Additional processes and diagnostic variables in the CSIRO ...
Transcript of Additional processes and diagnostic variables in the CSIRO ...
Additional processes and diagnostic variables in the CSIROEnvironmental Modelling Suite
Mark E. Baird1, Matthew P. Adams2, John Andrewartha1, Mike Herzfeld1, EmlynJones1, Nugzar Margvelashvili1, Mathieu Mongin1, Farhan Rizwi1, Barbara
Robson5, Jenny Skerratt1, Andrew D. L. Steven1, Karen A. Wild-Allen1, andMonika Soja-Wozniak1
1CSIRO, Oceans and Atmosphere, Hobart, Australia2School of Chemical Engineering, The University of Queensland, Brisbane,
Australia3Plant Functional Biology and Climate Change Cluster, Faculty of Science,
University of Technology Sydney, Sydney, Australia4CSIRO, Land and Water, Canberra, Australia
5Australian Institute of Marine Science6Bureau of Meteorology
October 5, 2020
1
0.1 Outline of document
A comprehensive description of the EMS optical and biogeochemical version B3p0 is given in Bairdet al. (2020), as published in 2020. This document provides scientific description of processes addedsince this work (Sec. 1), as well as the calculation of diagnostic variables that were not describedin Baird et al. (2020), including simulated true color (Sec. 2).
1 Processes added after GMD submission
1.1 Anammox
Anaerobic ammonium oxidation (anammox) is a globally important microbial process of the nitrogencycle that takes place in many natural environments. It is the combination of nitrite and ammoniumto make water and di-nitrogen gas. As the model does not explicitly represent nitrite, we assumethat anammox is the combination of nitrate and ammonium to make water and di-nitrogen gas:
Anammox : 2H+ + NH+4 + NO−3 −→ N2 + 3H2O (1)
The excess oxygen molecule in nitrate compared to nitrate results in the generation of 3 watermolecules, and the need for two hydrogen ions in the reactants. Because neither hydrogen ions noroxygen bound in water molecules are included in the model’s mass budget, anammox results in aloss of DIN and oxygen from the model.
We assume that anammox proceeds at a rate proportional the minimum concentration of the tworeactants:
∂DIN
∂t= ranamin [[NH4], [NO3]] exp (−[O2]/KO2,ana) (2)
where rana is the temperature-dependent rate coefficient for anammox. The exponential term onthe right side of Eq. 2 results in the anammox proceeding at its maximum rate when the dissolvedoxygen concentration is zero. The parameter KO2,ana is the oxygen concentration at which anammoxproceeds at 1/e, or 37 % of its maximum rate. At oxygen saturation anammox is greatly reduced.
2
Variable Symbol UnitsAmmonium concentration [NH4] mg N m−3
Nitrate concentration [NO3] mg N m−3
Dissolved oxygen concentration [O2] mg O m−3
Table 1: State and derived variables for anammox.
Description Symbol UnitsMaximum rate of anammox in the water column rana,wc 0.1 d−1
Maximum rate of anammox in the sediment rana,sed 0.1 d−1
Oxygen half-saturation constant for nitrification KO2,ana 500 mg O m−3
Table 2: Constants and parameter values used in the anammox process.
(3)
∂[NH4]
∂t= −1
2ranamin [[NH4], [NO3]] exp (−[O2]/KO2,ana) (4)
∂[NO3]
∂t= −1
2ranamin [[NH4], [NO3]] exp (−[O2]/KO2,ana) (5)
∂[O2]
∂t= −3
2
16
14ranamin [[NH4], [NO3]] exp (−[O2]/KO2,ana) (6)
Table 3: Equations for Anammox.
3
2D
iagnost
icoutp
uts
Nam
eU
nit
sD
escr
ipti
onA
bso
rpti
onat
440
nm
(at
440)
m−1
Tot
alab
sorp
tion
due
tocl
ear
wat
er,
CD
OM
,m
icro
alga
ean
dsu
s-p
ended
sedim
ents
at44
0nm
.Sca
tter
ing
at55
0nm
(bt
550)
m−1
Tot
alsc
atte
ring
due
tocl
ear
wat
er,
mic
roal
gae
and
susp
ended
sed-
imen
tsat
550
nm
.V
erti
cal
atte
nuat
ion
at49
0nm
(Kd
490)
m−1
Ver
tica
lat
tenuat
ion
(alo
ngz
axis
not
alon
gze
nit
han
gle)
ofligh
tat
490
nm
.A
vera
geP
AR
inla
yer
(PA
R)
mol
phot
onm−2
s−1
Mea
nsc
alar
phot
osynth
etic
ally
avai
lable
radia
tion
(400
-70
0nm
)w
ithin
the
laye
r.D
ownw
elling
PA
R(P
AR
z)m
olphot
onm−2
s−1
Mea
ndow
nw
elling
phot
osynth
etic
ally
avai
lable
radia
tion
(400
-70
0nm
)at
the
top
ofea
chla
yer
(i.e
.on
zgr
idnot
zce
ntr
e)L
ight
inte
nsi
tyab
ove
seag
rass
(EpiP
AR
sg)
mol
phot
onm−2
d−1
Mea
ndow
nw
elling
phot
osynth
etic
ally
avai
lable
radia
tion
(400
-70
0nm
)ab
ove
seag
rass
canop
y.L
ight
inte
nsi
tyab
ove
sedim
ent
laye
r(E
piP
AR
)m
olphot
onm−2
d−1
Mea
ndow
nw
elling
phot
osynth
etic
ally
avai
lable
radia
tion
(400
-70
0nm
)at
the
sedim
ent-
wat
erin
terf
ace.
Ver
tica
lat
tenuat
ion
ofhea
t(K
hea
t)m−1
Ver
tica
lat
tenuat
ion
(alo
ngz
axis
not
alon
gze
nit
han
gle)
ofhea
ten
ergy
.R
emot
e-se
nsi
ng
reflec
tance
(RX
)sr−1
Rem
ote-
sensi
ng
reflec
tance
atw
avel
engt
hX
XX
nm
,Rrs,X
.Rrs
isth
era
tio
ofw
ater
-lea
vin
gra
dia
nce
(i.e
.p
erso
lid
angl
e)to
the
inco
min
gso
lar
irra
dia
nce
(i.e
.fr
omal
ldir
ecti
ons)
.Sat
ellite
chlo
rophyll
(OC
*)m
gm−3
Outp
ut
ofban
dra
tio
chol
orop
hyll
algo
rith
ms
(OC
3M,
OC
3V)
Tab
le4:
Lon
gnam
e(a
nd
vari
able
nam
e)in
model
outp
ut
file
s,unit
s,an
ddes
crip
tion
ofop
tica
ldia
gnos
tic
vari
able
s.U
nle
ssot
her
wis
est
ated
quan
titi
esar
ece
llce
ntr
edve
rtic
alav
erag
es.
4
Nam
eU
nit
sD
escr
ipti
onB
icar
bon
ate
(HC
O3)
mm
olm−3
Con
centr
atio
nof
bic
arb
onat
eio
ns
[HC
O− 3
]ca
lcula
ted
from
carb
onch
emis
try
equlibra
atw
ater
colu
mn
valu
esofT
,S
,DIC
andAT
.C
arb
onat
e(C
O3)
mm
olm−3
Con
centr
atio
nof
carb
onat
eio
ns
[CO
2−
3]
calc
ula
ted
from
carb
onch
emis
try
equlibra
atw
ater
colu
mn
valu
esofT
,S
,DIC
andAT
.A
rago
nit
esa
tura
tion
stat
e(o
meg
aar
)-
Ara
gonit
esa
tura
tion
stat
e(Ω
a)
calc
ula
ted
from
DIC
,AT
,T
and
Sas
sum
ing
the
carb
onat
esy
stem
isat
equilib
rium
.O
xyge
n%
satu
rati
on(O
xy
sat)
%D
isso
lved
oxyge
nco
nce
ntr
atio
nas
ap
erce
nta
geof
the
satu
rati
onco
nce
ntr
atio
nat
atm
ospher
icpre
ssure
and
loca
lT
andS
.pH
(PH
)lo
g10
mol
m−3
pH
bas
edon
[H+
]ca
lcula
ted
from
carb
onch
emis
try
equlibra
atw
ater
colu
mn
valu
esofT
,S
,DIC
andAT
.Sea
-air
CO
2flux
(CO
2flux)
mg
Cm−2
s−1
Flu
xof
carb
onfr
omse
ato
air
(pos
itiv
efr
omse
ato
air)
.T
he
valu
eis
give
nin
the
laye
rin
whic
hit
was
dep
osit
ed(m
ust
be
thic
ker
than
20cm
),but
still
repre
sents
anar
eal
flux.
Sea
-air
O2
flux
(O2
flux)
mg
Om−2
s−1
Flu
xof
oxyge
nfr
omse
ato
air
(pos
itiv
efr
omse
ato
air)
.T
he
valu
eis
give
nin
the
laye
rin
whic
hit
was
dep
osit
ed(m
ust
be
thic
ker
than
20cm
),but
still
repre
sents
anar
eal
flux.
Del
tapC
O2
(dco
2sta
r)ppm
vP
arti
alpre
ssure
ofC
O2
inth
eoce
anm
inus
that
ofth
eat
mos
pher
e(3
96ppm
v).
Oce
anic
pC
O2
(pco
2surf
)ppm
vP
arti
alpre
ssure
ofC
O2
inth
eoce
an.
Tab
le5:
Lon
gnam
e(a
nd
vari
able
nam
e)in
model
outp
ut
file
s,unit
s,an
ddes
crip
tion
ofga
sdia
gnos
tic
vari
able
s.U
nle
ssot
her
wis
est
ated
quan
titi
esar
ece
llce
ntr
edve
rtic
alav
erag
es.
5
Nam
eU
nit
sD
escr
ipti
onT
otal
C,
N,
P(T
C,
TN
,T
P)
mg
m−3
Sum
ofdis
solv
edan
dpar
ticu
late
C,
N,
and
P.
Tot
alch
loro
phyll
a(C
hl
asu
m)
mg
m−3
Sum
ofch
loro
phyll
conce
ntr
atio
nof
the
four
mic
roal
gae
typ
es(∑ n
c iVi)
.E
colo
gyF
ine
Inor
ganic
s(E
FI)
kg
m−3
Sum
ofin
orga
nic
com
pon
ents
(see
ecologysetup.txt
for
thos
esp
ecifi
edin
the
code)
use
dfo
rT
SS-d
epen
den
tca
lcu-
lati
ons
such
asphos
phor
us
abso
rpti
on.
Eco
logy
Par
ticu
late
Org
anic
s(E
PO
)kg
m−3
Wei
ght
ofca
rbon
inm
icro
alga
e,zo
opla
nkto
n,
and
par
ticu
late
det
ritu
s.L
arge
phyto
pla
nkto
nnet
pro
-duct
ion
(PhyL
Npr)
mg
Cm−3
d−1
Rat
eof
larg
ephyto
pla
nkto
nor
ganic
mat
ter
synth
esis
from
inor
ganic
const
ituen
ts.
Sm
all
phyto
pla
nkto
nnet
pro
-duct
ion
(PhyS
Npr)
mg
Cm−3
d−1
Rat
eof
smal
lphyto
pla
nkto
nor
ganic
mat
ter
synth
esis
from
inor
ganic
const
ituen
ts.
Trichodesmium
net
pro
duct
ion
(Tri
cho
Npr)
mg
Cm−3
d−1
Rat
eof
Trichodesmium
orga
nic
mat
ter
synth
esis
from
inor
-ga
nic
const
ituen
ts.
Mic
rophyto
ben
thos
net
pro
-duct
ion
(MP
BN
pr)
mg
Cm−3
d−1
Rat
eof
mic
rophyto
ben
thos
orga
nic
mat
ter
synth
esis
from
in-
orga
nic
const
ituen
ts.
Tot
alphyto
pla
nkto
nnet
pro
-duct
ion
(Ppro
d)
mg
Cm−3
d−1
Rat
eof
tota
lphyto
pla
nkto
nor
ganic
mat
ter
synth
esis
from
inor
ganic
const
ituen
ts.
Lar
gezo
opla
nkto
nre
mov
alra
tefr
omsm
all
zoop
lankto
nm
gC
m−3
d−1
Rat
eof
carn
ivor
yof
larg
ezo
opla
nkto
non
smal
lzo
opla
nkto
n.
Sm
all
zoop
lankto
nre
mov
alra
tefr
omsm
all
phyto
pla
nkto
nm
gC
m−3
d−1
Sec
ondar
ypro
duct
ion
ofsm
all
zoop
lankto
n.
Lar
gezo
opla
nkto
nre
mov
alra
tefr
omla
rge
phyto
pla
nkto
nm
gC
m−3
d−1
Sec
ondar
ypro
duct
ion
ofla
rge
zoop
lankto
n.
Zoop
lankto
nto
tal
graz
ing
(Zgr
azin
g)m
gC
m−3
d−1
Tot
alse
condar
yp
elag
icpro
duct
ion
-fe
edin
gof
bot
hzo
o-pla
nkto
ncl
asse
son
phyto
pla
nkto
n.
N2
fixat
ion
(Nfix)
mg
Nm−3
s−1
Nit
roge
nfixat
ion
byTrichodesmium
.T
rich
odes
miu
mse
ttling
velo
c-it
y(T
rich
osv
)m
s−1
Ver
tica
lsi
nkin
gra
teof
Trichodesmium
.
Tab
le6:
Lon
gnam
e(a
nd
vari
able
nam
e)in
model
outp
ut
file
s,unit
s,an
ddes
crip
tion
ofp
elag
icdia
gnos
tic
vari
able
s.
6
Nam
eU
nit
sD
escr
ipti
onT
otal
epib
enth
icC
,N
,P
(EpiT
C,
EpiT
N,
EpiT
P)
mg
m−2
Sum
ofC
,N
,an
dP
inep
iben
thic
pla
nts
and
cora
ls.
Halop
hila
pro
duct
ion
(SG
HN
pr)
gN
m−2
d−1
Gro
sspro
duct
ion
ofnit
roge
nin
Halop
hila
,w
her
eN
fluxes
are
not
acco
unte
dfo
rin
resp
irat
ion.
Halop
hila
grow
thra
te(S
GH
Ngr
)s−
1T
urn
over
tim
eof
abov
e-gr
oundHalop
hila
bio
mas
s.
Sea
gras
sgr
owth
rate
(SG
Ngr
)s−
1T
urn
over
tim
eof
abov
e-gr
oundZostera
bio
mas
s.
Sea
gras
spro
duct
ion
(SG
Npr)
gN
m−2
d−1
Gro
sspro
duct
ion
ofnit
roge
nin
Zostera
,w
her
eN
fluxes
are
not
acco
unte
dfo
rin
resp
irat
ion.
Sea
gras
ssh
ear
mor
t.(S
Gsh
ear
mor
t)d−1
Los
sra
teof
seag
rass
due
tosh
ear
stre
ssm
orta
lity
.
Dee
pSG
shea
rm
ort.
(SG
Dsh
ear
mor
t)d−1
Los
sra
teof
dee
pse
agra
ssdue
tosh
ear
stre
ssm
orta
lity
.
Hal
ophila
shea
rm
ort.
(SG
Hsh
ear
mor
t)d−1
Los
sra
teof
Halop
hila
due
tosh
ear
stre
ssm
orta
lity
.
Dee
pse
agra
ssgr
owth
rate
(SG
DN
gr)
s−1
Turn
over
tim
eof
abov
e-gr
ound
dee
pse
agra
ssbio
mas
s.
Dee
pse
agra
sspro
duct
ion
(SG
DN
pr)
gN
m−2
d−1
Gro
sspro
duct
ion
ofnit
roge
nin
dee
pse
agra
ss,
wher
eN
fluxes
are
not
acco
unte
dfo
rin
resp
irat
ion.
Mac
roal
gae
grow
thra
te(M
AN
gr)
s−1
Turn
over
tim
eof
mac
roal
gae
bio
mas
s.
Mac
roal
gae
pro
duct
ion
(MA
Npr)
gN
m−2
d−1
Gro
sspro
duct
ion
ofnit
roge
nin
mac
roal
gae,
wher
eN
fluxes
are
not
acco
unte
dfo
rin
resp
irat
ion.
Tab
le7:
Lon
gnam
e(a
nd
vari
able
nam
e)in
model
outp
ut
file
s,unit
san
ddes
crip
tion
ofb
enth
icdia
gnos
tic
vari
able
s.
7
Nam
eU
nit
sD
escr
ipti
onC
oral
inor
ganic
sup-
ply
(Cor
alIN
up)
mg
Nm−2
s−1
Flu
xof
dis
solv
edin
orga
nic
nit
roge
nin
toco
ral
pol
yps
from
the
wat
erco
lum
n,
ab-
sorb
edby
zoox
anth
ella
e.N
ote
that
the
flux
isav
erag
edov
erth
ew
hol
egr
idce
ll,
while
the
frac
tion
ofth
eb
otto
mth
atis
resp
onsi
ble
for
this
upta
keis
give
nbyAeff
(Eq.??)
Cor
alor
ganic
supply
(Cor
alO
Nup)
mg
Nm−2
s−1
Flu
xof
par
ticu
late
orga
nic
nit
roge
nin
toco
ral
pol
yps
from
the
wat
erco
lum
n,
con-
sum
edby
the
cora
lhos
t.See
cora
lor
ganic
supply
.N
etca
lcifi
cati
on(G
net
)m
gC
m−2
s−1
Net
calc
ifica
tion
(cal
cifica
tion
min
us
dis
solu
tion
)at
the
sedim
ent-
wat
erco
lum
nin
-te
rfac
ele
adin
gto
ach
ange
inth
ew
ater
colu
mn
pro
per
ties
.R
ubis
coac
tivit
y,a∗ Q
ox
(CS
tem
pfu
nc)
-A
ctiv
ity
ofth
eR
ibulo
se-1
,5-b
isphos
phat
eca
rbox
yla
se/o
xyge
nas
e(R
uB
isC
O)
en-
zym
ebas
edon
the
seab
edte
mp
erat
ure
anom
aly
from
asp
atia
lly-v
aryin
gm
axim
um
sum
mer
tem
per
ature
.It
take
sa
valu
eof
1b
elow
the
sum
mer
tim
em
axim
um
,an
d0
at2
Cab
ove
(Eq.??).
Cor
alble
achin
gra
te(C
Sble
ach)
d−1
Los
sra
teof
zoox
anth
ella
edue
tore
acti
veox
yge
nst
ress
.
Cor
alm
ucu
sre
leas
e(m
ucu
s)m
gN
m−2
s−1
Rel
ease
rate
ofor
ganic
mat
ter
(at
the
Red
fiel
dra
tio)
due
top
olyp
and
sym
bio
nt
mor
atlity
term
s,ex
cludin
gth
ose
due
tore
acti
veox
yge
nst
ress
dri
ven
expuls
ion.
Tab
le8:
Lon
gnam
e(a
nd
vari
able
nam
e)in
model
outp
ut
file
s,unit
san
ddes
crip
tion
ofco
ral
dia
gnos
tic
vari
able
s.
8
Nam
eU
nit
sD
escr
ipti
onP
oros
ity
(por
os-
ity
sed,φ
)-
Fra
ctio
nof
ala
yer
by
volu
me
that
isocc
upie
dby
wat
er.
Cal
cula
ted
from
the
the
sum
ofth
em
ass
div
ided
by
Sed
imen
tden
sity
ofea
chof
the
par
ticu
late
com
pon
ents
,co
mpar
edto
that
ofw
ater
.Sea
-wat
erse
d*
flux
(*se
dflux)
mg
m−2
The
sea-
wat
er(+
vein
tow
ater
)flux
ofdiff
use
dtr
acer
*ca
lcula
ted
by
the
ecol
ogic
alpro
cess
diffusionepi.
This
diff
usi
onra
te,
typ-
ical
lym
uch
quic
ker
than
the
sedim
ent
model
,ac
counts
for
the
fast
diff
usi
onb
etw
een
the
top
sedim
ent
laye
ran
dth
eb
otto
mSea
-wat
erse
d*
flux
(*fluxse
di
inst
)m
gm−2
The
sea-
wat
er(+
vein
tow
ater
)flux
ofdiff
use
dtr
acer
*ca
lcula
ted
by
the
sedim
ent
model
.T
his
calc
ula
tion
incl
udes
,en
trai
nm
ent
ofp
orew
ater
sduri
ng
sedim
ent
dep
osit
ion,
loss
ofp
orew
ater
duri
ng
eros
ion,
and
asl
owdiff
usi
onra
te.
Den
itrfi
cati
onflux
(Den
fl)
mg
Nm−2
d−1
laye
r−1
Den
itri
fica
tion
rate
,sp
ecifi
edp
erla
yer
per
m2
(rat
her
than
per
m3)
soth
atth
edep
th-i
nte
grat
edden
itri
fica
tion
rate
can
be
calc
ula
ted
by
sum
min
gea
chla
yer
wit
hou
tknow
ing
the
thic
knes
sof
each
laye
r(t
hat
vari
esw
ith
tim
e).
Tab
le9:
Lon
gnam
e(a
nd
vari
able
nam
e)in
model
outp
ut
file
s,unit
san
ddes
crip
tion
ofse
dim
ent
model
dia
gnos
tic
vari
able
s.
9
2.1 Diagnostic age tracer
Tracer ’age’ is a diagnostic tracer (Monsen et al., 2002; Macdonald et al., 2009) use to quantify thespatially-resolved residence time of water in different regions. The age tracer, τ , is advected anddiffused by the hydrodynamic model using the same numerical schemes as other tracers such assalinity. When inside the region of interest, the age increases at the rate of 1 d d−1. When the agetracer is outside the region of interest, its age decays (or anti-ages) at the rate of Φ d−1. Thus, thelocal rate of change over the whole domain is given by:
∂τ
∂t= 1 (7)
∂τ
∂t= −Φτ outside ageing region (8)
(9)
In some applications the age is held to zero outside the region to represent time since water movedwithin the area of interested (such as the surface mixed layer (Baird et al., 2006)). For moreinformation see Mongin et al. (2016).
2.2 Remote-sensing reflectance
2.3 Backscattering
Backscattering. In addition to the IOPs calculated above, the calculation of remote-sensing re-flectance uses a backscattering coefficient, bb, which has a component due to pure seawater, and acomponent due to algal and non-algal particulates. The backscattering ratio is a coarse resolutionrepresentation of the volume scattering function, and is the ratio of the forward and backwardscattering.
The backscattering coefficient for clear water is 0.5, a result of isotropic scattering of the watermolecule.
For inorganic particles, backscattering can vary between particle mineralogies, size, shape, and atdifferent wavelengths, resulting, with spectrally-varying absorption, in the variety of colours thatwe see from suspended sediments. Splitting sediment types by mineralogy only, the backscatteringratio for carbonate and non-carbonate particles is given in Table 10.
The backscatter due to phytoplankton is approximately 0.02. To account for a greater backscatter-ing ratio, and therefore backscatter, at low wavelengths (Fig. 4 of Vaillancourt et al. (2004)), we
10
Wavelength [nm]412.0 440.0 488.0 510.0 532.0 595.0 650.0 676.0 715.0
Carbonate 0.0209 0.0214 0.0224 0.0244 0.0216 0.0201 0.0181 0.0170 0.0164Terrestrial 0.0028 0.0119 0.0175 0.0138 0.0128 0.0134 0.0048 0.0076 0.0113
Table 10: Particulate backscattering ratio for carbonate and non-carbonate minerals based onsamples at Lucinda Jetty Coastal Observatory, a site at the interface on carbonate and terrestrialbottom sediment (Soja-Wozniak et al., 2019).
linearly increased the backscatter ratio from 0.02 at 555 nm to 0.04 at 470 nm. Above and below555 nm and 470 nm respectively the backscatter ratio remained constant.
The total backscatter then becomes:
bb,λ = bwbw,λ+b∗bphy,λn+ bb,NAPnon−CaCO3,λc1NAPnon−CaCO3 + bb,NAPCaCO3
,λc2NAPnon−CaCO3NAPCaCO3
(10)where the backscatter ratio of pure seawater, bw, is 0.5, n is the concentration of cells, and forparticulate matter (NAP and detritus), bb,NAP,λ, is variable (Table 10) and the coefficients c1 andc2 come from the total scattering equations above.
The particulate component of total scattering for phytoplankton is strongly related to cell carbon(and therefore cell size) and the number of cells (Vaillancourt et al., 2004):
b∗bphy,λ = 5× 10−15m1.002C (R2 = 0.97) (11)
where mC is the carbon content of the cells, here in pg cell−1.
2.3.1 Water column contribution to remote-sensing reflectance
The ratio of the backscattering coefficient to the sum of backscattering and absorption coefficientsfor the water column, uλ, is:
uλ =N∑z′=1
wλ,z′bb,λ,z′
aλ,z′ + bb,λ,z′(12)
where wλ,z′ is a weighting representing the component of the remote-sensing reflectance due to theabsorption and scattering in layer number z′, and N is the number of layers.
11
The weighting fraction in layer z′ is given by:
wλ,z′ =1
z1 − z0
(∫ z1
0
exp (−2Kλ,z) dz −∫ z0
0
exp (−2Kλ,z) dz
)(13)
=1
z1 − z0
∫ z1
z0
exp (−2Kλ,z) dz
(14)
where Kλ is the vertical attenuation coefficient at wavelength λ described above, and the factorof 2 accounts for the pathlength of both downwelling and upwelling light. The integral of wλ,z toinfinite depth is 1. In areas where light reaches the bottom, the integral of wλ,z to the bottom isless than one, and bottom reflectance is important (see Sec. 2.3.2).
The below-surface remote-sensing reflectance, rrs, is given by:
rrs,λ = g0uλ + g1u2λ (15)
where g0 = 0.0895 sr−1 (close to 1/4π) and g1 = 0.1247 sr−1 are empirical constants for the nadir-view in oceanic waters (Lee et al., 2002; Brando et al., 2012), and these constants result in a changeof units from the unitless u to a per unit of solid angle, sr−1, quantity rrs,λ.
The above-surface remote-sensing reflectance, through rearranging Lee et al. (2002), is given by:
Rrs,λ =0.52rrs,λ
1− 1.7rrs,λ(16)
At open ocean values, Rrs ∼ 0.06u sr−1. Thus if total backscattering and absorption are approxi-mately equal, u = 0.5 and Rrs ∼ 0.03 sr−1.
2.3.2 Benthic contribution to remote-sensing reflectance
In the water column optical (Sec 2.3.1), the calculation of remote-sensing reflectance required thecontribution to water-leaving irradiance from bottom reflection. In order to calculate the importanceof bottom reflectance, the integrated weighting of the water column must be calculated (Sec. 2.3.1),with the remaining being ascribed to the bottom. Thus, the weighting of the bottom reflectance asa component of surface reflectance is given by:
wλ,bot = 1− 1
zbot
∫ zbot
0
exp (−2Kλ,z′) dz′ (17)
where Kλ is the attenuation coefficient at wavelength λ described above, the factor of 2 accountsfor the pathlength of both downwelling and upwelling light.
12
The bottom reflectance between 400 and 800 nm of ∼ 100 substrates (including turtles and giantclams!) have been measured on Heron Island using an Ocean Optics 2000 (Roelfsema and Phinn,2012; Leiper et al., 2012). The data for selected substrates are shown in Fig 3 of Baird et al. (2016).When the bottom is composed of mixed communities, the surface reflectance is weighted by thefraction of the end members visible from above, with the assumption that the substrates are layeredfrom top to bottom by macroalgae, seagrass (Zostera then Halophila), corals (zooxanthellae thenskeleton), benthic microalgae, and then sediments. Since the sediment is sorted in the simulationby the sediment process, the sediments are assumed to be well mixed in surface sediment layer.Implicit in this formulation is that the scattering of one substrate type (i.e. benthic microalgae)does not contribute to the relectance of another (i.e. sand). In terms of an individual photon, itimplies that if it first intercepts substrate A, then it is only scattered and/or absorbed by A.
Calculation of bottom fraction.
The fraction of the bottom taken up by a benthic plant of biomass B is Aeff = 1−exp(−ΩBB), withexp(−ΩBB) uncovered. Thus the fraction of the bottom covered by macroalgae, seagrass (Zosterathen shallow and deep Halophila) and corals polyps is given by:
fMA = 1− exp(−ΩMAMA) (18)
fSG = (1− fMA) (1− exp(−ΩSGSG)) (19)
fSGH = (1− fMA − fSG) (1− exp(−ΩSGHSGH)) (20)
fSGD = (1− fMA − fSG − fSGH) (1− exp(−ΩSGDSGD)) (21)
fpolyps = (1− fMA − fSG − fSGH − fSGD) (1− exp(−ΩCHCH)) (22)
Of the fraction of the bottom taken up by the polyps, fpolyps, zooxanthellae are first exposed. As-suming the zooxanthellae are horizontally homogeneous, the fraction taken up by the zooxanthellaeis given by:
fzoo = min[fpolyps,π
2√
3nπr2zoo] (23)
where πr2 is the projected area of the cell, n is the number of cells, and π/(2√
3) ∼ 0.9069 accountsfor the maximum packaging of spheres. Thus the zooxanthellae can take up all the polyp area. Thefraction, if any, of the exposed polyp area remaining is assumed to be coral skeleton:
fskel = fpolyps −min[fpolyps,π
2√
3nπr2zoo] (24)
The benthic microalgae overlay the sediments. Following the zooxanthellae calculation above, thefraction taken up by benthic microalgae is given by:
fMPB = min[(1− fMA − fSG − fSGH − fSGD − fpolyps),π
2√
3nπr2MPB] (25)
13
Finally, the sediment fractions are assigned relative to their density in the surface layer, assumingthe finer fractions overlay gravel:
M = MudCaCO3 + SandCaCO3 +Mudnon−CaCO3 + Sandnon−CaCO3 + FineSed+Dust (26)
fCaCO3 = (1− fMA − fSG − fSGH − fSGD − fpolyps − fMPB)
(MudCaCO3 + SandCaCO3
M
)(27)
fnon−CaCO3 = (1− fMA− fSG− fSGH − fSGD − fpolyps− fMPB)
(Mudnon−CaCO3 + Sandnon−CaCO3
M
)(28)
fFineSed = (1− fMA − fSG − fSGH − fSGD − fpolyps − fMPB)
(FineSed+Dust
M
)(29)
with the porewaters not being considered optically-active. Now that the fraction of each bottomtype has been calculated, the fraction of backscattering to absorption plus backscattering for thebenthic surface as seen just below the surface, ubot,λ, is given by:
ubot,λ = wλ,bot(fMAρMA,λ (30)
+fSGρSG,λ
+fSGHρSGH,λ
+fSGDρSGD,λ
+fzoobzoo,λ
azoo,λ + bzoo,λ+fskelρskel,λ
+fMPBbMPB,λ
aMPB,λ + bMPB,λ
+fCaCO3ρCaCO3,λ + fnon−CaCO3ρnon−CaCO3,λ + fFineSedρnon−CaCO3,λ)
where the absorption and backscattering are calculated as given in Sec. ??, and ρ is the measuredbottom refectance of each end member (Dekker et al., 2011; Hamilton, 2001; Reichstetter et al.,2015).
For the values of surface reflectance for sand and mud from Heron Island (Roelfsema and Phinn,2012; Leiper et al., 2012), and microalgal optical properties calculated as per Sec. ??, a ternaryplot can be used to visualise the changes in true colour with sediment composition (Fig 13 of Bairdet al. (2016)).
It is important to note that while the backscattering of light from the bottom is considered in themodel for the purposes of calculating reflectance (and therefore comparing with observations), it is
14
not included in the calculation of water column scalar irradiance, which would require a radiativetransfer model (Mishchenko et al., 2002).
15
2.4 Simulated hue angle and the Forel-Ule colour comparator scale
Hue is the ’colour’ of light, and is often reported as an angle, the hue angle, around a colour wheel.Hue angle is calculated from weighting the primary colours (red, green and blue) by the standardhuman perception of colour (CIE 1931 standard colourimetric). The intensity of red and green areboth normalised to the sum of red, green and blue, to provide a two dimensional quantification ofcolour. The two dimensions are then reduced to one by determining the angle generated by thedivision of the normalised red and green dimensions (y and x) relative to white (yw, xw, i.e. equalred/green/blue) (van der Woerd and Wernand, 2015), where the normalised values relative to whitevary between -1 and 1. Thus the hue angle is calculated as α = arctan(y− yw, x− xw) modulus 2πand varies along a continuum from blue (240) to green (130) to brown (20).
One advantage of hue angle as a measure of normalised water leaving radiance is that it indepen-dent of saturation or intensity so is more likely to be consistent across sensor types and viewingconditions. Hue angle can be calculated directly from the remote-sensing reflectance at three ormore wavelengths. We use the coefficients determined by van der Woerd and Wernand (2018) forSentinel-2 MSI-60 to convert remote-sensing reflectance into the primary colours. Fig. 1 shows animage of hue angle for the GBR.
Another approach to quantifying colour is a 21-integer colour scale, called the Forel-Ule (FU) colourscale. Table 6 of Novoa et al. (2013) provides a conversion between the hue angle and the FU scale.Here the continuum is from blue (1) to brown (21) (Fig. 1). The FU scale has had broad, noexpert use as an indicator of water colour and can be determined by simple colour card matchingor smartphone apps.
2.5 Simulated true colour
True color images are often used in the geophysical sciences to provide a broad spatial view of aphenomenon such as cyclones, droughts or river plumes. Their strength lies in the human experienceof natural colors that allows three ’layers’ of information, and the interaction of these informationstreams, to be contained within one image. Spectacular true color images of, for example, theGreat Barrier Reef (GBR), simultaneously depict reef, sand and mud substrates, sediment-ladenriver plumes and phytoplankton blooms. Further, the advection of spatially-variable suspendedcoloured constituents reveals highly-resolved flow patterns.
For these reasons and more, true color imagery has become a valued communication tool within boththe geosciences and wider community. Here we demonstrate that the power of true colour imagescan be harnessed for interpreting geophysical models if the model output includes remote-sensingreflectance, or normalised water leaving radiance, at the red, green and blue wavelengths.
16
Figure 1: Simulated hue angle (left), Forel-Ule colour scale (centre) and MODIS true colour (right)calculated from simulated remote-sensing reflectance on the Great Barrier Reef on 1 Jan 2011.
17
In order to interpret simulated true colour images, a palette of true colours from the model opticalrelationships has been painted (Fig. 2), for varying values of water column CDOM, NAP and chloro-phyll concentration. The green hues produced for varying IOPs are quite similar, demonstratingthe difficulty in ocean colour analysis. The most obvious trend in the image is that Chl producesa greener hue than NAP (top panels vs bottom panels). The difference between increasing CDOMand chlorophyll (right panels vs left panels) is more subtle, which is a significant challenge in remotesensing of chlorophyll in high CDOM coastal waters (Schroeder et al., 2012).
The important role of scattering can be seen in Fig. 2. At low chl and / or NAP, the colour becomesdark at moderately low CDOM concentration. In reality CDOM is usually associated with eitherchlorophyll or NAP, hence natural waters rarely appear black.
As the model calculates remote-sensing reflectance at any wavelength, it is possible to reproducesimulated true colour images using the same algorithms as the MODIS processing (Fig. 3). Thecomparison of observed and modelled true colour images is particular insightful because the impactof multiple spectrally-distinct events (such as a sediment plume and a phytoplankton bloom) canbe viewed on one image (Fig. 4). Furthermore, the colour match-up provides an intuitive andchallenging test for parameterisation of the spectrally-resolved optical coefficients.
True colour images use intensity in the red, green and blue wavebands. We use the centre wave-lengths of MODIS bands 1 (645 nm), 2 (555 nm ) and 4 (470 nm). Simulated true colour imagebrightness is adjusted using the MODIS approach by linearly scaling the above surface remote-sensing reflectance at each wavelength so that the brightest band has an intensity of approximately1. This requires multiplication of between 20 to 30. The scaled intensity is intensified in the darkerbands by scaling [0 30 60 120 190 255]/255 onto [0 110 160 210 240 255]/255. The final image isrendered in Matlab.
18
Figure 2: Palette of true colours from IOP relationships. The true colours are produced from theMODIS algorithm (Sec. 2.5), and IOP relationships in Sec. ??. The centre box is the colour producedby clear water absorption and scattering alone. From the centre, moving right is increasing CDOM(as quantified by salinity), down is increasing NAP, and left and up are increasing chlorophyll incells package by 0.35 and 0.73 respectively. The line contours in the top left panel are 0.5, 1, 1.5and 2 mg Chl m−3. Note that packaged chlorophyll is not plotted with NAP and unpackaged is notplotted with CDOM.
19
Figure 4: Observed (top) and simulated (bottom) true colour from simulated remote-sensing re-flectance at 670, 555 and 470 nm in the GBR4 model configuration in the region of the BurdekinRiver. A brightening of 10 (left) and 20 (right) was applied for comparison.
21
2.6 Estimates of chlorophyll using the OC3 algorithm
The ratio of above-surface remote-sensing reflectance as a combination of three wavelengths, R′, isgiven by:
R′ = log10 (max [Rrs,443, Rrs,488] /Rrs,551) (31)
The ratio R′ is used in the MODIS OC3 algorithm to estimate surface chlorophyll, ChlOC3, withcoefficients from the 18 March 2010 reprocessing:
ChlOC3 = 100.283+R′(−2.753+R′(1.457+R′(0.659−1.403R′))) (32)
obtained from http://oceancolor.gsfc.nasa.gov/REPROCESSING/R2009/ocv6/.
2.7 Estimates of vertical attenuation using the Kd,490 algorithm
The ratio of above-surface remote-sensing reflectance as a combination of two wavelengths, R′, isgiven by:
R′ = log10 (Rrs,488/Rrs,547) (33)
The ratio R′ is used in the MODIS algorithm to estimate vertical attenuation at 490 nm, Kd,490,with coefficients from the 18 March 2010 reprocessing:
Kd,490 = 10−0.8813+R′(−2.0584+R′(2.5878+R′(−3.4885−1.5061R′))) (34)
obtained from http://oceancolor.gsfc.nasa.gov/REPROCESSING/R2009/ocv6/.
2.8 Estimates of particulate organic carbon using the POC algorithm
The ratio of above-surface remote-sensing reflectance as a combination of two wavelengths, R′, isgiven by:
R′ = log10 (Rrs,488/Rrs,555) (35)
The ratio R′ is used in the MODIS algorithm to estimate concentration of particulate organiccarbon, POC, with coefficients from the 18 March 2010 reprocessing:
POC = 308.3R′−1.639 (36)
obtained from http://oceancolor.gsfc.nasa.gov/REPROCESSING/R2009/ocv6/.
22
2.9 Simulated turbidity from bb,595
Turbidity is a measure of water clarity. The units of turbidity from a calibrated nephelometer arecalled Nephelometric Turbidity Units (NTU). Turbidity is often directly related to total suspendedsolids. However, as turbidity is measured optically (in the case of the three-wavelength Wetlabssensor the scattering at 700 nm at a measured angle of 124), it is better to produce a simulatedturbidity - a calculation of what the optical model predicts an optical turbidity sensor would mea-sure. Simulated turbidity, like simulated remote-sensing reflectance, is an example of ”taking themodel to the observations”.
Thus, simulated turbidity (NTU) is given by:
T = 47.02 (bb,595 − 0.5bb,clear,595) + 0.13 (37)
where bb,595 is the backscattering coefficient at 595 nm, and the linear coefficient and offset wereobtained from comparison of a BB9 backscattering sensor and a Wetlabs NTU at Lucinda Jetty,north Queensland. The coefficients are an empirical fit to the factory calibration of NTU standards.The clear water backscatter is removed because it was also removed in the output of the BB9. Inpractice
2.10 Estimates of total suspended matter using 645 nm
We use a local relationship between the remote-sensing reflectance at 645 nm, Rrs,645 [sr−1], andtotal suspended matter, TSM (Petus et al., 2014):
TSM =(12450R2
rs,645 + 666Rrs,645 + 0.48)/1000 (38)
23
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.080
10
20
30
40
50
60
70
80
90
100Nov 2012 (circles), Sep 2013 (squares)
Remote−sensing reflectance at 645 nm [sr−1]
Tot
al S
uspe
nded
Mat
ter
[g m
−3 ]
Petus et al. (2010)Miller and McKee (2004)
Figure 5: Relationship between the remote-sensing reflectance at 645 nm, Rrs,645, [sr−1], and totalsuspended matter, TSM , in the samples taken during November 2012 (circles) and September 2013(squares), compared to the empirical relationship of Petus et al. (2014) for the Bay of Biscay andMiller and McKee (2004) for the Mississippi. The symbols are coloured by true color, showingbrowner sites with higher TSM and remote-sensing reflectance. The black circle represents the siteused for the model parameterisation of TSM optical properties.
24
2.11 Simulated fluorescence
The chlorophyll a molecule absorbs light at a range of wavelengths, and can, under certain cir-cumstances, fluoresce strongly (Falkowski and Raven, 2007; Huot et al., 2005). As a result, activefluorescence is commonly used as a measure of chlorophyll a concentration. For this measurement,the stimulus for fluorescence comes from the sensor. For example, the Wetlabs ECOPuck emits lightat 470 nm, and detects excitation at 695 nm. The fraction of the light that is returned dependsprimarily on the concentration of chlorophyll, but can also be affected by the size of the phytoplank-ton, through the packaging of pigments, and on the physiological state of the cells (Alvarez et al.,2017; Baird et al., 2013). As the model calculates spatially- and temporally- resolved measures ofpigment packaging and cell physiology for each phytoplankton type, it is possible to calculate anew diagnostic, simulated fluorescence, that represents florescence that would be measured by afluorometer exciting at 470 nm, measuring at 695 nm, and sampling a phytoplankton communitywith the properties (number, cell size, physiological state) of the phytoplankton in the model.
Simulated fluorescence uses the model predicted cell number, cell size, chlorophyll content, and thepigment-specific absorption coefficient at the excitation wavelength (470 nm, γ = 0.03 m2 mg−1)to calculate a simulated fluorescence. Specifically, the per unit of pigment absorption effective-ness, a∗, of a spherical cell r, pigment-specific absorption coefficient, γ, and intracellular pigmentconcentration ci is given by:
a∗ = πr2(
1− 2(1− (1 + 2ρ)e−2ρ)
4ρ2
)/(nγ470ciV ) (39)
where ρ = γ470cir, n is the concentration of cells and V is the cell volume (4/3πr3). Because weare using the above equation to calculate the packaging effect, we sum the product γ470ci for allpigments, recognising that the presence of any pigment can shade the chlorophyll a molecule.
Cell physiology also affects fluorescence. The photosystem II (PSII) fluorescences when light isabsorbed by the cell, but not used for either photosynthesis or dissipated as heat by photoprotectivepigments (called non-photochemical quenching, NPQ; see coral zooxanthellae photosystem modelin (Baird et al., 2018)). To capture reduced fluorescence due to NPQ, we use the normalised carbonreserves of the cell, R∗C .
The simulated fluorescence from all phytoplankton is given:
FL = CF3
4
P∑p=1
a∗pcp(1− 0.5R∗C,p
)(40)
where the sum is applied across all P cell types (small and large phytoplankton, Trichodesmium,microphytobenthos and dinoflagellates), cp is the water column concentration of chlorophyll-a fromeach cell type, and CF is a calibration factor and depends on the initial factory calibration, which is
25
usually undertaken using a monoculture of a large diatom. The factor of 3/4 has been pulled out ofCF to compare absorption cross-sections and volume specific attenuation (see Baird et al. (2013)).The term (1− 0.5R∗C), accounts for the heat dissipation by the xanthophyll cycle, and takes a valueof 1 when the carbon reserves are deplete (i.e. at depth in low light), and a value of 0.5 when thecells are carbon replete. The factor 0.5 accounts for the absorption of photosynthetic pigments evenwhen the cell is energy replete, and can be thought of as the ratio of photoabsorbing pigments toxanthophyll pigments.
2.12 Simulated normalised fluorescence line height (nFLH)
In complex coastal waters, absorption in the blue is often dominated by CDOM, so band ratioalgorithms like OC3M do not correlate well with in situ chlorophyll (Schroeder et al., 2012). Analternate, remotely-sensed, indirect measure of chlorophyll concentration is normalised fluorescenceline height (nFLH, Behrenfeld et al. (2009)). nFLH is a measure of the light emitted by fluorescenceat a red wavelength (678 nm for MODIS), as a result of solar radiation absorbed by photosyntheticpigments throughout the spectrum. Thus while simulated fluorescence described above (Sec. 2.11)is a measure of active fluorescence, nFLH is due to passive fluorescence. The passive fluorescence,PF (photon m−3), at 678 nm due to absorption by P phytoplankton types is given by:
PF =P∑p=1
np(1− 0.5R∗C,p
)R∗C
(109hc)−1
AV
∫αEo,λλdλ (41)
where the term(1− 0.5R∗C,p
)is discussed in Sec. 2.11, the multiplication by R∗C removes the com-
ponent of absorption that is used for photosynthesis (that was not applied in Sec. 2.11 for simulatedfluorescence due to the saturating pulse of the active fluorometer), and Eo,λ is the scalar irradianceat wavelength λ [W m−2], np is the concentration of cell type p [cell m−3], and h, c and Av are thePlanck constant, speed of light in a vacuum and Avagadro number respectively.
The passive fluorescence is assumed to be emitted isotropically (equal in all directions) and is quan-tified per unit of volume. For the purposes of calculating the impact on remote-sensing reflectance,we need to consider only the upwelling component. The upwelling irradiance from a 1 m3 spherical
volume occurs through the area of a circle within the 1 m3 volume, π(
3√
3/(4π))2∼ 1.21 m2,
with only half of the volumetric radiance upwelled. Thus, the upwelling irradiance due to passivefluorescence is given by PF
2π(
3√
3/(4π))2 .
To obtain remote-sensing reflectance we need to compare the upwelling component to downwellingirradance at 678 nm. This is already undertaken for the absorption and scattering of light in thebalance of backscatter over absorption plus backscatter, u. Passive fluorescence at 678 nm adds to
26
this balance, with the component due at 678 nm:
uPF =1
2π(
3√
3/(4π))2PF/((109hc)−1
AV678Eo,678
)(42)
This ratio of upwelling PF to downwelling solar radiation can then be added to u, the ratio ofbackscattering to absorption plus backscattering, at 678 nm (Eq. 12). The revised u then results inan updated Rrs,678, with the depth weighting based on the vertical attenuation coefficient.
Finally, nFLH is determined from the normalised water leaving irradiance (nLw) at the fluorescingband, and the bands either side of it. The term nLw is normalised to the sun zenith, Fo = 148.097mW cm−2 µm−1,
nFLH =Foπ
(Rrs,678 −
70
81Rrs,667 −
11
81Rrs,748
)(43)
where π has units of sr−1 to convert between planar flux of nLw to a solid angle of remote-sensingreflectance (Rrs).
Note that passive fluorescence will contributes to the underwater light field in all directions, butbecause it is at 678 nm where there is strong absorption by clear water we have assumed that itdoes not lead to photosynthesis. Thus we have only considered the effect of passive fluorescence onRrs,678 and therefore nFLH.
2.13 Simulated bioluminescence
Bioluminescence is represented as a source of light per unit volume, much like passive fluorescence,except that the source is a function of cell number, not light absorption. The wavelength of emissionis centred around 470 nm, and occurs over a range of approximately 20 nm.
Mobley (1994) suggested the intensity of the light emitted by a 1 m3 volume at 470 nm is given by:
BLP = Bn
(hc
λ
)(44)
where BLP is the bioluminescence potential (W m−3), n is the number of cells and B is the rate ofphotons emitted per cell and varies with species. For this formulation, a dedicated phytoplanktonclass is necessary to capture the abundance of cells.
An alternate formulation (Ondercin et al., 1995) uses a bioluminescence cross-section of chlorophyll,α∗BL = 1.27× 10−8 mol photon (mg chl)−1 s−1:
BLP = α∗BLnciVhc
λ(45)
27
where ci is the internal concentration of chlorophyll, and V is cell volume.
Like passive fluorescence (Eq. 42), the contribution to remote-sensing is obtained through an addi-tional term to the local balance of backscatter and absorption:
uBL =1
2π(
3√
3/(4π))2BLP/((109hc)−1
AV470Eo,470
)(46)
2.14 Secchi depth
Secchi depth is often calculated using:
ZSD = 1.7/Kd (47)
where Kd is a vertical attenuation of light, typically energy-weighted PAR. At a particular wave-length the constant varies from 1.7. In a recent study (Lee et al., 2015), it was shown that theconstant has a value of approximately 1.0 when the colour of human perception of the disk surfaceis used as the wavelength. This varies in different waters, from 486 to 475 nm, as the water clarityimproves. Even though this value is slightly bluer than 490 nm, we found a good match withobservations for:
ZSD,490 = 1/Kd,490 (48)
The light level at the Secchi depth is ESD,490 = E0,490 exp(−ZSD,490Kd,490). Substituting Eq. 48,ESD,490/E0,490 = exp(−1).
Depth-varying vertical attenuation. If Kd,490 varies with depth (as it does in the simulations), thenthe Secchi depth is found by looking for the layer in which ESD,490/E0,490 drops below exp(−1).The top of this layer has a depth Ztop. Given a light level at the top of this layer of Etop, and anattenuation within the layer of Kd,490, the Secchi depth is given by:
ZSD,490 = Ztop + log
(exp(−1)/Etop−Kd,490
)(49)
Over 5000 measurements of Secchi depth have been undertaken in the GBR and used to developa satellite algorithm for Secchi depth (Weeks et al., 2012). We adopt this approach to define asimulated remotely-sensed Secchi depth, ZSD:
ZSD,sim,rs = 10(log10(2.303/Kd,490)−0.529)/0.816 (50)
where 2.303/Kd,490 gives the 10 percent light depth and the empirical constants come from Weekset al. (2012).
28
Finally these can be compared to the observed in situ Secchi depth, ZSD,obs, and the remotely-sensedSecchi depth, ZSD,rs.
2.15 Optical plume classification
Optical properties have been used to classify the extend of plumes (Devlin et al., 2013; Alvarez-Romero et al., 2013). Here we use a similar approach on simulated plumes. First we determinefrom observations the spectra of 6 standard plume classes (Fig. 6), as adopted by Alvarez-Romeroet al. (2013). Using these standard plume classifications, we determine the dissimilarity betweenan observed or simulated spectra and the spectra of each standard class. The cosine dissimilaritybetween standard class c and the observed or simulated spectra, S(c), is determined by:
S(c) = cos−1
∑Wλ=1Rrs,c,λRrs,sim,λ√∑W
λ=1R2rs,c,λ
√∑Wλ=1R
2rs,sim,λ
(51)
where Rrs,c,λ is the remote-sensing reflectance of class c at wavelength λ, Rrs,sim,λ is the remote-sensing reflectance of the simulation at wavelength λ andW is the number of wavelengths considered.The observed or simulated spectra is then assigned to the standard class c with the minimumdissimilarity, S, between the standard class and the observed or simulated spectra.
A simpler spectra matching scheme using the rms difference between the spectra is also employed.
Using this classification technique we can compare the extent of plumes in an observed scene (Fig. 7left), and a simulated scene (Fig. 7 right) of the same day. The areal extent of the observed andsimulated plume classes can also be calculated, noting that the observed area is less due to clouds.
As a metric of the recent history of plume exposure, we propose a metric the plume exposure, P ,which is given by:
P =
∫ t
t−tc<6
(1/c) dt (52)
where c is the plume class determined from Eq. 51. The plume exposure is calculated for eachmodel grid cell. The metric is calculated from the most recent time that c for a grid cell was lessthan 6 (i.e. impacted by a plume class 1, 2, 3, 4 or 5). The metric is a weighted-running mean,such that exposure to plume class 1 for 10 days would give a value of 10, plume class 2 for 10 days5, plume class 3 for 10 days would give 10/3 etc.
An additional concept is the annual plume extent, or the mean plume area over the year.
29
Figure 6: Spectra of each of the optical plume classifications (1-6) and the open ocean. The areain km2 for each plume class within the region plotted is given on the left for both the observed andsimulated classes.
Figure 7: Observed and simulated optical plume classification on the 25 Jan 2011 in the BurdekinRiver region. See Fig. 6 for spectra of individual classes.
30
Des
crip
tion
Cal
cula
tion
Applica
tion
Mon
thly
bot
tom
ligh
t1
hou
rly
runnin
g-m
ean
seag
rass
via
bilit
y
(eqiv
al.
todai
lydos
e]1 τ
∫ t 0Ed,λ,t
exp(−
(t−t′
)/τ)dt′
[mol
m−2
d−1]
Ara
gonit
edis
solu
tion
exp
osure
∫ t t−t Ω
<3
(3−
Ω)dt′
acid
ifica
tion
stre
ss
rese
tti
me
1day
,[u
nit
day
s]M
onth
lynet
calc
ifica
tion
rate
1hou
rly
runnin
g-m
ean
calc
ifica
tion
index
1 τ
∫ t 0g net,t
exp(−
(t−t′
)/τ)dt′
cora
lac
cret
ion
Tem
per
ature
exp
osure
∫ t t−t T
>Tclim
(T−Tclim
)dt′
ther
mal
(deg
ree
hea
ting
wee
ks)
rest
tim
e7
day
,[
Cw
eeks]
cora
lble
achin
gW
eekly
inor
ganic
Nupta
ke1
hou
rly
runnin
g-m
ean
oxid
ativ
est
ress
by
cora
ls1 τ
∫ t 0Gtex
p(−
(t−t′
)/τ)dt′
Sal
init
yex
pos
ure
∫ t t−t S
<28
(28−S
)dt′
fres
hw
ater
rese
tti
me
1day
,[P
SU
day
s]co
ral
ble
achin
gW
eekly
net
dep
osit
ion
rate
1hou
rly
runnin
g-m
ean
cora
lsm
other
ing
(sin
kin
g/
resu
spen
sion
/diff
usi
on)
1 τ
∫ t 0Dtex
p(−
(t−t′
)/τ)dt′
[cm
d−1]
Hyp
oxic
exp
osure
∫ t t−t [
O2]<
2000
(200
0−
[O2])dt
low
oxyge
n
rest
tim
e1
hou
r,[m
gO
m−3
d−1]
stre
ssW
eekly
bot
tom
ligh
tat
tenuat
ion
1hou
rly
runnin
g-m
ean
pre
dat
orvis
ibilit
y
(appro
x.
from
490
nm
)1 τ
∫ t 0Kd,490,t
exp(−
(t−t′
)/τ)dt′
[m−1]
pre
yb
ehav
iour
Rem
ote-
sensi
ng
reflec
tance
see
Sec
tion
??
com
par
ison
wit
hRrs
412,
443,
488,
531,
547,
648,
667,
748
nm
oce
anco
lour
pro
duct
sM
OD
ISal
gori
thm
s-
Kd,
OC
3,T
SS,
PO
Cem
pir
ical
algo
rith
ms
usi
ng
com
par
ison
wit
h-
Kd,
OC
3,T
SS,
PO
Cre
mot
e-se
nsi
ng
reflec
tance
oce
anco
lour
pro
duct
sSim
ula
ted
true
colo
ur
RG
Bad
dit
ive
colo
ur
model
wat
erqual
ity
usi
ng
rem
ote-
sensi
ng
reflec
tance
non
-exp
ert
com
munic
atio
nP
lum
ecl
assi
fica
tion
cate
gori
cal
from
spec
tra
mat
chin
gplu
me
exte
nt
(1-6
)+
clea
rw
ater
usi
ng
RM
Ser
rors
onO
Cban
ds
1-7
Tab
le11
:D
iagn
osti
cva
riab
les.
Inad
dit
ion
to60
+st
ate
vari
able
s,der
ived
dia
gnos
tic
vari
able
shav
eb
een
dev
elop
edin
consu
ltat
ion
wit
hre
sear
cher
san
dm
anag
ers
that
pro
vid
em
etri
csfo
rim
pro
veunder
stan
din
gan
dap
plica
tion
.F
orth
ose
met
rics
wit
ha
tim
esc
ale,τ,
see
Sec
.2.
16.1
.
31
2.16 Diagnostic variables calculates by tracerstat.
The EMS software provides a library call tracerstat that calculates a diagnostic variable from acombination of prognostics and diagnostic variables, and can be the result of an integration in spaceor time. The EMS manual describes some of these. Here we concentration on options used in theBGC model.
2.16.1 Running mean
For a simulation the calculation of a running mean is problematic if not all values from whichthe mean is to be calculated are archived. A good example of this light, where a midday outputis useless as a daily mean. To overcome this, we introduce a running mean calculated from theweighted sum of the running mean at the previous time step, and the present value of the variable.Thus, for a 24 hour running mean, with outputs every hour, the running mean for downwellingirradiance, Ed, is given by:
Ed,t =23
24Ed,t−1 +
1
24Ed,t (53)
As this procedure is applied at each timestep, the running mean can be written as a time-integralwith an exponential decay of the importance of earlier values:
Ed,t =1
τ
∫ t
0
Ed,t exp(−(t− t′)/τ)dt′ (54)
where the running mean the coefficient, τ represents an exponential decay time. Thus for a weeklyrunning average, the impact of the value 1 week earlier on the running average is 100 × exp(−1),or 37 % of the impact of the present value. Or 63.2 % of the value is based on the last week, 86.5% on the last two weeks.
3 Acknowledgements
Many scientists and projects have contributed resources and knowhow to the development of thismodel over 15+ years. For this dedication we are very grateful.
Those who have contributed to the numerical code include (CSIRO unless stated):
Mike Herzfeld, Philip Gillibrand, John Andrewartha, Farhan Rizwi, Jenny Skerratt, Mathieu Mon-gin, Mark Baird, Karen Wild-Allen, John Parlsow, Emlyn Jones, Nugzar Margvelashvili, Pavel
32
Sakov, Jason Waring, Stephen Walker, Uwe Rosebrock, Brett Wallace, Ian Webster, Barbara Rob-son, Scott Hadley (University of Tasmania), Malin Gustafsson (University of Technology Sydney,UTS), Matthew Adams (University of Queensland, UQ).
Collaborating scientists include:
Bronte Tilbrook, Andy Steven, Thomas Schroeder, Nagur Cherukuru, Peter Ralph (UTS), RussBabcock, Kadija Oubelkheir, Bojana Manojlovic (UTS), Stephen Woodcock (UTS), Stuart Phinn(UQ), Chris Roelfsema (UQ), Miles Furnas (AIMS), David McKinnon (AIMS), David Blondeau-Patissier (Charles Darwin University), Michelle Devlin (James Cook University), Eduardo da Silva(JCU), Julie Duchalais, Jerome Brebion, Leonie Geoffroy, Yair Suari, Cloe Viavant, Lesley Clementson(pigment absorption coefficients), Dariusz Stramski (inorganic absorption and scattering coeffi-cients), Erin Kenna, Line Bay (AIMS), Neal Cantin (AIMS) and Luke Morris (AIMS).
Funding bodies: CSIRO Wealth from Oceans Flagship, Gas Industry Social & Environmental Re-search Alliance (GISERA), CSIRO Coastal Carbon Cluster, Derwent Estuary Program, INFORM2,eReefs, Great Barrier Reef Foundation, Australian Climate Change Science Program, University ofTechnology Sydney, Department of Energy and Environment, Integrated Marine Observing System(IMOS), National Environment Science Program (NESP TWQ Hub).
33
References
Alvarez, E., Nogueira, E., Lopez-Urrutia, A., 2017. In vivo single-cell fluorescence and size scalingof phytoplankton chlorophyll content. App. Env. Microbiol. 83, e03317–16.
Alvarez-Romero, J. G., Devlin, M., da Silva, E. T., Petus, C., Ban, N. C., Pressey, R. L., Kool,J., Roberts, J. J., Cerdeira-Estrada, S., Wenger, A. S., Brodie, J., 2013. A novel approach tomodel exposure of coastal-marine ecosystems to riverine flood plumes based on remote sensingtechniques. J. Enviro. Manage. 119, 194–207.
Baird, M. E., Cherukuru, N., Jones, E., Margvelashvili, N., Mongin, M., Oubelkheir, K., Ralph,P. J., Rizwi, F., Robson, B. J., Schroeder, T., Skerratt, J., Steven, A. D. L., Wild-Allen, K. A.,2016. Remote-sensing reflectance and true colour produced by a coupled hydrodynamic, optical,sediment, biogeochemical model of the Great Barrier Reef, Australia: comparison with satellitedata. Env. Model. Softw. 78, 79–96.
Baird, M. E., Mongin, M., Rizwi, F., Bay, L. K., Cantin, N. E., Soja-Wozniak, M., Skerratt, J.,2018. A mechanistic model of coral bleaching due to temperature-mediated light-driven reactiveoxygen build-up in zooxanthellae. Ecol. Model. 386, 20–37.
Baird, M. E., Ralph, P. J., Rizwi, F., Wild-Allen, K. A., Steven, A. D. L., 2013. A dynamic modelof the cellular carbon to chlorophyll ratio applied to a batch culture and a continental shelfecosystem. Limnol. Oceanogr. 58, 1215–1226.
Baird, M. E., Timko, P. G., Suthers, I. M., Middleton, J. H., 2006. Coupled physical-biologicalmodelling study of the East Australian Current with idealised wind forcing. Part I: Biologicalmodel intercomparison. J. Mar. Sys. 59, 249–270.
Baird, M. E., Wild-Allen, K. A., Parslow, J., Mongin, M., Robson, B., Skerratt, J., Rizwi, F., Soja-Woznaik, M., Jones, E., Herzfeld, M., Margvelashvili, N., Andrewartha, J., Langlais, C., Adams,M. P., Cherukuru, N., Gustafsson, M., Hadley, S., Ralph, P. J., Rosebrock, U., Schroeder, T.,Laiolo, L., Harrison, D., Steven, A. D. L., 2020. CSIRO Environmental Modelling Suite (EMS):Scientific description of the optical and biogeochemical models (vB3p0). Geoscientific ModelDevelopment 13, 4503–4553.
Behrenfeld, M. J., Westberry, T. K., Boss, E. S., O’Malley, R. T., Siegel, D. A., Wiggert, J. D.,Franz, B. A., McClain, C. R., Feldman, G. C., Doney, S. C., Moore, J. K., Dall’Olmo, G., Milligan,A. J., Lima, I., Mahowald, N., 2009. Satellite-detected fluorescence reveals global physiology ofocean phytoplankton. Biogeosciences 6 (5), 779–794.
Brando, V. E., Dekker, A. G., Park, Y. J., Schroeder, T., 2012. Adaptive semianalytical inversionof ocean color radiometry in optically complex waters. Applied Optics 51, 2808–2833.
34
Dekker, A. G., Phinn, S. R., Anstee, J., Bissett, P., Brando, V. E., Casey, B., Fearns, P., Hedley, J.,Klonowski, W., Lee, Z. P., Lynch, M., Lyons, M., Mobley, C., Roelfsema, C., 2011. Intercompar-ison of shallow water bathymetry, hydro-optics, and benthos mapping techniques in Australianand Caribbean coastal environments. Limnol. Oceanogr.: Methods 9, 396–425.
Devlin, M., da Silva, E., Petus, C., Wenger, A., Zeh, D., Tracey, D., Alvarez-Romero, J., Brodie,J., 2013. Combining in-situ water quality and remotely sensed data across spatial and temporalscales to measure variability in wet season chlorophyll-a: Great Barrier Reef lagoon (Queensland,Australia). Ecological Processes 2, 31.
Eyre, B. D., Cyronak, T., Drupp, P., Carlo, E. H. D., Sachs, J. P., Andersson, A. J., 2018. Coralreefs will transition to net dissolving before end of century. Science 359, 908–911.
Falkowski, P. G., Raven, J. A., 2007. Aquatic Photosynthesis, 2nd Edition. Blackwell Science.
Hamilton, L. J., 2001. Cross-shelf colour zonation in northern Great Barrier Reef lagoon surficialsediments. Aust. J. Earth Sci. 48, 193–200.
Huot, Y., Brown, C. A., Cullen, J. J., 2005. New algorithms for MODIS sun-induced chlorophyllfluorescence and a comparison with present data products. Limnol. Oceanogr. Methods 3, 108–130.
Lee, Z., Carder, K. L., Arnone, R. A., 2002. Deriving inherent optical properties from water color:a multiband quasi-analytical algorithm for optically deep waters. Applied Optics 41, 5755–5772.
Lee, Z., Shang, S., Hu, C., Du, K., Weidermann, A., Hou, W., Lin, J., Lin, G., 2015. Secchidisk depth: A new theory and mechanistic model for underwater visibility. Remote sensing ofenvironment 169, 139–149.
Leiper, I., Phinn, S., Dekker, A. G., 2012. Spectral reflectance of coral reef benthos and substrateassemblages on Heron Reef, Australia. Int. J. of Rem. Sens. 33, 3946–3965.
Macdonald, H. S., Baird, M. E., Middleton, J. H., 2009. The effect of wind on continen-tal shelf carbon fluxes off southeast Australia: a numerical model. J. Geophys. Res. 114,C05016,doi:10.1029/2008JC004946.
Miller, R. L., McKee, B. A., 2004. Using modis terra 250 m imagery to map concentrations ofsuspended matter in coastal waters. Remote Sens. Environ. 93, 259–266.
Mishchenko, M. I., Travis, L. D., Lacis, A. A., 2002. Scattering, Absorption and Emission of Lightby Small Particles. Cambridge University Press.
Mobley, C. D., 1994. Light and Water: Radiative Transfer in Natural Waters. Academic Press.
35
Mongin, M., Baird, M. E., Tilbrook, B., Matear, R. J., Lenton, A., Herzfeld, M., Wild-Allen, K. A.,Skerratt, J., Margvelashvili, N., Robson, B. J., Duarte, C. M., Gustafsson, M. S. M., Ralph, P. J.,Steven, A. D. L., 2016. The exposure of the Great Barrier Reef to ocean acidification. NatureCommunications 7, 10732.
Monsen, N., Cloern, J., Lucas, L., Monismith, S. G., 2002. A comment on the use of flushing time,residence time, and age as transport time scales. Limnol. Oceanogr. 47, 1545–1553.
Novoa, S., Wernand, M. R., van der Woerd, J. J., 2013. The Forel-Ule scale revisited spectrally:preparation protocol, transmission measurements and chromaticity. Journal of the EuropeanOptical Society - Rapid publications 8, 13057.
Ondercin, D. G., Atkinson, C. A., Kiefer, D. A., 1995. The distribution of bioluminescence andchlorophyll during the late summer in the North Atlantic: Maps and a predictive model. J.Geophys. Res. (Oceans) 100, 6575–6590.
Petus, C., Marieua, V., Novoac, S., Chust, G., Bruneau, N., Froidefond, J.-M., 2014. Monitoringspatio-temporal variability of the Adour River turbid plume (Bay of Biscay, France) with MODIS250-m imagery. Cont. Shelf Res. 74, 35–49.
Reichstetter, M., Fearns, P. R., Weeks, S. J., McKinna, L. I., Roelfsema, C., Furnas, M., 2015.Bottom reflectance in ocean color satellite remote sensing for coral reef environments. RemoteSens. 7, 15852.
Roelfsema, C. M., Phinn, S. R., 2012. Spectral reflectance library of selected biotic and abiotic coralreef features in Heron Reef. Tech. rep., Centre for Remote Sensing and Spatial Information Sci-ence, School of Geography, Planning and Environmental Management, University of Queensland,Brisbane, Australia, doi:10.1594/PANGAEA.804589.
Schroeder, T., Devlin, M. J., Brando, V. E., Dekker, A. G., Brodie, J. E., Clementson, L. A.,McKinna, L., 2012. Inter-annual variability of wet season freshwater plume extent into the GreatBarrier Reef lagoon based on satellite coastal ocean colour observations. Mar. Poll. Bull. 65,210–223.
Soja-Wozniak, M., Baird, M. E., Schroeder, T., Qin, Y., Clementson, L., Baker, B., Boadle, D.,Brando, V., Steven, A., 2019. Particulate backscattering ratio as an indicator of changing particlecomposition in coastal waters: observations from Great Barrier Reef waters. J. Geophys. Res.(Oceans) 124.
Vaillancourt, R. D., Brown, C. W., Guillard, R. R. L., Balch, W. M., 2004. Light backscatteringproperties of marine phytoplankton: relationship to cell size, chemical composition and taxonomy.J. Plank. Res. 26, 191–212.
36
van der Woerd, H. J., Wernand, M. R., 2015. True colour classification of natural waters withmedium-spectral resolution satellites: SeaWiFS, MODIS, MERIS and OLCI. Sensors 15, 25663–25680.
van der Woerd, H. J., Wernand, M. R., 2018. Hue-angle product for low to medium spatial resolutionoptical satellite sensors. Remote Sensing 10, 180.
Weeks, S., Werdell, P. J., Schaffelke, B., Canto, M., Lee, Z., Wilding, J. G., Feldman, G. C.,2012. Satellite-derived photic depth on the Great Barrier Reef: spatio-temporal patterns of waterclarity. Remote Sens. 4, 3781–3795.
A Additions to B3p0
The following changes in the model equations were made between versions B2p0 and B3p0:
• Optical properties of carbonate particles (absorption, scattering and backscattering coeffi-cients) are introduced using observations from the Lucinda Jetty Coastal Observatory.
• Spectrally-resolved phytoplankton backscattering replaces a wavelength-independent value.
• Coral bleaching processes added using new variables (Coral N, P, C reserves, xanthophyll pho-tosynthetic and heat dissipating pigments, oxidised, reduced and inhibited reaction centres,reactive oxygen species), parameter values (ROS threshold) and diagnostics (Rubisco activity,bleaching rate) following (Baird et al., 2018).
• Coral heterotrophic feeding fixed: reserves from grazed phytoplankton now returned to watercolumn.
• New diagnostic variables included in optical model: simulated fluorescence, simulated turbid-ity, simulated normalised fluorescence line height, simulated Secchi depth, downwelling lighton z-level interfaces, SWR bot abs (the PAR weighted bottom absorption calculated by themodel), OC4Me - chlorophyll algorithm for MERIS and Sentinel satellites.
• Optical code now has options of spectral absorption coefficients calculated from mass-specificabsorption laboratory experiments.
• Apply a relationship between aragonite saturation and sediment carbonate dissolution (Eyreet al., 2018).
• Spectrally-resolved light absorption by microphytobenthos in the sediment porewaters.
37
• Mass balance for oxygen includes oxygen atoms in nitrate, with stoichiometry changed forphotosynthesis / respiration.
• In the model description, energy reserves are more correctly referred to as carbon reserves.
• Quadratic term for seagrass mortality.
• Remineralisation rate of phosphorus has a separate rate to that of C and N, requiring 2 newparameters (ΦRDP
, ΦDOMP).
• Introduced new sediment-water exchange rate diagnostic variables for C, N, P, and O.
B Additions to B3p1
C Additions to B3p2
38