• The use of oxygen isotope variation in shells of estuarine ...

  • FileName: DettmanGCA.pdf [read-online]
    • Abstract: composition of bivalve mollusk shells that grew in the river-water/seawater mixing zone of the Colorado River. estuary. Sclerochronological techniques are used to identify tidally-induced, fortnight-scale bundles of daily. growth increments within shell cross-sections. ...

Download the ebook

Geochimica et Cosmochimica Acta, Vol. 68, No. 6, pp. 1253–1263, 2004
Copyright © 2004 Elsevier Ltd
Pergamon Printed in the USA. All rights reserved
0016-7037/04 $30.00 .00
The use of oxygen isotope variation in shells of estuarine mollusks as a quantitative record
of seasonal and annual Colorado River discharge
Department of Geological Sciences, University of Arizona, Tucson, AZ 85721, USA
Department of Invertebrate Zoology and Geology, California Academy of Sciences, San Francisco, CA 94118, USA
Geologisch-Palaontologischen Institute, Johann Wolfgang Goethe Universitat, D-60054 Frankfurt, Germany
¨ ¨
(Received September 10, 2002; accepted in revised form September 9, 2003)
Abstract—We describe a new method for the calculation of river flow that uses the oxygen isotope
composition of bivalve mollusk shells that grew in the river-water/seawater mixing zone of the Colorado River
estuary. Sclerochronological techniques are used to identify tidally-induced, fortnight-scale bundles of daily
growth increments within shell cross-sections. These fortnightly markers are used to establish a chronology
for samples taken for 18O analysis. A composite seasonal 18O profile derived from five shells that grew in
the absence of river-water flow is used as a baseline against which profiles of river-influenced shells are
compared. Because this comparison is between matched fortnights within a year, the temperature of shell
growth is likely to be very similar. The difference in 18O between the river-influenced shell and the “no-flow”
composite shell therefore represents the change in the 18O of the water due to the presence of river water in
the mixing zone. The river water end-member is also determined within a fortnightly context so that the
change in the 18O of mixing-zone water can be used to calculate the relative proportions of seawater and
fresh-water. The fresh-water end-member is calculated from the 18O of bivalves alive prior to the emplace-
ment of dams and water diversions on the Colorado River. The marine end-member is based on direct
measurements of the 18O of northern Gulf of California water during times of no Colorado River flow. The
system has been calibrated to absolute flow amounts using recent releases of known volume and
rate. Copyright © 2004 Elsevier Ltd
1. INTRODUCTION O profiles can be used to calculate Colorado River discharge
at its delta before upstream dam and diversions, and before the
Quantitative estimates of prehistoric river flow can provide oldest direct measurements.
valuable information on the natural range of variability in river
discharge and the response of the hydrologic cycle to climate
change. In most parts of the world, direct measurements of
river discharge have been made for less than 200 yr, so proxy
The Colorado River is one of the major sources of water for
indicators need to be employed to provide a long-term record of
six states in the southwestern United States and two in north-
drought or floods. Quantitative estimates of river discharge
western Mexico. Its annual discharge has been completely
based on tree-rings have been successfully used to provide
allocated within the U.S. and Mexico and for the last 40 yr
500 yr records of river flow (e.g., Stockton and Jacoby, 1978;
Colorado River water has discharged into the Gulf of California
Meko et al., 1995). Stable isotopes of strontium, oxygen, and
only during unusually wet years. The Colorado River compact
carbon from estuarine mollusk shells have been employed to
of 1922 allocated the water supplied by the Colorado River
extend the record back several millennia (see the pioneering
among California, Arizona, Colorado, Utah, Nevada, and New
work of Ingram and Sloan, 1992; Gagan et al., 1994; Ingram et
Mexico. The amount of water distributed by this pact (16.2
al., 1996), but calibration of isotopic variation with discharge
million acre-feet or 2.0 1010 m3) was based on approximately
variation has been very difficult. Other proxy indicators (typi-
15 yr of historical flow records covering an interval that turned
cally based on proportions of particular estuarine species) have
out to be a time of unusually high flow in the century-long
provided only semiquantitative estimates of salinity, and thus,
historical record of discharge (USGS, 1954; Hely 1969; Stock-
indirectly, river discharge.
ton and Jacoby, 1978). Thus, there is great interest in develop-
In this paper, we describe a method based on comparing
ing longer records of river flow to see if the twentieth century
seasonal isotopic variation in shells grown in the absence of
was typical or anomalous relative to the last one or two mil-
any river water influence with shells grown in the presence of
lennia. Tree ring studies have been used to reconstruct the last
a known amount of river water. When coupled with informa-
450 yr of precipitation and river flow in the upper Colorado
tion on the 18O mixing relationship of seawater and river
River basin (Stockton and Jacoby, 1978; Meko et al., 1995).
water, the offset between river-influenced and “no-flow” shell
These studies have suggested that the allocated flow is probably
20% greater than the average flow of the last four centuries.
Our study takes a different approach to the question of
* Author to whom correspondence should be addressed
([email protected]).
estimating long-term flow by looking at the geochemical record

Present address: Dept. of Geology and Geog., Denison University, of flow reaching the mouth of the Colorado River (summing
Granville, OH 43023, USA. both the upper and lower Colorado River basins). Using shell
1254 D. L. Dettman et al.
chemistry has the advantage of applicability over a much
longer time interval because individual shells can be directly
dated using 14C or radiocarbon calibrated amino acid racem-
ization. A disadvantage of this approach in comparison to the
tree ring records is the inability to construct a continuous
annual time series greater than a few years in length. The
species used in this study, Chione fluctafraga and Chione
cortezi, can live up to 18 yr (Schone et al., 2003) but after three
or four years shell growth slows dramatically making detailed
sampling of complete years difficult. Because our fossil spec-
imens are dated to 50 yr using a 14C calibrated amino-acid
racemization chronology developed using Chione from the
Gulf of California (Kowalewski et al., 1998), we can only
sample one or two years from a shell within 50-yr intervals.
Our ultimate goal in this research is to reconstruct the amount
of water delivered to the Gulf of California over the last 2000
yr. The ancient record will be presented in another paper; this
paper describes in detail our method for reconstructing ancient
river flow.
3.1. Mollusk Sclerochronology and the Annual Record
Mollusks add new carbonate material to their shells throughout their
life spans. They can be thought of as chart recorders that store chemical
information in shells as they grow. However, these chart recorders do
not run continuously or at constant speed. Typically, bivalves grow
rapidly early in life with a considerable deceleration in growth rate with
increasing age. In addition, bivalve shells frequently show growth
bands in cross section with a hierarchy of periodicities (Barker 1964; Fig. 1. Study area. Shells were collected from Isla Montague, Saca-
Jones et al., 1978; Jones and Quitmyer, 1996). The most prominent and tosa, Isla Pelicano, and El Golfo de Santa Clara (primary collection
least frequent bands are usually thought to be annual markers, often localities indicated by circles). Water samples were collected from
caused by a slowing and cessation of shell deposition due to environ- many locations between Puerto Penasco and Campo Don Abel.
mental stress (i.e., temperature extremes or turbid conditions). These
annual bands can be very useful in guiding the sampling of the shells
to recover clear seasonal cycles of environmental change from the
shell. In addition to annual bands, the genus of bivalve used here, different intervals of low sediment input (Thompson, 1968; Kow-
Chione, also shows clear daily growth bands whose widths are mod- alewski et al., 1994). The shells in a particular chenier are older than the
ulated by the fortnightly tidal cycle (Schone et al., 2002). This internal
¨ age of chenier formation, although there is an extremely wide range of
shell sclerochronology allows us to place our samples in the context of ages represented in each shell accumulation. Mulinia coloradoensis
a fortnightly calendar within the annual bands. makes up the vast majority of the shells in the cheniers (up to 95% —
Cool temperatures lead to cessation of Chione shell growth during Kowalewski et al., 1998), but this species is now nearly extinct in the
the winter months, from approximately mid-December to mid-March gulf (Rodriguez et al., 2001). We focused our collecting efforts on
(Goodwin et al., 2001), thus part of the annual environmental record is Chione species (C. fluctifraga, C. cortezi), because they are currently
lost. On average, stable isotope data for twenty fortnights are recorded living in the study area and provide the opportunity for stable isotope
in the shell. To estimate river discharge during the remaining six calibration studies relative to the modern environment. Collecting trips
fortnights, we use the US Geological Survey’s river flow record for the occurred at infrequent intervals over the last 8 yr. The Baja California
Colorado (US Geological Survey, 1954). Thirty years of data exist (western) side was usually visited in November and Isla Montague was
before the construction of the first dam. The average amount of flow usually visited in February. Open gulf water samples were collected on
during the six fortnights from mid-December to mid-March is 12% each of these trips.
3% of the total annual flow. Allowing for 1 fortnight uncertainty
in the dating of the fortnights in the sclerochronology leads to an
3.3. Methods
additional error of 3%. Thus, the unrepresented six fortnights account
for 12% 6% of the total annual river flow. Therefore adding 12% to Salinity of seawater samples was measured with an optical salinom-
our calculated discharge yields an annual discharge. eter with a 1 ppt precision. Stable isotope composition of water was
measured on a Finnigan MAT Delta-S gas ratio mass spectrometer
3.2. Sampling using two automated sample preparation devices. Oxygen isotope com-
position was measured using CO2 equilibration at 15°C for a minimum
Live mollusks were collected from tide pools, usually low in the of eight hours. Samples were not distilled before equilibration. Hydro-
intertidal zone. Live animals were collected at Sacatosa, Isla Montague, gen isotope composition was measured by reducing the water on
Isla Pelicano, and El Golfo de Santa Clara (Fig. 1). Water samples were chromium at 750°C and direct transfer of H2 into the mass spectrom-
usually collected from the tide pools along with the live-collected eter. Samples were normalized to V-SMOW and V-SLAP based on
animals. Subfossil shells were collected from cheniers along the west- secondary standards. Repeated standards had a standard deviation of
ern side of Isla Montague and in the Sacatosa area. Cheniers are 0.06‰ for 18O and 0.8‰ for D.
wave-formed shell islands created during times of low sediment input Shells were sectioned along the axis of maximum growth, mounted
to the gulf. During times of low sediment input, fines are transported to as thick sections ( 1 mm) on glass slides, and polished. Growth bands
deeper waters and shells are winnowed and gathered into long linear in cross-sectioned shells were subsampled using either a stabilized
islands paralleling the shore line. Multiple generations of cheniers mark dental drill and 0.5 or 0.3 mm diameter dental burs or a computer-
Oxygen isotope ratios and Colorado River discharge 1255
controlled micro-mill capable of recovering shell material with a 20
micron spatial resolution (Dettman and Lohmann, 1995). Carbonate
samples were processed in four different labs, University of Arizona,
University of Michigan, University of California at Santa Cruz, or
University of California at Davis. All were reacted with dehydrated
phosphoric acid under vacuum. The measured 18O and 13C values
were normalized to NBS-19 based on internal lab standards. Precision
of repeated standards is 0.1‰ for 18O and 0.06‰ for 13C (1 ).
The Sr87/Sr86 ratios of waters were measured in the isotope geo-
chemistry lab of the Department of Geosciences, University of Ari-
zona, following the methods described in Patchett and Ruiz (1987) and
Gross et al. (2001).
4.1. Modern Colorado River Delta Environment
Before its damming and diversion, the Colorado River emp-
tied into the Gulf of California. The northern end of the gulf is
surrounded by the Sonoran Desert and is a very hot and arid
region; monthly average temperature ranges from 16 to 34°C.
The northern gulf is known for its very large tidal range, often
in excess of 5 m. Local rainfall amounts are 60 mm annually
(Hastings, 1964), local runoff is negligible, and the diversion of
the Colorado River has led to very little fresh-water input in the
last half century. Evaporation therefore dominates the present-
day salinity and stable isotope character of seawater in the
northern end of the Gulf of California. In this hot desert
environment the large tidal amplitude contributes to the evap-
orative signature by running a relatively thin layer of water
across intertidal zones that can be tens of kilometers in width.
Salinities are high, ranging from 35 to 42 (practical salinity
units) (Fig. 2). Waters trapped in tidal pools or those draining
ponded areas in the intertidal zone can reach salinities as high Fig. 3. (a) Water samples collected during times when the Colorado
as 240. When the Colorado River is not flowing into the Gulf River was not flowing. (b) Waters collected during or shortly after
releases of Colorado River water.
of California, the isotopic composition of seawater is always
positive, typically ranging from 0.3‰ to 0.8‰ SMOW. The
mean 18O for water collected offshore in the northern end of
the gulf is 0.60‰ 0.16 (1 ) SMOW. Our measurements of River water into the Gulf of California (Fig. 3b). This is seen in
the 18O of water from tide pools ranged from normal seawater the salinity and isotopic composition of water samples during
values up to 7.6‰ SMOW. The D– 18O relationship shows months when significant amounts of Colorado River water
that both open Gulf of California seawater and tide pool waters entered the northern end of the Gulf of California. The trends
are on evaporative trends with slopes of 5 and 4 respectively in both salinity and 18O or D show simple linear mixing
(Fig. 3a). between gulf water (salinity 38, 18O 0.60‰, D
Water samples with 18O values less than 0‰ SMOW are 2.8‰) and Colorado River water (average 18O 12.0‰,
primarily the result of recent controlled releases of Colorado D 95‰).
Two types of temperature records show that the shallow
water in the northern end of the Gulf of California is warm to
hot throughout the year. Monthly average sea surface temper-
ature from satellite data ranges from 11 to 31°C, integrated over
a 14 km grid, for the grid cell immediately south of Isla
Montague (NOAA-CIRES, 2000). Water temperatures in the
low intertidal zone were recorded every two hours from mid-
February 1999 to mid-February 2000 (using a HOBO-Temp®
temperature logger). Averaged monthly or fortnightly temper-
atures cover the same range as the satellite data, but show
extreme variability on short time scales, up to 15°C differences
between day and night (Fig. 4). High temperatures occur in-
variably in the late afternoon, no matter what phase of the tidal
cycle is active. This indicates that both ambient seawater tem-
Fig. 2. Salinity vs. 18O for water samples in the northern Gulf of perature and direct solar heating play roles in controlling water
California. Samples with salinity greater than 42 are tide pool samples. temperature in the intertidal zone.
1256 D. L. Dettman et al.
Fig. 4. Digital temperature recorder data from Isla Montague. Solid Fig. 6. Photograph of polished cross section of a small Chione
line is a continuous temperature logger record from the intertidal zone cortezi ( 1 yr old) showing one fortnight of growth. Daily growth
from mid-February 1999 and to mid-February 2000. Squares are aver- increments are grouped into tidally controlled bundles.
age temperature of fortnights.
relatively easy to count in Chione sp. and in most cases are
4.2. Shell 18O in the Presence and Absence of Colorado
clearly grouped into fortnightly bundles that match the tidal
River Water
cycles (Fig. 6). The widest daily growth bands approach 0.25
The 18O of shell material depends on the temperature and mm. Growth in Chione slows and stops in the cooler winter
O of the water in which the specimen lives. Under today’s months resulting in an average of 20 fortnights of identifiable
conditions, where no river water reaches the gulf, the oxygen shell growth in a year. An exception to this is the first winter of
isotope profile in a shell is primarily the result of the seasonal the animal’s life, where growth often continues through the
temperature change in the gulf, possibly combined with a small winter. All shell 18O data presented here are from the first four
evaporation-driven cycle in seawater 18O. A comparison of full years of life of the animal. The sclerochronological uncer-
five annual profiles in the 18O of live-collected shells from tainty is most likely 1 fortnight (Schone et al., 2002).
no-flow years shows relatively good agreement, even though The total 18O range within the five shells from no-river-
the specimens lived in different years and at different locations flow conditions is approximately 0.6‰ to 2.2‰ PDB (Fig.
around the mouth of the river (Fig. 5). Figure 5 and all other 5). The differences between the curves are due to temperature
shell plots show 18O vs. fortnight number. The sclerochrono- and water 18O variation between the years and locations of
logical methods are discussed in detail in Schone et al. (2002)
¨ growth. Localized evaporation of seawater in tide pools plays a
and will be described only briefly here. Daily increments are role in offsetting the 18O values to more positive values on the
time scale of the tidal cycle. This probably explains the jagged
nature of some of the curves. There is also uncertainty in the
matching of fortnights, which may offset records horizontally.
This figure shows that the 18O variance between shells during
no-flow years is probably on the order of 0.5‰ PDB. Averag-
ing these records yields a relatively smooth curve that removes
evaporative anomalies of any single measurement. This aver-
aged 18O curve (line in Fig. 5) is what we call the “no-flow
record.” This represents the expected 18O cycle of Chione sp.
shell carbonate at the mouth of the river when no river water
reaches the Gulf of California.
Because the 18O of Colorado River water is much more
negative than seawater, its presence in the delta region should
lead to 18O values that are more negative than the no-flow
record of Figure 5. Many fossil shells show a large offset to
more negative 18O values (Fig. 7). Shell IM4-D2 (late 18th
century) has a minimum 18O value of 8.75‰ PDB and
reveals a sustained pulse of river water lasting at least six
fortnights. Shell NI2-D14 (late 11th century) has a much shorter
peak in river input and reaches a minimum of 7‰. Recent
Fig. 5. 18O records for five live-collected shells in years during
which no Colorado River flow occurred. Shell numbers beginning with
releases of known amounts of river water to the gulf have also
IM are from Isla Montague, IP are from Isla Pelicano, and EG are from been recorded in live-collected shells (Fig. 8). These anomalies
El Golfo. The “no-flow record” is the average of these by fortnight. can be matched to known amounts of water released and, as
Oxygen isotope ratios and Colorado River discharge 1257
discussed below, allow us to calibrate the relationship between
isotope offsets and river discharge.
4.3. An Aside on Sr/86Sr Ratios
Some studies quantifying river flow into estuarine systems
have used strontium isotope ratio gradients to measure the mix
of seawater and fresh-water (e.g., Ingram and Sloan, 1992). The
use of 87Sr/86Sr in the mixing zone has a major advantage over
oxygen isotope ratios because there is no temperature effect
and no measurable fractionation between the isotope ratio in
water and that preserved in shell. However, because there is
usually a much higher concentration of strontium in seawater
than in fresh-water, high precision 87Sr/86Sr analysis is needed
to calculate variations in salinity as the system approaches the
marine end-member (Bryant et al., 1995). In our study area,
seawater Sr concentration is 8.4 ppm at El Golfo and Colorado
River water is 1.2 ppm at Lake Mead (Gross et al., 2001). The
Sr/86Sr of water from the mixing zone of the delta during a
significant release of river water to the Gulf of California (Feb.
1999) shows moderate change as salinity varies from 36 to 29,
but because of the uncertainties on the 87Sr/86Sr measurement,
we cannot distinguish between salinities of 36 and 33 (Table 1).
The Ingram and Sloan (1992) study of waters from San Fran-
cisco Bay also failed to measure significant differences in
Sr/86Sr ratio across a range of salinities from 24.4 to 29.5,
Fig. 8. (a) Shell 18O profile from a shell collected in 1996 at Isla
Montague (IM2-EC2). The release of river water in early 1993 is
recorded in that year’s shell growth. (b) Shell 18O profile from a
Chione collected in November of 1998 at Sacatosa (ST8-A2R). A river
water release in February of 1998 is seen in the 2.2‰ offset during
first winter growth before fortnight 1 (see text).
although clear differences should have been present based on
their modeling. This suggests that oxygen isotope ratios may be
more effective in quantifying salinity variation when conditions
approach normal marine salinity. The major advantage of stable
oxygen isotopes over 87Sr/86Sr ratios in this study is practical.
Because river discharge varied greatly within the year, we use
oxygen isotope ratios, which are easy and inexpensive to ac-
quire, for extensive intra-annual documentation of this varia-
5.1. O Differences and the Mixing Zone
The comparison of the no-flow 18O profile with the profiles
of fossil shells allow us to track seasonal change in the 18O of
gulf water in the delta region through time. We minimize the
temperature effect on shell carbonate by comparing the 18O
values by fortnight. We assume that the average temperature of
a particular fortnight is similar for all shells, no matter what
year the carbonate was produced (see the discussion of error
estimates, below). If this is the case, then the only factor
Fig. 7. Shell 18O profiles from 2 shells dated using amino acid causing a difference in the 18O of the shells is the 18O of the
racemization (Kowalewski et al., 1998). Dates are tied to 14C ages with water in which the animal lived. The most important feature
uncertainties of approximately 50 yr. (a) Year 3 of a shell dated to controlling the isotopic composition of seawater in the mixing
1775 AD (IM4-D2) from Isla Montague. Note the strong river water
spike in fortnights 7 to 12 and the overall more negative values
zone is the amount of river water being delivered. Thus the
throughout the profile in comparison to the no-flow profile. (b) Year 2 O difference in a fortnight-by-fortnight comparison of fossil
of a shell dated to 1075 AD (NI2-D14) from Sacatosa. shell with the no-flow record gives us the change in the 18O of
1258 D. L. Dettman et al.
Table 1. Sr isotope ratios from waters collected in February 1999.
Location Salinity 84Sr/86Sr 2 87Sr/86Sr 2 87Sr/86Sr (norm)1
El Golfo 36 0.056516 19 0.709194 18 0.709174
Isla Montague 33 0.056493 12 0.709221 13 0.709201
The Y 29 0.056492 12 0.709282 11 0.709262
NBS-987 std 0.056500 13 0.710255 5
Colorado River (Cibola, Mar 96)2 0.710274 13 0.710254
Colorado River (L. Mead, Jan 99)2 0.710437 16 0.710417
Ratio after NBS-987 is normalized to 0.710235 (Hodell et al., 1991; Farrell et al., 1995).
Data from Gross et al., 2001.
gulf water due to the addition of Colorado River water at the fresh-water bivalve shells collected before the dams were built
location of bivalve growth. to calculate the seasonal variation in 18O of fresh-water being
Almost all of our fossil samples are more negative in 18O delivered to the Gulf of California by the river. Oxygen isotope
than the no-flow bivalve record. This offset to more negative profiles of two Anodonta dejecta, collected in the late 1800s
values is due to growth in less saline water in the delta region. from the river near Yuma show a large change in 18O (Fig. 9).
The magnitude of this difference allows us to quantify the Using average river water temperatures and stable isotope
amount of fresh-water at the place and year of shell growth fractionation factors calibrated for these fresh-water bivalves
with fortnightly resolution. Note that there are a few times (Dettman et al., 1999), we can calculate the 18O of the river
when the shell 18O is more positive than the no-flow profile water if seasonal shell growth can be modeled. Unfortunately,
(e.g., Fig. 8b). This is most likely due to shell growth in water daily bands are not visible in these shells and there are no tides
that is more positive in 18O than normal seawater due to to generate fortnightly growth markers. Note, however, that
evaporation, probably in a tide pool environment. These occur- uncertainties in temperature lead to only small uncertainties in
rences, attributable to very localized evaporation, are not used the calculated 18O of the water because it takes a 4.6°C change
in our calculations and we treat them as identical to the no-flow in temperature to change shell 18O by 1‰. Thus a mis-
profile. assignment of temperature by 4.6°C leads to a 1‰ error in the
Because salinity and 18O are conservative properties, the calculated 18O of the water.
mixing of waters yields a simple proportional sum of the two Growth in this subfamily of fresh-water bivalves is not
waters’ characteristics. If we know the 18O of end-members continuous throughout the year; some genera of unionid bi-
(pure river water and pure seawater) and the 18O change in the valves hibernate below temperatures of 12°C (Dettman et al.,
water due to the addition of river water, we can calculate the 1999). They are also very sensitive to suspended sediment and
percentage of river water at these locations during a particular may have stopped growing during the full flood-stage flow
fortnight. While the seawater end-member may change slightly triggered by snow melt in the Rocky Mountains. In each year
seasonally, our sampling has not been adequate to document (identified by the 18O cycle) there are two bands in the shell
seasonal or spatial variation. We therefore use the mean sea- where a thickened prismatic layer shows growth cessation. Our
water 18O value ( 0.60‰ SMOW) for the seawater end- growth model for these bivalves is based on these hiatuses: the
member of our mixing relationship. winter months are missing due to cold temperatures and are
Before dams were emplaced on the Colorado River, the associated with the growth break at the more positive 18O
seasonal 18O cycle in the river was very large. The isotopic values (approximately 6 to 8‰ PDB); the annual midsum-
cycle was mainly driven by a large

Use: 0.0991