Geochemistry, Geophysics, Geosystems

Reconstructing past upwelling intensity and the seasonal dynamics of primary productivity along the Peruvian coastline from mollusk shell stable isotopes

Authors


Abstract

[1] We present here a potential new method to evaluate past variations of the mean intensity of Peruvian coastal upwelling and of the seasonal timing of phytoplankton blooms. This method uses a combination of the monthly carbon and oxygen isotopic signals preserved in fossil mollusk shells, and a series of corrections to extract the variations of the dissolved inorganic carbon (DIC) δ13C. Based on the analysis of five shell samples (85 shells in total) from the southern Peruvian coast, we suggest that the mean coastal upwelling intensity can be determined from a linear relationship between average values of corrected shell δ13C and δ18O. This new potential proxy would bring additional independent information valuable to interpret paleoproductivity changes reconstructed from marine sediment of the nearby continental shelf. Results obtained on fossil samples from the middle Holocene show an increase in upwelling intensity during this period associated to a spatial reorganization of upwelling centers along the South Peruvian coast. At the seasonal scale, corrected shell δ13C enrichment indicates a phytoplankton bloom. Seasonal timing of phytoplankton blooms can be estimated by the lag with the annual temperature cycle reproduced by shell δ18O monthly variations. The results obtained with two modern shell samples indicate phytoplankton blooms occurring during summer and fall, consistently with in situ productivity observations. Our method relies on revisited assumptions about the influence of temperature and metabolism in mollusk shell δ13C. We further discussed the validity of these assumptions and the potential implications for the interpretation of similar data sets.

1. Introduction

[2] Marine phytoplankton performs approximately 50% of the net global carbon fixation through photosynthesis, and thus acts as a critical link between the organic and inorganic carbon reservoirs [Behrenfeld et al., 2006; Alvain et al., 2008]. This biological carbon fixation is strongly influenced by climate change in the low latitude oceans, with warmer waters increasing stratification and reducing productivity [Cox et al., 2000; Bopp et al., 2001; Behrenfeld et al., 2006; Steinacher et al., 2010]. The majority of marine primary productivity in the tropics is concentrated in upwelling zones where it is promoted through a supply of nutrients. Upwelling intensity is a major but complex parameter within the global carbon cycle. Deep waters saturated in CO2 are upwelled and act as a carbon source upon exchange with the atmosphere, while the high primary productivity increases the oceanic biological carbon pump. The response of primary productivity to the rate of upwelling may also be limited by alternate parameters such as terrestrial iron input [Martin et al., 1994; Hutchins et al., 2002; Dezileau et al., 2004] or light availability [Cole and Cloern, 1984]. A comparison of 30 model simulations against in situ data from the tropical Pacific by Friedrichs et al. [2009] revealed that the majority of models significantly underestimate current primary productivity. Estimates of the response of upwelling intensity and marine productivity to greenhouse warming remain highly uncertain.

[3] Reconstructions of past changes in upwelling intensity and marine productivity are required to fully understand their controls and to constrain model simulations. Many studies using marine sediment cores have been devoted to this issue. Despite their high interest, these studies often face two limitations: (1) paleoupwelling intensity are often inferred from paleoproductivity indicators based on the assumption that upwelling systematically promotes productivity, which is not the case in high nutrient low chlorophyll (HNLC) zones such as the Peruvian upwelling and (2) the mean annual carbon export budget simulated in biogeochemical models is an integration of seasonal scale processes that cannot be studied in sediment cores as a result of their low temporal resolution. Seasonal variability in the coastal Peruvian upwelling is recorded in mollusk shell Δ14C signals [Jones et al., 2009]. For ancient periods, variations of mean paleoupwelling intensity can be indicated by variations of marine reservoir age estimated by the radiocarbon analysis of contemporaneous marine shells and plant remains [Fontugne et al., 2004; Ortlieb et al., 2011]. However, this method has significant uncertainties because of potential old wood effect and intrashell variability of reservoir age [Jones et al., 2007, 2010]. We propose here to use an alternative paleoenvironmental archive that would yield independent estimates of the coastal upwelling intensity and insights into seasonal scale productivity processes.

[4] The potential of oxygen and carbon stable isotopes in mollusk shells as high resolution paleoupwelling archives has been recognized [Killingley and Berger, 1979; Krantz et al., 1987, 1988; Jones and Allmon, 1995], yet remains poorly exploited, primarily because of the multiple influences involved in mollusk δ13C. Here we propose a method combining oxygen and carbon stable isotope records (δ18O and δ13C respectively) in shells of the mollusk Mesodesma donacium resolved to monthly timescales to study the past variability of (1) the average coastal upwelling rate, and (2) the seasonal timing of coastal phytoplankton blooms, along the coast of Peru. We then applied the method to two fossil shell samples from the middle Holocene and the Inca period for a comparison against modern conditions.

1.1. Coastal Peru

[5] Coastal Peru is characterized by year-round southeasterly trade winds that generate strong Ekman transport and coastal upwelling [Brink et al., 1983; Huyer et al., 1987]. Because the continental slope is so close to the shore in Peru, the surface conditions over the Peruvian shelf are closely related to the coastal upwelling [Strub et al., 1998]. Deep waters are upwelled along the Peruvian coast in localized upwelling centers and then advected offshore. Therefore, a steep sea-surface temperature (SST) gradient is observed between cool coastal waters and relatively warmer waters offshore. Deep waters are upwelled throughout the year along the Peruvian coastline, with a maximum experienced during the austral winter (June to September).

[6] The vertical advection of deep water provides an annual supply of nutrients to the surface, supporting a highly productive phytoplankton bloom [Echevin et al., 2008; Taylor et al., 2008]. Average chlorophyll a concentrations decrease seaward from ∼10 mg/m3 on the coast to ∼1 mg/m3 ∼300 km offshore [Echevin et al., 2008]. Estimates of the paleoenvironmental conditions in this region are largely based on proxies derived from phytoplankton preserved within sediment cores from the continental shelf, approximately 30 to 140 km offshore, within the close influence of the coastal upwelling [Mohtadi et al., 2005]. Whereas low frequency temporal variability (≥101 yr) of average productivity is expected to be similar on the coastline and over the shelf, coastal and shelf phytoplankton blooms are characterized by different assemblages with different seasonal timings and strength [González et al., 2004; Ochoa et al., 2010]. Therefore the environmental conditions of the coastline recorded by mollusk shells may be significantly different, at the seasonal scale, from the conditions over the shelf recorded by sediment cores.

[7] Satellite data, in situ sampling and model simulations indicate that blooms along the Peruvian coastline currently occur during austral summer (Figure 1), in opposition to the seasonal phasing of upwelling intensity [Thomas et al., 1994; Chávez et al., 2008; Echevin et al., 2008]. Therefore, unlike the physically analogous Chilean region of the Humboldt Current [González et al., 2004], phytoplankton growth in the coastal Peruvian upwelling is not limited by the nutrient supply from the deep sea, but potentially by light (low stratus cloud cover is persistent from June to November) or the supply of bio-available iron such as observed in the HNLC equatorial Pacific [Coale et al., 1996].

Figure 1.

Monthly averages of chlorophyll a concentration along the Peruvian coastline from 1997 to 2005. Closed and open circles indicate IMARPE in situ and SeaWIFS surface chlorophyll a data respectively, averaged from the 250 km wide coastal zone between 4 and 15°S. The highest chlorophyll a concentrations were consistently observed during February, March and April, i.e., austral summer. Modified from the work of Echevin et al. [2008]. Monthly sea surface temperature were averaged for the 1925–2002 period from IMARPE in situ measurements in Puerto Chicama, Peru.

1.2. Environmental Information Potentially Reconstructed From Mollusk Stable Isotopes

[8] High resolution microsampling of M. donacium shells (see section 3.1) allows the evaluation of monthly variations in both δ13C and δ18O incorporated in shell aragonite (CaCO3). The oxygen isotopic fractionation between aragonite and seawater is a temperature dependent reaction; therefore the δ18O signal from M. donacium provides a proxy record for temperature reconstructions provided the seawater δ18O can be estimated [Carré et al., 2005a].

[9] The carbon isotopic composition of mollusk shells may have multiple influences: (1) the isotopic composition of the dissolved inorganic carbon (DIC), δ13CDIC, (2) the proportion of metabolic carbon in carbonate ions involved in the aragonite precipitation (metabolic effect) (see discussion in section 4.2), (3) kinetic disequilibrium effect (see McConnaughey and Gillikin [2008] for a detailed review) and (4) a temperature-related fractionation for which the mechanisms remain poorly constrained (see discussion in section 4.1.2). In section 4, we discuss a potential method for correcting shell δ13C so shell profiles only reflect δ13CDIC variations, which contain the desired environmental information.

[10] Photosynthesis reactions preferentially make use of isotopically lighter 12CO2. This has two implications: (1) at the ocean scale, as phytoplankton die this 12C enriched organic matter sinks to the seafloor and is mineralized at depth. Deep waters upwelled along the Peruvian coast are therefore characterized by 13C depleted DIC. (2) On a local scale, a phytoplankton bloom induces a 13C enrichment of the surface water DIC during the growth phase and a depletion of 13C during the following organic matter degradation phase. Thus, if the isotopic carbon ratios in mollusk shells are representative of marine DIC, they may provide insights into the variability of upwelling activity and primary production at monthly to millennial timescales [Killingley and Berger, 1979; Jones and Allmon, 1995]. Furthermore, δ13CDIC of surface waters is also influenced by air-sea CO2 exchange which depends on the difference of CO2 partial pressure (pCO2) between the atmosphere and the ocean [Lynch-Stieglitz et al., 1995]. The relative contribution of these environmental controls over the δ13CDIC depends on local oceanographic features and will be discussed. The isotopic influence of one possible complicating factor, the influence of terrestrial freshwater runoff, is not significant in Peru because of the extreme aridity (Figure 2) and the weak discharge of coastal rivers (Worldclim 1.4 data set; available at http://www.worldclim.org/current), as confirmed in situ by monthly salinity measurements [Carré et al., 2005a].

Figure 2.

The two sampling sites of Rio Ica and QLB in Southern Peru where both modern and fossil M. donacium shells were collected. Greyscale patterns over the continent indicate annual precipitation (Worldclim 1.4 data set). Color patterns in the ocean indicate mean annual sea surface temperature (NASA's MODIS data set). Cooler SSTs along the coast are related to the presence of upwelling, which appears stronger at 15°S near the Ica site than at 18°S near the QLB site.

2. Material and Study Sites

[11] The bivalve Mesodesma donacium inhabits the subtidal to intertidal zone of sandy beaches along the Peruvian and Chilean coastline, and provides typical individual records of approximately one to two years [Carré et al., 2005a]. Fossil mollusk shells were collected from two archeological sites along the Peruvian coastline: Rio Ica (14°52′S, 75°33′W) [Engel, 1957] and Quebrada de los Burros (QLB) (18°10′S, 70°38′W) [Lavallée et al., 1999; Carré et al., 2009] (Figure 2). Modern shell samples were collected from the respective nearby beaches in small mounds formed by fishermen activity during the second half of the 20th century. The sample size ranged from 13 to 20 shells (Table 1). Preservation of these historical shells was particularly high (as indicated by the remains of preserved organic matter) due to the extremely arid Peruvian coastline which receives precipitation of less than 20 mm yr−1 (Worldclim 1.4 data set). Sample horizons of the archeological sites were dated by radiocarbon AMS measurements on charcoal fragments recovered among shell material (Table 1). A total of 85 shells were analyzed.

Table 1. Radiocarbon AMS Dates Measured on Charcoal Samples Associated to the Sampled M. donacium Shellsa
SiteSample NameNumber of Shells14C Date (B.P.)Laboratory ReferencePeriod
  • a

    Three time periods were represented by the shells collected; the modern day, Inca period and middle Holocene.

  • b

    Dates published by Lavallée et al. [1999].

15°SIcamod15N/A Modern
 IcaInca20510 ± 40OS- 65628Inca
 IcamH205840 ± 35OS- 60543Middle Holocene
   5900 ± 40OS- 60564 
   5940 ± 45OS- 60556 
   6070 ± 30OS- 60544 
18°SQLBmod17N/A Modern
 QLBmHb136460 ± 60GifA-97287Middle Holocene
   6510 ± 60GifA-97288 
   6630 ± 70GifA-97289 

3. Methods

3.1. Shell Preparation and Microsampling

[12] Mollusk shells were embedded in polyester resin and radially sectioned using a diamond wire saw. Sections 1 mm thick were mounted onto glass slides and polished before sampling of the outer shell layer, along the direction of growth, with a Micromill™ automated microsampler. Each microsample constituted about 0.1 mg of powdered calcium carbonate, and represented about one month of shell growth based on shell inner growth lines (see Carré et al. [2005a, 2009] for details about sclerochronology). With this technique, occasional areas of recrystallized carbonate located on the shell surface can be identified by their color and avoided in the microsampling. Despite the high spatial control of the microsampler, an uncertainty remained over the identification of tidal growth lines and thus in the temporal resolution. The time span represented by a single microsample ranged between 0.5 to 2 months. Based on the δ18O profiles and the number of microsamples in a peak-to-peak 1 year cycle, the average time resolution of shell isotopic signals was re-estimated to be ∼1.1 months.

3.2. Isotopic Analyses

[13] Approximately 0.05 mg of aragonite powder was sub-sampled from every microsample and analyzed for carbon (δ13C) and oxygen (δ18O) stable isotope composition with a Kiel III carbonate device coupled to a Finnigan Delta Plus isotope ratio mass spectrometer. The carbonate device drops excess 100% phosphoric acid onto calcium carbonate samples held at 70°C. Measurement reproducibility was 0.04‰ for δ13C and 0.07‰ for δ18O based on routine analyses quality control calcite standards. Isotopic values were placed on the VPDB scale by measurement against internal laboratory calcite standards that were measured against the international standards NBS19 and LSVEC [Verkouteren, 1999] for δ13C and NBS19 and NBS18 for δ18O.

4. Results and Discussion

4.1. Influences on Shell δ13C

[14] Besides the seawater δ13CDIC, mollusk shell aragonite δ13C may also be influenced by kinetic fractionation, temperature, and the contribution of metabolic (also referred to as respired) carbon. We discuss below these influences in the conditions of our study.

4.1.1. Kinetic Effect

[15] A kinetic effect can be diagnosed by highly depleted values and a strong positive correlation between shell δ13C and δ18O [McConnaughey et al., 1997]. Since this was not observed in M. donacium shells, we discounted the influence of a significant kinetic effect in our study.

4.1.2. Temperature Influence

[16] From the analysis of seawater δ13CDIC and aragonite δ13C in several mollusk species, Grossman and Ku [1986] calculated empirical equations of the temperature dependent carbon fractionation in biogenic aragonite for Hoeglundinae and mollusks between 2.6 and 22°C:

equation image
equation image

Later, Romanek et al. [1992] studied carbon fractionation of inorganic aragonite and calcite and found no temperature dependence for the aragonite-HCO3 fractionation, suggesting that “aragonite-secreting taxa are influenced by temperature-dependent biological disequilibrium effects (“vital effects”).” After this temperature dependence observed by Grossman and Ku [1986] was qualified “vital effect,” most researchers studying carbon stable isotopes in mollusk shells used the fractionation factor calculated by Romanek et al. [1992] and considered that carbon isotopes in biogenic aragonite were not influenced by temperature [McConnaughey et al., 1997; Lorrain et al., 2004; McConnaughey and Gillikin, 2008; Gillikin et al., 2006a; Lartaud et al., 2010].

[17] In contrast to these researchers, we have returned to the Grossman and Ku [1986]equation (2) for three reasons: (1) the empirical data set indicates a highly statistically significant, yet currently mechanically undefined relationship between temperature and ε13mollusk-DIC; (2) as far as we know, no published data set using biogenic aragonite has invalidated this model, and (3) since we are studying biogenic aragonite, Grossman and Ku [1986] model is more appropriate than Romanek et al. [1992] model on inorganic aragonite. To support this latter reason, it seems important to remind that the mineralization processes for biogenic and inorganic aragonite are very different chemically and thermodynamically. First, aragonite does not spontaneously precipitate in the ocean and only exists because of shell organic matrix [Watabe and Wilbur, 1960; Suzuki et al., 2009]. The crystal growth in shells is fully biologically controlled (timing, place, rate, and directions) through complex interactions with a large range of multifunctional proteins that constitute the shell organic matrix [Watabe and Wilbur, 1960; Wheeler et al., 1981; Addadi and Weiner, 1992; Marin and Luquet, 2004; Marin et al., 2007; Suzuki et al., 2009]. The processes of organic aragonite precipitation involves the catalytic action of enzymes [Bevelander and Benzer, 1948; Freeman and Wilbur, 1948; Marxen et al., 2003, Marin and Luquet, 2004] which have a temperature dependent efficiency [Somero, 1969; Ikemoto, 1975; Hall, 1985; Pick and Karlish, 1982; Wolfenden et al., 1999]. Enzymatic reactions also induce a temperature dependent isotopic fractionation [Northrop, 1981; Schowen and Schowen, 1981; Liu and Warshel, 2007]. A temperature-dependent 13C fractionation between HCO3 and the other inorganic carbon species may also be involved. Studying the mechanisms behind a δ13C temperature effect in biogenic aragonite is beyond the scope of this paper, but it is clear that the question requires further research. The correlation between temperature and ε13mollusk-DIC may only be indirect and not causal but it still reveals the existence of a chemical and/or biological effect that can be estimated through temperature. The fact that Grossman and Ku [1986] calculated a similar equation for the foraminifera Hoeglunidae and twelve species of mollusks suggests that this effect may indeed be characteristic of a large range of aragonite-secreting organisms.

[18] In light of these observations and until further results suggest otherwise, correcting our mollusk shell δ13C data for an expected and quantified temperature effect using δ18O based temperature reconstructions [Carré et al., 2005a] was selected as the most suitable course of action.

4.1.3. Correcting Shell δ13C for the Temperature Effect

[19] We analyzed the oxygen isotopic composition of seawater samples from the southern coast of Peru collected between 2002 and 2007 (Table 2). In the Rio Ica area, average isotopic values were δ18Oseawater = 0.26‰V-SMOW (σ = 0.11, N = 8), whereas in the QLB area it was estimated at δ18Oseawater = 0.21‰V-SMOW (σ = 0.15, N = 20). The water δ18O on this coast displays little variation and no seasonality because it is not significantly affected by freshwater input. This was confirmed in modern conditions by seawater δ18O measurements (Table 2) and monthly salinity time series [Carré et al., 2005a]. Despite some variations of humidity, the coastal climate remained arid or semi-arid in southern Peru [Eitel et al., 2005; Kuentz et al., 2012] and precipitation was reduced in the Andes during the middle Holocene [Baker et al., 2001; Rein et al., 2005]. It can thus be considered constant in average throughout the year, and shell δ18O variations can be interpreted as reflecting changes in SST. The middle Holocene shell δ18O values were also corrected for changes in the mean ocean water δ18O caused by variations in ice volume. This was conducted using the sea level curve of Lambeck and Chappell [2001] with an estimate of a 1.05‰ decrease in oceanic δ18O between the Last Glacial Maximum and the present-day [Duplessy et al., 2002]. A correction of 0.049‰ and 0.079‰ was applied to mid-Holocene shells of Rio Ica and QLB respectively. Paleotemperature values were calculated using the Carré et al. [2005a] transfer equation. Temperature-related carbon fractionation was calculated using equation (2) and isotopic temperature estimates, and then subtracted from shell δ13C values (Figure 3). We therefore generated signal of corrected δ13C values (δ13Ccorr) which are primarily influenced by the variations of δ13CDIC and the contribution of metabolic carbon. The standard error of δ13Ccorr was estimated at 0.17‰ (1σ) and combines analytical errors on both shell δ13C and δ18O and the monthly variability of water δ18O. The 85 individual corrected δ13C profiles are represented in Figure 4 and span from ∼10 months to ∼24 months for the larger shells.

Figure 3.

δ13C (open diamonds) and δ18O (open squares) measured with a monthly resolution along the growth axis of a modern shell (brmd24) collected live in June 2003 on the Llostay beach (close to the QLB site). δ18O values were converted to temperature using Carré et al.'s [2005a] equation and a water δ18O value of 0.18‰(V-SMOW). δ13Ccorr values (closed diamonds) were calculated from shell δ13C corrected from the temperature fractionation effect and the Suess effect. Positive corrections appear in red and negative corrections in blue.

Figure 4.

δ13C profiles for all shells corrected for the temperature effect and for the Suess effect (modern shells only). Isotopic values were positioned on a monthly time axis after re-evaluating the mean microsampling resolution for every shell.

Table 2. Seawater δ18O Data
Coastal SiteLat.Long.Date (dd-mm-yy)δ18O (‰ V-SMOW)
Lagunillas13°55′S76°19′W09-08-030.28
Rio Ica beach14°52′S75°34′W11-02-070.39
   07-12-070.26
Puerto Caballas14°56′S75°29′W15-12-070.32
Puerto Lomas15°33′S75°50′W13-08-020.03
   19-12-070.38
Tanaka15°43′S74°28′W25-01-030.19
   16-12-070.25
Ilo17°38′S71°21′W22-08-020.34
   06-11-02−0.07
   04-12-020.10
   08-01-030.18
   07-02-030.04
   16-06-030.14
   16-07-030.26
   04-08-030.33
   03-09-030.35
   10-10-030.32
   14-11-030.08
   12-12-030.03
   12-01-040.47
   11-02-040.16
   17-03-040.3
   12-04-040.11
QLB18°01′°S70°50′W19-11-070.32
Llostay18°10′S70°39′W27-01-03−0.03
   08-11-070.33
   13-11-070.37

4.2. The Influence of Metabolic Carbon

[20] The relative proportion of carbon of metabolic versus environmental origin is an uncertain parameter that may strongly vary with species and environment [Tanaka et al., 1986; Klein et al., 1996; McConnaughey et al., 1997; Gillikin et al., 2006a, 2007; McConnaughey and Gillikin, 2008]. In early studies, authors proposed that shell δ13C variations mainly reflected δ13CDIC and could therefore be used as paleoenvironmental tracers [Mook and Vogel, 1968; Mook, 1971; Killingley and Berger, 1979; Arthur et al., 1983; Krantz et al., 1987, 1988; Bemis and Geary, 1996]. The contribution of metabolic carbon was estimated to be <10% in most marine calcified skeletons [McConnaughey et al., 1997; McConnaughey and Gillikin, 2008]. Because the δ13C of respired carbon is highly depleted (∼−20‰) compared to DIC (∼1‰), 10% variations of metabolic carbon contribution could have significant effects of ∼2‰ on shell isotopic signature [Gillikin et al., 2006a]. Besides an average contribution of metabolic carbon, δ13C variations within shells have also been attributed to variations in this metabolic component. Systematic decreasing trends in shell δ13C through ontogeny have been reported in mollusk shells [Romanek and Grossman, 1989; Keller et al., 2002; Elliot et al., 2003]. Lorrain et al. [2004] showed that in scallop shells, these ontogenetic δ13C trends were due to increasing contribution of respired carbon with the body size. An influence of food availability was also suggested by Van der Putten et al. [2000] because of decreasing δ13C values associated with Mn and Ba peaks which are believed to be related to phytoplankton blooms [Stecher et al. 1996; Gillikin et al., 2006b; Thébault et al., 2009]. However, the majority of the shell δ13C signal was associated with shell transplantation to a locality under larger influence of freshwater and thus with lower δ13CDIC values. Recently, Lartaud et al. [2010] observed decreased shell δ13C associated with food availability in cultured Crassostrea gigas, but a strong positive correlation with shell δ18O suggested the possible presence of kinetic effect, and these isotopic depletions were not observed in natural conditions. Based on the Pecten maximus isotopic records published by Lorrain et al. [2004], Chauvaud et al. [2011] proposed a model where intraseasonal shell δ13C variations were mainly due to metabolic effect modulated by food availability. This model, however, cannot be generalized since it did not fit the other P. maximus records published in the same article, where shell δ13C variations followed a classic δ13CDIC seasonal cycle, with low values during winter for the freshwater input and high values in spring during the peak of phytoplankton productivity. In summary, there remains little evidence for significant seasonal variations of metabolic carbon contribution in mollusk shells. The contribution of metabolic carbon in mollusk shells can be estimated using the following equation from McConnaughey et al. [1997]:

equation image

Where M is the fraction of metabolic carbon, δ13Cmeta, δ13CEnv, and δ13Cshell the isotopic value of the metabolic HCO3, environmental HCO3 and shell aragonite respectively. εarag-HCO3- is the enrichment factor between shell aragonite and HCO3 in the extrapalleal fluid. When using this equation, most authors assume that εarag-HCO3- is constant and equal to the enrichment factor calculated for inorganic aragonite (2.7 ± 0.6‰). Further research is needed to demonstrate the validity of this assumption. Because of the biological influence in the mineralization process, the average enrichment factor may be different for shell aragonite, and may also vary during the mollusk life with environmental or biological factors. Following the discussion of section 4.1.2, considering a temperature dependence of the enrichment factor may be a first improvement in the estimation of M.

[21] We used an approximate value of −20‰ for the metabolic carbon. The δ13CDIC value of sea surface water on the southern Peruvian coast was estimated at −0.18 ‰ (σ = 0.47‰) based on 17 measurements between 14°S and 18°S in the 2001–2004 period (Table 3). The average δ13Cshell value was 0.17‰ for the Ica modern shells and 0.58‰ for the Llostay modern shells. Using an enrichment factor calculated from the Grossman and Ku [1986] relationship for biogenic aragonite and an average temperature value of 17°C, the estimates of the metabolic contribution were 2% for Ica and 0.2% for Llostay.

Table 3. Seawater δ13CDIC Data
Coastal SiteLat.Long.Date (dd-mm-yy)δ13CDIC (‰ V-PDB)
Lagunillas13°55′S76°19′W23-11-01−0.15
   23-11-01−0.46
   20-06-03−0.46
   21-06-03−0.25
San Juan15°21′S75°10′W28-11-01−1.39
   28-11-010.17
Pocoma17°25′S71°23′W25-11-01−0.36
   25-11-010.07
   16-06-030.21
   16-06-030.21
   16-12-030.04
   17-01-040.22
   15-02-040.13
   28-03-040.21
   16-09-040.23
Ilo17°38′S71°21′W27-11-01−1.04
   27-11-01−0.37

[22] We concluded that δ13C variations in M. donacium shells are not significantly affected by inner variations of the metabolic effect because of three observations: (1) the estimates of the metabolic carbon percentages in M. donacium were low, (2) no systematic ontogenic decreasing trend was observed in the individual carbon isotopic signals, (3) no δ13C depleted values were associated to Mn and Ba peaks in M. donacium modern shells (shells brmd21 and brmd24 in the work of Carré et al. [2005a, 2006]). As a result we suggest that δ13Ccorr profiles in Figure 3 mainly reflect variations of seawater δ13CDIC.

4.3. Correcting for the Suess Effect

[23] Since the industrial revolution, the combustion of fossil fuels has provided an additional source of 12C enriched carbon into the atmosphere, decreasing the δ13C global value of the atmospheric CO2. This anthropogenic influence is referred to as the Suess effect. This additional 12C diffuses into the ocean and has reduced the surface ocean δ13C since the industrial revolution [Keeling, 1979; Quay et al., 1992; Bacastow et al., 1996]. To allow comparisons between the modern day and preindustrial periods, this anthropogenic effect should be removed from modern shell δ13C values. The amplitude of the Suess effect varies with local oceanographic characteristics. Based on estimations of anthropogenic carbon concentration in the South East Pacific [Goyet et al., 2009], we calculated a Suess effect correction of 0.5‰ for the Peruvian coast. This value is lower than the Suess effect measured in the Caribbean (0.9‰) and in New Caledonia (0.7‰) by Böhm et al. [1996], which is consistent with a larger contribution of deep water in Peru. We tested whether the seasonal variation of the coastal upwelling strength could induce seasonal variations of the Suess effect that would dampen the shell δ13C amplitude. Since no significant difference was found between δ13C annual amplitude of modern and fossil shells, we assumed this potential effect was not significant. An example of a modern shell δ13C signal corrected for the Suess affect and the temperature effect is shown in Figure 3.

4.4. Stable Isotopes and the Intensity of Coastal Upwelling

[24] Considering the assumptions discussed in sections 4.1 and 4.2, shell δ13Ccorr reflect δ13CDIC. The full data set plotted together (Figure 5) shows a strong linear correlation between mean δ13Ccorr and mean δ18O values, with a slope value of −0.80. Mean values of both δ13CDIC and δ18Ocarbonates are expected to provide an indication of the background signal generated by the presence of deep water. Deep water brought to the surface by coastal upwelling is characterized by lower temperatures (corresponding to higher δ18O carbonates values) associated with lower δ13CDIC readings.

Figure 5.

Mean value of corrected δ13C versus mean value of δ18O (corrected for ice volume effect in mid Holocene samples) for individual shells (small symbols) and for the five whole samples (large symbols). Bars indicate the range of mean values of individual shells within a sample. The least square linear regression was calculated from the individual shells mean values (N = 85).

[25] To test the hypothesis that the linear relationship between mean δ13Ccorr and δ18Ocarbonates which appears in our data set is related to the upwelling mean intensity, we compared it to the relationship between δ13CDIC and δ18Ocarbonates expected along a depth profile within the 100m surface layer using the following equation:

equation image

where D is the water depth (m), and T is the water temperature (°C). δ13CDIC depth profiles measured by Quay et al. [2003] provide an average estimate of 6.10−3 ‰m−1 for ∂ (δ13CDIC)/∂ (D) in the 100m surface layer. The temperature-depth dependence ∂(T)/∂(D) is ∼0.03°Cm−1 in the surface layer of the coastal Peru based on in situ measurements of IMARPE. ∂(T)/∂(δ18Ocarb.) in aragonite was estimated at −3.66°C/‰ for M. donacium by Carré et al. [2005a]. A value of −0.73 is obtained from equation (4), very close to the slope value of −0.80 obtained from our data set (Figure 5). This result supports our hypothesis that the linear relationship between δ13Ccorr and δ18Ocarb mean values obtained from mollusk shells defines a gradient of influence of deep water, equivalent to a gradient of upwelling intensity. Our interpretation is also supported by temperature instrumental data indicating a stronger upwelling today at 15°S (sample ICAmod) than at 18°S (sample QLBmod) (Figures 1 and 5).

[26] The linear relationship between average values of shell δ18O and δ13Ccorr defines a coastal upwelling intensity axis (Figure 6). So far, and because of the complex influences of mollusk δ13C signal, the local variations of paleoupwelling intensity were inferred from shell δ18O [Carré et al., 2005b, 2012] based on the negative relationship between sea surface temperature and upwelling rate. Although this approach is generally accurate, additional influences on shell δ18O, such as evaporation or fresh water input, may lead to incorrect interpretations. The combination of shell δ18O and δ13Ccorr and the position of average values relatively to the upwelling axis allow us to detect environmental influences. A position close to the upwelling δ18O − δ13Ccorr relationship suggest that upwelling rate is the major control of the average water geochemistry, whereas evaporative conditions (for instance in a close embayment) are indicated by a deviation toward δ18O enrichment, and fresh water influence is indicated by a combined depletion of δ18O and δ13Ccorr (Figure 6). The upwelling δ18O − δ13Ccorr relationship proposed here relies on a δ13CDIC-depth relationship, a temperature-depth relationship and a specific paleotemperature relationship (equation (3)); it is thus only valid for southern Peru and M. donacium shells although the method could be similarly applied with other species in other regions.

Figure 6.

Schematic model of the upwelling-related relationship between mean shell δ18O values and mean shell δ13Ccorr values. Deviations from this relationship may be due to evaporation effect or to continental fresh water influence.

[27] The δ13C and δ18O signals confirm that coastal upwelling conditions were significantly stronger during the middle Holocene (Figure 5), which had been already suggested by independent paleoenvironmental results [Bétancourt et al., 2000; Holmgren et al., 2001; Placzek et al., 2001; Rech et al., 2003; Fontugne et al., 2004; Koutavas and Sachs, 2008; Ortlieb et al., 2011]. Furthermore, during the middle Holocene, the strongest coastal upwelling was experienced at 18°S (Figure 5), whereas under modern conditions, upwelling intensity at 18°S is currently weaker than at 15°S. This result suggests a spatial reorganization of the intensity of coastal upwelling centers, possibly as a response to a different wind field.

4.5. Seasonal Signals of Stable Isotopes

[28] For modern samples, the average annual amplitudes of shell δ13Ccorr range from 0.72‰ (ICAmod) to 0.85‰ (QLBmod). In fossil samples, average annual δ13Ccorr amplitude is 0.80‰ for IcaInca, 0.83‰ for IcamH, and 0.99‰ for QLBmH. The difference between these values is not statistically significant. For the whole Peruvian South coast, we obtain an average annual δ13Ccorr amplitude of 0.83‰. Trends in δ18O follow the seasonal patterns of temperature, with low δ18O values representing the warmer temperatures of austral summer (Figures 3 and 8). Although δ13Ccorr monthly signals exhibit a less pronounced seasonal pattern than δ18O, seasonal δ13Ccorr variations are generally in phase with, or lag slightly behind, temperature. Enriched δ13Ccorr values are generally recorded in summer and depleted values are recorded during winter. If trends in shell δ13Ccorr represent δ13CDIC variations (section 4.3), then mollusk shells are recording seasonal fluctuations of δ13CDIC in coastal waters. At the seasonal scale, δ13CDIC is controlled by a combination of (1) primary productivity, (2) upwelling seasonality, and (3) sea-atmosphere CO2 exchange.

[29] Carbon isotope fractionation occurs during sea-atmosphere CO2 exchange. 12CO2 is favored in the outgassing reaction that induces an enrichment of δ13CDIC within surface waters. This effect increases with the CO2 flux, which depends on the pCO2 difference between the atmosphere and the surface water. Carbon isotopic equilibration takes approximately 10 years therefore δ13CDIC is never in isotopic equilibrium with the atmosphere along the Peruvian coast where the surface water is constantly replaced by deep water through upwelling. At the seasonal scale, carbon fractionation in sea-atmosphere exchange is mainly due to kinetic effects [Lynch-Stieglitz et al., 1995]. The net CO2 flux of the Peruvian coast is toward the atmosphere due to the very high pCO2 values that characterize upwelled waters. The CO2 flux, and therefore the fractionation effect, is strongest during the winter when ΔpCO2 may reach 1000 ppm and weakest during the summer when ΔpCO2 is closer to 0 ppm. Therefore, the effect of air-sea CO2 exchange on the δ13CDIC is enrichment during the winter and depletion during the summer. Since this is opposite to most trends recorded in mollusk shells, we conclude that air-sea CO2 exchange have a minor control on seasonal δ13CDIC variations.

[30] In contrast to air-sea exchange, seasonal variability in the intensity of coastal upwelling would produce δ13CDIC in phase with the temperature annual cycle. This effect can be estimated from the seasonal variations of the thermocline depth and the trend in δ13CDIC-depth relationship. In situ measurements reveal that the amplitude of the thermocline seasonal vertical movement in non-El Niño years is ∼30m. Considering a 6.10−3 ‰ m−1 depletion for δ13CDIC with depth [Quay et al., 2003], the effect of upwelling seasonality is estimated to be about 0.2‰, which cannot explain the 0.83‰ of variations observed in shell samples.

[31] The primary control for δ13CDIC seasonal variations would therefore be the remaining variable: primary productivity. Productivity peaks induce δ13C enrichment of the DIC in the surface layer, which would generally occur during summer as indicated by in situ and satellite chlorophyll a measurements [Echevin et al., 2008]. The phasing of the productivity effect is thus in agreement with δ13CDIC variations measured in mollusk shells. To further test this hypothesis, we used the relationship between δ13CDIC and PO4 in the ocean published by Broecker and Maier-Reimer [1992], which allows for an estimate on the productivity effect on δ13CDIC without air-sea exchange:

equation image

The seasonal variation in PO4 at 15°S on the Peruvian coast is ∼0.8μg.L−1 (measured by IMARPE in oceanographic cruises between 1964 and 2008, excluding El Niño years), which corresponds to a δ13CDIC seasonal variation of ∼0.9‰ based on equation (5). This value is very close to the amplitude of δ13CDIC recorded in mollusk shells, which supports our hypothesis that the dominant control on seasonal variations of δ13CDIC recorded by shell δ13Ccorr is primary productivity.

[32] In order to quantitatively estimate the seasonal timing of phytoplankton blooms and its variability with sites and periods, lag periods between δ13Ccorr and δ18O were evaluated using cross-correlation. Correlations between δ13Ccorr and δ18O trends of individual shells were evaluated between lag periods of ±6 data points. The lag corresponding to the highest positive correlation between isotopic signals, multiplied by the microsampling resolution, was interpreted as the monthly time difference between the productivity bloom (δ13Ccorr maximum) and the February temperature maximum/August temperature minimum for positive and negative correlations respectively (Figure 7). Compared to a simple peak-to-peak distance between δ13Ccorr and δ18O, this technique takes into account the full shape of the signal and is thus more reliable for estimating a time lag. The seasonal timing of blooms determined from each shell were then classified into austral spring (October–December), summer (January–March), autumn (April–June) or winter (July–September).

Figure 7.

The δ13Ccorr (diamonds) and δ18O (gray line) profiles of two shells of the same area: a modern shell from ICAmod (ICA1) and a fossil shell from ICAInca (IS2 12.1). Mean δ13Ccorr values are indicated by dashed lines. The difference between mean values is due to the mean intensity of the coastal upwelling. Summers (February) and winters (August) were identified from the δ18O curves. At the seasonal scale, δ13Ccorr maxima correspond to 13C enrichment phases of the DIC due to phytoplankton blooms. Lags between δ13Ccorr and δ18O are estimated by cross-correlation. A 2.2 month lag was estimated for the fossil shell, indicating a phytoplankton bloom in fall, while no lag was found in the modern shell, indicating a phytoplankton bloom in summer.

[33] The seasonal timings of phytoplankton blooms evaluated by this new technique were plotted as frequency distributions for the middle Holocene, the Inca period and modern day (Figure 8). Modern shell samples from both locations reveal that phytoplankton blooms were most common during austral summer, in agreement with chlorophyll a data (Figure 1) [Thomas et al., 1994; Echevin et al., 2008], which supports the reliability of the technique for paleoenvironmental analysis.

Figure 8.

Frequency distributions of the season of occurrence of the main phytoplankton bloom. The bloom dates were determined by the lag between shell δ13Ccorr and δ18O, assuming δ18O seasonal minima correspond to February (see Figure 7). Frequencies were defined as the percentage of shells in a sample. A comparison was made between modern (mod) and fossil (Inca and mH) samples for both study sites.

4.6. Past Seasonal Dynamics of Phytoplankton Bloom

[34] Based on our assumptions and method, the lag between δ13Ccorr and δ18O seasonal signals in individual shells provides an indication of the season of the main phytoplankton bloom (Figure 7). In our case study, we show that modern and fossil samples show similar coastal phytoplankton dynamics at 15°S: bloom development appears to coincide mainly with summer, which corresponds to the warmer SSTs and autumn, which correspond to the onset of upwelling strengthening (Figure 8). This would imply that light is the first limiting factor for phytoplankton growth and upwelled nutrients are a second limiting factor. At 18°S (QLB site), the modern and middle Holocene samples also show a predominance of summer blooms. However, an increase in spring blooms and an absence of fall blooms are also observed. This result suggests that light was the only limiting factor at that time and that upwelled nutrients were not at all limiting, which is consistent with stronger upwelling mean conditions.

[35] The influences of gas air-sea exchanges, upwelling seasonality, and biological effects, were likely secondary in our case study, but they still exist and might in some circumstances be predominant in the δ13Ccorr signal, inducing thus some noise in the final result. The precision of this technique for evaluating the seasonal timing of phytoplankton blooms is also limited by the variability of the temperature profile and the uncertainties in the δ18O signal. Finally, the method is based on an isotopic model that could be improved by comparing mollusk shell profiles with in situ weekly measurements of marine δ13CDIC, productivity, and wind stress over an annual cycle. Such calibration work may allow for more detailed reconstructions. Nevertheless, mollusk stable isotopes could potentially provide new valuable insights into long-term variability of coastal seasonal biogeochemical processes.

5. Conclusions

[36] In this study we proposed a new method to evaluate past variations (1) of the mean intensity of the Peruvian coastal upwelling and (2) of the seasonal timing of phytoplankton blooms. This method uses a combination of the monthly carbon and oxygen isotopic signal preserved in fossil shells of M. donacium. Our method involves a re-evaluation of the paradigm that considers that the 13C enrichment factor between shell aragonite and HCO3 is constant and equal to the enrichment factor of inorganic aragonite. We considered that it is more likely to vary with temperature following Grossman and Ku's [1986] empirical model, although the mechanism remains unknown. After being corrected for temperature and Suess effects, shell δ13Ccorr variations are believed to primarily reflect δ13CDIC variations. At the centennial to millennial scale, δ13CDIC mean values are primarily affected by the mean intensity of the upwelling, while primary productivity is the main factor controlling seasonal δ13CDIC variations.

[37] Five shell samples from the southern Peruvian coast were analyzed and show a strong linear correlation between average values of shell δ13Ccorr and δ18O, similar to the expected relationship along a gradient of influence of deep water, which represents a gradient of upwelling intensity. We then proposed a proxy model to estimate the past intensity of the coastal upwelling in Peru. This method could bring additional independent information valuable to interpret paleoproductivity changes reconstructed from marine sediment of the nearby continental shelf. Results obtained on fossil samples from the middle Holocene suggest an increase of the upwelling during this period, consistent with existing independent paleoenvironmental reconstructions. Additionally, fossil shells show a spatial reorganization of upwelling centers along the South Peruvian coast. At the seasonal scale, shell δ13Ccorr enrichment was interpreted as the indication of a phytoplankton bloom. Bloom seasonal timings may thus be estimated by the lag between shell δ13Ccorr and the annual temperature cycle reproduced by shell δ18O monthly variations. The results obtained with two modern shell samples indicate phytoplankton blooms occurring during summer and fall, consistently with in situ productivity observations. The timing of productivity blooms was similar in fossil samples suggesting little change in limiting factors of phytoplankton growth. One strength of this study lies in the large amount of data which averages out a significant part of random uncertainty and provides statistical robustness to the results. Although this work is only valid for southern Peru and the mollusk species M. donacium, this approach could be developed in other similar areas like the Benguela current system where the upwelling maintains arid conditions on the coast, and where numerous fossil shell middens are preserved.

[38] However, while our interpretation in this case study is arguably the most parsimonious hypothesis with regards to actual knowledge of carbon isotopes in mollusk shells, it is clear that further investigation is needed to understand the factors that influence the carbon isotopic fractionation in biogenic aragonite. We estimated here that the enrichment factor εarag-HCO3- is more likely to be not constant in biogenic aragonite, which should lead to a re-evaluation of the contribution of the metabolic carbon in the shell formation.

Acknowledgments

[39] We are grateful to D. Lavallée, M. Julien and the whole team of the Pérou-Sud archeological project (P.I., D. Lavallée) who have been studying the QLB site since 1995 and provided us with the fossil shells. We thank R. Kaandorp and H. Vonhof for their assistance with isotopic analyses at the Vrije Universiteit, Amsterdam. We are thankful to N. Mitma Garcia for assistance with fieldwork. We thank J. C. Duplessy and O. Kawka for their help. We also thank D. P. Gillikin and an anonymous reviewer for their comments that helped improve this article. This material is based upon work supported by the Pérou-Sud archeological project (MAE, France; P.I., D. Lavallée), the National Geographic Society grant 8122-06 (P.I., M. Carré), by a postdoctoral fellowship of the Joint Institute for the Study of the Atmosphere and Ocean (JISAO) under NOAA cooperative agreement NA17RJ1232, by the U.S. National Science Foundation under grant NSF-ATM-0811382 (P.I., J. P. Sachs), and the U.S. National Oceanic and Atmospheric Administration under grant NOAA-NA08OAR4310685 (P.I., J. P. Sachs). We thank L. Ortlieb and the PNEDC-CONCHAS project who contributed with DIC isotopic analyses. This is ISEM contribution 2011-196.