Differential cell size optimization strategies produce distinct diatom richness–body size relationships in stream benthos and plankton


Sophia I. Passy (e-mail sophia.passy@uta.edu).


  • 1The relationship between species richness and body size is one of the most thoroughly studied subjects in animal ecology; however, this relationship and its underlying mechanisms are largely unknown in photosynthetic organisms, especially protists.
  • 2In this continental study, I first examined the number of diatom species across the cell size spectrum in benthic and planktonic stream habitats. The relationship was right-skewed unimodal and was significantly different between the benthos and the plankton; larger sizes were more speciose in the benthos, and smaller sizes in the plankton. The species richness peaks were explained with allometric trade-offs between maximum nutrient uptake rate and dispersal in the benthos but maximum nutrient uptake rate and sinking resistance in the phytoplankton.
  • 3I also explored the cell size similarity among species and across environments. Small diatoms were significantly more similar in size than large diatoms, and benthic diatoms were significantly more similar than planktonic diatoms.
  • 4This is the first continental study on the richness–body size relationship in algae, which suggests that the environmental differences between benthic and planktonic habitats generated allometric trade-offs that have driven the cell size optimization towards larger species in the benthos but smaller in the plankton. The patterns of cell size similarity revealed a higher niche overlap in the benthos than in the phytoplankton and among small species than among large species. These findings indicate that interspecific competition in stream diatoms, which is a function of niche differentiation, is habitat-specific and inversely related to cell size.


The relationships of species richness with sampling area, productivity and body size are probably the three most extensively studied topics in ecology. While the first two relationships have been examined by animal and plant ecologists alike, the third one has been grossly overlooked in the plant literature despite a long and inspiring history of research in animal ecology (for an exhaustive review see Allen et al. 2006). The change in species richness with body size, i.e. the richness–body size relationship, most often conforms to a right-skewed unimodal distribution (May 1986; Siemann et al. 1996). Still, what mechanisms generate the richness–body size relationship and whether it is consistent with scale, latitude and habitat heterogeneity and across taxonomic categories are subjects of active research (Blackburn & Gaston 1994; Bakker & Kelt 2000; Gaston et al. 2001; Kozłowski & Gawelczyk 2002; Niklas et al. 2003; Knouft 2004; McClain 2004).

Traditionally, the unimodal shape of the richness–body size relationship has been explained with various allometric trade-offs, whereby small and large sizes are advantageous under different conditions and the richness peaks at intermediate sizes, which can be beneficial under most conditions. However, as the number of intermediate size species increases so does the competition among them, which in turn forces species towards sizes of lower adaptive value but comparatively less intense negative interspecific interactions. For example, a trade-off between speciation and extinction, with speciation favouring small species and extinction, small or large species, depending upon environmental stability, generates unimodal distributions (Dial & Marzluff 1988). A unimodal distribution could be the result of a trade-off between energy uptake and energy conversion to offspring, with large species being superior in resource acquisition and small species superior in conversion of resources into reproductive work (Brown et al. 1993). A trade-off between food patch size increasing with consumer body size and tolerance to low resource concentration in the patch decreasing with consumer body size provided the theoretical basis of the synthetic theory of biodiversity, which predicted a unimodal but left-skewed richness–body size relationship and higher body size similarity in larger species (Ritchie & Olff 1999). A model incorporating allometric scaling laws of the number of individuals and speciation rate decreasing with body size and active dispersal increasing with body size produced maximum species richness at intermediate body sizes (Etienne & Olff 2004). A trade-off between production and mortality has also been invoked to explain the unimodal body size distribution (Kozłowski & Weiner 1997; Kindlmann et al. 1999).

A completely different suite of driving forces has been proposed in elucidating the unimodal right-skewed richness–body size relationships in plants. It has been suggested that large plant species were selected against because they exhibit lower niche differentiation and fecundity allocation, and require environmental conditions that were uncommon throughout evolutionary history (Aarssen et al. 2006). The richness–body size relationship in algae has received very little attention and primarily in plankton studies (Havlicek & Carpenter 2001; Scheffer & van Nes 2006). In the periphyton this relationship remains unknown at continental scales, although slightly right-skewed and symmetrical distributions have been reported at regional and watershed scales, respectively (Soininen & Kokocinski 2006).

Diatoms differ from both animals and plants and it is to be expected that the mechanisms generating their richness–body size relationships are also different. For example, diatom dispersal involves the mature organism (like in animals) and not the reproductive structures such as spores or seeds (like in plants). However, like plants, diatoms are passive dispersers despite the ability of many raphe-bearing species to move in their immediate vicinity; therefore dispersal should scale negatively with body size. The resources used by diatoms, unlike animals, do not form discrete units such as prey or patches of vegetation; consequently, the size optimization in diatoms is not driven by the size of their food. Unlike both animals and plants, diatom mortality is a response to unfavourable environmental changes or grazing but not a natural completion of their lives when cell division, i.e. reproduction, takes place. Therefore, mortality can affect but cannot govern the body size distributions in diatoms. Finally, streams are highly heterogeneous ecosystems where, even within a single reach, there are areas of low and high disturbance (e.g. the margins vs. the thalweg) and areas of lower and higher nutrient levels (e.g. hard vs. soft substrates). Hence, in each stream reach there are conditions of low disturbance and higher nutrient levels that would support large diatoms. This contrasts with terrestrial systems, where undisturbed and nutrient-rich habitats required by large plants were historically rare (Aarssen et al. 2006).

Important properties of algal communities, such as taxonomic organization and biomass–richness relationships, were shown to be significantly different between planktonic and benthic stream algae (Passy & Legendre 2006a,b). Planktonic communities comprised fewer but more distantly related species and reached peak biomass at lower species richness than benthic communities. These differences were attributed to the existence of fewer and shorter environmental gradients in the phytoplankton, forcing a stronger niche differentiation among species. Taxonomic and biomass-richness structures of a community correlate with organismal body size, which is thus expected to exhibit different distributional patterns in the plankton and benthos. To test this hypothesis, I examined the richness–body size distributions of planktonic and benthic diatoms collected from all major watersheds in the United States by the National Water-Quality Assessment (NAWQA) Program. I hypothesized that the richness–body size relationship would be governed by similar physiological but different ecological factors in the two habitats. Nutrient uptake rate, which is determined by algal cell size irrespective of the habitat, i.e. freshwater or marine (Grover 1989; Aksnes & Egge 1991; Stolte et al. 1994; Wen et al. 1997), is expected to drive the cell size optimization in both planktonic and benthic diatoms. However, the environmental conditions in these two habitats are vastly different and dispersal can be hindered in the heterogeneous benthic environments.

Successful dispersal must include: dislodgement from the substrate (emigration), entraining and remaining in the water flow (suspension), and reaching suitable substrates (immigration), where individuals can establish and start reproducing. Dispersal is of much lesser importance in the plankton, where algae dwell and reproduce in their characteristic habitat or are constantly imported from the benthos. Sinking, on the other hand, which is strongly dependent on cell size, can interfere with life-supporting processes in the phytoplankton such as photosynthesis and growth. Therefore, it is suggested here that cell size distributions can be determined by the combined effect of nutrient uptake rate and dispersal in the benthos, and by nutrient uptake rate and sinking in the phytoplankton. While data exist on algal nutrient uptake and sinking velocity (discussed below), to my knowledge there is no comprehensive research on diatom dispersal but only partial studies on immigration and emigration, which do influence dispersal but do not equate with it. Here, dispersal was specifically examined in an independent data set, derived from an extensive 2-year spatial survey of three reaches in Batavia Kill, a highland stream in New York State (Passy & Blanchet 2007).

The objectives of the present study were: (i) to determine the continental richness–body size relationships in planktonic and benthic diatoms; (ii) to assess whether these relationships could be explained with the interplay of physiological and ecological allometries, derived from independent investigations; and (iii) to determine how similar in biovolume diatoms are along their cell size spectrum, which has implications for their niche differentiation and consequently competition strength.

Materials and methods

algal data

The NAWQA data set

The NAWQA data set analysed here contains 4778 diatom samples from more than 50 major river basins and aquifers across the USA, including sites in Alaska and Hawaii. Three habitat types, defined by the NAWQA, were quantitatively sampled for algae: hard substrates in richest targeted habitats (RTH), soft substrates in depositional targeted habitats (DTH), and phytoplankton (for details see http://water.usgs.gov/nawqa/protocols/OFR02-150/OFR02-150.pdf). RTH maintain the taxonomically richest community and encompass the following habitats: (i) shallow riffles in areas with coarse-grained substrates (epilithon); (ii) woody snags in reaches with fine-grained substrates (epidendron); and (iii) macrophytes where riffles or woody snags are absent (epiphyton). DTH microalgae are found in organically rich or sandy depositional areas along the stream margins, including epipelic and epipsammic habitats. Both RTH and DTH were sampled from a defined area of substrate. Phytoplankton is a community of suspended algae collected from 1 L of water in nutrient-rich streams or 5 L of water in unproductive, nutrient-poor streams. More information on the habitats and sampling techniques is given in Passy & Legendre (2006b). The NAWQA data set comprises 2699 RTH, 1682 DTH and 397 phytoplankton samples collected year round between March 1993 and September 2003.

Sample processing and algal enumeration and identification, followed by assessment of algal biovolumes, were carried out by specialized phycology laboratories (for details see http://diatom.acnatsci.org/nawqa/protocols.asp). Briefly, soft algae and diatoms were enumerated, followed by diatom identification in permanent mounts prepared from acid-digested samples. Biovolume was calculated for all taxa in a sample after approximation to simple geometric figures. In this study only the diatoms were considered because they are truly single celled algae and diatom cell biovolume is equivalent to organismal body size. Many diatoms form colonies but there is no exchange of materials among cells and therefore the colonial habit is unlikely to directly influence the nutrient uptake rate. Colonies and cell shapes deviating from a sphere can greatly reduce the sinking velocity of planktonic diatoms and this was partially taken into consideration in the calculation of sinking resistance (see below).

The Batavia Kill data set

Spatial surveys of algae were carried out in 2001 and 2002 in three 100-m cobble-bottom reaches within a 5-km stretch of Batavia Kill, an upland stream within the New York City Watershed (Passy & Blanchet 2007). The three reaches differed in geomorphic status and canopy cover. The upstream reach was geomorphically stable, with significantly greater depth and larger particle size than the other two reaches. The midstream and downstream reaches were geomorphically unstable. After the first survey, the downstream treatment reach was subjected to a large-scale restoration, including the re-channelling and bank stabilization of 1.6 km of stream. Canopy cover, measured in the three reaches in both years, averaged 82% in the stable reach and 17% in the midstream reach. The treatment reach was completely deforested before and after the first year of restoration. Algae were collected in a regular spatial grid approximately every 50 cm across the stream and every 10 m along the stream from 7.55 cm2 of rock surface. Samples were preserved in 4% formaldehyde. The surveys were conducted during 11–13 July 2001 and 9–11 July 2002, with 248 and 223 samples collected, respectively. For diatom identification, samples were digested with acids and mounted in Naphrax® (PhycoTech Inc., St Joseph, MI, USA). At least 300 frustules were counted with a 100 × 1.35 NA oil immersion objective. Biovolume data for the diatom species were obtained from the NAWQA data set, which includes numerous sites in New York State.

statistical analyses

The NAWQA data set

Body size frequency distributions were explored in the following way. The average cell biovolume for each species across all samples was calculated and ln-transformed. In each habitat, the ln-biovolume spectrum was subdivided into 12 classes (bin size of one unit on a ln-scale) and the number of species in each class was counted. The results were generally robust to increases in bin size; a decrease in bin size allowed the emergence of small local peaks but did not alter the position of the major peak, which remained invariant. The three habitats were analysed with a contingency table testing the null hypothesis that species richness distributions were independent of habitat type. The significance of skewness and kurtosis in ln-biovolume was determined by dividing the absolute value of the respective statistic by its standard error; if the ratio was greater than 2 then the statistic was significantly different from zero (systat 11 2004).

For all samples in a habitat, the ratio of larger-to-smaller diatom species adjacent in size (henceforth referred to as biovolume ratio) was examined as a function of the ln-biovolume of the larger diatom using a LOWESS smoother. The smoothing technique suggested a threshold behaviour of the dependent variable, which showed a different linear response on both sides of a threshold value of the predictor. This behaviour was modelled with a piece-wise linear regression using systat 11. The model was defined in systat as follows:

ŷ = b0 + b1x + b2(x – z)(x > z)( eqn 1)
if x < z then ŷ = b0 + b1x
if x > z then ŷ = b0– b2z + (b1 + b2)x

where ŷ = estimate of biovolume ratio, b0 = intercept of the first regression line, b1 = slope of the first regression line, b0b2z = intercept of the second regression line, b1 + b2 = slope of the second regression line, z = a threshold value of x where the slope changes, x = ln-biovolume of the larger species in the ratio. In order to standardize the variables across habitats and remove the influence of extreme values biovolume ratios larger than 10 were removed from all analyses.

The Batavia Kill data set

The frequency of each species in the samples from each of the three reaches was used as a measure of dispersal, i.e. the likelihood of a particular species reaching all habitats. Thus, species with high frequencies are good dispersers and vice versa. While dispersal has traditionally been measured in actively dispersing species in terms of the distance they can travel away from a source, to my knowledge there are no investigations that explicitly measured the dispersal capabilities of diatom species in nature. Diatoms are passive dispersers and their successful dissemination depends on their size. Smaller cells have lower probability of sinking (see eqn 3 below) and can thus travel farther than large cells; moreover, they have a higher probability of benthic immigration, especially at high current velocities (Stevenson 1996). In addition, dispersal is controlled by the population size (Finlay et al. 2002), i.e. the higher the number of potential colonists the higher the chance that at least one individual would reach a location. However, the number of individuals is an allometric function (Etienne & Olff 2004; S. I. Passy, unpublished data) and therefore the examination of species frequency as a function of body size would account for the number of individuals as well. Species frequencies within and across individual reaches in the 2 years studied were examined with a curve-fitting program (TableCurve 2D 5.01). The following parsimonious equation, which produced consistently good fits of the frequency–biovolume relationships, was used:

ŷ = b0 + b1x−1( eqn 2)

where ŷ = species frequency, b0 = intercept, b1 = slope, and x = ln-biovolume. Homogeneity of slopes across reaches was tested with ancova, which ensured that this relationship was not dependent on reach type. The dispersal in the benthic samples from the NAWQA data set was subsequently calculated as the inverse of ln-biovolume and standardized to range between 0 and 1.

nutrient uptake rate

Maximum nutrient uptake rate, ρmax (µmol cell−1 h−1), which reflects the sustained nutrient flux required to support maximal growth, is a power function of cell biovolume, V (µm3): ρmaxV0.66 (Irwin et al. 2006). Although maximum uptake rate per unit biovolume has been suggested to decrease with cell size in algae (Irwin et al. 2006 and references therein), in the case of diatoms, which have specific cell structure, it is expected to be size-independent for the following reasons. Diatom biovolume comprises two distinct parts: the cytoplasm with all cell organelles, including chloroplasts, and a large central vacuole, which contains mostly water but also ions, salts and sugars, and is used primarily as a depot for nutrients and storage products. The cytoplasm is metabolically active and requires nutrients to execute all anabolic processes, whereas the vacuole is comparatively inert. The cytoplasm is confined to a thin layer under the frustule, particularly pronounced in pennates, and being proportionate to the cell surface scales with the square of cell length (V2/3). Thus, expressing nutrient uptake per unit anabolically active biovolume, i.e. the part of the cell biovolume that needs nutrients, would give ρmaxV0.660V0.667 = V−0.007. An exponent of –0.007 is sufficiently close to zero to suggest size-independence of per biovolume nutrient uptake.

The maximum nutrient uptake rate was calculated as V0.66 for diatoms in all three studied habitats and standardized to range between 0 and 1.

sinking resistance

Sinking velocity for phytoplankton species is a function of their size and form resistance of non-spherical cells as indicated by the Ostwald's modification of Stoke's law (Graham & Wilcox 2000):

image( eqn 3)

where vs = sinking velocity, g = gravitational acceleration of the earth, r = radius of a spherical volume, equivalent to that of the algal cell, q′ = density of the algal cell, q = density of water, ν = viscosity of water, and φ = dimensionless form resistance. All terms in this equation except r and φ are considered constants and therefore with no effect on variable standardization. Using the NAWQA biovolume V (µm3) data, sinking velocity was calculated as V 2/3, which is proportionate to the square of the cell radius. The sinking velocity was then converted to sinking resistance by calculating the inverse of vs, i.e. V2/3, and the resistance was standardized to a range between 0 and 1. Standardization was carried out separately for centrics and pennates, which have comparatively low and high φ, respectively (Reynolds 1984). This rationale was used to determine the lower and upper bound of vs for species differing in form resistance.


In all habitats, the biovolume frequency distributions, which were equivalent to biovolume–species richness distributions (measurements of ln(Biovolume) were equivalent to the numbers of species in each habitat), were unimodal, significantly right-skewed (|Skewness|/SE of Skewness > 2), and significantly leptokurtic (|Kurtosis|/SE of Kurtosis > 2) (Table 1, Fig. 1). While skewness was comparable among the three habitats, kurtosis was much more severe in the benthos than in the plankton, which had a flatter species richness–ln-biovolume distribution. Contingency table analysis revealed a significant difference between benthos and plankton (χ2 = 52.499, d.f. = 22, P = 0.0002), rejecting the null hypothesis of habitat-independent biovolume distributions. The maximum species richness was observed at ln-biovolume between 8 and 9 (at 8.9) in the benthos, while in the phytoplankton, the peak spanned several biovolume classes, from 6 to 9, with ln-biovolume between 6 and 7 being slightly more speciose than the rest (Fig. 1). These findings indicate that smaller sizes were more speciose in the plankton and larger sizes in the benthos. Species richness distributions were not significantly different in the two benthic habitats (χ2 = 2.939, d.f. = 11, P = 0.992). Notably, there were substantially more species to the left of the optimum than to the right in the benthos, i.e. smaller biovolume classes were much more speciose than larger ones. This trend also occurred in the phytoplankton but was much less pronounced (Fig. 1).

Table 1.  Basic statistics of ln-biovolume in the three habitats. Note that n equals the number of observations, which is also equal to the number of species as each observation of ln-biovolume was taken from a different species. SES, SEK = standard error of skewness and kurtosis, respectively
Basic statisticsDTHRTHPhytoplankton
95% CI7.29–7.487.27–7.456.78–7.05
Figure 1.

Diatom richness–biovolume distributions in benthic and planktonic stream habitats from all major watersheds in the US. DTH = depositional targeted habitats, RTH = richest targeted habitats.

The frequency, a proxy measure of dispersal, was calculated for all diatom species in the three studied reaches of Batavia Kill individually and altogether, within and across years. Frequency was a nega tive power function of ln-biovolume (Fig. 2) and this relationship was independent of reach type as demonstrated by ancova (Table 2). Biovolume explained 19% of the frequency variance in 2001 and 13% in 2002 and these results were highly significant (Table 2). Therefore, despite substantial differences in geomorphic and light conditions among the three reaches, the relationship between diatom dispersal and biovolume remained invariant. Pooling all reaches together produced a frequency–biovolume relationship that was statistically equivalent to the relationships in individual reaches. This indicates that dispersal was also insensitive to the scale of observation, i.e. local within-reach vs. broader between-reach. Dispersal in the two benthic habitats of the NAWQA data set was estimated as the inverse of species ln-biovolume. Nutrient uptake rate was calculated and plotted as a function of ln-biovolume together with the dispersal estimates (Fig. 3a,b). The largest ln-biovolume (14.88) in all habitats was identified as an outside value by stem-and-leaf plots and was not included in the calculations of nutrient uptake rate, dispersal, and sinking resistance (below). Optimal biovolume was expected at the intersection of dispersal and nutrient uptake, i.e. at ln-biovolume values where neither of these processes was limiting. Such intersection was observed at ln-biovolume of 8.9 in DTH and RTH, which exactly corresponded to the value of ln-biovolume with the highest number of species.

Figure 2.

Frequency of 49 diatoms in 2001 and 65 diatoms in 2002 within three 100-m reaches of Batavia Kill, NY, as a function of their ln-biovolume. Data fits were generated from the following regression model: frequency = b0 + b1(ln-biovolume)−1. In 2001 b0 = –59.37, b1 = 970.13; in 2002 b0 =–53.77, b1 = 752.67. These regression parameters were not significantly different between the 2 years and were statistically equivalent to the respective parameters in the individual reaches.

Table 2. ancova of species frequency as a function of ln (biovolume)−1, reach, and their interaction in the 2 years of study of Batavia Kill. To eliminate correlation between effects, the independent variable was centred by subtracting the mean. R2 = coefficient of determination, n = number of observations, SS = sum of squares, MS = mean square, and d.f. = degrees of freedom
2001, n = 135, R2 = 0.19
Biovolume × Reach323.032161.510.180.83
2002, n = 158, R2 = 0.13
Biovolume × Reach631.882315.940.530.59
Figure 3.

Nutrient uptake rate and dispersal as power functions of ln-biovolume in DTH (a) and RTH (b). Nutrient uptake rate and sinking resistance as power functions of ln-biovolume in phytoplankton (c).

Sinking resistance was determined separately for the two large diatom groups (centrics and pennates), which differ in form resistance. An optimal biovolume range was expected between the intersections of the standardized nutrient uptake rate and sinking resistance. Indeed, these intersections were observed at ln-biovolume of approximately 5.9 and 7.2 (Fig. 3c), which corresponded to the most speciose ln-biovolume values (Fig. 1).

The biovolume ratio of species adjacent in size increased significantly (P < 0.00001) and non-linearly with ln-biovolume of the larger species in all habitats (Table 3). Piece-wise linear regressions explained between 31 and 34% of the variance in biovolume ratio and showed that its rate of increase was not uniform along the biovolume gradient. This rate was very low (0.04 ≤ b1 ≤ 0.07) at lower values of ln-biovolume but changed abruptly at a threshold value of ln-biovolume, above which the biovolume ratio rapidly increased (1.96 ≤ (b1 + b2) ≤ 2.32). The biovolume ratio was close to 1 below the threshold but much higher than 1 above it (Fig. 4a–c), indicating that species adjacent in size had nearly identical biovolumes below the threshold but disparate ones above the threshold. The rate of biovolume ratio increase with ln-biovolume was significantly different across habitats; it was highest in phytoplankton and lowest in DTH (Table 3, Fig. 4d). The threshold value was significantly higher in phytoplankton and DTH than in RTH.

Table 3.  Coefficients of determination (R2), regression coefficients, and 95% confidence intervals (CI) of piece-wise linear regressions of biovolume ratio vs. ln (biovolume) in the three studied habitats. Regression parameters that are not significantly different at P < 0.05 across habitats are in italic. P < 0.00001 in all regressions. n = number of observations; b0 = intercept; z = threshold value of the independent variable, where the slope of the fitting curve changes from b1 to b1 + b2
n79 07295 82210 566
95% CI0.895–0.9150.850–0.8720.707–0.807
95% CI0.037–0.0410.052–0.0550.065–0.081
b1 + b21.9582.0102.320
95% CI1.933–1.9821.983–2.0372.213–2.428
95% CI9.286–9.3059.216–9.2419.262–9.353
Figure 4.

Biovolume ratio as a function of ln-biovolume of the larger species in the pair in DTH (a), RTH (b) and phytoplankton (c). The fits for the three habitats, produced by piece-wise linear regressions, are plotted together in (d).


The richness–biovolume relationship was unimodal right-skewed in both benthic and planktonic diatoms from the US running waters. However, there were significant differences between the two habitats. In the benthos, the biovolume distributions were more strongly leptokurtic, displaying a distinct peak of species richness at ln-biovolume between 8 and 9. The position of this peak was exactly predicted by an allometric trade-off between nutrient uptake rate and dispersal. Smaller species (to the left of the peak) were limited by their ability to acquire resources, whereas larger species (to the right of the peak) were limited by their dispersal capabilities. Here I examined only the physiological aspect of the resource uptake limitation but it should be noted that it is further exacerbated by the strong competition in the benthos where larger species, by virtue of their taller stature, gain better access to light and nutrients from the water column and can create a resource-depleted environment for the smaller species in the lower stories. The richness peak can be interpreted as an outcome of cell size optimization, where species were not severely limited by their ability to either sequester resources or disperse. Considering that the majority of species (61%) was observed to the left of the richness peak and only a minor portion of the species to the right of it (13%) suggests that constraints on dispersal in the benthos are much more important than physiological restrictions for diatom cell size organization. Tolerance for low nutrient levels, well documented in small diatoms (Cattaneo et al. 1997; Wunsam et al. 2002), and production of storage vacuoles can offset a poor nutrient sequestration and explain why there are so many species with suboptimal abilities to obtain nutrients. The ineffectiveness of large diatoms to reach all locations where they would be competitively superior due to mechanical constraints and lower numbers of individuals takes a toll on their diversification, which is evident in their minor contribution to the overall species richness. Additionally, high profile diatoms have a higher susceptibility to physical disturbance from shear stress or grazing (Steinman 1996; Stevenson 1996) and chemical disturbance from various forms of pollution (Cattaneo et al. 1998), which will further limit their diversification.

Phytoplankton had a flatter biovolume distribution with much higher relative richness of smaller size classes than the benthos. An allometric trade-off between nutrient uptake rate and sinking resistance accounted for part of the richness peak containing the most speciose biovolume, i.e. ln-biovolume between 6 and 7. However, unlike dispersal, sinking resistance varies not only with cell size but also with a number of other factors such as form resistance, density of the cell, water temperature and turbulence, and is therefore more adequately represented by a band rather than a curve. Here, I treated the centric and pennate diatoms as two homogeneous groups, implicitly differing in form resistance. This, however, is an oversimplification because there is substantial variability of form resistance and cytoplasmic density within each group, which, if accounted for, would have generated a much larger spectrum of sinking resistance variability. Smaller species (to the left of the peak) were limited by nutrient uptake and represented 15% of all planktonic diatoms, while large species (to the right of the peak) were controlled by sinking resistance and accounted for 11% of all species. As both tails of the biovolume distribution in the phytoplankton contained a comparable number of species, it is logical to assume that the two mechanisms described here exert similar impact on diatom cell size organization. The reason may be in their common cellular roots, e.g. the amount of nutrient per cell determines nutrient uptake but also sinking resistance, as the excess nutrients stored in the cell as oils reduce its density and consequently its sinking velocity. Selective grazing can also contribute to a reduced number of smaller species that are more vulnerable to herbivory than larger diatoms. Finally, large diatoms are more likely to be exported into the phytoplankton following disturbance events in the benthos.

The continental richness–biovolume relationships of freshwater diatoms (class Bacillariophyceae) were significantly right-skewed in all habitats, which is common for higher taxonomic levels (e.g. class) and at larger geographical scales (Kozłowski & Gawelczyk 2002). There are different hypotheses for the prevalence of right-skewed distributions, including size-biased speciation and extinction, macroevolutionary constraint on small sizes, the fractal nature of the environment and body size optimization (Kozłowski & Gawelczyk 2002). Probably a combination of these mechanisms is responsible for the shape of the present distributions; however, the macroevolutionary constraint on small sizes, also known as the ‘reflecting barrier’, is particularly relevant in diatoms. According to this hypothesis, distributions cannot extend below the natural barrier of the lower limit of body size, where physiological constraints reflect diversification toward larger sizes (Kozłowski & Gawelczyk 2002 and references therein). This is why richness–body size distributions are not simply left truncated but humped. While this mechanism may not be unequivocal across animal lineages of widely varying lower size, it is apparent in diatoms with a lower size limit set by the minimum size of the eukaryotic cell, which, on average, does not fall below 10 µm in diameter. Spherical shapes with the same biovolume as the smallest diatoms in the NAWQA data set would exhibit diameters as low as 2.4 µm, i.e. below the lower limit of most eukaryotic cells, where algal maximum nutrient uptake rate, and metabolic and nutrient storage abilities are expected to be diminished. Thus, the existence of a left barrier and an optimization mechanism (maximizing nutrient uptake rate) that drives diversification away from it can account for the right-skewed richness–biovolume distributions in diatoms.

In all three habitats, the biovolume ratio increased along the cell size gradient with a distinct change of rate at a threshold ln-biovolume value of around 9.3 (very close to the richness peak). The piece-wise response of biovolume ratio to diatom size indicated that there were two distinct groups of species. Below the threshold, the biovolume ratios approached 1, i.e. diatoms adjacent in size were nearly identical. Above the threshold, diatoms displayed large distances in size. This finding is exactly opposite to the predictions of the synthetic theory of biodiversity (Ritchie & Olff 1999), which postulated a decreasing body size ratio with organismal size in local communities. The reason for this controversy is in the underlying trade-offs. The synthetic theory assumes a decreasing number of available patches (only the large ones) but increasing tolerance for low resource concentration in large species. Although this theory was supported by empirical data on large mammals and terrestrial plants, its premises cannot apply to unicellular algae. Diatoms inhabit a continuously changing environment with pronounced gradients of nutrient and light (Borchardt 1996; Hill 1996) and not discrete patches (pertains mostly to benthic forms). In addition, algae encompass small species that are generally better adapted to low resource concentrations than large species (Cattaneo et al. 1997; Li 2002; Wunsam et al. 2002; Jiang et al. 2005; Irwin et al. 2006) (pertains to benthic and planktonic diatoms alike).

The analysis of biovolume ratios in this study reveals that small species form a continuum where there is a species for every possible size and the distance between species adjacent in size approaches zero. The lack of discontinuities in the biovolume ratios suggests that smaller species (below the biovolume threshold) form a single guild. The notion that species identical in size within trophic levels are also ecologically equivalent has become an ecological axiom. Therefore, small species with near nil differences in biovolume are expected to occupy very similar niches, and in coexistence to experience high niche overlap, the amount of which can be estimated from the size similarity among species. Thus, the size similarity was the greatest in DTH but the lowest in phytoplankton, suggesting a significantly higher niche overlap in the benthos than in the phytoplankton. This hypothesis was put forth to explain the higher taxonomic distances in the phytoplankton compared with the benthos (Passy & Legendre 2006a) and is fully corroborated by the present rigorous analysis of the cell size structure of diatom communities. The highest biovolume similarity in DTH was probably forced by environmental conditions of unstable substrates and sediment-borne contaminants favouring a comparatively taxonomically uniform community of low-profile, motile and pollution-tolerant species (Passy & Legendre 2006a).

The high number of small species in a community has been explained mechanistically with the corresponding number of suitable patches. For example, the complex behaviour of large species may involve the utilization of a combination of patches, each one being sufficient to maintain a different small species (Hutchinson & MacArthur 1959). Furthermore, small species can exploit patches that are unavailable to large species due to their inability to perceive them or fit within them (Morse et al. 1985; Ritchie & Olff 1999; Aarssen et al. 2006). The remarkable biovolume similarity of small species documented here offers a complementary, functional explanation of the high richness of small species, i.e. small species exhibit much higher niche overlap, which is counterbalanced by their high tolerance to resource limitation (and environmental adversity in general). Indeed, species sufficiently similar to possess equal competitive abilities have been shown in a theoretical study to maintain stable coexistence and high diversity (Scheffer & van Nes 2006). Additionally, the superior dispersal of small species prevents local extinctions by a constant re-supply of new colonists.

Larger species (above the biovolume threshold) displayed substantial differences in cell size, which suggests the existence of multiple guilds. The members of different guilds have differential resource requirements and experience less competition in sympatry than the members of the same guild. In addition, the superior capabilities of large species to acquire resources would further relax their resource competition. The piece-wise linear nature of the diatom biovolume similarity can also be interpreted as a competition gradient whereby competition remains strong up to a threshold value of ln-biovolume, beyond which the negative interspecific interactions sharply decline. The classical paper by Hutchinson (1959) explored the size ratio of co-occurring competitors within mammals and birds and suggested that a ratio of about 1.28 is necessary for niche differentiation. Across the three habitats in this study, the average biovolume ratio of species below the ln-biovolume threshold was between 1.14 and 1.21, but between 2.02 and 3.05 above that threshold. This indicates that small species of diatoms experience much stronger interspecific competition than multicellular organisms, whereas the competition among large diatoms is alleviated.

In conclusion, this is the first continental study on the richness–body size relationship in algae to suggest that the environmental differences between the benthos and the plankton are responsible for allometric trade-offs driving cell size optimization towards larger species in the benthos and smaller species in the plankton. The patterns of biovolume similarity revealed a higher niche overlap in the benthos than in the phytoplankton and among small species than among large species. Therefore, the interspecific competition in stream diatoms, being a function of niche differentiation, is habitat-specific and inversely related to cell size.


I thank Don Charles for kindly providing the data set and Mark Ritchie for his comments on this project. I am grateful to Jim Grover for stimulating discussions and an insightful review, Christer Nilsson, Barney Davies and two anonymous reviewers for their suggestions, which substantially improved the manuscript. Financial support under grant #C004307 from the New York State Department of Environmental Conservation is gratefully acknowledged. This manuscript is submitted for publication with the understanding that the United States Government is authorized to reproduce and distribute reprints for governmental purposes. The views and conclusions contained in this document are those of the author and should not be interpreted as necessarily representing the official policies, either expressed or implied, of the US Government.