全球变化知识资源中心

Climate change will unquestionably influence global ocean plankton because it directly impacts both the availability of growth-limiting resources and the ecological processes governing biomass distributions and annual cycles. Forecasting this change demands recognition of the vital, yet counterintuitive, attributes of the plankton world. The biomass of photosynthetic phytoplankton, for example, is not proportional to their division rate. Perhaps more surprising, physical processes (such as deep vertical mixing) can actually trigger an accumulation in phytoplankton while simultaneously decreasing their division rates. These behaviours emerge because changes in phytoplankton division rates are paralleled by proportional changes in grazing, viral attack and other loss rates. Here I discuss this trophic dance between predators and prey, how it dictates when phytoplankton biomass remains constant or achieves massive blooms, and how it can determine even the sign of change in ocean ecosystems under a warming climate.

Roughly one-half of the net primary production on Earth occurs in the ocean, yet the spatially integrated biomass of the dominant marine photoautotrophs, phytoplankton, is only around 1% of the plant biomass on land^{1, 2}. The major contribution of phytoplankton (order 50 Pg C yr⁻¹) to biospheric production in part reflects the far greater area of the ocean and its less extreme seasonal cycles in growth conditions, but the stark contrast in total production-to-biomass ratios between land and ocean indicates a fundamental difference in system functioning. On average, the turnover time for the entire global phytoplankton biomass is a mere 2 to 6 days (refs 1,3), and across the vast subtropical gyres daily production is nearly perfectly matched by consumption and other losses (for example, cell lysis, sinking) throughout the year (that is, turnover ~1 day). Accordingly, the plankton world can exhibit trophic feedbacks, carbon cycling rates and climate sensitivities that behave very differently than many more familiar terrestrial systems where production of plant biomass can be decoupled from consumption over seasonal (for example, leaves in temperate deciduous forests) to multi-century timescales (for example, California Redwood).

Here, I provide a perspective on the counterintuitive nature of phytoplankton biomass and its temporal variability, with a particular focus on ocean regions exhibiting seasonal blooms (that is, periods of high biomass concentration). Phytoplankton blooms are 'hotspots' for fisheries production and play a vital role in atmosphere–ocean carbon dioxide exchange and the export of organic carbon to the deep sea^{4, 5, 6, 7}. The 'blooming process' is directly dependent on the physical properties of the upper ocean (for example, temperature, mixed-layer depth, seasonal stratification of the surface layer) that will be strongly modified by climate warming in the coming century, particularly at higher latitudes. However, the common view of how ocean physics is linked to phytoplankton biomass has been challenged by recent satellite, in situ and modelling studies^{8, 9, 10, 11}. This issue is important because contrasting bloom hypotheses^{8, 11, 12, 13, 14} yield differences in even the sign of predicted future change in ocean biomass. One shortfall of the current debate has been its near-exclusive focus on bloom initiation, without sufficient context to biomass changes occurring before the initiation event or mechanisms connecting initiation to the bloom climax. Here, these issues are brought to the fore as critical constraints in identifying an emergent conceptual framework for understanding contemporary blooms and their susceptibility to climate change.

Before restricting this discussion to the blooming phenomenon, it is beneficial to 'step back' to the global domain and consider the basic properties of phytoplankton biomass, which herein refers to volumetric carbon concentrations (mg C m⁻³) unless specifically noted otherwise. Our ability to monitor phytoplankton properties at the global scale comes from satellite observations of ocean colour. These measurements detect spectral and intensity variations in light emerging from the surface ocean and allow assessment of basic ecosystem properties, including phytoplankton biomass and chlorophyll concentrations (mg m⁻³). Satellite data resolution is spatially (~1–5 km) and temporally (~1–8 day) coarser than field measurements, the detected signal does not register ecosystem properties below the actively mixed surface layer and little detail can currently be retrieved on the rich diversity of species comprising plankton communities. Nevertheless, understanding the observed temporal variability in these global satellite-derived properties can provide fundamental insights on how ocean ecosystems function and the processes guiding their temporal development.

The physical and chemical properties of the upper sunlit layer of the ocean are highly variable. In the central ocean gyres, this layer is clear, warm, strongly stratified, contains vanishingly low nutrient concentrations and receives high daily doses of sunlight throughout the year. By contrast, high-latitude regions in winter can have mixing depths of hundreds of metres, temperatures near 0 °C, little to no sunlight and very high nutrient concentrations. By mid-summer, these same regions may be converted to clear, warm and low nutrient waters with shallow surface-mixing depths and daily sunlight exposures comparable to low latitudes. Despite this tremendous diversity in upper-ocean conditions, surprisingly little variability is observed in the annual minima of phytoplankton biomass across the globe (Fig. 1a). Thus, the subarctic Atlantic annual minimum is comparable to equatorial regions, even though the subarctic population has experienced mixing to depths of hundreds of metres. This uniformity in biomass minima (Fig. 1a) is one indication that ecological processes, in concert with physical–chemical properties, are important determinants of phytoplankton biomass. Interestingly, although ocean ecosystem models can be tuned to reproduce these observed minima, we still do not fully understand the mechanisms defining the lower biomass threshold¹⁵. One possibility is that it represents a boundary below which further decreases in phytoplankton stocks become limited by contact frequencies between predators and prey.

a,b, Annual minima and maxima in phytoplankton biomass (mg C m⁻³), respectively. c, Number of biomass doublings required to increase from the minimum values in a to the maximum values in b. d, Total number of phytoplankton cell divisions between the times of minimum and maximum concentration (total divisions = daily division rate × days from minimum to maximum). Images are based on NASA MODerate-resolution Imaging Spectrometer (MODIS) 16-day average data for 2011. Phytoplankton biomass was calculated following ref. 66. Black lines in b correspond to annual mean sea surface temperatures of 15 °C for 2011, which roughly delineates the permanently stratified oceans from higher-latitude, strongly seasonal regions. Note that the high doubling values west of equatorial Africa in c (marked by an 'X') are due to anomalously low values in a and may reflect uncertainties in satellite ocean retrievals in this region of high atmospheric aerosol (dust) loads.

Although a phytoplankton bloom is incontrovertibly understood as the condition of elevated biomass, the process of 'blooming' (that is, increasing from a low to high concentration) is an issue of competing rates. Let me briefly elaborate this point. During the annual cycle, a blooming phase is characterized by a period when the division rate of phytoplankton (μ; d⁻¹) exceeds the rate of total losses (l; d⁻¹) from zooplankton grazing, viral lysis and sinking out of the mixed layer. Thus, μ > l during blooming and the biomass accumulation rate, r = μ − l, has a positive value. Thus, it is the sign of the rate term, r, that defines 'blooming', not the absolute value.

The most common explanation for temperate and high-latitude phytoplankton blooms is that they are initiated when light-driven increases in division rate surpass a critical threshold value²⁸. Because the depth of surface mixing is a primary determinant of daily light exposure for phytoplankton in these regions, this interpretation of blooms is referred to as the critical depth hypothesis (CDH). Its description can be found in most biological oceanography textbooks. The CDH's attribution of bloom initiation to a threshold value for μ arose during its original formalization from the simple assumption that loss rates can be treated as a constant²⁸. The appealing simplicity of the CDH is likely one reason for its long-standing popularity¹⁴, but as a scientific foundation for understanding phytoplankton blooms, it does not stand up to observations.

The CDH yields two very testable outcomes: (1) deep mixing and low incident sunlight cause phytoplankton biomass to decrease (that is, r is negative) before reaching the critical threshold and (2) the rate of increase in biomass following bloom initiation covaries linearly with the division rate (that is, r = μ − c, where c is a constant). Today, mixed-layer phytoplankton properties are continuously monitored through satellite ocean colour measurements and in situ autonomous sensors. Analysis of these data for the subarctic Atlantic, a classic bloom-forming ocean region, has revealed that phytoplankton concentrations do not show a springtime transition from decreasing to increasing biomass. Instead, they show that phytoplankton concentrations can begin rising as soon as the mixed layer stops deepening^{8, 9, 10}. In other words, these data do not support the notion of a critical mixing depth before which phytoplankton accumulation is prevented by light-limited cell division. Furthermore, the satellite data yield no correlation between phytoplankton division rate and the rate of change in mixed-layer phytoplankton abundance¹¹. These findings refute the CDH as a viable explanation for phytoplankton blooms and are supported by sustained in situ autonomous optical profiling float measurements and ecosystem modelling^{9, 10}.

Dismissal of the CDH has resulted in an active debate to establish a revised interpretation of blooms. One key shortcoming of the traditional hypothesis was its neglect of known variability in phytoplankton loss rates. A strong correlation between phytoplankton division and loss rates would imply that food web interactions play a major role in determining bloom emergence. However, the allure of interpreting increasing biomass as an expression of rapid phytoplankton division rates is hard to resist. Accordingly, one side of the current bloom debate upholds this fundamental belief and has provided an alternative interpretation of the satellite data^{12, 13, 29}. Specifically, it has been noted that the initial rise in phytoplankton concentration at the end of mixed-layer deepening also coincides with a switch between positive and negative heat flux from the surface ocean. At this transition, convective mixing of the upper ocean is dampened and turbulent transport of phytoplankton from the surface to depth can become sufficiently slow relative to the cell division rate that surface biomass can increase. In other words, bloom initiation may be triggered by cessation of deep convective mixing and the crossing of a critical threshold in turbulent mixing. This critical turbulence hypothesis (CTH) also yields a very testable prediction: that phytoplankton loss rates exceed division rates before the late-winter switch in net heat flux. Satellite and sustained in situ measurements have now shown that mixed-layer-integrated phytoplankton biomass in the subarctic Atlantic often increases while convection is still actively deepening the mixed layer^{8, 9, 10}. These findings are inconsistent with the CTH as a robust explanation for phytoplankton blooms.

In the subarctic Atlantic, the winter transition from mixed-layer deepening to shoaling occurs around February. At this time, chlorophyll concentrations are at their annual minimum and, seaward of the continental margins, show little spatial variability across the basin (Fig. 2a). What makes this uniformity remarkable is its stark contrast to patterns in physical properties, particularly the depth of winter mixing (Fig. 2b). During late autumn and winter, convection deepens the mixed layer and dilutes surface phytoplankton populations with essentially phytoplankton-free water from below. This process should result in lower chlorophyll concentrations where mixing is deepest, yet this correspondence is not observed (Fig. 2a,b). The reason for this discrepancy is that phytoplankton concentration is always influenced by the balance between division and loss rates, and this balance is also tied to physical processes such as mixing.

a, February surface-chlorophyll concentration. b, February mixed-layer depth (MLD). c, February water-column chlorophyll inventory integrated from the surface to the MLD. d, Maximum chlorophyll concentration achieved during the spring/summer bloom climax. MLD data are from http://www.science.oregonstate.edu/ocean.productivity/ and based on a data-assimilating physical ocean model, as described in ref. 10. White boxes in a show the location of data used in Figs 3 (Box 1) and 4 (Box 2). For reference, the location of the 1989 Joint Global Ocean Flux Study-North Atlantic Bloom Experiment (JGOFS-NABE)⁶⁷ is identified by the white star.

Arguably, the two most impactful developments in biological oceanography during the second half of the twentieth century were global satellite ocean colour measurements¹⁷ and the ability to conduct large-scale (on the order of 50 km²) purposeful manipulations of natural plankton assemblages^{36, 37}. These latter in situ experiments involved low-concentration additions of soluble iron to iron-limited surface phytoplankton populations. The initial outcome of such enrichments is a rapid rise in phytoplankton concentration, predominantly reflecting increases in species that were initially rare and presumably severely iron stressed^{36, 38}. The second important outcome is that this blooming phase induced by iron addition only lasts for ~5 to 10 days before the enhanced phytoplankton division rates are either matched or overcome by escalating loss rates (that is, phytoplankton concentration stabilizes or decreases, respectively)^{11, 36, 39, 40}. Recognizing that purposeful iron enrichments represent acute and major perturbations to plankton communities, this rapid response time of loss processes to changes in phytoplankton concentration is remarkable and in part reflects zooplankton with very rapid population growth rates²⁴. It also presents a challenging question for understanding natural blooms. If an increase in phytoplankton division rate can be overtaken by losses in 10 days or less, how can blooming in regions such as the subarctic Atlantic be sustained over periods of many months? The answer to this question is fundamental to predicting how climate-driven alterations of upper-ocean physical conditions, which vary slowly over the annual cycle relative to phytoplankton turnover rates, will impact plankton biomass stocks.

Physical processes can clearly interact with ecosystem feedbacks to initiate blooms (Fig. 2), but what happens when such forcings stop? When the disturbance ends, a critical consequence is that the existing difference between phytoplankton division and loss rates now causes phytoplankton concentrations to increase. This switch is documented in the satellite record for the subarctic Atlantic by the coincident rise in phytoplankton carbon and chlorophyll concentrations with cessation of mixed-layer deepening^{8, 10}. The new rise in concentration increases the encounter rate between phytoplankton and their predators (grazers, viruses), thereby enhancing predator populations and intensifying density-dependent losses^{10, 11}. Purposeful iron-enrichment experiments suggest that this rise in loss rates should quickly curtail any further increases in the phytoplankton. However, there is a vital difference between artificial and natural blooms. Enrichment experiments cause phytoplankton division rates to increase from a low value to a higher value in a single-step function, which is soon matched by consumption rates. In contrast, division rates during natural, high-latitude blooms continue to increase throughout spring stratification, albeit often slowly^{8, 9, 10, 11}. It is this continual rise in division rate that sustains the bloom to its climax^{11, 31}. In other words, phytoplankton concentrations do not increase because of high division rates, but rather because division rates are accelerating. Conversely, decelerating division rates favour decreasing biomass. And, it is likely that this basic principle applies to phytoplankton biomass dynamics across the global oceans (Fig. 1) and will continue to operate under a warming climate.

The distinction between the 'division-rate' and 'ecosystem-feedback' interpretations of phytoplankton biomass cycles is well illustrated by subarctic Atlantic satellite data. The division-rate view is illustrated in Fig. 3a, which compares a climatological annual cycle in phytoplankton division rate (μ, green symbols) with measured rates of change in chlorophyll concentration (Δchl, red symbols; note the scale difference between the y axes