Millennial-scale plankton regime shifts in the subtropical North Pacific Ocean

See allHide authors and affiliations

Science  18 Dec 2015:
Vol. 350, Issue 6267, pp. 1530-1533
DOI: 10.1126/science.aaa9942

Community changes centuries in the making

How might climate change affect the base of the marine food chain? Phytoplankton, the foundation of the marine ecosystem, depend on ambient oceanographic conditions such as temperature, salinity, and nutrient availability, which affect ocean chemistry and isotopic distributions. McMahon et al. report carbon isotopic composition changes in the North Pacific Ocean over the past 1000 years, which reflect changes in the community composition of phytoplankton in the region (see the Perspective by Vogt). An ongoing trend toward greater prevalence of nitrogen-fixing cyanobacteria that began 100 years ago might lead to a more efficient carbon pump and remove increasing amounts of CO2 from the atmosphere.

Science, this issue p. 1530; see also p. 1466


Climate change is predicted to alter marine phytoplankton communities and affect productivity, biogeochemistry, and the efficacy of the biological pump. We reconstructed high-resolution records of changing plankton community composition in the North Pacific Ocean over the past millennium. Amino acid–specific δ13C records preserved in long-lived deep-sea corals revealed three major plankton regimes corresponding to Northern Hemisphere climate periods. Non–dinitrogen-fixing cyanobacteria dominated during the Medieval Climate Anomaly (950–1250 Common Era) before giving way to a new regime in which eukaryotic microalgae contributed nearly half of all export production during the Little Ice Age (~1400–1850 Common Era). The third regime, unprecedented in the past millennium, began in the industrial era and is characterized by increasing production by dinitrogen-fixing cyanobacteria. This picoplankton community shift may provide a negative feedback to rising atmospheric carbon dioxide concentrations.

Gone is the paradigm of the oligotrophic subtropical gyres as vast oceanic deserts. In the recent instrumental record, a new picture has emerged of substantial dynamics in plankton community structure, biogeochemical cycling, and export production (14). Numerous lines of evidence suggest that shifts in phytoplankton community regimes are intimately connected to oceanographic conditions (2, 4, 5). For instance, the 1976 polarity reversal of the Pacific Decadal Oscillation caused a shoaling of the mixed layer and subsequent declines in available nutrients to the North Pacific Subtropical Gyre (NPSG); these conditions probably promoted the food-web regime shift from a eukaryotic to a prokaryotic, cyanobacteria-dominated system (2). Such changes are superimposed on secular shifts associated with the areal increase of subtropical gyres that has been ongoing for at least 25 years (6). In the face of increasing climate change, it is imperative to understand recent changes at the base of the NPSG food web in the context of longer-term trends. However, our understanding of how NPSG plankton communities have shifted on centennial time scales has been limited by a lack of available methods and paleoarchives of sufficient length and resolution.

Hawaiian gold corals (Kulamanamana haumeaae) are extraordinarily long-lived deep-sea organisms that record the biogeochemical signatures of recently exported production in their proteinaceous skeletons (7, 8). We generated millennial-length records of bulk stable carbon isotopes (δ13Cbulk) from specimens of K. haumeaae collected from the top of the mesopelagic zone at two sites in the Hawaiian archipelago (Fig. 1). The δ13Cbulk records showed remarkable congruence, characterized by a gradual increase of ~1.0 per mil (‰) from ~1000 to ~1850 CE, followed by a rapid decrease of –1.0‰ from ~1850 to the present, after correcting for the Suess effect (Fig. 2A) (911). These changes in δ13Cbulk imply multicentennial-scale shifts in δ13C values associated with primary production, which we hypothesize reflect major changes in plankton community structure over the past 1000 years.

Fig. 1 NPSG productivity distribution, with sample locations.

(A) Spatial extent of the oligotrophic NPSG, determined from spring 2012 chlorophyll a concentrations measured remotely by NASA’s Aqua/MODIS (Moderate Resolution Imaging Spectroradiometer). The white box indicates the area shown in (B). [Image courtesy of NASA Goddard’s Ocean Biology Processing Group] (B) K. haumeaae sampling locations at Makapuu and French Frigate Shoals relative to the oceanographic station ALOHA (indicated by the X), overlain on ocean surface nitrate concentrations (National Oceanographic Data Center, 2013;

Fig. 2 1000-year bulk and essential amino acid δ13C records from deep-sea corals in the NPSG.

K. haumeaae–derived records of (A) δ13Cbulk (solid lines show the 20-year average; analytical error, 0.05‰) and (B) δ13CPhe (Phe, phenylalanine; squares; analytical error, 0.2‰) from Makapuu live coral (purple), Makapuu fossil coral (magenta), and French Frigate Shoals live coral (brown). All records have been corrected for the oceanic Suess effect since 1860 (10, 11). Well-known Northern Hemisphere climate phenomena are overlaid for reference (18).

Bulk δ13C records integrate the combined influences of the δ13C value of inorganic carbon utilized during carbon fixation, shifts in plankton community structure, trophic changes, and biochemical fractionation. To isolate plankton source signatures within this signal, we applied a powerful fingerprinting approach to the sampled deep-sea corals, based on the normalized δ13C values of essential amino acids (δ13CEAA) in primary producers (11). These δ13CEAA fingerprints reflect the substantial metabolic diversity in EAA synthesis pathways and associated isotope effects among evolutionarily distinct primary producers (12, 13). We found diagnostic multivariate patterns in literature values of normalized δ13CEAA among four key source end-members relevant to the NPSG [eukaryotic microalgae, dinitrogen (N2)–fixing and non–N2-utilizing cyanobacteria, and heterotrophic bacteria] (fig. S2). Because animals cannot synthesize the carbon skeletons of EAAs (14), these δ13CEAA fingerprints are incorporated, virtually unmodified, into upper-trophic-level consumers, including gorgonin corals (15). Furthermore, δ13CEAA fingerprints are robust to the many factors affecting bulk δ13C values, such as environmental and growth conditions (13, 16).

A subset of the deep-sea coral samples spanning the entire 1000-year record was analyzed for δ13CEAA at ~20-year resolution (table S1). δ13CEAA values were strongly correlated with δ13Cbulk (fig. S1), indicating that trends in δ13Cbulk can be attributed to changes in source carbon at the base of the food web (15). For example, the δ13CEAA of phenylalanine (Fig. 2B) mirrored that of δ13Cbulk (Fig. 2A); however, the magnitude of change was 10 times larger than in δ13Cbulk. This suggests that isotopic contributions from other macromolecules had a strong muting effect on δ13Cbulk values, and thus δ13CEAA is likely a more sensitive record of changes in primary producer δ13C than δ13Cbulk is (15). Our results indicate that variability in δ13C values of exported production was much larger than would be inferred from coral δ13Cbulk records alone, and strongly suggest broad changes in the sources of exported primary production through time.

To reconstruct past shifts in the relative contributions of major phytoplankton groups to export production in the NPSG, we applied a Bayesian stable isotope–mixing model to published source end-member δ13CEAA fingerprints (11) and the δ13CEAA records of K. haumeaae, normalized to their respective means (fig. S2 and table S2). Over the entire 1000-year record, photoautotrophic carbon dominated the corals’ sinking particulate organic matter (POM) food source (mean, 87 ± 6%), with a relatively small heterotrophic bacterial contribution (13 ± 6%) (Fig. 3). Likewise, prokaryotic cyanobacterial sources dominated photoautrophic carbon (63 ± 14%), with a moderate mean contribution from eukaryotic microalgae (24 ± 10%) (Fig. 3). On centennial time scales, however, relative contributions from photoautotrophic end-members shifted dramatically. For example, between 1000 and 1850 CE, cyanobacteria decreased from 80 to 50% of total exported production, offset by an equivalent increase in eukaryotic microalgae up to a peak contribution of ~45% in the early 1800s. Although previous studies have noted enhanced diatom abundances in the NPSG associated with mesoscale oceanographic features (17), such a sustained high level of eukaryotic microalgal production has never been observed in the modern instrumental record. The most conservative explanation of our data is that the changes in the phylogenetic identity of sources contributing to export production reflect changes in the relative community composition of surface plankton through time. An alternate, albeit highly unlikely, hypothesis is that surface plankton community composition has remained relatively constant through time, and instead the degree of decoupling between surface and export production has undergone dramatic changes as a function of climate shifts (11).

Fig. 3 Millennial record of relative carbon contributions from different primary producers recorded in deep-sea corals.

Shown are the relative contributions of carbon from N2-fixing cyanobacteria (black), non–N2-utilizing cyanobacteria (yellow), eukaryotic microalgae (green), and heterotrophic bacteria (gray), recorded in deep-sea corals over the past 1000 years. Corals were collected at 450 m depth in the Hawaiian archipelago in the NPSG. Relative contributions are based on a compound-specific Bayesian stable isotope–mixing model of normalized phytoplankton end-member and coral δ13CEAA values.

To better constrain the patterns of changing plankton community composition, we applied a hierarchical cluster analysis to the normalized δ13CEAA data (11). This approach identified three distinct plankton community regimes that corresponded temporally to well-known Northern Hemisphere climate phenomena (Fig. 4). The first regime corresponded to the Medieval Climate Anomaly [MCA; 950–1250 CE (18)], with δ13CEAA fingerprints indicative of export production dominated by nitrate (NO3)–utilizing cyanobacteria. There is general consensus that the putative MCA in the northern mid-latitudes was similar to the climate of the mid-20th century (18, 19), implying relatively warm sea surface temperatures, weak winds, shallow mixed-layer depths, and resultant nutrient limitation, all favoring a microbial loop–dominated community (2). The second regime corresponded to the Little Ice Age [LIA; 1400–1850 CE (18)]. In this regime, the plankton assemblage contributing to export production transitioned from a cyanobacteria-dominated community to one far more strongly influenced by eukaryotic microalgae (Fig. 3). This shift probably reflects a transition in the LIA to cooler sea surface temperatures, a reduction in stratification, an increase in mixed-layer depth, and an inferred increase in the supply of inorganic nitrate from depth (4, 20).

Fig. 4 Phytoplankton regime shifts recorded in deep-sea corals.

Shown is a dendrogram of similarity in exported plankton carbon utilized by deep-sea corals over the past 1000 years, based on an average-link hierarchical cluster analysis. The dendrogram is separated into three significantly different clusters according to multiscale bootstrapping with approximately unbiased P values >0.95. The dates (CE) are colored based on overlap with well-known Northern Hemisphere climate phenomena (18).

The third and current regime began at the end of the LIA and at the onset of the modern industrial age (~1850 CE) (Fig. 4). This regime is distinguished by a transition back to a cyanobacteria-dominated system. However, unlike the MCA period, the current regime is characterized by a biogeochemically distinct group of cyanobacteria, the N2-fixing diazotrophs. Historically, the availability of inorganic nitrogen (N) and/or phosphorus (P) was thought to limit plankton production in the NPSG (21). Since ~1850 CE, however, sea surface temperatures have increased, accompanied by a likely decrease in the trade winds concomitant with gyre expansion, as a result of Northern Hemisphere warming. The resulting increase in stratification and decrease in nutrient availability may have selected for a N2-fixing cyanobacterial community, as observed in the instrumental record over the past ~20 years (2, 22). Currently declining P inventories and increasing N:P ratios in the mixed layer at the HOT-ALOHA (Hawaiian Ocean Time-series–A Long-Term Oligotrophic Habitat Assessment) oceanographic station are thought to reflect this decades-long increase in N2 fixation (1, 2, 8), an idea that is further supported by recent literature suggesting that canonical Redfield ratios in the NPSG may be more plastic than previously realized (23, 24).

Our δ13CEAA fingerprinting data, which show a 47% increase in N2-fixing cyanobacteria carbon in exported POM since the end of the LIA, correspond well with recent evidence of a 17 to 27% increase in NPSG N2-fixation since ~1850 CE, determined from amino acid–specific nitrogen isotopes (δ15NAA) in the same suite of K. haumeaae specimens as used in this study (8). These studies represent fully independent lines of evidence supporting the hypothesis that recent decreases in δ15Nbulk values of exported POM in the NPSG are related to increases in diazotrophic plankton within a microbial loop–driven system (11).

By offering the first direct phylogenetic context for long-term shifts in isotopic records of exported POM, our data provide a major new constraint in understanding the evolution of NPSG biogeochemistry. For example, a recently proposed alternate hypothesis contends that advection of 15N-depleted nitrate from the Eastern Tropical Pacific, associated with a reduction in denitrification (25), might explain recent low δ15N values in the NPSG; similarly, Kim et al. (26) suggested that atmospheric N deposition is the dominant factor driving increases in values of N* (a nitrate- and phosphate-based tracer of N2 fixation and denitrification) across the Pacific. However, the δ15N value of N entrained in the mixed layer should not, by itself, affect planktonic community structure. Our new evidence for a profound phylogenetic community shift is fully consistent with increasing N2 fixation, probably linked to overall increased stratification and reductions in upwelled nitrate, over the past 100 years.

Taken together, our data show that phytoplankton community structure in the NSPG is subject to multicentennial shifts that are broadly linked to climate conditions. They also reveal that the present-day cyanobacterial community, which is characterized by strongly enhanced N2 fixation, is unprecedented within at least the past 1000 years. The transition to the current cyanobacterial regime (<200 years) was much faster than the transition from cyanobacterial dominance during the MCA to eukaryotic dominance during the LIA (>600 years). Both the nature and the rate of change of the current dominant autotrophic assemblage strongly suggest continuing rapid changes in NPSG plankton community structure associated with anthropogenic climate change and are consistent with the predicted expansion of N2-fixing cyanobacteria habitat (27).

Regime shifts in plankton community composition have far-reaching implications for productivity, food-web dynamics, biogeochemical cycling, and the efficacy of the biological pump (22, 28, 29). The fact that dominant cyanobacterial signatures were recorded in deep-sea corals from the mesopelagic zone strongly suggests that continuing shifts to an N2-fixing prokaryotic regime have fundamentally altered the main sources of exported POM. These observations also support recent evidence (3, 30) that small-cell picoplankton production, free-living and/or in cyanobacteria-diatom symbioses (11, 22), may be a more important component of export production in oligotrophic gyres than traditionally recognized. Further, recent studies have shown that plankton elemental stoichiometry is more variable than previously assumed under the classical Redfield paradigm, with C:P ratios being several times higher in the oligotrophic gyres than in upwelling regions (23, 24). This suggests that carbon export could actually be more efficient (per mole of P) in the oligotrophic gyres, despite their lower overall productivity, and, furthermore, that increasing nutrient limitation in warmer and more stratified oceans over the past 100 years may have served as a major negative feedback on rising CO2 concentrations (23, 24). Our finding that the phylogenetic origin of export production in the NPSG has trended toward N2-fixing prokaryotes over the past century strongly supports this idea. If small-cell export does in fact act as a more efficient carbon pump, our new records suggest that this carbon cycle feedback has already been operating for the past 100 years. For this feedback loop to persist into the future, the system cannot become phosphate-limited.

Supplementary Materials

Materials and Methods

Figs. S1 and S2

Tables S1 and S2

References (3168)


  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: All methods, additional figures, and source data are available in the supplementary materials. K.W.M., T.P.G., and M.D.M. conceived the project. K.W.M. prepared samples, performed bulk and compound-specific δ13C analyses, and wrote the manuscript. O.A.S. and T.L. assisted in data analysis and commented on the manuscript. T.P.G. and M.D.M. supervised this project, discussed the results, and commented on the manuscript. We thank M. Hanson, S. Fauqué, and J. Liu for assistance in the laboratory. This work would not have been possible without the captain and crew of the research vessel Ka‘imikai-o-Kanaloa and the pilots and engineers of the Hawaii Undersea Research Laboratory's Pisces IV and V submersibles. We also thank three anonymous reviewers for valuable feedback on the manuscript. Funding for sample collection was provided by the National Oceanic and Atmospheric Administration’s National Undersea Research Program and the National Geographic Society (grant 7717-04). A portion of this work was performed under the auspices of the U.S. Department of Energy (grant DE-AC52-07NA27344). The majority of the work presented here was funded by NSF (grant OCE-1061689). T.L. was supported by The Future Ocean, a program funded by the German Research Foundation.
View Abstract

Stay Connected to Science

Navigate This Article