supplementary material for - science · supplementary material for . holocene history of enso...
TRANSCRIPT
www.sciencemag.org/content/science.1252220/DC1
Supplementary Material for
Holocene history of ENSO variance and asymmetry in the eastern tropical Pacific
Matthieu Carré,* Julian P. Sachs, Sara Purca, Andrew J. Schauer, Pascale Braconnot, Rommel Angeles Falcón, Michèle Julien, Danièle Lavallée
*Corresponding author. E-mail: [email protected]
Published 7 August 2014 on Science Express
DOI: 10.1126/science.1252220
This PDF file includes:
Materials and Methods
Figs. S1 to S11
Tables S1 to S3
Supplementary Text
Full Reference List Other Supplementary Material for this manuscript includes the following: (available at www.sciencemag.org/content/science.1252220/DC1)
Data Tables S1 to S3 as a separate Excel file
Supplementary online material 1. Material
Figure S1. Regional map of archaeological sites Table S1. Shell samples
1.1. Ica modern sample 1.2. Llostay modern sample 1.3. Ancón 1.4. Asia 1.5. ICA-IN 1.6. ICA-IS2 1.7. Lomas 1.8. Chala 1.9. Quebrada de los Burros (QLB) Figure S2. Picture of Ancón shell midden
Figure S3. Picture of Asia shell midden Figure S4. Picture of ICA-IN shell midden Figure S5. Picture of ICA-IS2 shell midden Figure S6. Picture of Lomas shell midden Figure S7. Picture of Chala Viejo shell midden Figure S8. Picture of Quebrada de los Burros shell midden
Table S2. Radiocarbon dates 2. Methods
2.1. Diagenesis Figure S9. SEM observations of modern and fossil M. donacium shells Figure S10. Picture of a fossil shell from ICA-IN 2.2. Isotopic analyses 2.3. Producing monthly resolved isotopic signals 2.4. SST calculation 2.5. Determining seasonal ranges ∆T in shell records 2.6. Estimating ENSO variance from combined samples 2.7. Estimating uncertainties
Table S3. MoCo parameterization for uncertainty calculation 2.8. ENSO variance from coral records
3. Supplementary text 3.1. El Niño in coastal Peru today and in the past: a brief review 3.2. Recording ΔT distribution: strength and limits of the proxy
Figure S11. δ18O plots of 180 M.donacium shells 4. Supplementary Data Tables (available on Science online)
Data Table S1. Complete shell δ18O dataset Data Table S2. Isotopic seasonal amplitudes Δ(δ18O) Data Table S3. Statistics and uncertainties of isotopic data per sample
1. Material
Mesodesma donacium (Lamarck, 1818) is an endemic bivalve species of the Peruvian and
Chilean coast. The species forms dense banks from 10 m depth to the intertidal zone in high
energy sandy beaches. M. donacium shells are found in abundance in coastal archaeological
sites, referred to as shell middens when they are characterized by large anthropogenic mollusk
shell accumulations. M. donacium was collected for food by ancient fishermen since the first
trace of coastal human occupation ~13 ka (9). As today, shells were likely gathered manually
in the intertidal zone at depths of 0-1 m during low tide. For this reason, tidal growth lines are
well defined (see section 2.3), and the range of temperature owing to local scale variability
experienced by different shells is narrow.
Figure S1. Map of mean annual SST (color shade) and Peru topography (gray scale) with the location of Peruvian archaeological and modern shell middens studied here (circles), and sediment cores V21-30 (12) and SO147-106KL (squares) (25).
We analyzed two samples of modern Mesodesma donacium shells and 12 samples of fossil
shells from 7 distinct archaeological shell middens located along the Peruvian coast between
11.7°S and 18.1°S (Fig. S1). Shell samples are listed in Table S1, and the associated
radiocarbon dates in Table S2. Archaeological sites are described in sections 1.1 to 1.9 from
north to south.
Table S1. Shell samples
Shell sample Lat. (°N) Long.
(°E)
Number of
shells
Modern reference
Time interval (cal yrs BP)
Combined Sample
ICA -14.867 -75.567 13 -47 — -10 Modern Llostay -18.17 -70.658 13 -47 — -10
Chala -15.765 -74.206 19 ICA 528 — 445
Mill ICA-IS2 -14.872 -75.556 19 ICA 530 — 385 Lomas -15.565 -74.847 17 ICA 650 — 515
Asia3+5 -12.768 -76.599 9 ICA 3027 — 2785 3K Ancon -11.777 -77.175 6 ICA 3190 — 2958
QLB-H4 -18.018 -70.833 13 Llostay 4799 — 4527 H4
ICA-IN -14.867 -75.567 13 ICA 6897 — 6654 MH QLB-N2+N3* -18.018 -70.833 18 Llostay 7547 — 6737
QLB-N4* -18.018 -70.833 10 Llostay 8951 — 8049 QLB-45 QLB-N5* -18.018 -70.833 8 Llostay 9252 — 9006
QLB-N6* -18.018 -70.833 6 Llostay 9434 — 9304 QLB-67 QLB-N7* -18.018 -70.833 16 Llostay 9615 — 9307
1.1. ICA modern sample
A modern shell sample was collected in order to mimic archaeological shell samples. Shells
were collected from small modern shell mounds behind the sandy beach (14.87°S, 75.57°W)
close to the archaeological sites ICA-IN and ICA-IS2. These shell mounds were produced by
fishermen when M.donacium was seasonally exploited in the late 20th century, up until the
1998 El Niño caused mass mortality. Every shell was collected in a distinct shell mound to
minimize the chances of redundancy (7). The shell sample is thought to be representative of
three to four decades of fishing activity based on the testimonies of fishermen and the
provenance of rubbish associated with shells (e.g., newspaper, plastic, etc). This sample was
used as a modern reference for all archaeological samples except those from the Quebrada de
los Burros site. The description of methods for sclerochronologies, isotope analyses, SST and
ENSO calibrations were presented in previous studies (7,10).
1.2. Llostay modern sample
This sample was collected from modern shell mounds on Llostay beach (18.17°S, 70.66°W)
in southern Peru. The sampling method was the same as in Ica. This sample was used as a
modern reference for all the fossil samples from Quebrada de los Burros.
1.3. Ancón
This shell midden of the formative period (Fig. S2) is located just behind the archaeological
museum of Ancón. Shells were sampled in successive layers on a 2-m-high outcrop. Shell
preservation was generally good despite fog humidity and the high level of organic matter in
the surrounding sediment.
1.4. Asia
The sample Asia3+5 contains shells from two nearby shell middens (Asia3 and Asia5) of the
same cultural complex of the formative period in the lower Asia valley. Middens were
relatively small (less than 300 m2) and contained almost exclusively M. donacium shells. The
stratigraphy of Asia3 could be observed and sampled in a road cut (Fig. S3). Two charcoal
fragments, from the upper and the lower layers, were collected and radiocarbon dated (Table
S2). Asia5 was sampled on its surface and dated from one charcoal fragment. These shell
middens have not been excavated.
1.5. ICA-IN
This shell midden is located on the northern bank of Ica River, about 1 km from the shoreline
(Fig. S4). An early test pit suggested a first preceramic occupation with a later re-occupation
(34). Our radiocarbon dates (Table S1) confirmed that the shell material, collected in a new
2m deep test pit in the northwestern area of the shell midden, is from the preceramic period.
Preservation was excellent as attested by the presence of the proteic periostracum layer on
most shells. This site has also been referred to as Casavilca or L-1 but has not been
extensively excavated (35).
1.6. ICA-IS2
The site is an extended shell midden located on sand dunes on the southern bank of Ica River,
in front of ICA-IN (Fig. S5). House structures, ceramics and clothes visible on the surface
imply a late horizon occupation (36). Shells were collected from a 2.3m deep test pit in the
largest shell accumulation, characterized by more than 10 m in height. Shell content was
dense and almost exclusively composed of M. donacium. Shells were still bearing desiccated
muscle, indicative of exquisite, near-perfect preservation. This site has also been referred to as
La Yerba or L-3 but has not been extensively excavated (35).
1.7. Lomas
The site is located at the entrance of Puerto Lomas village. It is a 6m high midden containing
mixed archaeological material of the late intermediate period. The whole stratigraphy is
outcropping in a road cut. After cleaning the outcrop surface, shell samples were collected in
successive layers from the top to the bottom (Fig. S6). Five radiocarbon dates obtained from
charcoal and plant fragments indicated that the entire midden had accumulated in about a
century. The site has not been excavated but the outcrop showed the presence of varied
marine mollusk shells, crustaceans, fish bones, birds and mammals bones, and fragments of
cloth and ceramics, within a matrix of sandy clay. The stratigraphy showed little change from
the top to the bottom.
1.8. Chala
The archaeological site is located at ~11 kms from the coast, in front of the village Chala
Viejo. This small Inca town now used as a cemetery by modern inhabitants, has been briefly
described but has not been extensively excavated (37). Shells were collected in a 50-cm-high
midden in the southeastern part of the site (Fig. S7).
1.9.Quebrada de los Burros (QLB)
The Quebrada de los Burros archaeological site is an exceptional preceramic coastal site (Fig.
S8) that was the focus of investigation by the Pérou-Sud archaeological project from 1996 to
2010 (38-39). It is located at ~1.6 km from the actual shoreline in a narrow valley where a
small stream constantly flows. Six successive occupation levels (N2 to N7) were identified
that provided us with the shell samples QLB-N2+N3, QLB-N4, QLB-N5, QLB-N6, and
QLB-N7. Shells from N2 and N3 were combined in the sample QLB-N2+N3 because these
levels were chronologically indistinguishable. The site has been occupied by fishermen for
about 3000 years, with only one significant period of abandonment, characterized by a thin
sand layer between N4 and N3. The shell preservation was poor in the talweg but very good
in the upper part of the site. Shells were selected for their preservation state. The sample
QLB-H4 was collected on the surface of a shell accumulation located ~200 m north of the
Quebrada de los Burros archaeological site. It is the only shell sample directly dated with
radiocarbon because no charcoal was available. A local reservoir age deviation ΔR=226±98
yrs was used to calibrate the 14C age to a calendric age (40).
Figure S2. Picture of the outcropping archaeological layers of Ancón shell midden. (photo: Matthieu Carré)
Figure S3. Picture of Asia3 shell midden and outcropping shell-bearing layers. (Photo: Matthieu Carré)
Figure S4. Picture of ICA-IN preceramic shell middens. All white shells visible on the surface are Mesodesma donacium. (Photo: Nancy Mitma García)
Figure S5. Picture of ICA-IS2 shell midden. Shells on the surface are Mesodesma donacium. A wooden house structure is apparent on the left. (Photo: Matthieu Carré)
Figure S6. Picture of the outcropping layers of Lomas shell midden in a road cut. The outcrop was cleaned and sampled from the top to the bottom. (Photo: Matthieu Carré)
Figure S7. Picture of the shell midden in Chala Viejo. (Photo: Matthieu Carré)
Figure S8. Top : Picture of the Quebrada de los Burros excavation, level N6, 2007. Bottom : Quebrada de los Burros, level N5, area for M. donacium shells processing (39). (Photos: Perú-Sur French archaeological project)
Table S2. Radiocarbon dates
Shell sample
Dated Material
Lab. Ref. Age (14C yr BP) Error (±)
Chala Wood Charcoal
BETA-301049 OS-69950
410 520
30 30
ICA-IS2 Plant Plant Plant Plant Plant
Sac1 26484 Sac1 26486 Sac1 26485 Sac1 26487 OS-65628
445 460 470 510 510
30 30 30 30 40
Lomas Plant Plant
Charcoal Charcoal
Plant
OS-70125 OS-70126 OS-70127
CAMS-144813 OS-65629
545 545 615 675 680
25 25 30 30 30
Asia 3
Asia 5
Charcoal Charcoal Charcoal
SacA25581 SacA25582 SacA25584
2785 2825 2905
35 40 35
Ancon Charcoal Charcoal Charcoal
SacA25587 SacA25585 SacA25586
2960 2985 3005
35 35 30
QLB-H4 Mollusk OS-70085 4700 45 ICA-IN Charcoal
Charcoal Charcoal Charcoal
OS-60543 OS-60564 OS-60556 OS-60544
5840 5900 5940 6070
35 40 45 30
QLB-N2+N3*
Charcoal Charcoal Charcoal Charcoal Charcoal Charcoal Charcoal Charcoal Charcoal
GifA-99341 GifA-100142 SacA 7586
10623/GifA-972 11284/GifA-995 10625/GifA-972
Gif-11735 GifA-100346 GifA-100347
6050 6090 6260 6460 6520 6630 6415 6430 6500
80 110 35 60 90 70 80 80 80
QLB-N4* Charcoal Charcoal Charcoal Charcoal
SacC 7591 GifA-100342 GifA-100344 GifA-100345
7390 7880 7970 7980
35 90
140 90
QLB-N5* Charcoal GifA-102445 8220 90 QLB-N6* Charcoal SacA 10218 8400 40 QLB-N7* Charcoal
Charcoal GifA-102446 SacA 7589
8460 8675
80 40
* Dates published in Ref. 38.
2. Methods
2.1. Diagenesis
Shell diagenesis was evaluated by comparing X-Ray diffraction spectra of 6 test shells from
the QLB site with those of 2 modern shells (Llostay, and Valparaiso). Powdered shell
aragonite samples (1 g) and inorganic aragonite samples were analyzed with a MZ VI Seifert
diffractometer. No trace of recrystallized calcite was detected in any of the analyzed shells.
The effect of diagenesis on aragonite microstructure and shell organic matrix was further
investigated with Scanning Electronic Microscope (SEM) at University Montpellier 2. Since
the entire surface of every shell cannot be imaged by SEM before microsampling of the
carbonate, the aim of these observations was to build a diagnostic set so diagenetic patterns
could be identified by direct binocular microscope observations.
The shell aragonite microstructure was crossed-lamellar in the outer layer of modern shells
(Fig. S9A). Fossil shells from QLB, which are the oldest in our study, had crossed-lamellar
aragonite microstructure similar to that observed in the modern specimen (Fig. S9B).
Recrystallization could only be observed on the outer-most surface and was generally limited
to less than 50 μm in thickness (Fig. S9C). Larger recrystallized areas could be easily
identified on polished shell sections from one or more of the following: (i) lack of daily
growth lines, (ii) translucent appearance, (iii) brown to orange coloration and darker shading
than the rest of the shell. These areas can thus be avoided when microsampling for isotopic
analyses.
Friable chalky shells are sometimes found when humidity is present in the surrounding
sediment, as in the talweg under the QLB excavation. SEM observations showed that this
state is due to a diagenetic process characterized by a preservation of the aragonite
microstructure and a degradation of the shell inter-crystalline organic matrix that ensured
shell mechanical properties (Fig. S9D). Although it is not known whether the oxygen isotopic
signal would be affected by this diagenetic process, chalky shells were excluded from isotopic
studies.
We aimed to observe the shell inter-crystalline organic matrix to clearly assess its
preservation state in fossil shells. Eight polished shell sections were etched with 0.8% formic
acid for 30 seconds, immersed in successive ethanol baths with increasing concentration, and
then immersed in a HDMS bath as described in ref. 41. After this treatment, organic matrix
fibers emerged from the surface that could be observed with SEM (Fig. S9E). The organic
matrix was exceptionally well preserved in fossil shells (Fig. S9F).
Figure S9. SEM observations of modern and fossil M. donacium shells. Direction toward shell outer surface is indicated by white arrows. (A) Aragonite crossed-lamellar microstructure observed in a modern shell from Llostay, on a radially broken surface. (B) Same for a fossil shell from QLB. (C) Outer surface in a radially broken fossil shell from QLB showing recrystallization on the surface above crossed-lamellar structure. The recrystallized layer was here less than 50μm thick. (D) Crossed-lamellar microstructure in a chalky fossil shell from QLB. The organic matrix has disappeared and left aragonite fibers in place but without cohesiveness. (E) Polished and etched radial section in a modern shell. Organic matrix fibers in the crossed-lamellar structure can be observed. Daily growth lines also appear as organic rich parallel lines. (E) Same as in (E), but here from a fossil shell from QLB. The organic matrix fibers are more erect, yet appear perfectly preserved.
These diagnostic observations made on a set of modern and fossil shells from QLB were used
in all our samples to (i) select the best preserved shells for isotopic analyses, and (ii) avoid
altered parts when microsampling shell sections for isotopic analyses. This approach proved a
posteriori to be reliable since no aberrant value was obtained for oxygen or for carbon stable
isotopes, and a clear seasonal signal was obtained.
In most shell middens of the central coast, the general preservation conditions were such that
organic matter, such as plants and clothes were preserved. Indeed, the conditions were
sufficiently arid as to occasionally preserve the soft tissues of humans and animals. Mollusk
shells were generally found with their proteic periostracum layer intact, occasionally with
desiccated muscle tissue attached (Fig. S10). This high degree of organic preservation all but
guarantees perfect preservation of biogenic carbonate, since organic matter is degraded prior
to the associated carbonate (42).
Figure S10. (A) Picture of the left valve of a M. donacium shell from the ICA-IN shell middens (~6800 yr BP). The yellow shiny layer on the outer surface is the shell periostracum. It was only slightly altered by sand abrasion close to the umbo. (B) The inner face of the same valve is also well preserved. Some sand is stuck to the surface. Attached to the shell is intact muscle tissue (protein), preserved by the extreme aridity of the coastal Peruvian desert. (C) Close-up picture of preserved muscle tissue on the ~6800-year-old shell. Scale bar is 1 cm.
2.2. Isotopic analyses
Shells were prepared and microsampled using an automated MicroMill® following the method
described in ref. 7. The outer shell layer was microsampled following the growth axis from
the umbo to the ventral margin, so that the full life of the individual mollusk was analyzed.
Only the first year was generally incomplete because the layer was too thin and/or physically
degraded. Microsamples were contiguous. The microsampling resolution was maintained as
close as possible to 1 lunar month using tide-related fortnightly growth lines. Powdered
samples were manually collected with razor blades. The polished shell sections were cleaned
with compressed air before drilling the next sample. The oxygen isotopic composition of
powdered aragonite microsamples (~50 µg) was analyzed at the University of Washington
Isolab using a Finnigan Delta Plus isotope ratio mass spectrometer coupled to a Kiel III
carbonate device. Aragonite samples were digested in 100% phosphoric acid at 70°C. The
standard deviation for repeated measurements of the internal standard was better than 0.08 ‰.
Raw δ18O values were corrected as recommended in Ref. 43. δ18O values were reported with
respect to the Vienna Pee Dee Belemnite (VPDB) scale using NBS19 (δ18O = -2.2 ‰) and
NBS18 (δ18O = -23.01 ‰).
2.3. Producing monthly resolved isotopic signals
An accurate estimate of the isotopic signal temporal resolution involves back and forth
evaluation between growth lines and the isotopic data. The identification of fortnightly growth
lines yields a first estimate of the shell inner chronology, but it implies uncertainties. A
second estimate of the resolution is obtained a posteriori using the isotopic record based on
the number of values between seasonal extrema. It showed that the temporal resolution in the
first shell isotopic records was about 1 month in average and ranged from ~0.5 to ~3 months.
In most shells, resolution was close enough to one month (0.7 – 1.5 months) so that no further
analyses were required. In case of doubts in the identification of the seasonal extrema (warm
seasons in Peru commonly have two maxima), shell growth lines were analyzed a second time
to select the most likely option. If time resolution in an annual cycle was ~0.5 month,
consecutive data points were averaged in pairs so that the obtained value would integrate
about 1 month. If time resolution in an annual cycle was ~3 month or more, the shell was re-
sampled to reach monthly resolution. If better resolution could not be obtained, the isotope
record was partially or completely discarded.
The isotopic composition of shell microsamples corresponding to seasonal extrema was re-
analyzed to reduce the uncertainty on these strategic data points, except when extrema already
resulted from the averaging of two consecutive data points. Replicate analyses were
performed either on a remaining sub-sample or on a re-micromilled aragonite sample.
Replicates were averaged. Monthly shell δ18O values are available in the supplementary Data
Table S1.
2.4. SST calculations
δ18O values measured in shells were converted to SST using a paleotemperature equation
specifically calibrated with modern M. donacium shells from Llostay (10). Based on field
measurements, we used a modern seawater δ18O value of 0.2±0.1‰ for all sites (9). Seawater
δ18O was corrected for the effect of ice volume using the sea level reconstruction from (44),
and assuming a maximum ice volume effect of 1.05‰ for the last glacial maximum (45). The
same ice volume correction δiv was applied to every shell of sample interval and was
calculated for the middle of the corresponding time period. The error bar on δiv was estimated
from the range of values corresponding to the 1σ calibrated time interval.
The mean annual SST for an individual shell record was calculated as the mean of all the
monthly values from that shell (7,11). The mean annual SST for a sample of N shells was
calculated by averaging the mean values of the N shell records. The mean value of a shell
record can be slightly biased if the record does not contain an equal number of summers and
winters, but this effect is then averaged out when calculating the mean annual SST for the
consortium of individual shells comprising a depth / age interval. Mean SST values, ice
volume corrections and uncertainties are available in the supplementary Data Table S3.
2.5. Determining the seasonal range of SST (∆T)
The SST seasonal range, or ∆T, is proportional to the isotopic seasonal range ∆(δ18O) (10):
∆T = 3.66×∆(δ18O)
Normalized mean values, normalized variance, and skewness are therefore equivalent for both
∆T and ∆(δ18O). Determining ∆(δ18O) values is central to our study since these values are the
basis for characterizing ENSO. As discussed above (Sec. 2.3), maximum and minimum δ18O
values from each shell were replicated to improve precision. ∆(δ18O) values were calculated
as the absolute difference between consecutive seasonal extrema in shell δ18O records. They
include |summer(i)-winter(i)| and |summer(i+1)-winter(i)|. A complete annual cycle thus
yields two values of the seasonal range ∆(δ18O). Because consecutive ∆(δ18O) values are not
independent, we propagated uncertainty as follows. Using the number of degrees of freedom
N for a sample of n shells:
N=∑floor[(ki+1)/2]i=1,n
with ki being the number of ∆(δ18O) values in the ith shell record.
Past changes in the mean seasonal amplitude of SST were estimated from the average of
∆(δ18O) values in shell samples, normalized by the average value obtained in the closest
modern sample (Table S1). This way, the estimated relative changes are independent of the
SST-δ18Oshell relationship, and the influence of local variability is minimized. All ∆T values
are provided in the supplementary Data Table S2.
2.6. Estimating ENSO variance from combined samples
The variance of the Niño1+2 index is strongly correlated (R=0.85) to ENSO-related
interannual variability on the coast of Peru, which in turn is estimated from the variance of
∆(δ18O) values in a sample of shells (7). A larger dataset (typically 20 or more shells) is
required for a robust estimate of the variance and skewness than for the mean annual SST or
the mean seasonality (typically 5-10 shells) (11). We thus combined shell samples from
adjacent time periods (e.g. Table S1) to derive estimates of ENSO variance and skewness
with greater precision. ∆(δ18O) values of shell samples were centered before being combined
so that the variance and skewness of the combined data set was not biased by the difference in
the mean values of the samples.
The two modern samples (ICA and Llostay) were combined and used to normalize the results
obtained from fossil shells. Normalization is necessary to minimize archive (i.e., M.
donacium)-specific biases and facilitate comparison with normalized variance values from
corals and/or other biogenic carbonates.
It is important to note that var[∆(δ18O)]fossil/ var[∆(δ18O)]modern, our indicator of past ENSO
variance in the eastern Pacific, (1) is independent from the SST-δ18Oshell relationship and its
inherent uncertainties, and (2) is not influenced by decadal-to-centennial modulation of mean
annual SSTs.
We analyzed an 87-year instrumental SST time series (1925-2012) with monthly resolution
from Puerto Chicama, Peru, to estimate the relative contribution of interannual versus low-
frequency variability to total ∆T variability. For this period, the interannual frequency domain
(2-9 years) accounted for 75% of the total ∆T variance, or 67% excluding extreme El Niño
events. The decadal frequency domain (9-30 years) only represented 13% of the variance, or
18% excluding the 1982-83 and 1997-98 El Niño events. This analysis shows that the
variance of ∆T estimated from M. donacium shell samples in Peru is very likely to be
dominated by interannual variability.
Centennial scale variability may also partly contribute to the variance of the seasonal
amplitude in shell samples, especially in the shell accumulations that cover a long time
period, but the instrumental record is too short to estimate this variability. We minimized this
influence by centering ∆(δ18O) values of each shell group before combining the datasets for
variance and skewness calculations.
2.7. Uncertainty estimates
Standard errors of reconstructed SST statistics were calculated using the MoCo program,
which involves pseudo-proxy and Monte Carlo analyses described in Ref. 11. Standard errors
calculated by MoCo are the standard deviation of the population of errors generated by 5000
pseudo-reconstructions of the statistics of an instrumental SST time series. We used the
instrumental SST time series from Puerto Chicama IMARPE coastal laboratory (1925-2002)
without extreme El Niño years (1982-1983, 1997-1998), since shells record ENSO variance
without extreme events. Uncertainty sources simulated by MoCo include random sampling
within modern climate variability and within local scale variability, monthly seawater δ18O
variability, analytical error, and biological “noise” related to the mollusk archive (temperature
tolerance, random growth breaks, intra-shell heterogeneity).
The fact that we estimate the sampling-related uncertainty of fossil shells from modern
variability is a limitation on the accuracy of the standard errors we computed. However, since
modern variability appeared a posteriori larger in the modern sample than in any fossil
sample, standard errors may actually have been overestimated rather than underestimated.
The potential systematic error of mean annual SST reconstructions include (1) a potential
systematic error Ecal related to uncertainties in the SST-shell δ18O calibration (11), and (2) a
potential systematic error Eice related to uncertainties in the ice volume δ18O correction. Eice
was defined for a shell sample by the range of ice volume correction in the 1σ age interval of
our calibrated radiocarbon dates.
Although MoCo includes a wide range of uncertainty sources, MoCo estimates for skewness
standard errors are lower than simple parametric estimates considering random sampling from
a Gaussian distribution. This results from the fact that M. donacium has a temperature
threshold that results in the most extreme warm values being exluded from the assemblage of
shells during any time period. Since skewness is strongly sensitive to outliers, excluding
extreme values makes the estimate from a random sample more stable, and less sensitive to
sampling, explaining the low standard deviation of reconstructed skewness values. This
biological limitation of M. donacium growth therefore increases the fidelity with which fossil
shells capture ENSO skewness.
Parameterization of the MoCo program is described in Table S3. For normalized seasonality
(variance) estimates, standard errors were divided by the corresponding modern seasonality
(variance) value. Standard errors and potential systematic errors are reported in the
supplementary Data Table S3.
Table S3. MoCo parameterization Parameters Description Value Monte Carlo analysis N Number of specimen per sample See online dataset NY Number of years spanned by individual records 1 TS Biological superior limit for skeletal growth 20 °C TI Biological inferior limit for skeletal growth 6 °C gbi, i=1 to 12 Does skeletal growth occur during month i ? 1 gap How many random 1-month growth gaps per year ? 1 σS Standard deviation of spatial T variations 1.5 °C σW standard deviation of weather monthly noise 0.1 ‰ σC standard deviation of carbonate micro-heterogeneity 0.1 ‰ σA Analytical error (1σ) 0.08 ‰ Niter Number of iteration of the Monte Carlo analysis 5000
Proxy model calibration Alpha Slope of the linear proxy model -3.66 °C/‰ Beta Intercept of the linear proxy model 17.41 °C σT Standard deviation of T in calibration dataset 1.23 °C σP Standard deviation of shell δ18O in calibration dataset 0.43 ‰ R Pearson’s correlation coefficient in calibration dataset 0.89 NC Number of datapoints in calibration dataset 138 T0 Average value of T in calibration dataset 16.7 °C
2.8. ENSO variance from coral records
For a meaningful comparison between corals and mollusks, ENSO-related variance contained
in Central Pacific coral records was extracted slightly differently than in the original work (5).
We used monthly interpolated δ18O coral records from Palmyra and Line islands in Fig. 3D
(5, 46), available at http://www.ncdc.noaa.gov/data-access/paleoclimate-data/datasets/coral-
sclerosponge. Seasonal variability and decadal trends were removed by subtracting an 11-yr
sliding window of the averaged annual cycle. The detrended time series were then smoothed
with a 3-month running mean. The total variance of the detrended mollusk series was then
normalized by the total variance of the modern detrended coral series from the same island
(Palmyra, Fanning, or Christmas). ENSO anomalies were thus defined as anomalies of the
annual cycle, independent of decadal variations in the climatic mean state, as in mollusk
shells.
3. Supplementary text
3.1. El Niño in coastal Peru today and in the past: a brief review
ENSO determines 75% of the year-to-year sea temperature variability on the Peruvian coast,
but from the Peruvian perspective, El Niño refers to catastrophic events. It is now well
understood that these events are largest when they are centered in the Eastern tropical Pacific
(canonical mode), as opposed to the Central tropical Pacific. The largest ENSO events in the
instrumental record occurred in 1982-83 and 1997-98. SST anomalies reached ~8°C during
several consecutive months, with a major impact on coastal ecosystems, rainfall regime, and
the Peruvian economy. Traces of past catastrophic events were sought in archaeological
faunal remains and in coastal sediments.
The occurrence of warm-water mollusk species was studied in archaeological shell middens
on the central coast of Peru where only temperate water species occur today (47). Faunal
assemblages defined as tropical were observed prior to ~5.8 ka, and mixed (temperate and
tropical) assemblages were found after that time. This was interpreted as evidence of
permanently warm conditions north of 10°S in the early and middle Holocene, and the onset
of ENSO variability around 5.8 ka (47-49, 16). These authors also suggested that ENSO
activity started with a low frequency between 5.8 to 3.2-2.8 ka and reached a higher level of
activity after 2.8 ka that might have been responsible for the abandonment of temples in Peru
(48). An alternative interpretation of thermally anomalous mollusk assemblages based on
geomorphological and isotopic evidence suggests that the occurrence of these warm species
was facilitated by the presence of shallow bays protected by beach ridges (13, 50-52). Support
for the hypothesis of mid-Holocene tropical conditions in central Peru was provided by
isotopic analyses in fossil fish otoliths 6.4-6 ka (53), but issues with the fish species ecology
and with the origin of modern specimen used for reference made the interpretation
controversial (54).
Extreme El Niño events produce flood deposits in the coastal valleys. In the Casma valley, 13
flood events were radiocarbon dated to have occurred in the last 3200 years, and a total of 18
such events were identified in the entire Holocene (27). On the southern coast of Peru, close
to the QLB site, 6 El Niño-related flood deposits were identified 12-8.4 ka, zero flood event
occurred 8.4 to 5.3 ka, and 4 events occurred after 5.3 ka (17, 55), implying to a strong EP
mode of ENSO before 8.4 ka. The influence of winter rainfall events due to increased
southerly winds has also been proposed to explain these flood deposits (55).
A record of coastal floods was also obtained from the analysis of the concentration of lithic
grains in the SO47-106KL sediment core, collected at 12°S off the Peruvian coast (27). This
core is well located to record large river discharge related to extreme El Niño events.
However, the freshwater flow generating the lithic flux has a mixed origin. The regular yearly
lithic flux is driven by precipitation in the nearby high Andes. Rainfall events related to
extreme El Niños are characterized by large lithic flux anomalies larger than 4 standard
deviations in the detrended record (28). Eight flood events occurred prior to 8 ka, and 14
occurred after 4 ka, a result consistent with the coastal geomorphological reconstructions.
Despite these rare floods and a relative increase of humidity in the early Holocene, the coastal
climate remained arid throughout the Holocene (16, 56).
3.2. Recording ΔT distribution: strength and limits of the proxy
The distribution of seasonal ΔT values in Peru reflects ENSO activity. Since mollusk shells
do not record the full range of ΔT values, the impact on ENSO reconstruction needs to be
assessed.
First, very low seasonal amplitudes may not be detected if they are similar to the monthly
noise. Such years of very low ΔT values would be identified because the chronology of the
records is controlled by the analysis of shell growth lines. However, ΔT values lower than the
monthly uncertainty cannot be estimated. Here, the monthly uncertainty in SST
reconstructions is due to monthly seawater isotopic variability (σ=0.1 ‰), analytical error
(0.08 ‰), and micro-scale biocarbonate heterogeneity (0.1 ‰), and is estimated at 0.6 °C.
Although ΔT values lower than 0.6 °C may occur in very strong La Niña conditions, the
lowest ΔT value in the instrumental record is 1.2 °C, observed in 1985 in Callao. The
detection limit is therefore low enough to detect the large majority of ΔT values.
The largest ΔT values cannot be recorded because of shell mortality during extreme El Niño
events. The proxy may thus underestimate ENSO variability in periods of strong ENSO
activity. The significance of this potential bias can be estimated using the sedimentological
records of flood events (28). 22 El Niño-related flood events were recorded in the Holocene
(Fig. 3B), representing an average frequency of 1 event in ~450 years, and a maximum
frequency of 1 event in 250 years in the late Holocene. 250 to 500 years is the typical time
period represented by a shell midden in which a random shell sample yields 42 years on
average. This means that, even if shells were able to record these extreme events, the
probability of missing them is 83 to 92%, which is also the level of confidence for our records
to be unbiased by the mortality effect.
In cooler periods like the early Holocene, a relatively higher upper temperature limit for shell
growth may have allowed shells to record larger ΔT values, which may lead to an over-
estimation of ΔT mean and variance compared to warmer periods like the late Holocene.
However, the largest ΔT values (individual or average) were not observed in the cooler
periods but in the warmer modern sample, which implies that this bias does not affect our
dataset.
Moderate to strong El Niño events may generate rainfall events on the northern coast of Peru,
but the correlation with rainfall on the central and southern coast where our sites are located is
not significant (29, 57). When they do occur, coastal rainfall in the southern coast related to
moderate to strong El Niño events, though exceptional, are limited to short showers that are
too weak to have any significant impact on the seawater isotopic composition (57). It is thus
very unlikely that the shell isotopic record presented here was biased by El Niño-related
rainfall events.
Figure S11. Ontogenic δ18O records measured in M. donacium shells grouped by sample (Table S1). Shells within each sample are ordered as in Data Table S1.
References and Notes 1. A. Santoso, S. McGregor, F. F. Jin, W. Cai, M. H. England, S. I. An, M. J. McPhaden, E.
Guilyardi, Late-twentieth-century emergence of the El Niño propagation asymmetry and future projections. Nature 504, 126–130 (2013). Medline doi:10.1038/nature12683
2. A. C. Clement, R. Seager, M. A. Cane, Orbital controls on the El Niño/Southern Oscillation and the tropical climate. Paleoceanography 14, 441–456 (1999). doi:10.1029/1999PA900013
3. W. Zheng, P. Braconnot, E. Guilyardi, U. Merkel, Y. Yu, ENSO at 6ka and 21ka from ocean–atmosphere coupled model simulations. Clim. Dyn. 30, 745–762 (2008). doi:10.1007/s00382-007-0320-3
4. Y. Luan, P. Braconnot, Y. Yu, W. Zheng, O. Marti, Early and mid-Holocene climate in the tropical Pacific: Seasonal cycle and interannual variability induced by insolation changes. Clim. Past 8, 1093–1108 (2012). doi:10.5194/cp-8-1093-2012
5. K. M. Cobb, N. Westphal, H. R. Sayani, J. T. Watson, E. Di Lorenzo, H. Cheng, R. L. Edwards, C. D. Charles, Highly variable El Niño-Southern Oscillation throughout the Holocene. Science 339, 67–70 (2013). Medline doi:10.1126/science.1228246
6. T. Lee, M. J. McPhaden, Geophys. Res. Lett. 37, L14603 (2010).
7. M. Carré, J. P. Sachs, A. J. Schauer, W. E. Rodríguez, F. C. Ramos, Palaeogeogr. Palaeoclimatol. Palaeoecol. 371, 45 (2013). doi:10.1016/j.palaeo.2012.12.014
8. Materials and methods are available as supplementary materials on Science Online.
9. D. H. Sandweiss, H. McInnis, R. L. Burger, A. Cano, B. Ojeda, R. Paredes, M. C. Sandweiss, M. D. Glascock, Quebrada jaguay: Early south american maritime adaptations. Science 281, 1830–1832 (1998). Medline doi:10.1126/science.281.5384.1830
10. M. Carré, I. Bentaleb, D. Blamart, N. Ogle, F. Cardenas, S. Zevallos, R. M. Kalin, L. Ortlieb, M. Fontugne, Stable isotopes and sclerochronology of the bivalve Mesodesma donacium: Potential application to Peruvian paleoceanographic reconstructions. Palaeogeogr. Palaeoclimatol. Palaeoecol. 228, 4–25 (2005). doi:10.1016/j.palaeo.2005.03.045
11. M. Carré, J. P. Sachs, J. M. Wallace, C. Favier, Exploring errors in paleoclimate proxy reconstructions using Monte Carlo simulations: Paleotemperature from mollusk and coral geochemistry. Clim. Past 8, 433–450 (2012). doi:10.5194/cp-8-433-2012
12. A. Koutavas, S. Joanides, El Niño-Southern Oscillation extrema in the Holocene and Last Glacial Maximum. Paleoceanography 27, PA4208 (2012). doi:10.1029/2012PA002378
13. M. Carré, M. Azzoug, I. Bentaleb, B. M. Chase, M. Fontugne, D. Jackson, M.-P. Ledru, A. Maldonado, J. P. Sachs, A. J. Schauer, Mid-Holocene mean climate in the south eastern Pacific and its influence on South America. Quat. Int. 253, 55–66 (2012). doi:10.1016/j.quaint.2011.02.004
14. J. Sadler, M. Carré, M. Azzoug, A. J. Schauer, J. Ledesma, F. Cardenas, B. M. Chase, I. Bentaleb, S. D. Muller, M. Mandeng, E. J. Rohling, J. P. Sachs, Reconstructing past upwelling intensity and the seasonal dynamics of primary productivity along the Peruvian
coastline from mollusk shell stable isotopes. Geochem. Geophys. Geosyst. 13, Q01015 (2012). doi:10.1029/2011GC003595
15. H. V. McGregor, M. J. Fischer, M. K. Gagan, D. Fink, S. J. Phipps, H. Wong, C. D. Woodroffe, A weak El Niño/Southern Oscillation with delayed seasonal growth around 4,300 years ago. Nat. Geosci. 6, 949–953 (2013). doi:10.1038/ngeo1936
16. M. Fontugne, P. Usselmann, D. Lavallée, M. Julien, C. Hatté, El Niño Variability in the Coastal Desert of Southern Peru during the Mid-Holocene. Quat. Res. 52, 171–179 (1999). doi:10.1006/qres.1999.2059
17. D. K. Keefer, M. E. Moseley, S. D. deFrance, A 38 000-year record of floods and debris flows in the Ilo region of southern Peru and its relation to El Niño events and great earthquakes. Palaeogeogr. Palaeoclimatol. Palaeoecol. 194, 41–77 (2003). doi:10.1016/S0031-0182(03)00271-2
18. D. H. Sandweiss et al., in Climate Change and Cultural Dynamics: A Global Perspective on Mid-Holocene Transitions D. G. Anderson, K. A. Maasch, D. H. Sandweiss, Eds. (Academic Press, Elsevier, 2007) pp. 25-50.
19. D. T. Rodbell, G. O. Seltzer, D. M. Anderson, M. B. Abbott, D. B. Enfield, J. H. Newman, An approximately 15,000-year record of El Nino-driven alluviation in southwestern ecuador. Science 283, 516–520 (1999). Medline doi:10.1126/science.283.5401.516
20. C. M. Moy, G. O. Seltzer, D. T. Rodbell, D. M. Anderson, Variability of El Niño/Southern Oscillation activity at millennial timescales during the Holocene epoch. Nature 420, 162–165 (2002). Medline doi:10.1038/nature01194
21. J. L. Conroy, J. T. Overpeck, J. E. Cole, T. M. Shanahan, M. Steinitz-Kannan, Holocene changes in eastern tropical Pacific climate inferred from a Galápagos lake sediment record. Quat. Sci. Rev. 27, 1166–1180 (2008). doi:10.1016/j.quascirev.2008.02.015
22. D. T. Rodbell, G. O. Seltzer, B. G. Mark, J. A. Smith, M. B. Abbott, Clastic sediment flux to tropical Andean lakes: Records of glaciation and soil erosion. Quat. Sci. Rev. 27, 1612–1626 (2008). doi:10.1016/j.quascirev.2008.06.004
23. A. W. Tudhope, C. P. Chilcott, M. T. McCulloch, E. R. Cook, J. Chappell, R. M. Ellam, D. W. Lea, J. M. Lough, G. B. Shimmield, Variability in the El Niño-Southern Oscillation through a glacial-interglacial cycle. Science 291, 1511–1517 (2001). Medline doi:10.1126/science.1057969
24. H.-Y. Kao, J.-Y. Yu, Contrasting Eastern-Pacific and Central-Pacific Types of ENSO. J. Clim. 22, 615–632 (2009). doi:10.1175/2008JCLI2309.1
25. S. I. An, F. F. Jin, Nonlinearity and Asymmetry of ENSO. J. Clim. 17, 2399–2412 (2004). doi:10.1175/1520-0442(2004)017<2399:NAAOE>2.0.CO;2
26. B. Dewitte, J. Vazquez-Cuervo, K. Goubanova, S. Illig, K. Takahashi, G. Cambon, S. Purca, D. Correa, D. Gutierrez, A. Sifeddine, L. Ortlieb, Change in El Niño flavours over 1958–2008: Implications for the long-term trend of the upwelling off Peru. Deep Sea Res. Part II Top. Stud. Oceanogr. 77-80, 143–156 (2012). doi:10.1016/j.dsr2.2012.04.011
27. B. Rein, A. Lückge, L. Reinhardt, F. Sirocko, A. Wolf, W.-C. Dullo, El Niño variability off Peru during the last 20,000 years. Paleoceanography 20, n/a (2005). 10.1029/2004PA001099 doi:10.1029/2004PA001099
28. B. Rein, How do the 1982/83 and 1997/98 El Niños rank in a geological record from Peru? Quat. Int. 161, 56–66 (2007). doi:10.1016/j.quaint.2006.10.023
29. L. E. Wells, Holocene history of the El Niño phenomenon as recorded in flood sediments of northern coastal Peru. Geology 18, 1134 (1990). doi:10.1130/0091-7613(1990)018<1134:HHOTEN>2.3.CO;2
30. A. T. Wittenberg, Are historical records sufficient to constrain ENSO simulations? Geophys. Res. Lett. 36, L12702 (2009). doi:10.1029/2009GL038710
31. P. Braconnot, Y. Luan, S. Brewer, W. Zheng, Impact of Earth’s orbit and freshwater fluxes on Holocene climate mean seasonal cycle and ENSO characteristics. Clim. Dyn. 38, 1081–1092 (2012). doi:10.1007/s00382-011-1029-x
32. K. Takahashi, A. Montecinos, K. Goubanova, B. Dewitte, ENSO regimes: Reinterpreting the canonical and Modoki El Niño. Geophys. Res. Lett. 38, L10704 (2011). doi:10.1029/2011GL047364
33. R. W. Reynolds, N. A. Rayner, T. M. Smith, D. C. Stokes, W. Wang, An Improved In Situ and Satellite SST Analysis for Climate. J. Clim. 15, 1609–1625 (2002). doi:10.1175/1520-0442(2002)015<1609:AIISAS>2.0.CO;2
34. F. Engel, Southwest. J. Anthropol. 13, 54 (1957).
35. F. Engel, Les amas de coquillages de la côte péruvienne (Ancón-Río Ica). J. Soc. Am. 44, 39–47 (1955). doi:10.3406/jsa.1955.2594
36. J. H. Rowe, Archaeological Explorations in Southern Peru, 1954-1955. Am. Antiq. 22, 135 (1956). doi:10.2307/276816
37. D. Lavallée et al., Chungara. Revista de Antropología Chilena 43, 333 (2011).
38. D. Lavallée, M. Julien, Prehistoria de la Costa Extremo-Sur del Perú. Los Pescadores Arcaicos de la Quebrada de los Burros (10000 –7000 a.P.) (Travaux de l'Institut Français d'Etudes Andines, Lima, 2012), pp. 478.
39. L. Ortlieb, G. Vargas, J.-F. Saliège, Marine radiocarbon reservoir effect along the northern Chile–southern Peru coast (14–24°S) throughout the Holocene. Quat. Res. 75, 91–103 (2011). doi:10.1016/j.yqres.2010.07.018
40. B. Schöne, D. Bentley, Nautilus 116, 25 (2002).
41. C. P. Glover, S. M. Kidwell, Influence of Organic Matrix on the Post-Mortem Destruction of Molluscan Shells. J. Geol. 101, 729–747 (1993). doi:10.1086/648271
42. T. S. Tobin, A. J. Schauer, E. Lewarch, Alteration of micromilled carbonate δ18O during Kiel Device analysis. Rapid Commun. Mass Spectrom. 25, 2149–2152 (2011). Medline doi:10.1002/rcm.5093
43. K. Lambeck, J. Chappell, Sea level change through the last glacial cycle. Science 292, 679–686 (2001). Medline doi:10.1126/science.1059549
44. J.-C. Duplessy, L. Labeyrie, C. Waelbroeck, Constraints on the ocean oxygen isotopic enrichment between the Last Glacial Maximum and the Holocene: Paleoceanographic implications. Quat. Sci. Rev. 21, 315–330 (2002). doi:10.1016/S0277-3791(01)00107-X
45. K. M. Cobb, C. D. Charles, H. Cheng, R. L. Edwards, El Niño/Southern Oscillation and tropical Pacific climate during the last millennium. Nature 424, 271–276 (2003). Medline doi:10.1038/nature01779
46. D. H. Sandweiss, in Case Studies in Environmental Archaeology E. J. Reitz, L. A. Newsom, S. J. Scudder, Eds. (Plenum Press, New York, 1996).
47. D. H. Sandweiss, K. A. Maasch, R. L. Burger, J. B. Richardson III, H. B. Rollins, A. Clement, Variation in Holocene El Niño frequencies: Climate records and cultural consequences in ancient Peru. Geology 29, 603 (2001). doi:10.1130/0091-7613(2001)029<0603:VIHENO>2.0.CO;2
48. D. H. Sandweiss, Terminal Pleistocene through Mid-Holocene archaeological sites as paleoclimatic archives for the Peruvian coast. Palaeogeogr. Palaeoclimatol. Palaeoecol. 194, 23–40 (2003). doi:10.1016/S0031-0182(03)00270-0
49. T. J. De Vries, L. E. Wells, Thermally-anomalous Holocene molluscan assemblages from coastal Peru: Evidence for paleographic, not climatic change. Palaeogeogr. Palaeoclimatol. Palaeoecol. 81, 11–32 (1990). doi:10.1016/0031-0182(90)90037-8
50. C. Perrier, C. Hillaire-Marcel, L. Ortlieb, Paléogéographie littorale et enregistrement isotopique (C, O) d’événements de type El Niño par les mollusques holocènes et récents du nord-ouest péruvien. Geogr. Phys. Quat. 48, 23 (1994). doi:10.7202/032970ar
51. T. J. De Vries et al., Determining the Early History of El Nio. Science 276, 965–967 (1997). doi:10.1126/science.276.5314.965
52. C. F. T. Andrus, D. E. Crowe, D. H. Sandweiss, E. J. Reitz, C. S. Romanek, Otolith delta 18O record of mid-Holocene sea surface temperatures in Peru. Science 295, 1508–1511 (2002). Medline doi:10.1126/science.1062004
53. P. Béarez, T. J. DeVries, L. Ortlieb, Comment on “Otolith delta18O record of mid-Holocene sea surface temperatures in Peru”. Science 299, 203a (2003). Medline doi:10.1126/science.1076173
54. G. Vargas, J. Rutllant, L. Ortlieb, ENSO tropical–extratropical climate teleconnections and mechanisms for Holocene debris flows along the hyperarid coast of western South America (17°–24°S). Earth Planet. Sci. Lett. 249, 467–483 (2006). doi:10.1016/j.epsl.2006.07.022
55. B. Mächtle, I. Unkel, B. Eitel, B. Kromer, S. Schiegl, Molluscs as evidence for a late Pleistocene and early Holocene humid period in the southern coastal desert of Peru (14.5°S). Quat. Res. 73, 39–47 (2010). doi:10.1016/j.yqres.2009.05.007
56. S. Rome-Gaspaldy, J. Ronchail, Bull. Inst. Fr. Etudes Andines 27, 675 (1998).