Space-for-time substitution in predicting the state of picoplankton and nanoplankton in a changing Arctic Ocean



[1] The Arctic Ocean is changing rapidly but there are no long-term time series observations on the state of the phytoplankton community that could allow a link to be made from physical/chemical pressures to the impact on marine ecosystems. Here, we test the idea that space-for-time (SFT) substitution might predict temporal change in the Canada Basin premised on differences in the present state of phytoplankton in other geographic zones, specifically the ratio in the abundance of picophytoplankton to nanophytoplankton (Pico:Nano). We compared the change in Pico:Nano observed in the Canada Basin from 2004 to 2012 to the different average states of this ratio in 26 other ocean ecological regions. Our results show that as upper ocean nitrate concentration changed in the Canada Basin from year to year, the concomitant change in Pico:Nano was statistically commensurate with the difference that this ratio exhibits between Longhurst ecological provinces in relation to nitrate concentration. Lower average concentration of nitrate in the upper water column is associated with a higher value of Pico:Nano, a result consistent with resource control of phytoplankton size structure in the ocean. We suggest that SFT substitution allows an explanation of temporal progression from spatial pattern as a test of mechanism, but such statistical prediction is not necessarily a projection of future states.

1. Introduction

[2] The driving force of climate change exerts strong pressures on the physical environment, changing the state of marine ecosystems which impact biogeochemistry, biodiversity, as well as socioeconomics, and eliciting responses of mitigation and adaptation [Carmack et al., 2012]. Although there is incontrovertible evidence of strong physical pressures in the Arctic Ocean, there is only a weak description of the time-dependent state of its ecosystems because many observational time series in this ocean are short in duration and conflated by seasonality [Carmack et al., 2010]. As an alternative to a long biotic record which is necessary for robust temporal trend analysis, it is tempting instead to invoke space-for-time (SFT) substitution [Pickett, 1988] for predicting a possible future state of arctic phytoplankton. In other words, assuming a similarity between patterns of spatial zones and temporal sequences, one might arguably infer the future state of the Arctic from the present state of other ocean zones. We shall see that this perhaps may not be the promise of SFT even if we find the similarity.

[3] The traditional application of SFT is in chronosequence analysis of ecological succession in which it is assumed that there is homologous similarity between sequences of spatial zones and temporal sequences [Pickett, 1988]. Increasingly, SFT is being adopted for use beyond traditional chronosequence application, in particular to predictions of climate change effects on biodiversity in which sequences in space and time are tested for analogous similarity [Williams and Jackson, 2007; Blois et al., 2013]. From these studies, it seems that SFT is most effective when focused on structural and compositional features of ecological communities that integrate environmental variance, and not on functional dynamics that are more sensitive to short-term environmental fluctuations. The coarse scale averaging effect of SFT for elucidating regional trends therefore lends it well to a macroecological approach of projections based on spatial zones [Kerr et al., 2007; Fisher et al., 2010]. The spatial zones in the pelagic ocean that are of macroecological interest are those partitioned by physical oceanographic processes controlling the annual cycles of phytoplankton production and consumption [Longhurst, 2007].

[4] In this paper, we examine a short (nine year) record of phytoplankton community structure in the Arctic Ocean in which the passage of time is a surrogate for the operation of functional variables that are the currency common to both space and time in a mechanistic model. A key consideration will be whether there is a similarity between temporal change in a beta ocean which is stratified mainly by salinity and spatial patterns in alpha oceans which are stratified mainly by temperature. Oceans with haline stratification allow only very shallow convection near the poles compared to oceans with thermal stratification, and so the low efficiency of nutrient replenishment to the photic zone is a strong constraint on annual new production in the Arctic Ocean [Carmack, 2007].

2. Methods

2.1. Canada Basin

[5] Hydrographic observations from conductivity, temperature, depth (CTD) profiling system and phytoplankton sampling from Niskin bottles were carried out in the Arctic Ocean on the CCGS Louis S. St-Laurent in the Joint Ocean Ice Study (JOIS), which is an ongoing multinational collaboration involving the Joint Western Arctic Climate Study (JWACS), the Beaufort Gyre Exploration Project (BGEP), and also included the 2007 Canadian International Polar Year (IPY) program Canada's Three Oceans (C3O). In this study, we focused our analysis on 13 deep-water stations in the Canada Basin that were sampled consistently every year from 2004 to 2012 (Figure 1). Over the 9 year study, sampling was conducted mainly from late July to late August, except in 2009 and 2010 when the program was carried out from mid-September to mid-October. The collection and analysis of physical (temperature, salinity, and density), chemical (nitrate), and biological (picophytoplankton and nanophytoplankton) data used in this study have been described elsewhere [McLaughlin et al., 2012]. The vertical distribution of nitrate in the upper 150 m layer was empirically fit to a Weibull distribution (SigmaPlot, SPSS Inc.) and the nitracline was designated as the depth at which nitrate first exceeded 0.5 mmol m−3.

Figure 1.

Map of the Arctic Ocean showing hydrographic station locations (circles) and 13 of those stations in the Canada Basin (stars, CB station numbers 2, 3, 4, 7, 8, 9, 15, 16, 17, 18, 21, 27, 29) from which the data were used to construct the annual time series 2004–2012.

[6] There is a complex layering of water masses in the Canada Basin: the deepest feature of Pacific origin is found at 150 m (water modified in the Chukchi Sea in winter) [Coachman and Barnes, 1961], below which is Atlantic origin water [Coachman and Barnes, 1963]. We combined all measurements at each station above 150 m to give an upper ocean average, and then combined the upper ocean average from each of the 13 focus stations to give an overall regional average for the Canada Basin. Some aspects of the first 5 years of this time series have been discussed elsewhere [Li et al., 2009].

2.2. Longhurst Ecological Provinces

[7] The microbe database at the Bedford Institute of Oceanography consists of flow cytometric measurements of the abundance (cells ml−1) of picophytoplankton, nanophytoplankton, and bacterioplankton from many parts of the global ocean. The largest portion of the data arise from repeated station occupations of the Atlantic Zone Monitoring Programme (AZMP) on the Nova Scotian Shelf since 1997, and the Atlantic Zone Off-shelf Monitoring Programme (AZOMP) in the Labrador Sea since 1994 [Li et al., 2006]. Geographic coverage of the data is extended greatly by a cruise of the Sedna IV in March and April 2003 from the west coast of Canada (Victoria) to the east coast (Dartmouth) via the Panama Canal; by a cruise of the R/V Nathaniel Palmer from May to July, 2004 in the South Atlantic from the South Sandwich Islands to Capetown, South Africa [Doolittle et al., 2008]; and by a 7 month global circumnavigation of the southern hemisphere by the R/V Mirai of JAMSTEC during the Blue Earth Global Expedition [Bouman et al., 2006]. Selecting data from the upper 25 m without regard to the date of sampling, we binned the measurements into 26 ecological provinces delineated by Longhurst [2007] (Figure 2).

Figure 2.

Map of hydrographic station locations extracted from the Bedford Institute of Oceanography microbial database showing the assignment of stations to 26 Longhurst ecological provinces that are variously colored. White areas of the ocean indicate provinces that are not represented in the database.

[8] The distribution of data across provinces is highly uneven, ranging from 3606 observations in the Northwest Atlantic Shelves Province (NWCS) to a single observation in the Australia-Indonesia Coastal Province (AUSW) (Table 1). The Boreal Polar Province (BPLR) extends across most of the entire northern polar region, and thus includes not only the Arctic Ocean, but also the continental shelf of the Labrador Sea. For the purpose of this study, we separate these two subregions of the BPLR to maintain a focus on the Canada Basin as a beta ocean. Some aspects of the global pattern of microbial plankton binned into Longhurst provinces have been discussed elsewhere [Li, 2009a], extending an earlier analysis limited to the North Atlantic [Li and Harrison, 2001].

Table 1. Average Microbial Abundance (cells ml−1) in the Upper 25 m Depth Layer of 26 Longhurst Ecological Provinces, the Ratio of Pico:Nano, and the Number of Samples Included in Each Provincea
  1. a

    The Canada Basin is shown separately from the rest of BPLR to which it belongs.

Canada Basin (BPLR)74.44−144.442885270622662272432413730

2.3. World Ocean Atlas

[9] To examine the global distribution of physical and chemical properties, we extracted the objectively analyzed one-degree climatological fields for temperature, salinity, and nitrate composited for an annual period at standard depth levels from the World Ocean Atlas, WOA ( We calculated seawater density [Fofonoff and Millard, 1983] and constructed a simple index for stratification (kg m−4) which was the arithmetic difference of density at 150 m and 0 m, divided by 150 m.

[10] The spatial dimension in our SFT analysis is meridional, therefore we collapsed the longitudinal dimension of each variable into 5° latitude bins to yield an annual average for zonal bands. This climatology provides the global average upon which our nonuniform sampling of the 26 Longhurst provinces may be placed in context. We did this in two steps. First, the geographical centroid of actual observation locations in each province (Table 1) was used to query WOA for values of the variables at those centroids. In a small number of cases in which the centroid was located on land because of ocean stations distributed across a geographical mosaic of land and water, the centroid was reassigned to the closest location of water. The returned values of such a WOA query indicate whether our particular collection of sampling locations in a Longhurst province is representative of a zonal band. Second, the values of actual physical and chemical measurements at our particular sampling locations are averaged and assigned to the geographical centroid to indicate whether our observations (at nonuniform times of the year) are representative of the annual climatological value at the centroid location obtained in step 1.

3. Results

3.1. Canada Basin

3.1.1. Average State

[11] The average state of the Canada Basin is described by the grand mean of 13 focus stations over the past 9 years, with observations binned at 10 m depth intervals to give characteristic vertical profiles of properties from the sea surface to 150 m (Figure 3). The upper 50 m is cold, fresh, strongly stratified, well-oxygenated, nitrate-deplete, poor in phosphate and silicate, and microbially abundant. The concentration of chlorophyll a at the surface is very low (0.05 mg m−3) and attains a subsurface maximum value of 0.22 mg m−3 at the top of the nutricline in the 50 to 60 m depth bin, within which the nitrate concentration is 1.6 mmol m−3. Nanophytoplankton also attain a subsurface maximum in this depth interval, but the picophytoplankton maximum is slightly shallower, by one depth bin.

Figure 3.

Average state of the Canada Basin in summer showing 10 m depth-binned profiles of (a) temperature, (b) salinity, (c) sigma-theta, (d) dissolved oxygen, (e) nitrate, (f) phosphate, (g) silicate, (h) heterotrophic bacteria, (i) chlorophyll a, (j) picophytoplankton, (k) nanophytoplankton, and (l) Pico:Nano ratio. Error bars indicate standard deviation across stations and years.

[12] The abundance ratio of picophytoplankton to nanophytoplankton (Pico:Nano) is a scale-free index of community structure. Its value is greater than 10 in the upper 50 m, and less than 10 at greater depths. At 150 m, there are no detectable phytoplankton in the aphotic nutrient-rich reservoir.

3.1.2. Changing State

[13] Our previous report [Li et al., 2009] on conditions from 2004 to 2008 indicated that in the upper 150 m of the water column, temperature increased, salinity decreased, stratification increased, nitrate decreased, picophytoplankton increased, and nanophytoplankton decreased. In the following 4 years from 2009 to 2012, our measurements do not indicate a continuation of the earlier trajectories, even though the extent of sea ice in the Arctic continues to decrease (Figure 4a) and the nitracline continues to deepen (Figure 4f). It must be noted that what is nominally an August time series is now less robust because of observations made later in the annual cycle in both 2009 and 2010. As an indication of the 4 year change since our last report, we find that the 2012 values for many variables are approximately the same as the 2008 values, except picophytoplankton which is 40% lower (Figure 4g). Pico:Nano started at a value of 4 in 2004, and ended at a value of 8 in 2012, with a high value of 18 in 2011 (Figure 4i).

Figure 4.

The change in state of the Canada Basin, 2004–2012. (a) Sea ice extent in August, data downloaded from the National Snow and Ice Data Center (; (b) temperature; (c) salinity; (d) stratification index; (e) nitrate concentration; (f) depth of the nitracline; (g) picophytoplankton abundance; (h) nanophytoplankton abundance; (i) ratio of picophytoplankton to nanophytoplankton. Except for sea ice extent and nitracline depth, all variables are upper ocean averages (z ≤ 150 m), with error bars indicating among-station standard deviation.

3.2. Longhurst Ecological Provinces

[14] The latitudinal patterns of temperature, salinity, stratification, and nitrate based on our nonuniform sampling of Longhurst provinces are generally fair representations of the annual climatologies in those provinces, and also fair representations of the global zonal bands within which those provinces are found (Figure 5). The abrupt transition from weak to strong stratification at the northern polar boundary (Figure 5c) is a mark of the alpha/beta ocean distinction [Carmack, 2007], also discussed by Longhurst [1995]. Notably, the BPLR-Canada Basin subprovince shows significant departure from the 74.5°N zonal band as a whole, with the subprovince characterized by even lower salinity, stronger stratification, and lower nitrate than the northern beta ocean as a whole.

Figure 5.

Latitudinal variation of (a) temperature, (b) salinity, (c) stratification index, and (d) nitrates. Lines are annual values of 5° latitude zonal bands averaged from the 0, 10, and 20 m World Ocean Atlas standard depths. Open circles are WOA values at the geographical centroid of the collection of sampling locations in our 26 Longhurst provinces. Filled circles are our in situ measurements (z ≤ 25 m) averaged from the sampling collections.

[15] With some confidence that our sampling of the Longhurst provinces does not grossly distort the global latitudinal pattern for physical and chemical variables, we then constructed the latitudinal patterns for picophytoplankton and nanophytoplankton (Figure 6). Picophytoplankton comprise three components. First, the picoeukaryotic algae which have a bimodal distribution with peaks at about 50°N and 30°S (Figure 6a). Second, the cyanobacterium Synechococcus which exhibits a high abundance between 50°N and 30°S, but decreases sharply poleward (Figure 6b). Third, the cyanobacterium Prochlorococcus which is undetectable poleward of 45° latitude in both hemispheres (Figure 6c). The sum of these components results in a broad unimodal distribution for picophytoplankton as a group (Figure 6d). The distribution of nanophytoplankton (Figure 6e) resembles that of picoeukaryotes, being bimodal with peaks of abundance in the subpolar oceans.

Figure 6.

Latitudinal variation in the abundances of (a) picoeukaryotic algae, (b) Synechococcus, (c) Prochlorococcus, (d) picophytoplankton, (e) nanophytoplankton, and (f) the ratio of picophytoplankton to nanophytoplankton. Only samples collected in the upper 25 m are included. Units are base 10 logarithms of cells per milliliter or of the scale-free ratio. Lines are fourth-order polynomial curves intended as visual aids.

[16] In a gradient from the equator to the subpolar oceans, picophytoplankton generally decrease while nanophytoplankton generally increase. Thus, except in the polar oceans, these phytoplankton size classes are distributed as approximate mirror opposites. As such, the ratio Pico:Nano is meridionally unimodal and decreases from low to high latitudes (Figure 6f). The three-order magnitude variation of Pico:Nano is primarily due to the two-order magnitude variation in Pico (Figure 6d) and secondarily due to the one-order magnitude variation in Nano (Figure 6e).

3.3. Functional Associations

[17] Pairwise correlations of variables linking physical/chemical pressures to biological state are examined in a common presentation without reference to the dimensions of time or space. At the global scale of the Longhurst provinces, surface Pico:Nano is well predicted by surface temperature (r2 = 0.82, p < 0.001, Figure 7a), slightly less well by surface salinity (r2 = 0.33, p < 0.01, Figure 7b), surface density (r2 = 0.42, p < 0.01, Figure 7c), and surface nitrate (r2 = 0.55, p < 0.01, Figure 7e), but much less well by the stratification index (r2 = 0.14, p < 0.1, Figure 7d). Temperature, salinity, density, stratification, and surface nitrate are each useful in predicting Pico:Nano within the Canada Basin, but all the functional associations at regional scale differ from those at global scale, as evident in nonoverlapping data clusters in bivariate domains (Figure 7).

Figure 7.

Pairwise associations between Pico:Nano of the surface ocean (z ≤ 25 m) versus (a) surface temperature (z ≤ 25 m), (b) surface salinity (z ≤ 25 m), (c) surface density (z ≤ 25 m), (d) stratification index (z ≤ 150 m), (e) surface nitrate (z ≤ 25 m), and (f) column-average (z ≤ 150 m) nitrate concentration. Data are from the geographical survey of Longhurst provinces and World Ocean Atlas (open circles) and from the multiyear time series in the Canada Basin (filled circles).

[18] The association between Pico:Nano in the surface layer (z ≤ 25 m) and nitrate in the upper ocean (z ≤ 150 m) is the only one examined that is statistically indistinguishable at the global scale of the Longhurst provinces and at the regional scale of the Canada Basin (Figure 7f). Analysis of covariance of these two relationships (df = 1) indicates that the slopes are not statistically different (F = 0.006, p = 0.94). However, the adjusted means are significantly different (F = 16.9, p < 0.001). The relationship combining both global (spatial) and regional (temporal) scale measurements is: log[Pico:Nano] = 2.11 − 0.83 log[nitrate]; n = 35; r2 = 0.35; p < 0.001. In other words, although there is a bias shown in the adjusted means, the state of Pico:Nano in the Canada Basin can be statistically predicted, with a 31% coefficient of determination, by the present state of this community indicator in other regions by reference to nitrate concentrations.

4. Discussion

4.1. Space and Time

[19] The interdependence of space and time is an enduring theme in physical and biological oceanography [Vance and Doel, 2010]. The two domains generally share commensurate scales of variability such that long time scale phenomena are manifest over large spatial scales and short-lived events occur over small areas. This means that climate-related signals must, by definition, be sought over appropriately large spaces of the ocean and over appropriately long periods of time. It is also because of this interdependence of temporal and spatial variability that SFT might be seen as a useful approach, not as a substitute for long-term monitoring of the oceans, but as a test of our understanding of ecological mechanisms common to both spatial differences and temporal change.

[20] We chose Longhurst provinces as the spatial unit for a study of phytoplankton ecology at large scale because they provide a rational organization for macroscopic pattern based on oceanography [Longhurst, 2007]. This set of 51 nonoverlapping regions is a partition of the global ocean based on regional differences in the seasonal evolution of phytoplankton growth and ecologically significant physical oceanographic processes. Thus, provinces do not represent a chronosequence of different stages in ecological succession. Instead, each province is a characteristic unit of large ocean space within which regional oceanographic processes constrain variable behavior in local areas.

[21] The global distributions of picophytoplankton and nanophytoplankton have been previously described [Li, 2009b; Buitenhuis et al., 2012], but mostly using a continuum approach in which ecological continuity from place to place is assured through the unchanging operation of a universal mechanistic model. In essence, the projection of biotic distribution into future climate conditions using present-day realizations of ecological niches [Flombaum et al., 2013] is an untested assumption of the continuum approach. As an alternative to continuity, the case for an ecological geography using a partition approach is compelling but so far, only little explored for these small phytoplankton [Li, 2009a]. Here in the present paper, we show how this partition approach also appears to be continuous when the parameters of a suitable model remain the same everywhere.

4.2. Macroecology of Pico:Nano

[22] Picoplankton and nanoplankton are groupings of different plankton cells by size [Sieburth et al., 1978], which is a meaningful way to explore macroscopic patterns in the sea because of the pivotal role that body mass occupies in macroecological theory [Brown et al., 2004]. The abundance ratio of these adjacent plankton size classes is potentially informative because it is a simple scale-free indicator of the community that integrates a nest of molecular, cellular, and population processes. The ratio is a macroscopic indicator because pattern regularity is observed at a hierarchical scale higher than that of the interacting units [Li and López-Urrutia, 2013]. As a primary constraint on the ratio, we first note that cell abundance (N) is inversely proportional to cell mass (M) according to allometric law, and second that the M ratio of Pico:Nano is on order of 1/1000 (i.e. picogram to nanogram), therefore the N ratio of Pico:Nano is on order of 1000. Actual observed values of Pico:Nano are usually greater than 1 and less than 1000 (Figure 6f). A secondary feature of this ratio is that its value is higher near the top of the upper water column than the bottom (Figure 3l), suggesting that the advantage enjoyed by picophytoplankton over nanophytoplankton diminishes with depth as nutrient availability increases and/or photon availability decreases.

[23] The apparent simplicity of this system indicator conceals many levels of complexity. Not only do each of the component size classes contain many genotypes, phenotypes, and ecotypes, but the balance of the size classes is also affected by trophic interactions with resource competitors, grazers, predators, parasites, viruses, symbionts, allelopaths, and others. A reductionistic dissection of the processes giving rise to observed values of Pico:Nano seems intractable, even if all the rules that govern the interactions were completely known. Pragmatically, we take the approach of predictive emergence in which system state is predictable by virtue of the supervenience of system properties on constituent properties [Pigliucci, 2013]. Thus, the appearance of regularity in Pico:Nano emerges from constituent interactions that are known (or knowable) in principle, but prediction of this pattern is made at the macroscopic level.

4.3. SFT Substitution for Pico:Nano

[24] The average state of the Canada Basin in summer (Figure 3) is similar to that of the Canadian Arctic Archipelago in general, and that of the adjoining southeast Beaufort Sea in particular [Martin et al., 2010; Tremblay et al., 2009], except that it is more oligotrophic in character, as Carmack and McLaughlin [2011] have portrayed in a section plot of chlorophyll a across the entire region. Indeed, on this basis, the Canada Basin is as hyper-oligotrophic as the South Pacific gyre [Morel et al., 2007], which has long since been considered the most oligotrophic of ocean waters.

[25] However, the average state conceals a systematic change over the period of observation [McLaughlin et al., 2011]. The water has been freshening due to increased sea ice meltwater and river runoff extending into the Basin [Yamamoto-Kawai et al., 2009]; the near-surface temperature maximum has been increasing due to sea-ice retreat and decreased albedo [Jackson et al., 2010]; both the nitracline and the chlorophyll maximum have been deepening due to increased Ekman pumping and downwelling, with the result that nitrate concentration at the depth of the chlorophyll maximum has been decreasing [McLaughlin and Carmack, 2010]. These are the changing pressures that have been imputed as the cause of the changing state of the community of small phytoplankton, indicated by Pico:Nano. Since these pressures are interrelated, what is the proximate factor that might predict Pico:Nano? The answer provides the supervenient mechanism for SFT substitution.

[26] In the upper ocean, here defined as the water column above 150 m, the nitrate reservoir at the bottom of this layer supplies the demand by the phytoplankton at the top of this layer. The average concentration in the layer is therefore a net outcome of supply and demand. This simple average is easily calculated from our measurements in the Canada Basin, but of equal importance, it is also available for global regions from the WOA climatology. Nitrate in the upper ocean depends on its vertical flux, which in turn depends on stability, density, salinity, and temperature. All of these ultimate factors, to greater or lesser degrees, are useful predictors of Pico:Nano, but it is only the putative proximate factor (average upper ocean nitrate concentration) that makes the same statistical prediction over space and time (Figure 7f). It should be noted that neither the residue of nitrate left in the surface layer (Figure 7e), nor the depth of the nitracline (Figure 4f), nor the reservoir of nitrate in the deep layer (not shown) are good proximate predictors. Physical mixing connects supply to demand but the depth of the mixing layer differs markedly by region; thus deep reservoirs of the same nitrate concentration do not necessarily signify the same supply regimes. Secondarily, it is possible that some portion of the nitrate in the Arctic is supplied not by physical flux, but by microbiological nitrification in the upper euphotic zone [Tremblay et al., 2008].

[27] Our result indicates a similarity between temporal change and spatial difference in the community balance between the two smallest size classes of photoautotrophs that is statistically explained in part by the supply and demand of nitrate. This is a macroscopic relationship that supervenes on the microscopic interactions between nutrient molecules and cellular processes.

[28] In principle, explanation (which we have just shown here in a statistical form) can only be turned into prediction (of future states) if there is perfect symmetry between the inferences that can be made between causes and effects [Hull, 1974]. Almost always, process laws that permit inference of past and future states given a present state pertain only to closed systems, such as astronomical laws of planetary motion. In contrast, open biological systems are governed by casual laws which are temporally asymmetric because effects always follow from an open web of casual connections. In an open plankton ecosystem, causes are usually neither severally necessary nor jointly sufficient for their effects. In other words, we cannot say that each of the causes (e.g., temperature, salinity, density, stratification, nitrate) taken separately is necessary for the effect on Pico:Nano; and neither can we say that all of them taken together is sufficient for the effect. Therefore, from the cause we cannot deductively infer the effect, and from the effect we cannot deductively infer the cause. Yet cause and effect are not unrelated. The relationship between upper ocean nitrate concentration and Pico:Nano embodies statistical explanatory power but not strength of inference.

[29] The statistical explanatory power of the Pico:Nano versus nitrate relationship lies in understanding that the relative success of different phytoplankton size classes depends primarily on resource availability, and not on temperature [Marañón et al., 2012]. Small cells have a large surface area to volume ratio, which allows them to better cope with nutrient diffusion limitation and gravitational settling; and small cells have photosynthetic pigments localized in smaller intracellular packages, which allows them to better harvest photons [Raven et al., 2005].

4.4. Impacts

[30] On a pan-arctic scale, there is a great diversity of phytoplankton. A recent authoritative assessment [Poulin et al., 2011] reports 1874 species, with most of these named species being in the microplankton size class (>20 µm), comprising centric diatoms, pennate diatoms, dinoflagellates, and other taxa. Despite this diversity, the contribution of microplankton to the standing biomass of total phytoplankton is substantially less than the aggregate of smaller phytoplankton. In the Beaufort Sea, picoplankton alone may represent 50% of the chlorophyll a, with an even greater representation (>70%) when the size class is extended to 5 µm [Tremblay et al., 2009]. Earlier results from the Canada/United States 1994 Arctic Ocean Section reported that small phytoplankton (<5 µm) accounted for 59–88% of the total biomass [Gosselin et al., 1997]. In our own recent studies in the Canada Basin [Comeau et al., 2011], we found 80% of the chlorophyll a in the <3 µm fraction. Taken together, these results strongly suggest that when all picoplankton and nanoplankton cells are considered together, they account for most of the phytoplankton biomass in the Canada Basin. Thus, sustained year-to-year changes in the abundance, biomass, and productivity of picoplankton and nanoplankton are likely to have a measurable impact on the pelagic food web.

[31] On ecologic impact, Tremblay et al. [2012] discussed some possible consequences of such food web changes. In particular, they showed by inverse modeling that residual carbon flow from the plankton is insufficient to sustain commercial fish catches in Amundsen Gulf. Any decrease in the ecological efficiency of carbon transfer from primary producers to higher trophic levels would presumably further reduce the existing residual carbon flow.

[32] On genealogic impact of a higher Pico:Nano ratio, one might speculate that the evenness component of diversity may be lower because of increased dominance by picoplankton. This may already be incipient with Bacteria at the subsurface chlorophyll layer of the Canada Basin. Here, the number of bacterial Operational Taxonomic Units has recently decreased, and the smaller number of OTUs account for higher percentages of the populations [Comeau et al., 2011].

4.5. Conclusion

[33] We conclude that as upper ocean nitrate concentration changes in the Canada Basin from year to year, the change in Pico:Nano is commensurate with the difference that this ratio exhibits between one Longhurst province and another in relation to nitrate concentration. This is a prediction only in the sense that the pattern embodies some explanatory power rather than some strength of inference. It is not a recipe for forecasting the future, not least because of the low coefficient of determination and the systematic bias between the spatial and temporal models; but more importantly because the recipe is predicated on known stable states without accommodation for transitions to alternate states at points of criticality. For a hyper-oligotrophic ocean, what is a possible next state? Would the future ecological community be one that has no present-day analogue [Williams and Jackson, 2007]? In the Arctic, as elsewhere, long-term and fine-scale observations remain necessary to provide evidence rooted in time and place. This is a way for humans to adapt and, if necessary, transform the social-ecological system with which they live.


[34] We thank chief scientists Sarah Zimmermann and Jane Eert and the sampling and analysis teams for their dedication in collecting high-quality time-series data. We also thank the captains and crews of the CCGS Louis S. St-Laurent, as well as Jean Lemire and the captain and crew of the Sedna IV. The Arctic ship-based program was supported by Fisheries and Oceans Canada, the US National Science Foundation Office of Polar Programs (grant OPP-0424864) and the Canadian International Polar Year office. We are grateful to those who have worked in the flow cytometry laboratory over the years, including Kelly Haussler, Peter Sykes, Diane Horn, Karen Scarcella, Chantal Giroux, and Gillian Forbes. We appreciate the constructive reviews of the journal referees who made their identities known to us. This work was supported in part by DFO Maritimes Atlantic Zone Monitoring Programme and Labrador Sea Monitoring Programme; and also NSERC 249994-02 to RJN.