
m) accoun ing o 63% o he o al abundance. In e ms o biomass, hey ep esen a small bu
signi ican p opo ion o he o al biomass.
3.1. Re e ence Simula ion; Expe imen A
3.1.1. Physics
We ha e e alua ed bo h he model capaci y o ep oduce he seasonal changes in he physical
en i onmen as well as sub-daily changes associa ed wi h he semi-diu nal ides. Ini ial calib a ion
o he model and simula ions o e alua e he sensi i i y o he model o a ia ions in o cing ha e
been pe o med o he yea s 2009–2010, when da a om he oceanog aphical buoy a e a ailable and
can help o educe unce ain y in he me eo ological o cing. In Augus 2010, we deployed a bo om
moun ed Teledyne RD Ins umen s 600 kHz B oadband Acous ic Dopple Cu en P o ile (ADCP) a
L4 o 20 days [61].
The seasonali y in hyd og aphy is well ep oduced by he model (Figu e 2a). This a ises pa ly
om he ac ha we a e using a basic nudging echnique o elax he model empe a u e (T) and
salini y (S) o he obse ed CTD p o iles wi h a elaxa ion ime o wo weeks. The co ela ion o
nea -su ace (1.5 m) empe a u e be ween he independen obse a ions om he L4 moni o ing buoy
and he model was la ge han 0.9 and signi ican a he 99.9% con idence le el ( ull s a is ics can be
ound in Table A4).
The abili y o he model o ep oduce he eloci y s uc u e was e alua ed du ing a simula ion o
he Augus ADCP deploymen . The esul s indica e ha he model se up ep oduces he ba o opic
componen o he idal eloci y well conside ing he 1D app oach. The co ela ion be ween he dep h
mean a e aged eloci y be ween 2 and 30 m was 0.7 a he 99.9% con idence le el. The dominan E–W
componen is be e esol ed albei wi h a small o e es ima ion o he lood peak eloci y (co ela ion
o 0.9 a 99.9% con idence le el). The N–S componen shows a la ge ampli ude in magni ude han
he obse a ions which esul ed in lowe co ela ions o 0.7. The eloci ies closes o he bo om
(a e aged be ween 1 and 2 m abo e he seabed, (Figu e 2b)) ha e a be e ag eemen be ween model
and obse a ions wi h an o e all co ela ion o 0.8 o he eloci y magni ude. As i is a idally
domina ed s a ion, his is key o ensu ing a ealis ic le el o mixing is imposed o he biological model.
Remo e Sens. 2020,12, 2038 9 o 29
Figu e 2.
Time e olu ion o (
a
) Six hou ly SST as simula ed wi h GOTM (solid line) and obse ed om
he L4 buoy (do s) and (
b
) eloci y magni ude a e aged be ween 1 and 2 m om he sea bed o he
model and obse a ions. The one o one plo wi h all da a used in calcula ing he s a is ics is included
in he le op co ne . The axes ha e he same ange as he main g aph.
3.1.2. Biogeochemis y
The model ou pu a six hou ly in e als in 2008–2009 a he sampling dep h o 10 m (Figu e 3a,b)
shows he model co ec ly ep oduces he app oxima e iming and du a ion o he sp ing bloom.
The model indica es a sho sha p sp ing bloom in 2008 (Figu e 3a) wi h subsequen sho -li ed
blooms du ing June–July in ag eemen wi h he a ailable obse a ions. The sp ing bloom in 2009 is by
con as longe (Figu e 3b) al hough he model o e es ima es he chl-a concen a ions. None heless,
he obse a ions sugges a subs an ial o e es ima ion o dia om biomass in 2008 (Figu e 4b) bu he
co ec o de o magni ude in May 2009 (Figu e 4b). The model unde es ima es o al chl-a om Janua y
o mid-Ap il and om Sep embe o Decembe in 2008 and 2009. The main disc epancy wi h he
obse a ions is he la e dino lagella e bloom (no shown) which he model unde es ima e in bo h yea s
al hough he miss-ma ch is mos e iden in 2009 (Figu e 3b).
Remo e Sens. 2020,12, 2038 16 o 29
Figu e 8.
Examples o nea -su ace p o iles o empe a u e o days 18–20 May 2009 o Expe imen
(
a
) A and (
b
) B. Du ing hose days, a e age nigh –day empe a u e di e ences we e o he o de o
0.5
◦
C. Wind speed a e aged 9
ms−1
. Black lines ep esen s nigh ime p o iles (03 and 21 h) while ed
lines ep esen day ime p o iles (09 and 15 h).
The de elopmen o he e y nea -su ace g adien esul s in changes o he ampli ude o he daily
luc ua ions o empe a u e such ha he nigh –day ange educes in Expe imen B, making nigh ime
minima wa me and day ime maxima coole han in Expe imen A as he nea -su ace g adien s limi
he exchange o hea wi h he a mosphe e.
pCO
2
shows a simila nigh –day e olu ion o he empe a u e p o iles wi h a consis en nega i e
g adien in he op 20 cm and he de elopmen o a day ime subsu ace maximum coinciden al wi h
he empe a u e maximum.
Expe imen B showed ha he ca bona e sys em model a iables a e sensi i e o he inc eased
esolu ion nea he su ace. The compa ison be ween expe imen s A and B (Table 1) shows he la ges
imp o emen s in 2008, in pa icula o pH, pCO
2
and DIC (a 10%, 19% and 18% imp o emen
espec i ely, Table 1). The e ec on he ai –sea lux is la ges and i is appa en h oughou he yea
sa e he mos s a i ied pe iods be ween June and Augus . The impac is impo an and esul s
in a 70% o 100% inc ease in he ne CO
2
up ake a L4. The modeled in eg a ed annual CO
2
lux
was 1.8 mo C/m
2
/yea in Expe imen A o 2008 inc easing o an in eg a ed annual CO
2
lux o
2.8 mo C/m
2
/yea in Expe imen B o he same yea . A simila esul is ob ained o 2009 wi h a change
om ne CO2sou ce o 0.3 mo C/m2/yea in Expe imen A o ne CO2sou ce o 0.6 mo C/m2/yea
in Expe imen B.
3.3. E alua ion o Su ace Slicks; Expe imen C
Expe imen C in ol ed a pa ame e iza ion o he e ec s o p ima y p oduc ion on he CO
2
ans e
eloci y (
KCO2
). This was in oduced h ough a scaling
KCO2
wi h he g oss p ima y p oduc ion
exc e ed by he phy oplank on as a esul o i s me abolic ac i i y. In p ac ice,
KCO2
was educed
o a maximum o 70% o i s wind dependen alue (calcula ed ollowing Nigh ingale e al.
[65]
)
when he g oss p oduc i i y su passes he 30% highes alues modeled a L4 in he yea s 2008–2009.
The esul s a e summa ized in Table 1and sugges a negligible impac when compa ed o Expe imen
A. The in eg a ed annual CO
2
lux suppo s hese esul s wi h changes smalle han 1% o bo h 2008
and 2009 yea s.

Remo e Sens. 2020,12, 2038 17 o 29
Table 1.
Pea son co ela ions o simula ion expe imen s A–D, be ween obse ed and modeled
a iables o 2008 and 2009 a he shallowes dep h a ailable. Signi ican co ela ions a he 95%
a e indica ed by *. Values in i alic signi y an imp o emen wi h espec o Expe imen A.
Yea To al Chl To al C Silica e Ni a e Phosph. Ammon. DIC pCO2Ph
Exp A
2008 0.43 * 0.70 * 0.82 * 0.90 * 0.82 * −0.23 0.46 * 0.52 * 0.53 *
2009 0.45 * 0.76 * 0.86 * 0.90 * 0.77 * −0.46 * 0.47 * 0.41 * 0.38
Exp B
2008 0.41 * 0.67 * 0.83 * 0.92 * 0.87 * −0.35 * 0.51 * 0.62 * 0.63 *
2009 0.48 * 0.75 * 0.86 * 0.88 * 0.90 * −0.46 * 0.54 * 0.35 * 0.33 *
Exp C
2008 0.43 * 0.70 * 0.82 * 0.90 * 0.82 * −0.23 0.46 * 0.52 * 0.53 *
2009 0.45 * 0.76 * 0.86 * 0.90 * 0.77 * −0.46 * 0.47 * 0.41 * 0.38
Exp D
2009 0.46 * 0.49 * 0.86 * 0.90 * 0.88 * −0.40 * 0.30 * −0.20 * −0.22
3.4. E alua ion o Da a Assimila ion; Expe imen D
Resul s om Expe imen D show ha assimila ion o EO chl-a imp o es some o he model
a iables bu i also de e io a es bulk a iables like o al biomass (Table 1and Table A1 in Appendix A).
As expec ed om p e ious wo k [
34
] he o ecas s ep always shows a la ge a iance o ensemble
sp ead han a e he analysis s ep sugges ing a educ ion in ensemble a iance as a esul . The modeled
mixed laye chl-a (Figu e 9a) shows an imp o ed ag eemen wi h obse a ions (Table 1) bu he model
ails o cap u e he sp ing bloom co ec ly. In pa icula , he model solu ion cons uc ed as he mean
o all he model ensembles o e es ima es chl-a in Ma ch and end o Ap il while he o e es ima ion
du ing he obse ed sp ing bloom in May imp o es. The summe e olu ion o chl-a is be e esol ed
as is he inc ease seen in Sep embe (Figu e 9a).
The imp o emen seen in chl-a do no howe e ansla e o a be e i in modeled o al biomass,
and he assimila ion o chl-a deg ades he model solu ion as a esul o he la ge miss- ep esen a ion o
he sp ing bloom (no shown). Indeed, when he esul s be ween 15 Ma ch and 15 Ap il a e emo ed,
he co ela ion o o al biomass inc eases o 0.61 a he 95% signi icance. One e ec o he assimila ion
is he o e es ima ion o all he nu ien a iables du ing he sp ing bloom (no shown). This e lec he
inabili y o he model o lowe he chl-a ield wi hou inc easing he nu ien alues du ing ha pe iod.
The disc epancies and excess nu ien s a e howe e co ec ed du ing he summe pe iod and ha e a
small impac on he o e all co ela ion wi h obse a ions (Table 1).
I is, howe e , no ewo hy ha while he co ela ions wi h obse a ions o all a iables in he
ca bona e sys em (pH, DIC and pCO
2
, Table 1) do no imp o e, he bias o RMSE is signi ican ly
educed (Table A4). The ensemble a iance was educed in he modeled chl-a as a esul o he
assimila ion bu i is appa en in Figu e 9b ha his is no he case o pCO
2
o DIC and pH (no shown),
al hough a e he ini ial inc ease du ing he i s h ee mon hs o he expe imen , i is also ue ha i
did no inc ease mono onically.
Remo e Sens. 2020,12, 2038 18 o 29
Figu e 9.
Time e olu ion o esul s om Expe imen D o (
a
) 10 m chl-a and in si u su ace chl-a
obse a ions ( ed do s) and (
b
) mixed laye pCO
2
and in si u obse a ions (black). The solid black line
co esponds o he model esul s om Expe imen A. The hicke blue line co esponds o he ensemble
mean while he hin blue line co esponds o one s anda d de ia ion.
Remo e Sens. 2020,12, 2038 19 o 29
4. Discussion
4.1. Ecosys em Models P edic i e Capabili ies o CO2
We ha e shown ha a 1D app oach using GOTM–ERSEM cap u es he seasonali y o he pelagic
ecosys em a L4, including he a iables d i ing he ai –sea exchange o CO2(e.g., yea 2008 Table 1).
The esul s om yea 2009 (Table 1) sugges ha he model’s abili y o simula e he o al biomass and
nu ien s is no a su icien condi ion o esol e he su ace e olu ion o DIC and pCO
2
. This ollows
he poo e simula ion o chl-a in 2009 and sugges s ha an adequa e simula ion o chl-a and he
associa ed plank on p oduc ion and espi a ion a es is c i ical o a be e es ima ion o he CO
2
dynamics. The seasonali y and ange o simula ed DIC and pCO
2
ag ee wi h ecen obse a ions [
45
]
and sugges bo h a physical and biological con ol. The high p oduc ion du ing he sp ing and summe
pe iods (Figu e 5) p oduces unde sa u a ion o CO
2
(Figu e 10a) esul ing in a ne sink o CO
2
in o
he wa e s. I is wo h no ing ha he sho - e m a iabili y in ai –sea CO
2
luxes is as high as he
seasonal signal (Figu e 9a) and ha he summe blooms can d i e a simila exchange o he longe
sp ing bloom.
Despi e he model’s abili y o cap u e he seasonal cycle on all he a iables o he ca bona e
sys em he model seems o unde es ima e he sho - e m a iabili y seen in he da a (Figu e 9b).
This can be ela ed o he coa se empo al esolu ion o he wind o cing (six hou ly) (e.g., [
64
]) and he
e ec o la e al ad ec ion which is in insic in he da a and no esol ed in he model. A L4, a ia ions
in DIC d i e he obse ed a iabili y in he ca bona e sys em and hus pCO
2
as o al alkalini y has
e y low a iabili y [
45
]. The model solu ion sugges s ha L4 can be conside ed a neu al si e wi h
espec o CO
2
luxes o a small sou ce (Table 2) in con as o he esul s epo ed by Ki idis e al.
[45]
.
Table 2. Ai -sea lux annual in eg a ed alues o each o he expe imen s.
Expe imen Ai -Sea Flux o CO2in molC/m2/day
2008 2009
Exp A −1.8 −0.3
Exp B −2.8 −0.6
Exp C −1.8 −0.3
Exp D 1.3 ± 1.7
4.2. Sensi i i y o Modeled CO2 o SST Diu nal Cycle
The inc eased e ical esolu ion nea he su ace in Expe imen B enabled he simula ion o
he diu nal cycle in SST [
64
] which o e all had a small impac on he abili y o GOTM–ERSEM o
simula e he bulk p ope ies o he pelagic ecosys em a L4 as obse ed a 10 m (Table 1) bu a la ge
impac on he ai –sea luxes o CO
2
(Figu e 10a,b). The compa ison wi h SST measu ed a he L4 buoy
yielded a small dec ease in co ela ion ( om 0.94 o 0.93 a he 99.9% con idence le el) and a small
dec ease in he bias (see Table A4). Expe imen B did howe e imp o e he simula ion o pCO
2
, DIC
and pH in 2008. The la ges di e ences in he ca bona e sys em be ween expe imen s A and B in 2009
concen a ed in he i s 5 mon hs (Janua y o May, Figu e 10a) and in Expe imen B esul ed in an
inc eased supe sa u a ion and esul ing lux o he a mosphe e om Janua y o Ma ch. The ampli ude
o he daily a ia ions in he luxes emained ela i ely unchanged om June o Decembe (Figu e 10b)
d i en p ima ily by he daily luc ua ions in p oduc i i y. The di e ence was la ges du ing he low
p oduc i i y mon hs o Janua y o Ma ch and i was pa ly esponsible o he change in alue o he
ne lux es ima es.
Remo e Sens. 2020,12, 2038 20 o 29
Figu e 10.
Time e olu ion o (
a
) nea -su ace pCO
2
, (
b
) ai –sea lux o CO
2
and (
c
) wa e column
a e age o G oss P oduc ion and Communi y Respi a ion a io o Expe imen s A, B and D. The black
line in a 385 in (a) ep esen s he a mosphe ic pCO2 alue used in he simula ions.
Remo e Sens. 2020,12, 2038 21 o 29
The changes wi h espec o he e e ence expe imen (A) a e mos ly loca ed in he op 10 m and
dec ease wi h dep h in a signi ican numbe o a iables (Figu e 11). These include a e a iables like
espi a ion and p oduc ion as well as s a e model a iables like biomass o plank on and zooplank on.
These di e ences will be discussed in ela ion o h ee pe iods in 2009 de ined as a win e non-g owing
season (Janua y o Ma ch wi h a consis en g oss p oduc ion o communi y espi a ion (G/R) a io less
han one, Figu e 10c), he summe g owing season om Ap il o Augus (G/R a e aging one o abo e)
and he au umn non-g owing season wi h G/R again less han one.
Mos o he a iables e alua ed changed be ween 0.1% o 6% excep o g oss p oduc ion and
communi y espi a ion ha showed inc eases be ween 26 and 34% in all seasons. Expe imen B
displayed a small inc ease in empe a u e in he op 10m h oughou he yea (0.1%) inc easing o 0.5%
du ing he g owing season. To al de i us ( he sum o he h ee classes in ERSEM, small, medium and
la ge) dec eased in he win e (4%), while on a e aged did no change du ing summe and dec eased
by 5% in he au umn in Expe imen B. I adiance le els also inc eased in Expe imen B (5% in all
seasons) in esponse o a dec ease in biomass o 2% in win e and 6% in au umn).
In gene al, hough, pCO
2
inc eased in Expe imen B in bo h g owing and non-g owing season by
2%. Du ing he non-g owing season, his ansla es in o an inc ease elease o CO
2
o he a mosphe e
in bo h 2008 and 2009. Du ing he g owing season, he e is no one speci ic end, as he CO
2
exchange
wi h he a mosphe e depends on bo h he biological up ake o CO
2
as well as he wind. Du ing he
wo yea s e alua ed, in 2009 he e was no o e all change while in 2008, he e was a la ge elease o
CO2in Expe imen B.
4.3. Sensi i i y o Modeled CO2 o Su ace Su ac an s
The educ ion o he gas ans e eloci y (
KCO2
) as a unc ion o sea-su ace su ac an s had a
e y small impac on he in eg a ed annual CO
2
lux (Table 2) and bulk ecosys em a iables (Table 1)
in ou modeled si e compa ed o he esul s o expe imen s B and D (Sec ions 3.2 and 3.4). This is
bo h a consequence o ou choice o pa ame e iza ion (Sec ion 3.3) o he p oduc ion o su ac an
ma e ial and he en i onmen al cha ac e is ics o ou si e. The la ge a iabili y in phy oplank on g oss
p oduc ion (simila o phy oplank on ne p oduc ion in Figu e 5) implies ha he educ ion o
KCO2
is
only ac i e du ing a small p opo ion o he ime (
<
5% du ing 2008–2009) be ween Ap il–Sep embe
and esul s in small educ ions o
KCO2
(
<
1%). The high a e age wind speed (~8
ms−1
obse ed a L4
in 2008–2009 also limi s he impac o he su ac an s on he annual in eg a ed luxes. This con as s
wi h he indings o Pe ei a e al.
[52]
who epo ed an o e all 9% educ ion in he ne CO
2
lux
in eg a ed o e he en i e A lan ic Ocean basin in 2014. None heless, hei analysis showed ha a ou
la i ude (al hough no es ima es a e p esen ed o he UK shel ) he dec eased su ac an supp ession
was consis en ly less han 5% excep o July–Sep embe when i inc eased o 10%.
4.4. Sensi i i y o Modeled CO2 o he Assimila ion o chl-a
The assimila ion o chl-a had a signi ican impac in he ca bona e sys em bu mos ly by educing
he bias in DIC, pH o pCO
2
. The co ela ions de e io a ed o all h ee a iables bu he concen a ion
o he obse a ions a ound he May o Sep embe mon hs p ecludes a de ini i e assessmen o he
e ec o su ace chl-a assimila ion on simula ing he seasonali y o he ca bona e sys em. Expe imen
D also co ec ed he simula ion o chl-a and he nu ien s silica e and phospha e (Table 1).

Remo e Sens. 2020,12, 2038 22 o 29
Figu e 11.
Dis ibu ion wi h dep h o he a e aged 2008–2009 oo mean squa e e o be ween
Expe imen A and B o selec ed a iables. The alues a e pe cen age change wi h espec o he
maximum alue o each a iable. The highe alues in he op 10 m indica es di e ences be ween he
wo models a e concen a ed he e.
Simila o Expe imen B, Expe imen D had a signi ican impac on he op 10 m a e aged G and
R, pa icula ly du ing he win e (G inc eased 450%, R 20%) and in au umn (G inc eased 60% R 23%).
O all he a iables e alua ed, only plank on biomass had simila changes wi h a 900% inc ease in
win e compa ed o 70% inc ease in he au umn. The second la ges co ec ions co esponded o
he zooplank on biomass, which dec eased in win e (28%) and in summe (10%) while i inc eased
in au umn (53%). Gene ally, he g owing season saw mo e modes changes ac oss all a iables
be ween 3 and 12%. I is wo h emembe ing ha al hough he pe cen age inc eases a e signi ican
hey ep esen a small numbe as hey co espond o he non-g owing seasons (win e in pa icula )
when biomass le els and ca bon up ake and espi ed a e e y small. This is a u he indica ion ha
win e ecosys em dynamics and au umn ones a e he leas well cap u ed by his 1D model app oach.
One way o isualizing how he in o ma ion associa ed wi h chl-a is sp ead ac oss all o he model
a iables du ing he analysis s ep is o calcula e he ime e olu ion o he c oss-co ela ion ma ix
among all o he ensemble membe s (Figu e 12). Chl-a is gene ally always co ela ed wi h model
es ima es o phy oplank on p oduc ion and espi a ion and i s co ela ion wi h he biomass o he
unc ional ypes (including phy oplank on and zooplank on) a ies in ime. Fo example, chl-a is highly
co ela ed wi h dia oms du ing he sp ing bloom, while he co ela ion wi h he picoplank on inc eases
du ing he sho -li ed blooms o his g oup du ing he summe pe iod. Equally, he co ela ions
be ween chl-a and DIC, pH o pCO
2
a e ela i ely low and indica es s ong non-linea dynamics and
eedback be ween chl-a o biomass and he ca bona e chemis y. This sugges ha di ec co ec ions o
hese a iables we e ela i ely small in he EnKF analysis. Assimila ion o chl-a indi ec ly co ec ed
hese a iables in he o ecas s ep, h ough he e-ini ializa ion o he o he model a iables in he
Remo e Sens. 2020,12, 2038 23 o 29
analysis. The low co ela ions also explain why we do no see a educ ion in he ensemble sp ead
o hese a iables (e.g., Figu e 9b) The co ec ions o nea -su ace pCO
2
(Figu e 10b) esul ing om
he assimila ion o chl-a we e signi ican ac oss he seasonal cycle excep o June o Augus when
all expe imen s ag ee in hei simula ion o chl-a and p ima y p oduc ion. The co ec ions we e as
high as 100
µ
a m and ended o educe he alues o nea -su ace pCO
2
esul ing in ne sink o CO
2
o
1.3 molCm
−2
yea
−1
in ag eemen wi h he es ima es o Ki idis e al.
[45]
. The la ges co ec ion wi h a
signi ican impac on he annual es ima es ela es o he p ima y p oduc ion co ec ions in Sep embe
o Decembe .
Figu e 12.
Time e olu ion o 2009 o he c oss-co ela ion ma ix ac oss ensemble membe s o
Expe imen D. The colo s indica e he co ela ion o each a iable wi h he op 10 m mean o o al chl-a.
5. Conclusions
We ha e shown ha he 1D GOTM–ERSEM model has skill in simula ing bo h he pelagic
ecosys em and he a iables ela ed o he dynamics o CO
2
a he coas al s a ion L4, al hough he
skill a ies om yea o yea . Resol ing he nea -su ace empe a u e g adien s induced by he
diu nal hea ing cycle (Expe imen B) has he mos e icien (lowes compu a ional cos and skill
equi emen s) impac a sho , seasonal and annual ime scales and esul s in imp o emen s ac oss
he e alua ed a iables ha a e compa able o he esul s om he assimila ion o chl-a Expe imen
(D). The co ela ion o ca bona e sys em a iables (DIC, pCO
2
and pH) imp o e by up o 19% while
annually in eg a ed alues o CO
2
exchange wi h he a mosphe e changed by up o 50% wi h espec
o he e e ence simula ion, inc easing he ne sou ce na u e o L4. A he L4 si e, he in oduc ion
o a dependence o he CO
2
gas ans e eloci y on g oss p oduc ion as a p oxy o he p oduc ion
o su ace slicks (Expe imen C) has a non-signi ican impac on he dynamics o CO
2
, including he
ai –sea annual exchange es ima es.
The assimila ion o su ace chl-a (Expe imen D) has an o e all mixed e ec on he abili y o he
model o ep oduce he obse a ions. As expec ed he simula ion o chl-a imp o es and while he
co ela ion o nu ien s do no signi ican ly change, hei bias dec eases. The e ec on he ca bona e
sys em a iables is simila ; he bias dec eases o DIC and pH bu so do he co ela ions. The la ges
impac o assimila ion is on he ne ai –sea gas exchange whe e he esul s indica e he po en ial o
L4 o be a weak ne sou ce o ca bon o he A mosphe e (1.3
±
1.7 molCm
−2
yea
−1
). Tha chl-a
Remo e Sens. 2020,12, 2038 24 o 29
assimila ion can deg ade he simula ion o some a iables possibly ela es o he equi emen o
succinc pa ame e changes o success ully simula e he speci ic yea s conside ed in his wo k.
The esul s o he h ee sensi i i y expe imen s indica e ha esol ing he empe a u e p o ile
nea he su ace appea s he mos impo an aspec o hose in es iga ed he e o a skilled simula ion
o he biogeochemis y a L4. This esul indica es ha his ine e ical scale should be included in
u u e shel sea models used in ai –sea lux modeling s udies.
Au ho Con ibu ions:
Concep ualiza ion, R.T. and J.S.; me hodology, R.T. and S.C.; so wa e, R.T., M.R.-V., Y.A.,
L.P.; alida ion, V.K., E.M.S.W., T.S., J.F., G.T. and V.M.; o mal analysis, R.T.; in es iga ion, R.T., M.R.-V.; esou ces,
R.T., V.K., E.M.S.W., C.W., J.F., T.S., V.M.; da a cu a ion, R.T., V.K., E.M.S.W., C.W., J.F., M.R.-V.; w i ing—o iginal
d a p epa a ion, R.T.; w i ing— e iew and edi ing, R.T.; p ojec adminis a ion, J.S.; unding acquisi ion, J.S.
and R.T. All au ho s ha e ead and ag eed o he published e sion o he manusc ip .
Funding:
This wo k was unded by he Eu opean Space Agency (ESA) h ough he OceanFlux G eenhouse Gases
p ojec (con ac numbe 4000104762/11/I-AM), NERC Na ional Capabili y Modeling and he UK Na ional Cen e
o Ea h Obse a ion. GT was suppo ed by he Eu opean Union unded con ac In o ma ion Sys em on he
Eu ophica ion o ou Coas al Seas (ISECA) (Con ac no. 07-027-FR-ISECA) unded by INTERREG IVA 2 Me s
Seas Zeeen C oss-bo de Coope a ion P og amme 2007–2013.
Acknowledgmen s:
We would like o hank he c ew o he RV Plymou h Ques o hei assis ance du ing
ieldwo k and all o hose ha con ibu e o he ou ine analysis o da a om he Wes e n Channel Obse a o y.
Thanks o Ca olyn Ha is o he nu ien analysis o L4 samples. All in si u da a a e a ailable om he B i ish
Oceanog aphic Da a Cen e (www.bodc.ac.uk). Sa elli e ocean colo and sea-su ace empe a u e (SST) we e
ob ained om he NERC NEODAAS se ice hos ed a PML.
Con lic s o In e es : The au ho s decla e no con lic o in e es .
Appendix A
Table o all s a is ics pe o med in he expe imen s
Table A1.
Pea son co ela ions o expe imen s A–D, be ween obse ed and modeled a iables o
2008–2009 a he shallowes dep h a ailable. Signi ican co ela ions a he 95% a e indica ed by *.
Exp Yea Va iable Dep h Co pValue RMSE Mean S d N
Exp A 2008 Biomass 10 0.70 * 0.000 99.59 21.91 42.81 40
Exp A 2009 Biomass 10 0.76 * 0.000 119.83 17.85 33.54 41
Exp B 2008 Biomass 10 0.67 * 0.000 98.91 19.95 37.18 40
Exp B 2009 Biomass 10 0.75 * 0.000 120.41 17.36 34.76 41
Exp C 2008 Biomass 10 0.70 * 0.000 99.71 21.99 43.05 40
Exp C 2009 Biomass 10 0.76 * 0.000 119.83 17.85 33.54 41
Exp D 2009 Biomass 10 0.49 * 0.001 119.58 23.32 37.84 42
Exp A 2009 Chl-a 0 0.47 * 0.003 1.28 0.78 0.95 38
Exp B 2009 Chl-a 0 0.50 * 0.002 1.29 0.81 1.00 38
Exp C 2009 Chl-a 0 0.47 * 0.003 1.28 0.78 0.95 38
Exp D 2009 Chl-a 0 0.46 * 0.003 1.39 0.95 1.15 40
Exp A 2008 Chl-a 10 0.43 * 0.005 2.42 0.65 0.87 41
Exp B 2008 Chl-a 10 0.41 * 0.008 2.42 0.66 0.88 41
Exp C 2008 Chl-a 10 0.43 * 0.005 2.42 0.66 0.88 41
Exp A 2009 Chl-a 10 0.45 * 0.004 1.27 0.76 0.95 39
Exp B 2009 Chl-a 10 0.48 * 0.002 1.28 0.79 1.00 39
Exp C 2009 Chl-a 10 0.45 * 0.004 1.27 0.76 0.95 39
Exp D 2009 Chl-a 10 0.50 * 0.002 1.29 0.95 1.09 37
Exp A 2009 Chl-a 50 0.63 * 0.000 0.45 0.29 0.46 33
Exp B 2009 Chl-a 50 0.63 * 0.000 0.45 0.28 0.45 33
Exp C 2009 Chl-a 50 0.63 * 0.000 0.45 0.29 0.46 33
Exp D 2009 Chl-a 50 0.32 0.054 0.46 0.30 0.21 36
Exp A 2009 P ima y p oduc ion 10 0.40 * 0.015 23.51 17.50 23.95 36
Exp B 2009 P ima y p oduc ion 10 0.39 * 0.018 17.22 10.34 15.05 36
Exp C 2009 P ima y p oduc ion 10 0.40 * 0.015 23.51 17.50 23.95 36
Exp D 2009 P ima y p oduc ion 10 0.42 * 0.010 32.16 23.81 32.09 38
Remo e Sens. 2020,12, 2038 25 o 29
Table A2.
Pea son co ela ions o expe imen s A–D, be ween obse ed and modeled a iables o
2008–2009 a he shallowes dep h a ailable. Signi ican co ela ions a he 95% a e indica ed by *.
Exp Yea Va iable Dep h Co pValue RMSE Mean S d N
Exp A 2008 Dia oms 10 0.66 * 0.000 53.38 16.11 38.63 40
Exp A 2009 Dia oms 10 0.62 * 0.000 27.65 12.21 30.43 41
Exp B 2008 Dia oms 10 0.64 * 0.000 49.85 14.37 33.20 40
Exp B 2009 Dia oms 10 0.61 * 0.000 29.48 12.18 32.00 41
Exp C 2008 Dia oms 10 0.66 * 0.000 53.55 16.18 38.84 40
Exp C 2009 Dia oms 10 0.62 * 0.000 27.64 12.21 30.43 41
Exp D 2009 Dia oms 10 0.13 0.415 33.21 14.93 32.09 42
Exp A 2008 Flagella es 10 0.29 0.067 6.34 3.49 5.06 40
Exp A 2009 Flagella es 10 0.76 * 0.000 7.23 3.02 3.81 41
Exp B 2008 Flagella es 10 0.29 0.068 6.18 3.22 4.54 40
Exp B 2009 Flagella es 10 0.76 * 0.000 7.38 2.83 3.56 41
Exp C 2008 Flagella es 10 0.29 0.071 6.35 3.51 5.09 40
Exp C 2009 Flagella es 10 0.76 * 0.000 7.23 3.02 3.81 41
Exp D 2009 Flagella es 10 0.48 * 0.001 7.34 5.12 7.45 42
Table A3.
Pea son co ela ions o expe imen s A–D, be ween obse ed and modeled a iables o
2008–2009 a he shallowes dep h a ailable. Signi ican co ela ions a he 95% a e indica ed by *.
Exp Yea Va iable Dep h Co pValue RMSE Mean S d N
Exp A 2008 Phospha e 1 0.84 * 0.000 0.17 0.39 0.16 39
Exp A 2009 Phospha e 1 0.90 * 0.000 0.19 0.39 0.16 41
Exp B 2008 Phospha e 1 0.87 * 0.000 0.19 0.42 0.16 39
Exp B 2009 Phospha e 1 0.90 * 0.000 0.21 0.42 0.16 41
Exp C 2008 Phospha e 1 0.84 * 0.000 0.17 0.39 0.16 39
Exp C 2009 Phospha e 1 0.90 * 0.000 0.19 0.39 0.16 41
Exp D 2009 Phospha e 1 0.88 * 0.000 0.24 0.47 0.18 43
Exp A 2008 Ni a e 1 0.90 * 0.000 2.29 5.04 2.68 40
Exp A 2009 Ni a e 1 0.90 * 0.000 2.54 4.99 2.79 42
Exp B 2008 Ni a e 1 0.92 * 0.000 2.66 5.50 2.65 40
Exp B 2009 Ni a e 1 0.88 * 0.000 2.89 5.37 2.75 42
Exp C 2008 Ni a e 1 0.90 * 0.000 2.29 5.04 2.68 40
Exp C 2009 Ni a e 1 0.90 * 0.000 2.54 4.99 2.79 42
Exp D 2009 Ni a e 1 0.90 * 0.000 2.40 4.98 3.00 44
Exp A 2008 Ammonium 1 −0.23 0.156 0.59 0.46 0.35 39
Exp A 2009 Ammonium 1 −0.46 * 0.002 0.63 0.46 0.38 41
Exp B 2008 Ammonium 1 −0.35 * 0.031 0.70 0.51 0.45 39
Exp B 2009 Ammonium 1 −0.46 * 0.003 0.71 0.51 0.48 41
Exp C 2008 Ammonium 1 −0.23 0.156 0.59 0.47 0.36 39
Exp C 2009 Ammonium 1 −0.46 * 0.002 0.63 0.46 0.38 41
Exp D 2009 Ammonium 1 −0.40 * 0.008 0.67 0.61 0.40 43
Exp A 2008 Silica e 1 0.82 * 0.000 1.95 4.42 2.04 37
Exp A 2009 Silica e 1 0.86 * 0.000 2.17 4.47 1.97 41
Exp B 2008 Silica e 1 0.83 * 0.000 1.99 4.49 1.94 37
Exp B 2009 Silica e 1 0.86 * 0.000 2.15 4.45 1.98 41
Exp C 2008 Silica e 1 0.82 * 0.000 1.95 4.42 2.03 37
Exp C 2009 Silica e 1 0.86 * 0.000 2.17 4.47 1.97 41
Exp D 2009 Silica e 1 0.86 * 0.000 1.73 4.15 2.04 43