全球变化知识资源中心

Climate change is altering oceanic conditions in a complex manner, and the concurrent amendment of multiple properties will modify environmental stress for primary producers. So far, global modelling studies have focused largely on how alteration of individual properties will affect marine life. Here, we use global modelling simulations in conjunction with rotated factor analysis to express model projections in terms of regional trends in concomitant changes to biologically influential multi-stressors. Factor analysis demonstrates that regionally distinct patterns of complex oceanic change are evident globally. Preliminary regional assessments using published evidence of phytoplankton responses to complex change reveal a wide range of future responses to interactive multi-stressors with <20–300% shifts in phytoplankton physiological rates, and many unexplored potential interactions. In a future ocean, provinces will encounter different permutations of change that will probably alter the dominance of key phytoplankton groups and modify regional productivity, ecosystem structure and biogeochemistry. Consideration of regionally distinct multi-stressor patterns can help guide laboratory and field studies as well as the interpretation of interactive multi-stressors in global models.

Multiple lines of evidence, ranging from time-series observations to climate modelling experiments, demonstrate the ongoing role of climate change in modifying many ocean properties such as temperature, salinity and pH (refs 1, 2, 3). Coupled ocean–atmosphere–land Earth system models link present-day evidence of a changing ocean³ with that of a future ocean by providing detailed projections of how climate change will continue to alter concurrently a range of characteristics, for example, enhanced vertical density stratification in the upper ocean, over the coming decades⁴. The effect of changing conditions on marine life has been explored in detail using manipulation experiments in which individual oceanic properties such as pH are perturbed on the basis of future climate change modelling projections⁵. Initial global modelling studies concentrated on the potential impact of a subset of processes on planktonic organisms, for example, changes in temperature, nutrients and stratification on phytoplankton growth⁴ or ocean acidification on calcification⁶. Recently, coupled Earth system model studies have begun to focus more on the complexity of these climate-change-mediated environmental changes, including the overlapping effects of warming, acidification and/or hypoxia and their influence on ocean biogeochemistry^{2, 7}. The goal of our study is a new framework for interpreting and visualizing coupled Earth system model results and helping design future laboratory and field experiments.

Two complementary approaches have been taken by modellers investigating how changing oceanic conditions will alter phytoplankton productivity and the resulting biogeochemical signatures^{8, 9}. A number of coupled Earth models incorporate phytoplankton–zooplankton dynamics by representing either a single generic phytoplankton or several phytoplankton functional groups^{7, 8} (for example, size classes, biominerals). At the other end of the spectrum are simulations that use a phytoplankton community of ~100 ‘species’ and allow for emergent behaviour⁹. Despite these previous advances in addressing the complex nature of plankton communities, neither modelling approach (which includes our model, see later), as of yet, resolves the full complexity of phytoplankton physiological responses to multiple co-varying stressors evident from the rapid recent advances in laboratory and field manipulation studies^{10, 11}.

A growing body of evidence from time-series observations¹² and manipulation experiments^{10, 11} reveals that biota such as phytoplankton will be significantly influenced by such concurrent and complex change, termed here oceanic multi-stressors. The effects of multi-stressors can be demarcated into independent and interactive (synergistic or antagonistic)^{11, 13}. The former are where individual stressors each alter phytoplankton physiology but do not interact, whereas in the latter case, the interplay between multi-stressors results in amplification or diminution of phytoplankton processes relative to the combined effects of the individual stressors alone. Laboratory and field studies have shown up to fourfold physiological amplification due to the interplay of multi-stressors, such as iron and temperature on polar diatoms¹⁴. Hence, there is a need for models to move beyond their reliance on simple and numerically rigid representations in contrast to the complexity and sensitivity suggested in laboratory multi-stressor studies. Models must further incorporate this widespread interactive facet of multi-stressors to investigate to what extent cumulative environmental stress may be exerted on oceanic biota.

Although coupled Earth system model experiments have been pivotal to better understanding the ramifications of climate change on the ocean, this rich source of information has been under-used so far as the basis for designing targeted process studies. Model projection results for different climate scenarios are often displayed as two-dimensional global maps of the change for each ocean property. Although qualitatively valuable for visualizing large-scale patterns (Fig. 1), these maps are usually considered in isolation from maps of trends in other oceanic properties⁴. Innovative visualization methods have been recently proposed for mapping simultaneously potential hotspots, such as the equatorial Pacific, where specific thresholds (relative to the global mean trends) are crossed for one or more environmental stressors (for example, sea surface temperature (SST), subsurface oxygen) in a future ocean⁷.

a, Temperature (°C). b, pH. c, Log₁₀ SiO₃ (mmol m⁻³). d, Log₁₀ NO₃ (mmol m⁻³).

The first objective of the study is to identify and summarize regional variations in the strengths and relationships of the climate change signals among a range of different physical and biogeochemical forcing factors (or stressors). The model analysis was conducted on climate change scenario output across 14 provinces (Supplementary Fig. 1) from the fully coupled Community Earth System Model^{15, 16} (CESM1(BEC), detailed in Methods and Supplementary Methods).

The climate-change-mediated alteration of individual upper-ocean properties can be assessed visually from conventional two-dimensional maps of the difference (temporal anomaly) between the future and present-day projections (Fig. 1 and Supplementary Fig. 2). However, rather than focus on changes in model fields individually, the aim here is to explore how multiple properties change concurrently with one another. To begin to address the question of ‘What does the pattern for climate change multi-stressors look like for the global ocean?’ we used the data presented in Fig. 1 to provide insights into which properties ‘change together’ globally (Supplementary Fig. 3). Table 1 shows statistical analyses for each model variable including the global mean temporal anomaly (future–present), where log normalization is used for some variables. As in previous studies^{4, 7, 9}, the future upper ocean is projected, on average, to be warmer with lower nutrient concentrations and reduced pH.

Full table

Changes to the properties within each region will exert both individual and multiple (independent versus interactive)^{11, 13} stresses that together will result in cumulative physiological^{10, 11} and/or biogeographical^{4, 18} effects on phytoplankton. Assessment of the effects of these regionally distinctive multi-stressor patterns (Table 2) on phytoplankton requires information on the composition of the resident phytoplankton in each province, along with data sets on the range of responses to environmental forcing by each major phytoplankton group from manipulation experiments¹⁰ and/or time-series/survey observations¹⁹ (Table 3 and Supplementary Table 4). Such an analysis enables the model data to be transformed into a preliminary appraisal of the effects of complex environmental change on the dominant phytoplankton groups within provinces (Supplementary Fig. 4). Our approach provides insights into the nature (independent versus interactive effects) and degree (that is, amplification versus diminution) of future changes to phytoplankton processes to be expected. In doing so, it reveals many of the challenges that lie ahead for both the experimental manipulation and modelling communities as they attempt to better address oceanic multi-stressors.

Full table

The CESM1(BEC) marine ecosystem module (BEC, Biogeochemistry/Ecosystem/Circulation) includes NPP and the explicit representation of three phytoplankton functional types (PFT)—diatoms, diazotrophs and small phytoplankton⁴². Model evaluation studies indicate that the CESM1(BEC) exhibits comparable skill against observations of present-day ocean physical and biogeochemical metrics and similar patterns of climate change to other CMIP5 Earth system models⁷. The discussion (in the main text) of the ramifications of the phytoplankton responses to multiple stressors (both individual and interactive effects) is focused only on a compilation of laboratory and field evidence, as opposed to the CESM1(BEC) model outputs for each PFT, as the PFTs as parameterized do not, as of yet, take into account the individual biological influence of each of the altered stressors (for example, the effects of increased CO₂ on diazotrophs²⁷), nor their fully interactive effects^{10, 11} (with some exceptions such as interlinked temperature, nutrient and light effects on phytoplankton growth³⁸).

As with all plankton functional type models, the CESM1(BEC) ecosystem model contains a number of assumptions regarding the biological responses of phytoplankton and zooplankton to environmental conditions (for example, temperature, nutrient and light response functions on photosynthesis and growth; elemental stoichiometry of biomass and detritus; rate functions for growth and mortality)^{38, 39}. The CESM(BEC) model is broadly similar in construction to other PFT models, although specific aspects of the climate change solutions will of course reflect the choices in functional form and parameters.

Two model time periods are used for comparison: the present day (1981–2000) and a future projection (2081–2100) following a high-emission climate change scenario (RCP8.5) with a rapid rise in atmospheric CO₂. Spatial maps of 15 ocean variables are computed for both present and future conditions by averaging over 20-year periods to reduce the effect of interannual variability; difference maps (ΔT, ΔpH and so on) are then calculated by subtracting the future minus the present. Statistical and factor analysis, summarized in Supplementary Fig. 3, are conducted on the original model grid (~1 degree horizontal resolution). Regional binned products are also produced for illustrative purposes by averaging the model output into 14 standard provinces that approximately match large-scale ocean physical and biogeochemical boundaries (for example, subpolar upwelling versus subtropical downwelling gyres; Supplementary Fig. 1).

Factor analysis was performed on the temporal anomaly fields (ΔT, ΔpH and so on) on the original model grid (~1 degree horizontal resolution; Fig. 3). Twelve variables, chosen as a subset of the ocean variable fields, were standardized by removing the spatial mean and dividing by the spatial standard deviation (Supplementary Fig. 3 and Methods). Factor loadings (contributions of different upper-ocean variables to each factor) and factor scores (spatial patterns for each factor) were computed using singular value decomposition of the standardized variable covariance matrix⁴³. On the basis of the communalities (a measure of variable representation, by factor), the 6 factors with the highest corresponding eigenvalues were kept for varimax rotation⁴³ and the resulting rotated factor loadings and scores are presented in Fig. 3. The spatial patterns for any particular standardized variable due to an individual factor can be recovered by multiplying the factor loading by the factor score map. The factor analysis helps visualize the multi-stressor patterns and characterize the main relationships among the stressor variables that contribute to these patterns. The major aspects of the regionally distinctive multi-stressor patterns can be reconstructed compactly at the model grid scale using the spatial patterns of the leading factors that explain the largest fraction of the global variance in the stressor variables.

1. Doney, S. C. The growing human footprint on coastal and open-ocean biogeochemistry. Science 328, 1512–1516 (2010).
   • CAS
   • ISI
   • PubMed
   • Article
2. Gruber, N. Warming up, turning sour, losing breath: Ocean biogeochemistry under global change. Phil. Trans. R. Soc. A 369, 1980–1996 (2011).
   • CAS
   • PubMed
   • Article
3. IPCC Climate Change 2013: The Physical Science Basis (eds Stocker, T. F. et al.) (Cambridge Univ. Press, 2013).
4. Sarmiento, J. L. et al. Response of ocean ecosystems to climate warming. Glob. Biogeochem. Cycles 18, GB3003 (2004).
   • CAS
   • Article
   URL: