Sixty Years of Sverdrup: A Retrospective of Progress in the Study of Phytoplankton Blooms

. One of the most dramatic large-scale features in the ocean is the seasonal greening of the North Atlantic in spring and summer due to the accumulation of phytoplankton biomass in the surface layer. In 1953, Harald Ulrik Sverdrup hypothesized a now canonical mechanism for the development and timing of phytoplankton blooms in the North Atlantic. Over the next 60 years, Sverdrup’s Critical Depth Hypothesis spurred


Sixty Years of Sverdrup
Sixty years later, we (a group of graduate students) were challenged by our instructor (coauthor Sosik) to explore how this hypothesis influenced the rate of progress in understanding bloom dynamics. To fuel our debate, we traced bloom formation theories through time back to Sverdrup (1953). In this process, we realized that the path of these ideas through time was not as clear as we initially imagined, and defining the ensuing "progress" was one of the most challenging aspects of our task.
To start, we evaluated a naive definition of progress-one study makes a discovery, a later study builds upon it, this process iterates, and progress is made. We then looked for how and why new insights into phytoplankton blooms deviated from this "linear" type of progress. By examining the literature citing Sverdrup (1953) Sverdrup (1953)  -Henry Bryant Bigelow, 1926 The drivers and the timing of phytoplankton blooms in the ocean have puzzled and captivated scientists since the advent of the field of biological oceanography (Mills, 1989 Most notably, these cases demonstrate that the evolution of our understanding of phytoplankton blooms was paced by access not only to technology but also to concurrent insights from several disciplines. This exploration of the trajectories and successes in bloom studies highlights the need for expanding interdisciplinary collaborations to address the complexity of phytoplankton bloom dynamics.  Sverdrup's Critical Depth Hypothesis (a,b) and the other hypotheses (c,d,e) discussed in this paper. (a) Production is governed solely by light and thus declines exponentially with depth, while loss/respiration is constant with depth. The "critical depth" is the bottom of the layer within which the integrated production equals integrated respiration (dashed line). (b) When mixing is deeper than the critical depth, there is net loss, so a bloom is not expected (left ellipse); if the mixing depth is shallower than the critical depth, there is excess production and a bloom can occur (right ellipse). Modified from Sverdrup (1953) (c) The Critical Turbulence Hypothesis. Even if mixing is deeper than the critical depth, a bloom can form if the rate of mixing is slow enough that phytoplankton are retained in sunlit waters for suitably long periods (right ellipse). (d) Bloom initiation due to Stratification from Mixed Layer Eddies. The left-right gradient indicates light to heavy water from north to south (Northern Hemisphere). As the Coriolis effect induces eddies, lighter water is pushed above heavier water, creating shallow mixed regions where phytoplankton can grow. Adapted from Mahadavan et al. (2012) (e) The Dilution Recoupling Hypothesis. A deepening of the mixed layer (right ellipse) dilutes the predatory pressure on the phytoplankton, allowing a bloom to begin even if mixing is deeper than the critical depth. irradiance, the coefficient of light extinction, and mixed layer depth. By incorporating the concepts of Gran and Braarud (1935), Sverdrup proposed that in the North Atlantic, deep mixed layers in winter months keep the phytoplankton in an unfavorable light environment and therefore limit production. A "critical depth" is defined as the bottom of a layer in which the total production of organic matter by the phytoplankton community-from this depth to the surface-is equal to its destruction by respiration ( Figure 1a). If phytoplankton are mixed evenly to depths that exceed the critical depth, loss exceeds production and there is a net loss of biomass.

THE CRITICAL DEPTH HYPOTHESIS
Conversely, when the mixed layer depth is shallower than the critical depth, phytoplankton have the potential to bloom because the whole community experiences sufficient light levels to support net growth (Figure 1b). The Critical Depth Hypothesis is a simple, quantitative model that has provided a working and testable framework for the ensuing theoretical and empirical experiments over the years.

PROGRESS FROM SVERDRUP'S FR AMEWORK
The model of bloom formation described in Sverdrup (1953) laid the foundation for several studies, both immediately after its publication and during the ensuing decades, that showed the type of straightforward advancement we might naively expect and describe as "linear" progress. For example, Semina (1960) demonstrated that bloom formation in the Bering Sea near Kamchatka was better explained by stability, nutrients, and grazing than light limitation, and Menzel and Ryther (1961) invoked the critical depth model to explain how the Sargasso Sea can sustain high growth despite a deep mixed layer (i.e., high water clarity). On the theoretical front, models by Steele (1962) and Steele and Menzel (1962) Sverdrup (1953) citations over time. (b) Timeline of seminal achievements pertinent to the case studies highlighted in this paper and based on an extensive literature review (see Supplementary Bibliography). The achievements selected are all burgeoning theories or technologies that spurred research leading to a greater understanding of specific physical, chemical, and biological processes controlling phytoplankton bloom formation. Citations, in chronological order, are (for nutrients) Michaelis and Menten (1913), Redfield (1934), Browne (1942), Droop (1974), Huntsman andSunda (1980), Fitzwater et al. (1982), and Martin et al. (1990); (for upper ocean physics) Grant et al. (1962), Ozmidov (1965), Nasmyth (1973, Thorpe (1977), and Oakey (1982); and (for microbial and grazer community) Pomeroy (1974), Hobbie (1977), Landry and Hassett (1982) and Fenchel (1982), Chisholm et al.(1988), Davis et al. (1996), and Scholin et al. (1998). this study occurred nearly 50 years after Sverdrup (1953) and over 20 years after the advent of ocean color satellites might seem to suggest a lag after all the necessary pieces existed. While ocean color satellites had been available for decades, until SeaWiFS was deployed, the quality of those data was not sufficient to support the findings of Siegel et al. (2002). these and other studies showed that the characteristic spring bloom is a mid-to high-latitude phenomenon (e.g., Cushing, 1959;Yoder and McClain, 1993;Obata et al., 1996;Siegel et al., 2002), and that Sverdrup's critical depth could not explain all bloom formations.  (Heimdal, 1974;Schei, 1974;Townsend et al., 1992;Eilertsen, 1993;Backhaus et al., 1999;Dale et al., 1999;Körtzinger et al., 2008), led to insights into different bloom formation processes.
Historically, many researchers have looked to the role of nutrients in spring blooms. For example, Steele and Menzel (1962) conducted a theoretical study that used mixed layer depth and nitrogen uptake to describe a winter bloom.
Anderson (1964) (Hollibaugh et al., 1981;Eilertsen, 1993;Eilertsen et al., 1995;Hansen and Eilertsen, 1995)   Many scientists were interested in turbulence in the 1950s. Skellam (1951) and Kierstead and Slobodkin (1953) present investigations of the critical size of a patch in which phytoplankton could bloom in the face of turbulent diffusion away from that area. These studies paved the way for a growing interest in plankton patchiness (Ōkubo, 1980) and the interaction between turbulence and expected plankton distributions in the horizontal dimension (Platt, 1972;Ōkubo, 1980). During the late 1970s into the mid-1980s, attention was directed toward the relationship between exposure to variable light, due to vertical mixing within the water column, and phytoplankton physiology. For example, Marra (1978a,b) investigated how fluctuation in the light environment can lead to different rates of photosynthesis. This led to questions on how natural vertical turbulence (or lack thereof) in the mixed layer would affect production (Gallegos and Platt, 1982) or support the acclimation of phytoplankton to different light levels (Tilzer and Goldman, 1978).
Throughout this time, technological developments were important in advancing knowledge of turbulence in the mixed layer and its effect on phytoplankton. Through the 1960s, stratification was assessed at a few depths by either density or temperature differences (e.g., Aron, 1959;Nival, 1965;Walsh, 1971;Coste et al., 1972). These measurements, however, gave no information on how quickly mixing within this homogeneous layer occurred. Estimates of mixing were scarce and often calculated from atmospheric and water properties and not measured directly (e.g., Thomas, 1966). Before 1960, because turbulence could not be measured in places less energetic than tidal channels, little information existed for other parts of the ocean (Gregg, 1991). The development of new measuring devices and methods in the late 1960s and 1970s-hot film anemometers, airfoil probes, and finescale conductivity-temperature-depth (CTD) measurements-facilitated quantification of turbulence in less-energetic regions (Grant et al., 1968a,b;Osborn, 1974;Thorpe, 2005). It was not until the 1980s, however, that turbulent energy could be routinely measured in the upper ocean (Dillon and Caldwell, 1980;Oakey and Elliott, 1982). These results allowed Denman and Gargett (1983) to estimate the temporal and spatial scales for vertical displacements that phytoplankton could undergo in turbulent motion.
These advances in turbulence studies were not applied to the study of blooms, as the field's focus was still on photoacclimation. For example, the influential study by Lewis et al. (1986) simultaneously measured the rates of turbulent dissipation (with a free-fall microscale profiler) and the photo-  (Fogg, 1991;Stramska and Dickey, 1994;Huisman et al., 1999Huisman et al., , 2002Ebert et al., 2001;Ghosal and Mandre, 2003 and Gargett, 1983;Venrick et al., 1987;Owen, 1989)

DILUTION RECOUPLING HYPOTHESIS
In the 1940s, it was realized that blooms were the consequence of subtle imbalances between phytoplankton division rates and loss rates (Riley, 1946;Riley and Bumpus, 1946). These ideas formed the foundation for quantitative modeling of planktonic ecosystems. In particular, the Critical Depth Hypothesis is a simplification of this framework, as the loss term in Sverdrup's model combines the effects of predation (i.e., grazing), respiration, and-as extrapolated by others-vertical export of sinking particles (e.g., Siegel et al., 2002;Behrenfeld, 2010) into a single rate assumed to be constant at all depths and times. Despite In the 1970s, the alternative technique of direct microscopic counting with an epifluorescence microscope (Francisco et al., 1973;Hobbie et al., 1977) led to reassessment and the conclusion that earlier methods grossly underestimated the large concentration of bacteria in the sea. Pomeroy, Azam, and colleagues (Pomeroy, 1974;Azam et al., 1983) challenged the canonical view of the  Chisholm et al. (1988Chisholm et al. ( , 1992 to discover a novel picoplankter (Prochlorococcus) that is now considered the most abundant autotroph in the world. The advent of dilution techniques allowed Landry and Hassett (1982) and Fenchel (1982) to determine that grazing by protists was responsible for holding bacteria and picoautotroph populations at relatively constant values.
These rapidly growing micrograzers respond quickly to increases in abundance of their phytoplankton prey but never "overgraze" because of feeding threshold effects that make it energetically unprofitable for the grazers to feed when the prey density drops below a given value (Strom et al., 2001). These grazing thresholds, multiple trophic levels, and patchiness are used to explain the lack of blooms in high nutrient, low chlorophyll areas (Strom et al., 2000).
Other work on grazing rates (Landry and Hassett, 1982;Landry et al., 1995Landry et al., , 1997 showed that grazer control of pico-  (1946), Nielsen (1958), andCushing (1959) , 1978-1986), Banse (1992Banse ( , 2002 showed that the annual North Atlantic phytoplankton bloom results from a mismatch between growth and loss processes, where growth is temporarily higher than loss. Additionally, the bloom terminates with either exhaustion of surface nutrients or overgrazing by heterotrophs, such that the rates of growth and loss are once again in balance. To investigate whether this initial decoupling was a result of increased phytoplankton growth rates or decreased losses, Behrenfeld (2010) used a satellite record of phytoplankton biomass in the North Atlantic and merged concepts originally formulated by Cushing (1959), Evans and Parslow (1985), Banse (1992Banse ( , 2002, and Marra and Barber (2005) into the Dilution Recoupling Hypothesis ( Figure 1e). The synthesis of these fundamental ideas, combined with new technology and perspective, enabled the interplay among phytoplankton growth, grazing, and seasonal physical processes to be evaluated in greater detail. Behrenfeld (2010) argues that bloom initiation occurs in the winter when the mixed layer is deepest, contrary to the Critical Depth Hypothesis.
Deep mixing replenishes the surface ocean with essential nutrients for bloom formation and dilutes phytoplankton cell density, thus reducing the encounter rate between predator and prey during winter (Cushing, 1959;Strom et al., 2001).
As the mixed layer depth shoals and increasingly concentrates phytoplankton and grazer populations, the lagging grazer population eventually "recouples" with the phytoplankton population and can limit bloom extent and duration (Cushing, 1959 Miki and Jacquet, 2008;Breitbart, 2012). Today, the development of novel in situ biological sensors provides opportunities to explore the relationships between phytoplankton community growth and loss rates and physical processes at previously unprecedented resolution levels. For example, the Imaging FlowCytobot (Olson and Sosik, 2007) and the Environmental Sample Processor (Scholin et al., 1998)