Geochimica

et Cosmochimica
Acta,1977,
Vol.41,pp. 1313lo 1324. Pergamon
Press.Printedin GreatBritain

The mechanism of iron removal in estuaries

E. A. BOYLE* and J. M. EDMOND
Department of Earth and Planetary Sciences, Massachusetts Institute of Technology,
Cambridge, MA 02139, U.S.A.
and

E. R. SHOLKOVITZ
Grant Institute of Geology, University of Edinburgh, Edinburgh, Scotland

(Received 4 January 1977; accepted in revised form 18 May 1977)

Abstract--A survey of U.S. east coast estuaries confirms that large-scale rapid removal of iron from
river water is a general phenomenon during estuarine mixing. The river-borne ‘dissolved’ iron consists
almost entirely of mixed iron oxide-organic matter colloids, of diameter less than 0.45 pm, stabilized
by the dissolved organic matter. Precipitation occurs on mixing because the seawater cations neutralize
the negatively charged iron-bearing colloids allowing flocculation. The process has been duplicated
in laboratory experiments using both natural filtered and unfiltered river water and a synthetic colloidal
goethite in 0.05 ktrn filtered water. The colloidal nature of the iron has been further confirmed by
ultracentrifugation and ultrafiltration. A major consequence of the precipitation phenomena is to reduce
the effective input of ‘dissolved’ iron to the ocean by about 90% of the primary river value, equivalent
to a concentration of less than 1 pmol per liter of river water.

INTRODUCTION counted entirely, it is more probable that insoluble

organic-hydroxy-iron species are present in stable
THE OCEANIC ‘residence time’ of dissolved iron relative
colloidal dispersion.
to the river input appears to be only a few hundred
SHOLKOVITZ (1976) reported a series of experiments
years (GARRELSand MCKENZIE,1971). This extraordi-
on the chemical composition of the precipitates
narily short ‘cycling’ time is much less than the mix-
formed when filtered river waters and seawater were
ing time of the ocean, so removal of the river-borne
mixed under a variety of conditions. The amount of
iron must be localized to a high degree. It is possible
precipitated material increased with the salinity of the
that the major sink is the estuarine mixing zone given
mixture; Fe, Mn, Al, P, and organic carbon all
the marked gradients in chemical and physical
behaved similarly. The reaction was rapid, most of
properties that occur there. Evidence in fact exists
the precipitate forming within l/2 hr of mixing. Subse-
for removal of iron from solution in several estuaries
quently ECKERT and SHOLKOVITZ (1976) found that
(COONLEYet al., 1971; WINDOMet al., 1971; BOYLE
at equimolar concentrations the effectiveness of preci-
et al., 1974; SUBRAMANIAN and D’ANGLEJAN,1976;
pitation by different sea-salt cations increased in the
YEATSand BEWERS,1976; HOLLIDAYand LISS, 1976).
sequence NaCl @ MgC12 < CaCl,. These experimen-
If this phenomena prevails generally, then the input
tal results were interpreted as a cation-induced coagu-
of dissolved iron to the open ocean must be consider-
lation of negatively-charged humic colloids contain-
ably less than the primary river flux. The geochemical
ing bound iron and other elements. The experiments
cycles of other elements also may be affected, es-
establish a definite association between iron and
pecially if the process involves precipitation of hyd-
organic matter and provide important constraints on
rated ferric oxide, an efficient chemical scavenging
hypotheses concerning the mechanism of iron remo-
agent. Therefore an understanding of the chemical
val in estuaries.
nature of iron in rivers and its reactivity in estuaries
The present investigation was undertaken in order
is essential to the construction of the marine geo-
to understand further the nature of the estuarine iron
chemical cycle of it and other trace elements.
precipitation process and to assess its generality. A
Iron concentrations typically found in filtered or
combination of field studies, laboratory mixing ex-
centrifuged fresh-water samples exceed by far the
periments, and synthetic model experiments were
levels expected for inorganic dissolved iron species
used in this effort.
in equilibrium with solid phases (JONESet al., 1974).
This has been explained either as an effect of organic METHODS
complexation or of the presence of colloidal iron (a) Field work
oxide particles. STUMM and MORGAN (1970) reviewed In the summer and fall of 1973 a number of east coast
the problem and concluded that while soluble organic U.S. estuaries were surveyed to determine whether iron
complexes of iron in natural waters could not be dis- removal processes are widespread. Subsequent surveys
were undertaken to clarify aspects of these processes.
* Present address: Grant Institute of Geology, Univer- Surface water samples were collected throughout the
sity of Edinburgh. salinity range (e.g. Fig. 1) and immediately filtered through
1313
1314 E. A. BOYLE.J. M. EDMONDand E. R. SHOLKOVITZ

acid-leached glass fiber filters. Within 24 hr after collection
the samples were acidified to pH = 2 with hydrochloric
acid. For the 10/13/75 Mullica estuary samples (Table 5,
Fig. 3) a Whatman GF/F glass fiber filter (nominal pore
size 0.7 pm) was used; for the rest of the samples Mead
glass fiber filters (nominal pore size 1.2 pm) were used.
Glass fiber filters were chosen over membrane filters
because their greater flow velocity allows for immediate
filtration of turbid estuarine waters and avoids the storage
effects noted by LEWIN and CHEN (1973). This advantage
comes at the cost of precise pore size definition, a problem
which will be discussed later.
(b) Analysis
Salinity was determined on an unacidified aliquot by
conductivity, chloride selective electrode, or chloride titra-
tion using a Buchler/Cotlove chloridometer. The iron con-
tent of the samples was determined calorimetrically follow-
ing reduction by hydroxylamine for 1 hr at room tempera-
ture (RAINWATER and THATCHER,1960). An acetate buffer
was used with bipyridyl or ferrozine (STOOKEY,1970) as
the color forming reagent. A water color blank correction
was measured for each sample. Only that fraction of iron
originally present as Fe”(aq) or converted to Fe” by
the reduction procedure is dete~ned by this method and
is referred to in the text as “calorimetric iron”. The preci-
sion of the determination is approximately +O.l pmol/l or
&-3”/, whichever is greater.
Organic matter interferes with the determination of fer-
rous iron in natural water samples by reducing ferric iron
(SHAPIRO, 1966; MCMAHON 1967). An upper limit for
Fe’+(aq) was estimated by immtdiate calorimetric analysis
of filtered samples buffered at pH 5 with elimination of
prior acidification and the reducing reagent.
Total iron was determined by direct injection into a
heated graphite atomizer for atomic absorption spectro-
photometry (SEGARand CANTILLO,1975) using an internal
addition on each sample for standardization. The esti-
mated precision of this analysis is iO.05 pmol/l or +Sy<,
whichever is greater. A Perkin--Elmer HGA-2100 graphite
furnace and model 403 s~ctrophotometer with mode1 56
chart recorder were used.
Anaiysis of precipitates formed in the mixing experi-
ments described below was by thin-film X-ray fluorescence
and CHN elemental analyzer (SHOLKOVITZ,1976). Milli-
pore membrane filters were used in the laboratory experi-
ments unless otherwise noted.
RESULTS AND DISCUSSION
Data from the preli~nary survey of east coast U.S.
estuaries vable 1) are displayed as plots of iron vs
salinity in Fig. 1. The estuaries studied show pro-

Table 1. Calorimetric iron and salinity data for some east coast United States estuaries. [For Merrimack
10/30/73, F-8, see BOYLEet al. (1974)]
Connecticut Merrimack Connecticut Merrimack Parker
7/5/73,B-12* -l/26/73,0-5 8/10/73,B-7 8/14/73,X?-3 6,'21/73,R-26
s Fe S
___?!%
S Fe
--C
S -Fe, --S Fe,
a;66 & 5.05 4.0 0.29 0.3 0.07 3.5 0.38 11.9
0.08 1.1 0.05 4.0 0.38 1.0 0.38 3.3 0.44 12.5
0.45 8.5 1.3 3.1 2.1 0.0 2.1 3.2 1.05 12.5
2.1 5.5 5.3 2.5 4.1 0.3 4.E 2.5 2.88 11.5
2.5 5.3 2.3 6.9 0.0 7.3 1.6 3.73 6.6
12.5 :*; 14.4 1.2 9.9 0.0 10.4 1.5 5.76 5.8
12.5 0:a 20.6 0.8 10.9 0.1 13.1 4.0 7.61 4.8
25.6 0.7 12.6 0.4 15.8 0.8 11.59 3.0
25.6 0.6 1.9.4 0.0 17.6 0.8 16.61 2.5
27.5 0.1 23.8 0.3 21.9 1.0 20.60 1.9
23.8 0.3 27.5 0.5 26.76 1.3
31.3 0.2 33.8 1.0 27.94 1.0

Rappahannock Mullica Connecticut Parker

9/3/73, B-22 9/5/'73, F-153 11,'11,'73,F-3 10/14/74, F-5
s s __Es Fet S Fe, S Fe_ - Fe+ DH
I !!I.% Et_ L A- L
0.26 3.3 11.0 0.25 23.4 24.9 0.16 3.07 0.09 3.59 4.86 6.5
0.30 16.4 0.31 22.9 28.7 0.16 3.08 0.17 4.22 5.75
0.56 13.3 0.83 18.8 20.4 0.17 2.94 0.58 4.91 6.92 6.5
1.23 1.8 6.5 1.70 14.6 16.7 0.17 2.5E 0.93 5.18 8.29
2.47 2.54 10.9 16.4 0.31 2.71 1.18 6.05 9.27 6.4
2.56 0.5
0.6 ',:; 3.82 8.0 9.2 1.19 2.67 1.82 5.00 7.06
2.93 0.5 2.2 4.06 8.0 10.0 1.39 2.61 2.32 4.35 6.80 6.6
3.19 0.2 0.64 8.57 3.4 3.4 2.04 2.43 3.04 4.43 6.65
3.20 0.3 1.4 9.81 2.4 2.5 2.52 2.30 4.22 3.73 6.1
8.97 11.73 1.9 1.9 2.90 2.09 4.42 3.48 5.21
10.47 ::1" 0.28
0.23 13.32 1.4 1.e 3.06 2.23 5.76 3.05 6.9
13.85 0.1 0.23 15.84 1.1 1.4 3.70 2.03 7.45 2.91 4.20
20.13 0.7 4.26 2.02 8.53 2.74 6.9
21.89 0.6 4.37 1.94 10.48 2.31
22.12 0.6 5.78 1.62 11.79 1.89 2.52 7.1
22.51 0.5 10.43 1.00 13.43 1.73
23.82 0.5 10.66 0.68 14.86 1.32 7.2
25.34 0.9 13.12 0.37 15.84 1.28 1.75
25.73 0.4 15.27 0.68 17.48 1.25 7.4
2 .03 0.4 15.98 0.70 20.61 1.11 7.5
18.41 0.52 21.86 0.95
23.48 0.26 23.31 0.75
29.34 0.07 25.31 0.57 0.90
30.46 0.05 27.33 0.58
21.55 0.51

* Analytical code: B = bipyridyl reagent, F = ferrozine reagent; followed by number of days between
acidification and analysis. S = Salinity. Y&,.Fe, = calorimetric iron, pmol/l. Fe, = total iron, pmol/l.
 The me~h~~jsm of iron removal in estuaries 1315

6-
\
,
I
I
:
CONNECTICUT 1 7”5?3
l E$lO,73
4\

3
\1.
‘\\
.‘\
\
MERRIMACK 7 2673

‘:
\
,
4- ‘L
* \\,

‘\
‘%
a- ‘<
. .
‘.._
. -;
“-_~----___~-- ---.--.-~_ - -._.-._

0 5 b IS 20 2’5 35

Jr
,
6

*,
3-f
,
RAPPAHANNCfJK 9.3 73
L MERRIMACK 814 73

% -- --- -_* --_-__ *

IO 15 20

. .
lI,I l

‘\
5- 8
-\
‘.
‘.
;=_.
*‘1-w.__
---.___
--“-‘*‘,

% J IO IS 20 Is 3b

t
1
20 :
i
: MULLICA 9573

\
:
1
:
i
I@ :

‘?,
‘.
‘.
‘1
.L.
l.
-*-‘.__
01 , --a... --*--II)_*____ 1 ___.
0 5 IO I5 20 25 30

SALINITY, %o
Fig. I. Calorimetric iron vs salinity for severa east coast U.S. estuaries.

nounced negative curvature from a two end-member calcufated using the flux model of BOVLEet al. (1974)
iron-salinity dilution fine indicating that iron is (Table 2). Removal of iron ranges from 5Oa/z,to greater
removed from solution in the mixing zone. Although than 95% among the rivers studied. The various river
low-iron tributaries or a short-term change in the end members had iron concentrations from as low
river end-members could mimic remova1, none of the as 3 pm/l (Connecticut) to as high as 24 pmol/l (Mul-
estuaries have significant tributaries in the region lica): rivers of higher iron have a greater total remo-
sampled, and temporal changes are just as likely val in their estuaries. This process has been abserved
to result in apparent production, which is never now in at least 10 estuaries on 16 occasions and is
observed. Lower limits on the degree of removal were therefore established as a common phenomenon.
1316 E. A. BOYLE.J. M. EDMONDand E. R. SHOLKOVID

Tahlc 2. Iron removal from some east coast estuaries

Zero-salinity High-salinity
Estuary end member Intercept % Removal
Iron, umol/Y. Iron, Lmlol/r.

Connecticut -i/5/73 8.5 2.5 71

11/11/73 3.1 0.9 71

Merrimack 7/26/73 4.0 1.5 63

8/14/73 3.5 1.0 71

10/30/73 3.7 1.7 54

Parker B/21/73 11.9 4.0 66

Mullica g/5/73 23.4 1.0 96

Rappahannock g/3/73 3.3 0.3 91

with hydrochloric acid to pH = 6.5. Splits of each

Laboratory mixing experiments were performed to
constrain the factors controlling the removal of iron
when fresh and saline waters meet. Water was taken
from the Merrimack River at Lowell, Mass. (3/6/74).
Unfiltered and Mead-glass-fiber-filtered aliquots were
mixed with equal volumes of (1) distilled water, (2)
0.7 N NaCI, (3) seawater diluted to lO“,, salinity, (4)
seawater at 30?;, salinity, and (5) seawater acidified
of the mixtures were allowed to stand for 4 and 24 hr
before being refiltered through Mead glass fiber fil-
ters. Iron analyses were then performed colorimetri-
tally (Table 3). The precipitation reaction is seen to
be rapid, with little additional removal occurring after
4 hr. The extent of the reaction depends on the
salinity of the mixture, with less precipitation taking
place at lower salinity. The effect is not simply related

Table 3. Initial mixing experiments (Merrimack River 3/6.!74)

Filtered or
Mixing Solution Mixing Time, hr. % Removal*
_______ unfiltered

Distilled water F 4 11

F 24 8

II 4 16

u 24 12

F 4 27

F 24 32

u 4 38

U 24 40

300/,, seawater, F 4 60
pH=8.2 F 24 66

U 4 72

U 24 78

30°/,o seawater, F 4 53
p~=6.5
F 24 58

u 4 48

U 24 42

0.7 M N.&l F 4 21

F 24 26

U 4 25

u 24 25

* Iron determined calorimetrically using ferrozine 4 days after acidification.

Measured iron concentration in the initial filtered river water was
3.36 pmol/liter.
 The mechanism of iron removal in estuaries 1317

r
I”

l l
/

15
RAPPAHANNOCK ESTUARY MULLICA ESTUARY
9) w
\ 913173 \ 9/5/73
E 14 E
/
3 / a20 -

0
0 I 2 3 4 0 IO 20
COLORIMETRIC IRON, pm/k! COLORIMETRIC IRON, pm/t

l ,/’
- PARKER ESTUARY MULLI CA ESTUARY
10/14/74 ? lO/l3/75
i’
%f E 20-
i

0 I 2 3 4 5 6 0 IO 20
COLORIMETRIC IRON, pm/k? COLORIMETRIC IRON, pm/e

Fig. 2. Total iron vs calorimetric iron for filtered estuarine samples.

to ionic strength, since 0.7 M sodium chloride solu-
tion is less effective than seawater. There must be a
specific interaction with seawater species other than
sodium or chloride ions, as also observed by ECKERT
and SHOLKOVITZ(1976). Changes in solution pH over
the range of values likely to be encountered in
estuaries have no appreciable influence on the precipi-
tation: seawater acidified to pH = 6.5 (the same pH
as the river water) is only slightly less effective than
at 8.2. The results also demonstrate that the presence
of river-borne particulate matter has a minor effect
on the extent of precipitation for this river.
(3) Calorimetric us total iron
The calorimetric method used in the preliminary
field and laboratory experiments determines only that
fraction of the iron converted to Fe’+ by the rela-
tively mild conditions of room temperature hydroxy-
lamine reduction. Some forms of iron may not re-
spond to this procedure. Both calorimetric and total
iron determinations were therefore made on samples
from four surveys (Tables 1 and 5); the analyses are
compared in Fig. 2. In two of the cases (Mullica
9/S/73 and 10/13/75) the results by the two methods
are not significantly different for most samples, allow-
ing for the estimated analytical accuracies. For the
other estuaries (Rappahannock 9/33/73 and Parker
10/14/74) the total iron concentrations are higher
than the calorimetric results. The total iron/colori-
metric iron ratio is very nearly constant for each estu-
ary, so the same curvature is shown by a plot of total
iron vs salinity as in the plots shown in Fig. 1. The
conclusion that iron is removed from solution during
estuarine mixing is still valid; however, in some cases
the absolute amount of iron removed is higher than
suggested by the calorimetric analyses.
These analytical differences provide information
about the chemical nature of the iron. For example,
two of the Parker 10/14/74 samples (originally ana-
lysed 5 days after collection and acidification) were

Table 4. Comparison of calorimetric iron analyses 5 and 59 days after collection

with total iron analyses (Parker Estuary, 10/14/74).
I?XXl, ~InOl/L

Salinity 5 Day Analysis 59 Day Analysis Total

0.09 3.59 3.81 4.86

1.18 6.05 7.46 9.27

1318 E. A. BOYLE J. M. EDMOND and E. R. SHOLKOVITZ

reanalyzed by the colorimet~~ method an additionai
54 days later (Table 4). The new analyses fall in
between the original analyses and the total iron con-
centration. Evidently the refractory iron component
slowly converts to a more labile form upon prolonged
acidification. One of the samples from the Mullica
10/13/75 survey showed much less calorimetric iron
than total iron (Fig. 2). The sample was therefore sub-
jected to a harsher reduction procedure by boiling
in the presence of hydroxylamine for l/2 hr. This pro-
cedure increased the calorimetric result to near the
level observed for total iron: 50’1;,of the total iron
in this sample is converted to Fe’ ’ only by the
harsher reduction. These analytical differences must
be reconciled with the mechanistic interpretation of
the estuarine processes and are discussed further in
the conclusions section.
The possible role of ferrous iron in the estuarine
removal process should be considered. The oxidation
rate of ferrous iron is inversely proportional to the
square of the hydrogen ion concentration. At pH
values below 6.5 the oxidation rate is slow enough
to give Fe’+ a half-life of days to weeks (STUMM and
LEE, 1961); when a low-pH river mixes with high-
er-pH seawater, Fe *+ should oxidize and precipitate
as hydrated ferric oxide. GJESSING(1964) reported that
;I high percentage of the total iron in some Norwegian
freshwater samples was in the ferrous state but
SHAPIRO(1966) and MCMAHON(1967) demonstrated
that this observation was an artifact of the analytical
procedure: under acidic conditions natural organic
matter will reduce ferric iron, the reaction being
accelerated by heat and bright light. An upper limit
for ferrous iron can be estimated by immediate colori-
metric analysis at intermediate pH. This method was
used in two of the estuaries studied (Parker. 10/30/74
and Mulhca 30~13174)and indicated that ferrous iron
comprised less than IO?;, of the total iron, The
absence of a pronounced pH effect in the mixing ex-
periments mentioned earlier, as well as the total--
calorimetric differences (Fig. 2) also demonstrate that
oxidation of ferrous iron is not a significant mechan-
ism for iron removal in estuaries.
A combined field and laboratory study of the Mui-
lica estuary was undertaken to clarify the mechanism

Table 5. Data from the Mullica estuary. lO;t3;75

Salinity Total Iron LT Color. Iron HT Color. Iron PH

"1.0 ulnol/P. \lrnOl/R** rEmol/P **

0.00 9.5* 8.3

0.00 7.1* 7.4
0.22 25.4 22.4 22.4
0.90 2L.9 19.7 19.4
1.61 14.1 5.8, 6 12.2, 12.8 -
2.20 18.7 16.6 16.5
3.23 14.6 14.9
5.27 10.6 10.1
5.60 10.6 9.6 9.6
7.13 7.6 6.7
9.09 6.3 4.6
9.61 3.6 3.3 3.3
12.07 2.1 2.0
12.29 2.2 2.0 1.0
13.84 1.7 1.6 1.5 7.6
16.27 1.1 1.0 7.6
18.54 0.9 0.6 7.7
21.17 0.3 0.3 7.8
22.66 0.2 0.3 7.9
24.59 0.2 0.2 0.2 7.9
27.38 3.1 0.0 8.0
28.44 0.1 0.0 8.0
29.77 0.0 0.0 8.1
30.34 0.0 0.1 a.2
-
*These samples were allowed a long filtration time through a highly
clogged filter; reduction of the pore size by the clogging is the most likely
explanation for these low values. as discussed in the conclusion.
** Calorimetric analyses using ferrozine 93 days after acidification.
LT = room temperature reduction with hydroxylamine (1 hr).
HT = boiling reduction with hydroxylamine (l/2 hr).
 The mechanism of iron removal in estuaries 1319

Three filter types were used for the purposes of differ-

MULLICA ESTUARY 10/13/75 ent analyses. The filtered river water was mixed in
varying proportions with filtered seawater and with
NaCl, MgCl, and CaCl, solutions. Each mixture was
then processed as described by the flow chart in Fig.
4. The combined filtrates (both 1 hr and 1 day) were
analyzed for total iron. Fe, P, Al, Ti and Si determina-
tions were performed on the Millipore filters; particu-
late organic carbon was determined on the glass fiber
filter. The preweighed Nuclepore filters were washed
with small portions of distilled water, dried in a dessi-
cator and reweighed. The low ionic strength of the
‘0.. wash solution tended to peptize the precipitate, but
0 I I -j-*--o_+*_ __.I_ _ _
the disaggregation was slow compared to the rinse
0 5 IO 15 20 25 30
time and little net weight was lost. The results of these
SALINITY, %o
analyses are listed in Table 6.
Fig. 3. Total iron vs salinity, Mullica Estuary, 10/13/75.
The laboratory mixing experiments (Table 6, Fig.
5) resemble the field observations. The differences in
of iron precipitation. This estuary was chosen because absolute concentration reflect the different filtration
extensive (>95”A) iron removal had already been procedures employed.
observed (COONLEYet al., 1970; Table 1, Fig. 1). The As observed previously, the precipitation is cation-
field data from this survey (Table 5, Fig. 3) confirm dependent. Solutions of major seawater salts were
the previous results. added individually to the river water up to levels
(n) Mixing experiments. The laboratory mixing ex- equivalent to their concentration in lo?;, estuarine
periments were performed on river water prefiltered water. At these concentration levels, magnesium is
successively through a Mead glass fiber filter, a What- more effective than calcium and sodium (Table 6).
man GF/F glass fiber filter, and twice through an The chemical composition of the particulates fol-
0.45 pm Millipore membrane filter. The experimental lows patterns similar to those observed by SHOLKO-
strategy was similar to that of SHOLROVITZ (1976). VITZ (1976). Organic matter and iron are the major

r-l r-l
FILTERED FILTERED
RIVER SEA
WATER WATER

‘-h. STAND l HOUR

FILTER THROUGH
FILTER THROUGH FILTER THROUGH WHATMAN GF/F
0.45,, MILL I PORE 0.4~ NUCLEPORE
GLASS FIBER
FILTER FILTER FILTER+

TCOMBINE FILTRATES
STAND I DAY

FILTER THROUGH
FILTER THROUGH FILTER THROUGH
WHATMAN GF/F
0.45~ MILLIPORE 0.4,, NUCLEPORE
GLASS FIBER
FILTER FILTER
FILTER*

I COMBINE FILTRATES

’ PRECOMBUSTED

NOTE: equal portions (*IO%) were filtered through each fitter

Fig. 4. Flow chart describing mixing experiments with filtered Mullica River water.
1320 E. A. BOYLE.J. M. EDMOND and E. R. SHOLKOVITZ

Table 6. Mixing experiments using filtered Mullica River water: the effects of salinity and various cations
on iron precipitation, the composition of the precipitates, and the effect of time on the extent of precipitation
FILTRATE PRECIPITATE __-~
"100 ug/liter of river water**

Salinity Total Total Ps P Al Ti Si Organic

of mixture Iron Weight carbon

0 15.5 o* o.o* 0.0% o.oo* o* 0*

2 13.9 145 1.5 3.6 0.38 0 0
5 8.3 367 4.5 9.9 0.69
+? 9 4.2 1602 414 7.6 13.4 0.13 12 287
.-i 14 0.63 3185 LO15 21. 33.1 0.52 86 564
18 0.28 2982 1044 27. 31.2 0.30 68 129
28 0.11 2121 1041 67. 44.0 0.00 328

0 16.0 a* o.o* o.o* o.oo* o* o*

2 11.8 98 0.5 0.0 0.85 11 0
8.2 ___ 24-I 3.7 2.7 0.26 27
2.0 853 381 13. 11.0 0.05 41
r( 14 0.57 140 35 14. 0.7 0.42 10
18 0.21 317 15. 4.2 0.00 13 415
28 0.06 a47 36. 10.0 0.00 41

3.2~1O-~M 1 hr 5.11*** 200 3.2 4.4 0.22 59

&Cl2 24 hr -- --_ 53 3.2 3.6 0.15 3
1.6~lO-~M L hr 3.36 680 13.0 31.2 0.52 92
MgCl2 24 hr -- 6 4.7 0.0 0.18 7
1.6x10 -1 1 hr 5.74 ___ 430 3.0 6.3 0.49 97
Nat1 24 hr -- --_ 49 0.3 0.0 0.03

* Zero-salinity particulate “blank” subtracted from the listed values (SHOI.KOVITZ,1976).Hence the zero
salinityresults are zero by definition. The difference between the initial “dissolved” concentration (15.5 16.0
~mo1~) and the maximum amount of iron precipi~ted (18.6 ~rnol/l~ is attributed to differences in the AAS
and XRF calibrations and is not considered significant.
** Total on filters divided by the volume of river water in the filtered mixture.
***The initial iron concentration of the river water was 12.3 pmol/l; the dilution due to addition of the
cation solutions was 0.69:;.

components. The appearance of iron in the precipi-
tates mirrors its loss from solution (Fig. 5). Ti, Al
and P are also incorporated (Table 6). Silicon is preci-
pitated, but the total amount is small relative to the
river dissolved concentration (100 pmol !I).
The kinetics of the precipitation reaction were
studied by filtering ~awater~river water mixtures at
known time intervals through a 0.45 pm filter. The
reaction is extremely rapid (Fig. 6) with a time scale
of a few minutes. The iron concentration decreases
to an apparent “‘equilibrium level” determined by the
salinity of the mixture.
(b) Preccipitute redissolution. A large quantity of
precipitate was collected by adding CaClz to the til-
tered river water. centrifuging and briefly washing the
excess CaCI, from the precipitate with distilled water.
Portions of this precipitate were mixed with various
s~water/distilled water mixtures, dispersed in an
ultrasonic bath for 15 min and then filtered through
a 0.45 m filter after standing for I, 2 and 9 days.
The filtrates were analyzed for total iron. The form
of the Fe-S curves (Fig. 7) is similar to those observed
for the precipitation process, but lower iron concen-
tration levels are observed which change with time.
Although the precipitation resembles an equilibrium
process in some aspects. strictly the reaction is not
reversible.
(c) Anionic properrks cfthr precipitutr. The mixing
experiments discussed above and those of ECKERT
and SHOLKOVITZ (1976) demonstrate that for a given
freshwater sample, the degree of precjpi~tion is deter-
mined by the nature and concentration of the added
cation. It has been estabiished that humic materials
contain abundant negatively-charged carboxyl groups
(e.g. SCHNITZER and SKINNER, 1965): a major factor
in the precipitation reaction may be the neutralization
of these negatively charged groups on the organic
component of the precipitate.
This possibility was investigated using an ion-
exchange experiment. The precipitate was formed by
adding Car& to the filtered river water and the mix-
ture refiltered through a 0.45 pm Milli~re filter. After
being washed free of excess CaC12 this was rinsed
through with several portions of 2 x 10e4 M MgC’l,
(thereby replacing exchangeable Ca by Mg), distilled
water (to remove excess MgClJ and 2 x tom4 M
CaCl, (replacing exchangeable Mg by Ca). The fil-
trates were analyzed for calcium and magnesium by
flame atomic absorption and the results corrected for
the filter exchange blank. Equivalent ion-exchange
 The mechanism of iron removal in estuaries 1321

SOLUTION
.-- I hr.
A- 24 hrs.

-SALINITY, %o

0 5 IO 15 20
1
SALINITY. O/am

Fig. 7. Redissolution experiment: total iron vs salinity.

tures allowed to stand for 45 min. They were then

refiltered through an 0.45,~m filter, and the filterate
analyzed for total iron. The experiment was per-
formed initially at final cation concentrations of
3 x 10s3 M. From these results an estimate was
made of the individual cation concentration required
to precipitate equal quantities of iron and the experi-
Fig. 5. Total iron vs salinity, Mullica River water mixing
ment redone at these new concentration levels (Table
experiments.
8). The alkaline earth metals display the sequence of
effectiveness Mg < Ca < Sr < Ba.
capacities were measured for Ca-replacement and
This type of ordering has been interpretated as a
Mg-repla~ment, 54 and 47me~l~g of precipitate
manifestation of the Schulz~Hardy rule for hydro-
respectively. Quantitative interpretation of this
phobic colloid coagulation (ONG and BISQUE,1968;
exchange capacity would require a study of the pH-
ECKERT and SHOLKOVITZ,1976). However, the di-
dependence, but the experiment does establish the
valent first row transition metals are increasingly
presence of negatively-charged cation-exchanging
effective in the sequence Zn < Mn < Co, Ni < Cu.
sites on the precipitate.
This demonstrates that there is a specific chemical
The effect of different cations on the degree of iron
interaction between the metal cation and the colloids
precipitation provides information on the nature of
in addition to the purely electrostatic effects, an inter-
the cation-organic interaction. Small portions of con-
action which is analogous to that controlling the
centrated solutions of chloride or sulfate salts of di-
stability constants of 1: 1 complexes of these metals
valent alkaline earth and upper row transition metals
(IRVINGand WILLIE, 1953). Therefore the precipi-
were added to the filtered river water and the mix-
tation reaction is controlled by chemical interactions
as well as the physical controls on colloid coagulation.
(d) Colloidaf nature qf iron in river water. Iron alone
accounts for at least one-third of the total weight of
the precipitate formed during the mixing of river
water with seawater making it unlikely that this is
formed from a simple soluble organic-iron complex,
If all of the precipitate was organic matter forming
a 1: 1 complex with iron, the average molecular
weight of the organic material would be 112. This
is an extreme upper limit, as other ~mponents are
present (Table 6). Most evidence suggests that the
;‘. molecular weight of freshwater organic matter is no

I ‘-n---.---___..______A______
J\..,
lower than 400 (SHAPIRO,1958). On the other hand
DEB (1949) demonstrated that humic materials can
h. .,. . w ‘,““...‘...,...............,.,.... I . . .. .. ..I...
L. ...
..
stabilize colloidal iron oxide at iron/carbon mole
I I I I ratios as high as 4; a similar effect was observed by
20 30 40 50
TIME, minutes SHAPIRO(1964). This type of “solubilization” mechan-
Fig. 6. Kinetics of precipitation of iron from Mullica River ism is more compatible with the high iron content
water/seawater mixtures. of the precipitate. To test directly for the presence
1322 E. A. BOYLE,J. M. EDMONDand E. R. SHOLKOVIT~

Table 7. Effect of ultrafiltration and ultracentrifugation on iron concentration of

0.45 pm filtered Mullica River water

Treatment Initial total iron* Final total iron % removed

~nlOl/P. ~imOl/L

0.10 urn filtration 15.8 4.3 73%

0.05 urn filtration 15.8 3.3 79%

3O,OOOg, 2 hr. 12.0 9.2 23%

9O,OOOg, 3 hr. 10.6 2.5 76%

_-~-- _~
* These experiments were performed on different batches of 0.45 pm filtered river
water.
Table 8. The effect of various cations on iron precipitation
~~_._.~
Salt Concentration, p10l/P. Total iron*, ~~mol/&

MgCl2 7.3 6.4

C&l2 3.3 7.8

SrC12 2.7 5.8

BaC12 1.7 5.7

MnC12 2.7 6.5

coc12 2.0 6.4

NiS04 2.0 6.5

cuso4 1.3 3.1

znso4 2.7 7.6

* The initial iron concentration of the river water was 12.3 pmol/l;
the dilution due to addition of the cation solutions was <0.6”/,.

of colloidal iron, Mullica river water was filtered 0.05 pm filter and then mixed with the colloidal sus-
through a 0.45 pm filter and subjected to ultra-centri- pension in a 33 : 1 ratio. When this mixture was passed
fugation and to refiltration through O.lOpm and through an 0.45 /*m filter almost all of the added iron
0.05 pm filters (Table 7). Filtration using successively remained in the filtrate. The river’s dissolved organic
smaller pore sizes removed progressively larger matter stabilizes the crystallites and retains the iron
amounts of iron, up to 79% at 0.05 pm. Increasingly in solution. This refiltered mixture was then used in
severe ultracentrifugation sedimented up to 767: of seawater mixing experiments as described previously.
the iron (9O,OOOg,3 hr); a brown pellet was observed A salinity dependent precipitation was observed simi-
in the bottom of the centrifuge tube. Most of the iron lar to that for the naturally-occurring iron (Fig. 8).
in these filtered water samples therefore consists of The similarity of behavior supports the hypothesis
colloidal particles of diameter less than 0.45 pm. that river-borne colloidal iron oxide is stabilized by
LAMAR (1968) PERHAC (1972) and KENNEDY et ~1.
(1974) have noted the presence of a substantial col-
loidal fraction of iron in other freshwater samples.
It is likely that “dissolved” iron is predominantly col-
loidal in most oxygenated freshwater systems.
(e) Colloidal iron model experiment. The iron-
organic colloid association was modeled using a
well-characterized colloidal suspension of goethite
(FREUNDLICHand WOSNESSENSKY,1923) with average
particle dimensions of 0.1 x 0.01 pm (HELLER and
PETERS, 1970). In a 1 :50 dilution with distilled water,
the iron in this suspension will not pass through an
0.45 q filter because the crystallites clump together
SALINITY, %.
to a slight extent. The use of distilled water filtered
Fig. 8. Colloid model mixing experiment: total iron vs
through a pre-rinsed 0.05 pm filter gives the same
salinity for the filtrates. The 0.05 pm filtered river water
result; residual detergents leaking out of the filter do contained 1.4 pmol/l Fe; addition of the colloid and filtra-
not stabilize the individual crystallites. A portion of tion through an 0.45 pm Millipore filter raised the iron
filtered river water was refiltered through a pre-rinsed concentration to 11.7 pmol/l.
 The mechanism of iron removal in estuaries
natural organic matter and that this combined iron-
organic colloid undergoes a salinity-dependent preci-
pitation in estuaries. Presumably the organic stabili-
zation is effected by the formation of an exterior
sheath of negatively charged carboxyl groups which
constitute an energy barrier to coagulation (STUMM
and MORGAN,1970) and which are neutralized by sea-
water cations. In natural systems the colloids are less
likely to contain crystalline goethite and are probably
amorphous mixtures with domains of hydrated ferric
oxide and coprecipitated organic matter (COEY and
READMAN,1973).
CONCLUSIONS
(1) Iron removal has now been observed in at least
10 estuaries on 16 occasions. This removal is prob-
ably a common phenomenon.
(2) In the Mullica river “dissolved” iron consists
of mixed iron oxide-organic matter colloids of dia-
meter less than 0.45 pm. The dissolved organic matter
stabilizes these colloids. When the river water mixes
with seawater in the estuary, these negatively charged
colloids react rapidly with seawater cations to form
a precipitate. The degree of coagulation increases with
increasing salinity; further work is required to under-
stand in detail the mechanism of this salinity-depen-
dence.
(3) A similar iron precipitation process has been
observed in several other estuarine situations and in
laboratory experiments. Given the widespread avail-
ability of organic matter and iron oxide in most fresh-
water systems, the same conclusions should apply to
these cases.
(4) The estuarine precipitation process can be
modeled in the laboratory using an organic-stabilized
goethite colloid.
(5) Mild acidification (pH = 2) slowly dissolves the
iron oxide; wet chemical analyses may produce low
iron analyses if dissolution is not complete. Boiling
briefly in the presence of the reducing agent
hydroxylamine dissolves the iron oxide (Table 5) and
should be employed in analytical procedures which
require the iron to be in a labile form (e.g.
calorimetric and solvent extraction methods). Iron
released by severe treatment such as autoclaving has
been interpreted in the past as “organic” iron. This
interpretation is correct to the extent that iron is
associated with organic matter; however, the chemical
environment of the iron is more strictly char-
acterizable as a hydrated ferric oxide phase. S&ice
phosphorus is probably bound in the colloids as
phosphate substituting for oxygen in the oxide phase
(LEVESQUEand SCHNITZER, 1967), similar caution
should be applied to the interpretation of the
difference between total phosphorus and “inorganic
phosphate” (as determined by most analytical
procedures for phosphate which may not completely
dissolve the iron colloids) as “organic phosphorus”.
(6) The reaction between metal cations and the
1323
coagulated iron-organic colloids may be regarded as
a type of organic complexation. Investigations on the
complexing ability of dissolved organic matter
frequently employ metal concentrations higher by
orders of magnitude than the naturally occurring
levels. The metal cations may in fact be bound by
the organic matter in precipitates like those observed
in this work rather than as soluble complexes.
Extrapolation of stability behavior observed in
complexing studies at high metal concentrations to
the lower natural levels assuming a dissolved
metal-ligand interaction may lead to erroneous
conclusions about the speciation of the trace metal.
Future work in this area should take care to
distinguish between the two effects.
(7) As demonstrated here and by SHOLKOVITZ
(1976) other element+notably, Mn, Al and P-are
bound in the iron-organic precipitate. These and
other species should show a removal effect in
estuaries. This removal may be overwhelmed by a
dominance of the soluble form of the element; for
example, in the ultracentrifugation experiment where
75% of the iron was sedimented, only 6% of the
manganese was removed. Some scavenging of soluble
constituents by the flocculating colloids may occur
but should not be as extensive as expected for freshly-
precipitated ferric oxide. The question of the degree
to which a constituent is non-conservative in a given
estuary must be resolved by field studies and labora-
tory mixing experiments.
(8) In his classic treatise on the chemical
composition of freshwater systems, L~VINGSTONE
(1963) discussed some of the ambiguities involved in
iron analysis, but nonetheless went on to compute
a world average dissolved iron concentration for
rivers of 12 pmolfliter. This number has been widely
quoted and used in the calculation of the oceanic
residence time for iron. The present work and that
of KENNEDYet al. (1974) make it clear that the iron
concentration of a filtered freshwater sample depends
highly on the effective pore size of the filter and the
degree to which clogging reduces this. In any event,
removal of a large fraction of the iron from solution
in estuaries rules out the calculation of a meaningful
oceanic residence time based on the primary river
input since on conversion of the colloids to larger
particles the dispersal of the iron is governed by
particulate transport mechanisms. Based on the
sampling of estuaries in this work the supply of iron
to the ocean is reduced from the primary river
concentration to less than 1 pmol per liter of river
water (Table 2). The precipitated iron may be
remobilized by reactions in reducing sediments on the
continental shelf and returned to the water column
by diffusion through the pore waters. Therefore the
above estimate can only be tentative. Hard water
rivers containing little iron or organic matter may
behave somewhat differently than those studied here,
but they do not contribute significantly to the net
iron flux.
1314 E. A. BOYLE. J. M. EUMOND and E. R. SHOLKOVITZ

A~kn~~~sle~~e~~e~i~- BUB COLLIER,JAMES RICHMAN, ROBERT
STALLARD and several undergraduates in the MIT
Environment Waters Seminar provided welcome assistance
with the field work. This research is based on part of the
Ph.D thesis of E.B. (MIT-WHO1 Joint Program in
Oceanography) who was supported during part of the
period of research by an NSF fellowship. Other support
for the work came from the MIT UROP office and a grant
from the Doherty Foundation. The work of E.R.S. was
supported by an NERC (Great Britain) grant. We thank
KARL TIJREKIAN. MICHAELBENDER.PETERBEW~R,GF~~RCE and lakes, USGS Professional Paper 440-G.
HARVEY and GEOFF THOMPSONfor their constructive
criticism of an early version of this manuscript, and W.
STUMM.P. Liss and R. MANT~URAfor constructive reviews.
REFERENCES
BOYLE E. A., COLLXRR., DENGLERA. T., EUMONDJ. M..
Nc; A. C. and STALLARD R. F. (1974) On the geochemical RAINWAT~KF. H. and THATC‘HER
mass balance in estuaries. Geochim. Cosmochim. Acfo 38,
1719-1728.
C~EY J. M. D. and READMAN P. W. (1973) Characterization SC-I~NITZ.EKM. and SKINNER S. I. M. (1965)
and magnetic properties of natural ferric gel. Earth
P&met. Sci. Lrrt. 21, 45. 51.
C~ONLEY L. S.. JR., BAKERE. B. and HOLLANI)H. D. (19711
Iron in the Mullica River and Great Bay. New Jersey.
Chem. Grol. 7, 51.-63.
DEB B. C. (1949) The movement and precipitation of iron
oxides in podzol soils. J. Soil Sci. 1, 112 122.
ECKER~ J. M. and SHOI.KOVITZE. R. (1976) The
flocculation of iron, aluminum, and humates from river
water by electrolytes. Geochim. Cosmochim. 4ctcr 40,
847m848.
FRF.UNDLICH V. H. and WOSNIXENSKY S. (1923) Ueber
Eisenoxydsole die aus Eisencarbonyl gewonnen werden.
Kofltriri Zeitscltr@ 33, 222 -227.
GARRELS R. M. altd MACKENZIE F. T. (1971) ~7~~~~~ti~~~~SHAPIRO J. flY66) On the measurement of ferrous iron in
@” Se[liment~jr~, Rocks. Norton.
GJE~~IN~ E. T. (1964) Ferrous iron in water. Lim&.
Ocecmoq. 9, 212-274.
HOLLI~AYL. M. and Lrss P. S. (1976) The behaviour of
dissolved iron, manganese, and zinc in the Beaulieu estu- STOOKEY L. L. (1970) Ferrozine: a new spectrophotometric
ary, S. England. Est. Coasfal Mur. Sci. 4, 349.-353.
H~~LLER W. and PETERSJ. (1970) Mechanical and surface
coagulation 1: surface coagulation of Z-FeOOH sols. J.
Co/. Intet$~<~ Sri. 32, 592-605.
IRVIN<; H. and WILLEAMS R. J. P. (lY53) The stability of
transition-metal complexes. d. Chern. Sot. 3192~-3210.
JONES B. F., KENNHIY V. C. and ZF.LLWEGER G. W. (1974)
Comparison of observed and calculated ~OIlcentrations
of dissolved Al and Fe in stream water. Writer Resources
Rex. f0, 791 -793.
KENNEDY V. C.. ZELLWEC;ER G. W. and Jokers B. F. (1974)
Filter pore-size effects on the analysis of Al, Fe. Mn.
and Ti in water. Water Resources Res. IO, 785 790.
LAMARW. L. (196X) Evaluation of organic color and iron
in natural surface waters. USGS Prof. Paper hC@-D,
D24--D29.
LEVE~QCJ~:M. and SC~+NITZEK M. (1967) Organo-metallic
interactions in soils: 6. Preparation and properties of
fulvic acid-metal phosphates. J. Soil Sci. 103.- 1X3--190.
L.~~IN J. and CHENC. (I 973) Changes in the concentration
of soluble and particulate iron in seawater enclosed in
containers. Limnol. Ocecmogr. 18, 59%596.
LIVINC;STOWFD. A. (1963) Chemical composition of rivers
M~MAHo~ J. W. (19671 The influence of light and acid
on the measurement of ferrous iron in lake water.
Lintnol. Oc<~<~noqr. 12, 437.-442.
OX; H. L. and B~syrlr: R. E. (1948) Coaguiation of humic
collolds by metal ions. 1. Soil Sci. 106, 220.-224.
PHCHAC R. M. (1972) Distribution of Cd, Co. Cu. Fe. Mn,
Ni. Pb, and Zn in dissolved and particulate solids from
two streams in Tennessee J. Hyd&. 15, 177~.186.
L. L. (19601 Methods
for collection and analysis of water sambles. U.S. Geol.
Survey. Paper 1454.
Organo-metallic interacti(~ns in soils: 4. Carboxyl and
hydroxyi groups on organic matter and metal retention.
J. Soii Sri. 99, 27x- 284.
%.~;AK D. A. and C‘ANTILLO A. Y. (1975) Direct
determination of trace metals in seawater by flameless
atomic absorption spectrophotometry. In ilnal~tical
Methods in Ocrcmogruphy (editor Gibb), pp. 56-81.
Advances in Chemistry Series 147. American Chemical
Society.
SHAPIROJ. (1958) Yellow acid” cation complexes in lake
water. Science 127, 702.-704.
SIIAPIROJ. (lY64) Effect of yellow organic acids on iron
and other metals in water. Anr. M/tr. Wrks. Assoc. J,
55, 1062 ~1081.
natural waters. Lintnol. Ocranoyr. 11. 293-298.
SHOLKOYITZ E. R. (1976) Flocculation of dissolved organic
and inorganic matter during the mixing of river water
and seawater. Gcochim. Cosmochim. Acta 40. X31..845.
reagent for iron Altal. Chem. 42. 779-781.
ST~JMMW. and LEC G. F. (1961) Oxygenation of ferrous
iron. Ind. Eq. Ckc>nr.52. 143..146.
STLIMMW. and MORGANJ. J. (1970) Aqcrutic Chemistry.
pp. 7
-8 2 294, pp. 492...493,and p. X0. Wiley-Interscience.
S~BRA~AN~A~ V. and DANGLEJANB. (1976) Water chemis-
try of the St. Lawrence Estuary. J. ~~~ro~. 29, 341-354.
W~NI~M H. L.. BKX K. C. and SMITH R. (1971) Transport
of trace elements to the Atlantic Ocean by three south-
eastern rivers. Sotrthtwst Gtd. 1lOY 1181.
YI:ATS P. A. and BEwtiRs J. M. (lY76) Trace metals in the
waters of Saguenay Fjord. Curl. J. Eurth Sci. 13.
1319 1327.