Physiological and ecological drivers of early spring blooms of a coastal phytoplankter

See allHide authors and affiliations

Science  21 Oct 2016:
Vol. 354, Issue 6310, pp. 326-329
DOI: 10.1126/science.aaf8536

Drivers of phytoplankton blooms

Despite decades of study, there is little evidence to link increases in phytoplankton growth in response to springtime warming with the dynamics of phytoplankton blooms. This lack of understanding makes it difficult to make predictions about global biogeochemical cycling in response to climate change. Hunter-Cevera et al. analyzed over a decade of data collected hourly from the New England shelf between 2003 and 2016 (see the Perspective by Worden and Wilken). Blooms now occur 20 days earlier than at the start of observations, because earlier springtime warming stimulates cell division earlier each year. Nevertheless, despite the shift in timing, predatory organisms in the food chain are still ready to consume the superabundance, which brings the blooms to an abrupt end each year.

Science, this issue p. 326; see also p. 287


Climate affects the timing and magnitude of phytoplankton blooms that fuel marine food webs and influence global biogeochemical cycles. Changes in bloom timing have been detected in some cases, but the underlying mechanisms remain elusive, contributing to uncertainty in long-term predictions of climate change impacts. Here we describe a 13-year hourly time series from the New England shelf of data on the coastal phytoplankter Synechococcus, during which the timing of its spring bloom varied by 4 weeks. We show that multiyear trends are due to temperature-induced changes in cell division rate, with earlier blooms driven by warmer spring water temperatures. Synechococcus loss rates shift in tandem with division rates, suggesting a balance between growth and loss that has persisted despite phenological shifts and environmental change.

Marine phytoplankton account for one-half of global primary production. Of considerable interest and concern is how climate change may affect this production. Increased temperature, ocean acidification, and altered nutrient delivery all have the potential to affect phytoplankton dynamics, including the timing and magnitude of blooms, which can dominate seasonal productivity (1, 2). There is evidence of current and ongoing changes in plankton phenology (35), with potentially substantial ecological consequences for marine systems (6).

Recently, there has been uncertainty about the detection of trends in phytoplankton biomass and how possible trends relate to climate change (79). The uncertainty arises in part from difficulties in species-level detection of phytoplankton. Many studies use bulk measurements that reflect a composite of the phytoplankton community (10). These measurements (such as chlorophyll concentration) can mask taxon-specific changes and obscure the mechanisms that govern responses to climate change. Another challenge lies in the need to observe and measure phytoplankton at appropriate time scales to elucidate those mechanisms. Ecological interactions and physiological responses of phytoplankton are rapid (on the order of minutes to hours). To adequately capture population dynamics, we must sample at this frequency, but also for extended durations because identification of seasonal, yearly, or decadal trends requires time series of these lengths.

We address this lack of temporal and taxonomic resolution for the picophytoplankter Synechococcus by using observations of individual cells and their properties from an automated submersible flow cytometer, FlowCytobot (FCB) (11) , deployed at the Martha’s Vineyard Coastal Observatory (MVCO). FCB has been deployed at MVCO since 2003, with year-round observations beginning in 2007. The data consist of a 13-year time series of hourly measurements of Synechococcus concentration and cell properties.

At MVCO, Synechococcus concentration exhibits a strong seasonal cycle, with low concentrations in winter and early spring, followed by a two- to three-order-of-magnitude bloom event in late spring (Fig. 1A). The population fluctuates around a slowly declining trend during summer and early fall and then declines sharply in late fall. Although this classic pattern (12) is stable from year to year, we found that the timing of the spring bloom varied by up to 4 weeks within our time series, and in particular we noted a trend of earlier blooms from 2003 to 2012 (~20-day advance) and later blooms from 2013 to 2015. We quantified these shifts by determining the day of the year at which the concentration first exceeds threshold concentration levels (Fig. 2B and fig. S1). Concurrent observations of temperature (Fig. 1D) show that earlier blooms coincide with warmer spring conditions (Fig. 2A and fig. S2). For each degree increase of the mean temperature in April, the spring bloom advances 4 to 5 days. The water at MVCO has been warming (fig. S3) in a manner consistent with the multidecadal trend in this region (13). Large seasonal and interannual variations are superimposed on these warming trends.

Fig. 1 Daily time series at MVCO from 2003 to 2016.

(A) Averaged Synechococcus cell concentration (cells per milliliter). (B) Division rates estimated with a matrix population model [per day (d)]. (C) Loss rates calculated by subtracting net growth rate [obtained from changes in (A)] from division rate (B). (D) Water temperature (degrees Celsius). Climatology (mean annual pattern for the time series) is shown on the right for (E) Synechococcus cell concentration, (F) division rate, (G) loss rate, and (H) water temperature.

Fig. 2 Multiyear trends showing spring temperature changes and Synechococcus bloom shifts from 2003 to 2016.

The data are shown by day of the year (vertical axis), with values denoted by color. (A) Temperature. Markers indicate the day in each year when water temperature first exceeds 6° (triangles), 9° (circles), 12° (stars), or 15°C (squares). (B) Synechococcus cell concentration. Markers indicate the day in each year when cell concentration exceeds 8 × 103 (triangles), 1.6 × 104 (circles), 4.8 × 104 (stars), or 9.6 × 104 (squares) cells ml−1. (C) Integrated division rate (cumulative summed division rate starting at year day 85). Markers indicate the day in each year when the integrated division rate exceeds values of 4 (triangles), 12 (circles), 24 (stars) or 36 (squares). (D) Integrated loss rate (cumulative summed loss rate starting at year day 85). Markers indicate the day in each year when the integrated loss rate exceeds values of 4 (triangles), 12 (circles), 24 (stars), or 36 (squares). Solid lines in each panel represent best fits from a piecewise linear regression model between year-day threshold crossings (as denoted by the markers) and year. Parallel lines reflect insensitivity to the choice of threshold and a shift in only the timing, not in the shape, of the spring bloom trajectory.

Numerous studies have identified correlations between temperature and Synechococcus concentration across a range of ocean conditions (12, 1417). In particular, there is evidence that the spring bloom begins in northeast U.S. and Canadian waters when the temperature exceeds ~6°C (12, 18). In addition, studies with cultured isolates have shown that Synechococcus physiology is temperature-dependent (19, 20). However, in situ evidence of a direct link between cell division rates and temperature is limited (12, 19, 20).

To understand the mechanism underlying the phenology shifts at MVCO, it is necessary to separate contributions from cell division and cell loss. Traditional methods for doing so require labor-intensive measurements (21, 22) that are impractical to sustain over months and years. Instead, we used a method that exploits the diel variations in the cell size of Synechococcus (17) (fig. S4). Cells typically increase in volume during daylight hours as they photosynthesize and then decrease in volume in the evening through cell division. By fitting a matrix population model to the size-distribution time series, it is possible to accurately estimate daily population division rates for Synechococcus (17). The same method also works well for other phytoplankton species (23, 24). The model relies only on changes in cell size and not on changes in cell concentration, which are often decoupled from division rate (fig. S4) as a result of cell loss or advection. Isolating the contribution of cell division is critical, because the net population growth rate is typically much lower than the division rate as a result of the tight coupling between primary producers and consumers in planktonic ecosystems (25). In combination with sustained, high-resolution in situ observations, this model makes it possible to produce the multiyear records of daily division rate (Fig. 1B) that are required to understand phenological shifts.

Model-estimated division rates reflect a composite response of the entire Synechococcus population. Natural populations are often composed of more than one Synechococcus type, and multiple genotypes are known to coexist at MVCO (26, 27). Types can differ in physiological characteristics, such as temperature growth response or tolerance (20). Nonetheless, we find that Synechococcus division rates exhibit a pronounced seasonality, suggesting strong environmental drivers for the entire population. The division rate is as low as 0.1 day−1 in winter and as high as 1.4 day−1 in spring (Fig. 1B), increasing rapidly after water temperature exceeds ~5°C (Fig. 3). The increase slows or stops once water temperature reaches ~16°C in spring. The shape of the relationship between temperature and division rate strongly resembles that of the relationship for cultured Synechococcus (19, 20). During spring, a single clade-I genotype makes up the majority of the population (26). A cultured MVCO isolate belonging to this clade exhibits the same relationship between division rate and temperature as we report here for the natural population (17) (fig. S5). No other environmental factor (light or nutrients) shows such a strong relationship with in situ division rate in spring (fig. S6). Together, these results support the view that temperature is the main factor limiting Synechococcus division during late winter and spring (17).

Fig. 3 Relationship between division rates and temperature between 1 February and 15 June for all data points (2003–2016).

Colors indicate day of the year. The solid line is the relationship between the weekly climatology of division rate and the weekly climatology of temperature. The dashed line is the relationship between the weekly climatology of loss rate and the weekly climatology of temperature.

Our second main result is that the phenological shifts of the spring bloom are caused by temperature-induced shifts in the timing of the division rate increase (fig. S2). Specifically, the date at which population growth potential (computed as the integrated division rate for each day in spring; methods and fig. S7) reached threshold values advanced over the warming period 2003–2012 but retreated when spring temperatures cooled in 2013–2015 (Fig. 2C). The record-warm conditions in 2012 mark a transition between periods of advancing and retreating that are coincident across threshold crossings for spring temperature, Synechococcus concentration, and integrated division rate. In conjunction with the temperature dependence of the division rate in spring (Fig. 3), this result supports the conclusion that shifts in the timing of spring blooms reflect a direct physiological response to shifts in the onset of seasonal warming.

If division rates were the only factor affecting bloom phenology, we would expect not only a shift in timing, but also a change in the rate of increase in cell concentration. In fact, only the timing has shifted, whereas the rate of increase in cell concentration has not changed systematically (Fig. 2B). The stability of the bloom trajectory suggests that population losses have also shifted in tandem with growth (Fig. 2D and fig. S7). To illustrate this, we calculated bulk Synechococcus loss rates by subtracting the division rate from the net growth rate. The loss rates closely track division rates in magnitude over the entire time series (Fig. 1C), but during the spring bloom, the loss rate is on average ~ 0.15 day−1 lower than the division rate (Fig. 4A). It is this slight imbalance persisting for several months that leads to a steady blooming phase. The narrow margin by which Synechococcus can essentially “outgrow” their losses in spring makes this a critical time for cells to accumulate.

Fig. 4 Relationships among division rate, loss rate, net growth rate, and time of year for 2003–2016.

(A) Relationship between division rate and loss rate for data in spring (1 March to 30 June). The solid black line represents a one-to-one relationship. Colors indicate day of the year. (B) Relationship between division rate and net growth rate for spring data. The solid line is the best fit from a linear regression model, and the dashed line is the zero horizon (for reference). Colors are as in (A). (C) Weekly climatology of division rate (bold black line) overlaid with weekly climatology of net growth rate (thin gray line). Periods of positive (negative) net growth rate are indicated with green (gray) shading.

Our third main result, that loss rates closely track division rates, suggests that the dominant losses are biological (e.g., viral lysis and grazing) rather than physical in nature. It is unlikely that advection of spatial patches, for instance, would produce this tight correspondence year after year.

Beyond the connection to climate, our findings provide important insights into the ways that physiology and environmental factors interact to control phytoplankton blooms, a classic problem that has been the subject of recent controversy (28). Considering bulk properties of phytoplankton, Behrenfeld and Boss (29) have recently argued that, despite decades of study, no evidence has emerged linking bloom dynamics to physiological factors that regulate phytoplankton division rates. This led them to conclude that bloom dynamics depend principally on the impact of consumers, a long-recognized control of phytoplankton abundance (30). Our findings for Synechococcus agree with those of Behrenfeld and Boss in so far as division rates (and, by inference, loss rates) are roughly 10 times the accumulation (net growth) rates. Our results differ, however, in that we find a significant positive correlation between division and accumulation rates over the course of the spring bloom (Fig. 4, B and C). This correlation was not detected by Behrenfeld and Boss, and perhaps should not be expected to be evident in the satellite-based observations of chlorophyll concentration that they analyzed (29). Those observations aggregate the entire phytoplankton community over a relatively large region of the ocean and mask individual responses of different taxa.

Our observations, made at a much smaller spatial scale and with much finer taxonomic and temporal resolution than that of satellite data, reveal a connection between division rates and the bloom dynamics of Synechococcus. Consumers (including grazers, viruses, and parasites) certainly play a major role in shaping the bloom’s trajectory, but the bloom is triggered by an environmental factor, the seasonal temperature rise, which leads to increases in the Synechococcus division rate (Fig. 3). The bloom persists until the division rate plateaus (Fig. 4B), at which point losses overtake division and the bloom begins to decline.

We were able to diagnose the importance of temperature in regulating the dynamics of a ubiquitous marine primary producer, Synechococcus, by exploiting a 13-year time series comprising data on millions of individual cells and their traits. This allowed us to not only quantify the relationship between temperature and cell division in a natural population, but also to document how that relationship is the basis for a dramatic phenological shift affecting both Synechococcus and their consumers. It remains to be seen whether this ecological coupling will hold as warming trends continue in the decades to come.

Supplementary Materials

Materials and Methods

Figs. S1 to S8

References (3133)

References and Notes

Acknowledgments: We thank E. T. Crockford, E. Peacock, J. Fredericks, the MVCO Operations Team, the captain and mate of the R/V Tioga, and P. Henderson of the Woods Hole Oceanographic Institution (WHOI) Nutrient Analytical Facility for logistical and analytical support. We thank S. W. Chisholm and J. Waterbury for discussions and comments on the manuscript. This work was supported by U.S. NSF grants OCE-0119915, OCE-0530830, OCE-1031256, DEB-1145017, and DEB-1257545; NASA grants NNX11AF07G and NNX13AC98G; Gordon and Betty Moore Foundation grant GGA#934; the Investment in Science Fund, given primarily by WHOI Trustee and Corporation Members; and a National Defense Science and Engineering graduate fellowship from the U.S. Department of Defense. Data used in this study were deposited in the Dryad Digital Repository (
View Abstract

Navigate This Article