Functional classifications and their application in phytoplankton ecology

Summary

1. Ecologists often group organisms based on similar biological traits or on taxonomic criteria. However, the use of taxonomy in ecology has many drawbacks because taxa may include species with very different ecological adaptations. Further, similar characters may evolve independently in different lineages.
2. In this review, we examine the main criteria that have been used in the identification of nine modes of classifying phytoplankton non‐taxonomically. These approaches are based purely on morphological and/or structural traits, or on more complex combinations including physiological and ecological features.
3. Different functional approaches have proved able to explain some fraction of the variance observed in the spatial and temporal distribution patterns of algal assemblages, although their effectiveness varies greatly, depending on the number and characteristics of functional traits used. The attribution of functional traits to single species or broad groups of species has allowed a few classifications (e.g. Functional Groups, FG) to be used in the assessment of ecological status.
4. We stress that the misuse of functional classifications (by applying them under conditions other than those intended) can have serious consequences for interpreting ecological processes. Assigning functional traits or groups cannot be considered a surrogate for the knowledge of species or ecotypes, and the use of specific traits must always be justified and circumscribed within the limits of ecological questions and hypotheses.
5. An important future challenge will be to integrate advances in molecular genetics, metabolomics and physiology with more conventional traits; this will form the basis of the next generation of functional classifications.

Introduction

Species in any one community may have similar ecological roles, therefore revealing some ‘redundancy’ in ecological functions. This has led ecologists to group organisms with similar ecological features, with the aim of obtaining a framework that potentially simplifies the complexity of real ecosystems. Ecological groups defined in this way are called adaptive syndromes or Functional Groups (Solbrig, 1993). At the ecosystem level, a grouping based on feeding relationships was one of the first attempts to link species into functional groups, opening broad research fields including ecosystem energetics, physiological ecology and trophic interactions (Odum, 1959; Cummins & Klug, 1979; Azam et al., 1983; Jørgensen & Kay, 2001). In plant ecology, functional classifications have been widely used (Körner, 1993) and progressively updated, leading to the development of new paradigms (Lavorel et al., 2007; Grime & Pierce, 2012). In animal ecology, several studies have been performed where organisms were grouped into functionally coherent clusters, generally called ‘guilds’ (e.g. Fauth et al., 1996; Barnett, Finlay & Beisner, 2007).

The phytoplankton is an extremely diverse, polyphyletic group of photosynthetic protists and cyanobacteria, which fuel food webs and drive biogeochemical cycling (Rousseaux & Gregg, 2014). Over the last few decades, various attempts have been made to categorise traits and functions in the phytoplankton (Lewis, 1976; Litchman & Klausmeier, 2008; Litchman et al., 2010), most recently opening new research perspectives in ‘chemotaxonomy’ (Descy, Sarmento & Higgins, 2009) and ‘ecometabolomics’ (Peñuelas & Sardan, 2009). Much understanding of the role of the phytoplankton comes from studies in culture, which, for instance, determined the growth and nutrient uptake kinetics of a series of taxa (Morris, 1981; Reynolds, 2006). However, the clustering of species according to physiological features is difficult (because data are not always available), leading many authors to rely on classifications based on other biological traits (Kruk et al., 2010).

‘Ecology is evolution in action’ (Krebs, 2009); thus, from an evolutionary perspective, functional criteria should comprise the biological processes and characters implicated in adaptation. The criteria used to define functional groups in phytoplankton include morphology, physiological features and, where appropriate, taxonomy. Besides biological and taxonomic traits, other criteria include ecological features, such as phenology, implicitly acknowledging that species showing similar seasonality respond similarly to a set of particular environmental conditions. In this respect, phytoplankton functional groups are arbitrary assemblages. Species could be classified taking into account their shape and the dimensions (Naselli‐Flores, Padisák & Albay, 2007) or specific physiological requirements (e.g. nutrient demands). In these two cases, planktonic diatoms forming long filaments (e.g. Aulacoseira), and planktonic filamentous green algae (e.g. Mougeotia) would be merged or placed in two separate groups, depending on the trait chosen, that is shape type or silica in the cell walls, respectively. The number of functional groups that can be devised is potentially very large. The choice of criteria encompasses the whole gradient of levels of organisation or biocomplexity (Fig. 1). At one extreme, modern phylogenetic analyses are revolutionising our view of relationships between taxa (Krienitz & Bock, 2012; Komárek, 2013). Similarly, diverse groups of algae are clearly circumscribed in their ability to produce specific metabolites, for example toxins in cyanobacteria (Metcalf & Codd, 2012; Newcombe et al., 2012). At the other extreme, phenological features have traditionally played an important role in the identification of ‘vegetation units’ (Reynolds, 1980). Intermediate levels of biocomplexity, which include both morphological and physiological traits, are those that are most easily seen as useful in the identification of functional groups.

Figure 1

The biocomplexity gradient in phytoplankton ecology. Physiological functions (e.g. mixotrophy, CO₂ concentrating mechanisms, N fixation) are carried out at the level of organisms, structures, cells and molecules.

The aims of this review are as follows: (i) to examine the criteria used in identifying phytoplankton functional groups, summarising and evaluating critically the main classifications proposed so far. The use of functional classifications should always take into account the range of circumstances in which they are intended to apply Therefore, emphasis will be put on (ii) the limitations in the application of classifications, based on the particular choice of discriminant criteria involved. The article concludes (iii) with a discussion of the potential future development of functional groups in phytoplankton ecology.

We do not review every article that has made use of some sort of functional classification but have tried to include those that have proposed well‐described, documented and widely applicable systems of classification (irrespective of the criteria used) and have contributed to the advance of functional classification in phytoplankton ecology. In Table 1, the different functional groups considered in this work have been roughly arranged based on the main criteria used for the classification. In particular, the work by Reynolds et al. (2002) set a milestone in the application of phytoplankton functional groups. This approach is considered here in detail, quantitatively testing the mutual relationships of the Functional Groups (FG) and their links with the main environmental constraints.

Table 1. Phytoplankton functional classifications were analysed in this work

Functional group	Acronym	Principal criteria	Main discriminant features	Reference (relevant to phytoplankton ecology)
r and K selection	r/K	Functional	Functional (growth) and morphometric attributes (see Pianka, 1970)	Margalef (1978); Reynolds (1988b)
Competitive, Stress‐tolerant and Ruderal strategists	CSR	Functional	Functional (growth) and morphological/morphometric attributes (see Reynolds, 2006)	Reynolds (1988a)
Biomass size spectrum; Normalised Biomass Size spectrum	BSS, NBS	Morphometrical	Size distribution	Platt & Denman (1978); Kamenir et al. (2004)
Traditional Taxonomic Size Spectrum	TTSS	Morphometrical	Size distribution	Kamenir et al. (2006)
Phytoplankton Geometric Shapes	PGS	Morphological	Shapes	Stanca et al. (2013)
Morphologically Based Functional Groups	MBFG	Morphometrical Structural	V, S, S/V, MLD, mucilage, flagella, aerotopes, heterocytes and siliceous exoskeletal structures	Kruk et al. (2010)
Functional Groups	FG	Phenological Ecological Functional	Phenology and ecological/functional attributes (tolerances to: low zmix, light, temperature, SRP, DIN, Si, CO2; high zooplankton grazing; see Table S1)	Reynolds (1980); Reynolds et al. (2002); Padisák et al. (2009)
Morpho‐Functional Groups	MFG	Morphometrical Structural Functional Taxonomic	Structural, functional and taxonomic characters: flagella, mixotrophy, cellular organisation, aerotopes, dimensions, shapes, mucilage	Salmaso & Padisák (2007)

• V, Volume; S, cell surface; MLD, maximum linear dimension; SRP, soluble reactive phosphorus; DIN, dissolved inorganic nitrogen; Si, reactive silica.

Taxonomic classifications

Species are the basic unit in ecosystem studies. Taxonomy at the species level brings the most complete level of information once the species niches are clearly defined. Higher taxonomic units were widely used to evaluate the distribution of phytoplankton (e.g. Wetzel, 2001), particularly along trophic and physical gradients. However, only a few generalisations are possible, including, among the others, the increase of cyanobacteria in eutrophic (Downing, Watson & McCauley, 2001; Jeppesen et al., 2005) and warmer lakes (Paerl & Huisman, 2008; Winder & Sommer, 2012), and the decrease of chrysophytes in eutrophic waterbodies (Kalff & Watson, 1986). Analyses based on finer taxonomic resolution (e.g. families; Salmaso et al., 2006) are more difficult to apply and interpret due to the large number of taxonomic units.

Once originally based on pigment composition and cellular structure, modern phytoplankton taxonomy is being strengthened by molecular techniques (Wilmotte & Herdman, 2001; Rajaniemi et al., 2005; Krienitz, 2009; De Clerck et al., 2013). DNA sequencing allows obtaining quantitative data matrices that can be analysed numerically, providing lineage relationships between species (Ciccarelli et al., 2006; Chakerian & Holmes, 2012). Nevertheless, the use of taxonomy in ecology has at least two severe drawbacks. On the one hand, many broader taxonomic groups include species with very different ecological properties (e.g. among diatoms, there are species forming large colonies and others with small single cells). On the other hand, distantly related species can share ecological attributes (e.g. mixotrophy) by convergent evolution, that is the independent evolution of analogous characters in different lineages (Wilson, 1992).

Classification of life history traits and the evolution of competitive abilities

The basics of competitive abilities: r and K selection (r/K)

The theory of r and K selection (Tables 1, 2) was first proposed by animal ecologists (MacArthur & Wilson, 1967). In this classification, populations are characterised by the relative importance of the parameters r (rate of increase) and K (carrying capacity) of the logistic equation for population growth (Pianka, 1970; Begon, Townsend & Harper, 2006). Organisms selected for a high rate of increase (r) rarely reach the asymptotic density (K), but spend most of the time on the rising portion of the logistic curve, responding quickly to the availability of environmental resources but collapsing in response to disturbance or superior competitors, for instance. K‐selected populations fluctuate near the asymptotic density for most of the time, have slower intrinsic rates of increase and use resources efficiently (thus being relatively tolerant of resource limitation).

Table 2. Criteria used to define the functional classifications (codes as in Table 1)

	Size	Shap	Stru	Func	Ecol	Habi	Taxo	TaxK	MisR
BSS/NBS	Y							None	None
TTSS	Y							None	None
PGS		Y						None	None
MBFG	Y	Y	Y					Basic	Low
rK	Y			Y	Y			None	Medium
CSR	Y	Y		Y	Y			Basic	Low
FG	Y	Y	Y	Y	Y	Y	Y	High	High
MFG	Y	Y	Y	Y			Y	Basic/High	Medium

• Size, dimensions; Shap, shape; Stru, structural characters; Func, functions (explicit use of physiological properties); Ecol, ecological attributes, including trophic preferences; Habi, habitat; Taxo, use of taxonomical criteria. TaxK indicates the level of taxonomical knowledge required to include the species in a group, while MisR indicates the risk of misplacement.

Margalef (1978) interpreted the two r and K extremes as a continuum of life history strategies that could be represented along a gradient of decreasing concentrations of nutrients and turbulence. Species that are r‐selected are small, with high surface area to volume ratios, while K‐selected species have large dimensions, either consisting of large cells or large colonies, both resistant to grazing, and often motile. The concept of r and K selection has been widely applied in phytoplankton ecology (Sommer, 1981; Reynolds, 1988a,b; Steinberg & Geller, 1993). A few modifications were proposed to accommodate species sensitive to physical mixing, opening the way to the application of the CSR classification to phytoplankton (Reynolds, 1988a).

The CSR model

Taking into account the two extremes of ‘stress’ (physical and chemical limitations) and ‘disturbance’ (e.g. grazing, diseases, wind and frost), Grime (1977) identified for terrestrial vegetation four possible permutations. For one of these (high stress and high disturbance), no strategy was possible. The three remaining combinations included the Competitive (C), Stress‐tolerant (S) and Ruderal (R) strategies (Grime, Hodgson & Hunt, 2007). Reynolds (1988a) hypothesised that a similar classification of strategies could be applied to phytoplankton. Species must be adapted to exploit environments saturated by light and nutrients (C), to develop under low‐nutrient conditions (S) or to endure turbulent transport through the light gradient (R). Species ascribed to a specific strategy were distinguishable by given morphometric physiological and metabolic features (growth rates, light harvesting, nutrient uptake, temperature optima and sinking).

Applications of the CSR classification are described in Reynolds (2006). For example, in several deep perialpine European lakes, it was possible to identify a vernal phase with R strategists (large diatoms: tolerant of vertical mixing and favoured by high nutrient availability), an early summer phase of C strategists (small flagellates; intolerant of mixing but good competitors for nutrients), a successive phase of S strategists (dinoflagellates and cyanobacteria; tolerant of low nutrients but not of vertical mixing), followed by a late‐summer mixing phase favouring R strategists diatoms and conjugatophytes. This classification was further applied by, among the others, Lindenschmidt & Chorus (1998), Elliott, Reynolds & Irish (2001), Hart (2006), Naselli‐Flores & Barone (2011), Barbosa, Barbosa & Bicudo (2013) and Naselli‐Flores (2014).

Classifications based on morphology and structure

Three morphologically and structurally based classifications (one of them subsuming three kinds of size spectra) have been proposed in the recent years (Tables 1, 2); the Biomass Size Spectrum (BSS), the Normalised Biomass Size Spectrum (NBS) and a Traditional Taxonomic Size Spectrum (TTSS); plus Phytoplankton Geometric Shapes (PGS) and Morphologically Based Functional Groups (MBFG).

Size spectra

Various size spectra (SS) have been studied both in marine and freshwater environments (Kamenir, Dubinsky & Zohary, 2004). To evaluate the BSS, phytoplankton cells were counted and measured and then distributed into geometric size classes according to their cell volume (V_i). The BSS was represented graphically, showing the distribution of biomass into classes of increasing cell volumes. Normalised BSS (NBS) were determined via normalisation of the total biomass in each V_i to the change in cell volume across the category (Platt & Denman, 1978). This way, NBS describes the mean cell density estimate (Kamenir et al., 2004). In the Traditional Taxonomic Size Spectrum (TTSS) (Kamenir, Dubinsky & Zohary, 2006), size classes were defined as in BSS, and a TTSS was created as the frequency distribution of the cumulative number of taxa (recorded during one specific period of time) in size classes based on their cell volume.

Analyses of BSS and TTSS allowed an evaluation of differences in the distribution of cell size in waterbodies of contrasting trophic state and of any changes during pronounced ecosystem shifts (Kamenir et al., 2004, 2008; Kamenir & Morabito, 2009).

Phytoplankton Geometric Shapes

Stanca, Cellamare & Basset (2013) used the geometric shapes of phytoplankton as the only criterion in studying the distribution of phytoplankton along the coast of the Salento peninsula (SE Italy). Phytoplankton species were allocated to the most similar geometric shape selected from those described by Hillebrand, Dürselen & Kirschtel (1999), Sun & Liu (2003) and Vadrucci, Cabrini & Basset (2007). At the same time, morphometric measurements (surface, volume and surface to volume ratios) were obtained from basic linear dimensions. Since no name was provided for this classification, we will refer to this approach as Phytoplankton Geometric Shapes (PGS) (E. Stanca, pers. comm.).

Stanca et al. (2013) argued that the high variability in PGS was related to morphological adaptations to the environment. Elongated shape maximises the cell surface exposed to light and was favoured (by mixing of the water column) during the winter. Rounded and combined shapes, mostly of mixotrophs, developed when the water column was thermally stable and nutrients depleted.

Morphologically Based Functional Groups

Kruk et al. (2010) classified phytoplankton into seven functional groups (MBFG) based on shape and structures. The classification was based on nine descriptors, namely volume, surface, surface to volume ratios, maximum linear dimension and the presence of mucilage, flagella, aerotopes, heterocytes and siliceous exoskeletal structures. Many functional and demographic features (which were excluded in the group definition) were differently distributed among the groups, suggesting a functional meaning in their separation and identification. The features tested included growth, sinking velocity, silicon half saturation constant for growth‐ and abundance‐related variables. The seven functional groups are characterised by a set of a priori features that allow the inclusion of new species. The MBFG classification has the advantage of simplicity and, not requiring specific knowledge about physiological traits and taxonomy, is also simple to apply in a variety of circumstances (Kruk & Segura, 2012).

Using data from 211 lakes, Kruk et al. (2011) showed that the occurrence of the various MBFG could be predicted from environmental conditions with an accuracy higher than for Functional Groups (FG, see below) and for the majority of species. Nevertheless, in a successive study of 83 lakes over a gradient from subpolar to tropical regions, Kruk et al. (2012) did not find systematic relationships between environmental gradients and phylogenetic affiliation or particular functional groups as defined by morphology.

Composite functional classifications

Besides biological traits, Functional Groups (FG) and Morpho‐Functional Groups (MFG) (Tables 1, 2) include also taxonomy and (for FG) ecology as discriminant attributes.

Functional Groups

The modern definition of Functional Groups (FG) by Reynolds et al. (2002) has its roots in the schemes, already available in the 1940s–1950s, where lakes were classified by the phytoplankton they supported (Reynolds, 1997). Using observations from a group of lakes in north‐west England, and applying traditional phytosociological methods (Braun‐Blanquet, 1964), Reynolds (1980) recognised 14 phytoplankton associations identified with alphanumeric labels (coda), each including species coexisting together and with similar seasonality. Successively, the use of ‘association’ was criticised, recognising that some species, although showing comparable adaptations and similar environmental optima, are not always found simultaneously. At present, the accepted term, FG, is intended to group together species with similar morphological and physiological traits, and with similar ecological features (Reynolds et al., 2002). While originally the groups (coda) were allocated in blocks ordered alphabetically to reflect seasonal chronological shifts in a set of temperate lakes (Reynolds, 1984), with the successive incorporation of information from lakes located at different latitudes, the alphabetical order has lost its significance. The system was therefore expanded to 31 coda accommodated based on expert judgment (Reynolds et al., 2002; Table 3). A subsequent review by Padisák, Crossetti & Naselli‐Flores (2009) recognised more than 40 coda, although not all of them yet sufficiently substantiated to be brought into the ‘final’ classification. Inclusion of new species in the FG coda requires a deep knowledge of the taxonomy and autecology of the species concerned. On the other hand, compared with the other classifications, the FG are well described in terms of habitat properties, environmental tolerance and trophic state (Reynolds, 2006; Padisák et al., 2009).

Table 3. For each functional groups (FG), the table reports the tolerances to different environmental conditions based on the data reported in Reynolds et al. (2002: their table III)

Codon (FG)	Representative species	zm	I	Temp	SRP	DIN	Si	CO2	fZoo	Trophic Code	Trophic Code
		<3	<1.5	<8	<10^−7	<10^−6	<10^−5	<10^−5	>0.4
A	Urosolenia, Cyclotella comensis	0	0.5	1	1	1	1	0	0	O	3
B	Aulacoseira subarctica, A. islandica	0	1	1	1	0	0	0	0	M	5
C	Asterionella formosa, Aul. ambigua, Stephanodiscus rotula	0	1	1	0	0	0	0.5	0	E	7
D	Synedra acus, Nitzschia spp., Stephanodiscus hantzschii	1	1	1	0	0	0	1	0	H	9
N	Tabellaria, Cosmarium, Staurodesmus	0	0	0	1	0	0.5	0	0.5	M	5
P	Fragilaria crotonensis, Aulacoseira granulata, Closterium aciculare, Staurastrum pingue	0	0	0	0	0	0.5	1	1	E	7
T	Geminella, Mougeotia, Tribonema	0	0.5	0	0.5	0	1	0.5	1	M	5
S1	Planktothrix agardhii, Limnothrix redekei, Pseudanabaena	1	1	1	0	0	1	1	1	H	9
S2	Spirulina, Arthrospira, Raphidiopsis	1	1	0	0	0	1	1	1	H	9
SN	Cylindrospermopsis, Anabaena minutissima	1	1	0	0	1	1	1	1	E	7
Z	Synechococcus, prokaryote picoplankton	1	0	1	1	1	1	0.5	0	O	3
X3	Koliella, Chrysococcus, eukaryote picoplankton	1	0	1	1	0	1	0	0	O	3
X2	Plagioselmis, Chrysochromulina	1	0	1	0.5	0	1	0.5	0	ME	6
X1	Chlorella, Ankyra, Monoraphidium	1	0	1	0	0	1	1	0	EH	8
Y	Cryptomonas	1	1	1	0	0	1	0.5	0	E	7
E	Dinobryon, Mallomonas (Synura)	1	1	1	1	0	1	0	0	OM	4
F	Colonial Chlorophytes, e.g. Chlorococcales	1	0	1	1	0	1	0	0	M	5
G	Eudorina, Volvox	1	0	1	0	0	1	1	1	E	7
Ja^a Codon J includes many Chlorococcales which are undergoing a wide taxonomical rearrangement (Krienitz & Bock, 2012).	Pediastrum, Coelastrum, Scenedesmus, Golenkinia	1	0.5	1	0	0	1	0.5	0	EH	8
K	Aphanothece, Aphanocapsa	1	0.5	0	0	0	1	1	0.5	E	7
H1	Dolichospermum flos‐aquae, Aphanizomenon/Chrysosporum	1	0	0	0	1	1	1	1	E	7
H2	Dolichospermum lemmermannii, Gloeotrichia echinulata	1	0	0	0	1	1	1	1	M	5
U	Uroglena	1	0	0.5	1	0	1	0	1	OM	4
LO	Peridinium, Woronichinia, Merismopedia	1	0	0	1	0	1	0	1	M	5
LM	Ceratium, Microcystis	1	0	0	0	0	1	1	1	E	7
M	Microcystis, Sphaerocavum	1	0	0	0	0	1	1	1	E	7
R	Planktothrix rubescens, P. mougeotii	1	1	0	0	0	1	0.5	1	M	5
V	Chromatium, Chlorobium	1	1	0	0	0	1	0	0	E	7
W1	Euglenoids, Synura, Gonium	1	1	1	0	0	1	0.5	0	H	9
W2	Bottom‐dwelling Trachelomonas	1	1	1	0	0	1	0.5	0.5	M	5
Q	Gonyostomum	–	–	–	–	–	–	–	–	–	–

• FG are identified with alphanumeric labels (coda). The original table reports the tolerance (+), no positive benefit (−) or different responses depending on the different species in a specific coda (+/−) to a set of environmental conditions. In this work, for each FG, the table has been coded into a numerical matrix substituting 0, 0.5 and 1 to −, +/− and +, respectively. Where tolerance was suspected, but not proven, occasional question marks (?) in the original table have been replaced with 0.5. Owing to the incomplete description, group Q (Gonyostomum) was excluded from the analysis. z_m, depth of the surface mixed layer (m); I, mean daily irradiance (mol photons m⁻² day⁻¹); Temp, water temperature (°C); SRP, soluble reactive phosphorus (mol L⁻¹); DIN, dissolved inorganic nitrogen (mol L⁻¹); Si, soluble reactive silicon (mol L⁻¹); CO₂, dissolved carbon dioxide (mol L⁻¹); fZoo, zooplankton grazing (proportion of the water processed daily by zooplankton). The trophic classes (ultraoligotrophy, oligotrophy, mesotrophy, eutrophy and hypertrophy and intermediate states) were coded numerically: U (1), UO (2), O (3), OM (4), M (5), ME (6), E (7), EH (8), H (9).
• ^a Codon J includes many Chlorococcales which are undergoing a wide taxonomical rearrangement (Krienitz & Bock, 2012).

Functional Groups represent the classical and the widest used system of classifying the phytoplankton. Nevertheless, relationships between the FG and their links with the main environmental constraints have never been tested quantitatively and confirmed. These points are briefly revisited in the next section.

Functional attributes of FG coda

A quantitative analysis of the relationships among the FG coda is presented here, based on their relative tolerance to different environmental conditions reported by Reynolds et al. (2002: their table III; for raw data and coding criteria see Table 3). Trophic classifications (coded from 1, ultra‐oligotrophy, to 9, hypereutrophy) were obtained from Reynolds et al. (2002: their table I), integrating material from Reynolds (1984) and Padisák et al. (2009). Functional Groups were analysed by non‐metric multidimensional scaling (NMDS) and cluster analysis (Ward's method), both applied to Euclidean distance matrices (Oksanen et al., 2013; Murtagh & Legendre, 2014). Environmental variables were related to the strongest gradients in FG composition by fitting environmental vectors to the NMDS configurations and by surface fitting (R Core Team, 2014; for R scripts see Table S1).

The cluster analysis and NMDS confirmed the close connection between some FG and their separation into five groups (Fig. 2a,b); for a description of coda see Table 3. The consistency of the data matrix used in the analysis was quite apparent in the relationships of groups 1–5 with the environmental variables (Fig. 2c). A high tolerance of low phosphorus (mostly oligotrophic environments) was contrasted with a high tolerance of low dissolved carbon dioxide and irradiance (mostly eutrophic environments). Tolerance of low water temperature (e.g. in winter) was contrasted with a tolerance of high filtration rates by zooplankton and low dissolved inorganic nitrogen (DIN, usually in early summer–autumn). The position near the origin in the NMDS suggested that tolerances of euphotic depth and low silica availability were not consistently linked to the pattern of FG in the analysis. The gradient Temperature/grazing DIN allowed the clearest separation of two broad groups of FG, that is 1–2 and 4–5, respectively, whereas group 3 takes an intermediate position along this gradient (Fig. 2b,c). The first group (1–2) includes many FG that were originally defined to accommodate diatoms and other taxonomic assemblages developing in the spring and early summer (Reynolds, 1984). In the second group (4–5), FG are composed by species developing almost exclusively in warmer and stratified conditions. These differences were substantiated by a greater tolerance of FG 4–5 to zooplankton grazing (with many large species and colonies) and low nitrogen concentrations (with all the dinitrogen‐fixing cyanobacteria in group 4). Orthogonal to this (i.e. ‘upper left to bottom right’), the gradient phosphorus–carbon dioxide/irradiance further divided groups 1 and 5 from 2 to 4 (Fig. 2b,c). This can be interpreted as a trophic gradient. This was further supported by the results of the vector and surface fitting in Fig. 2d, which show a strong linear relationship between the trophic state (nutrient availability) and functional groups. At the eutrophic extreme, FG representatives (Table 3) were cyanobacteria developing in warm epilimnia (S_N, S2) and turbid lakes (S1), purple and green sulphur bacteria (V), and small‐celled and fast‐growing diatoms (D). At the oligotrophic extreme, the diatoms were well represented with coda A and N, along with small and single‐celled cyanobacteria (Z).

Figure 2

Classification (a) and NMDS ordination (b) of the Functional Groups (FG) defined by Reynolds et al. (2002) (Table 3). The numbers 1–5 divide the main FG. The analyses were carried out taking into account the environmental tolerance, that is, depth of the surface mixed layer (z_m); mean daily irradiance (I); water temperature (Temp); soluble reactive phosphorus (SRP); dissolved inorganic nitrogen (DIN); soluble reactive silicon (Si); dissolved carbon dioxide (CO₂); zooplankton grazing (fZoo). (c) Ordination of tolerances as weighted averages of FG scores. (d) Vector and surface fitting of trophic state coded numerically from 1 (ultraoligotrophy) to 9 (hypereutrophy).

Morpho‐Functional Groups

Morpho‐Functional Groups were identified using a priori determined traits influencing functional processes and ecological characteristics (Salmaso & Padisák, 2007). Groups were classified based on the presence of flagella, the ability to obtain alternative sources of fixed carbon and nutrients, cellular organisation, dimensions, shapes, and, when ecologically relevant, taxonomy. Compared with the other classification systems, MFG do not make use exclusively of morphological/structural criteria in the definition of groups (as in MBFG), or even of phenological, habitat and trophic information (as in the FG classification). The criteria to define MFG were explicitly chosen as among the strongest drivers able to predict the best competitors under different environmental constraints (see Weithoff, 2003). Being based on an identification key, the inclusion of species in the system is quite straightforward. Conversely, the use of the classification requires, as a preliminary step, the ability to classify the species from the genus to the order. The system is flexible enough to accommodate a greater number of groups, depending on the characteristics of the habitat analysed. An update of MFG, including some new groups (e.g. Tolotti et al., 2012), is given in Table S2.

Following a similar approach, different morpho‐functional classifications were subsequently conceived for benthic diatoms (Morpho‐Functional Diatom Groups, MFDG; Centis, Tolotti & Salmaso, 2010) and river phytoplankton (Fraisse, Bormans & Lagadeuc, 2013).

Morpho‐Functional Groups were used for the first time to compare the phytoplankton in lakes Garda and Stechlin (Salmaso & Padisák, 2007) and in two reservoirs with contrasting hydrological regimes (Tolotti, Boscaini & Salmaso, 2010). Other applications of MFG are discussed below.

Applicability of functional groups in the derivation of water quality indices

The development of phytoplankton functional group systems coincided with that of the European Union's Water Framework Directive (EC Parliament & Council, 2000). Of the functional approaches discussed above, the FG, MFG and the TTSS were included in studies aiming at assessing ecological status.

The assemblage (Q) index is based on the relative share of FG coda in the total biomass, multiplied by a numerical factor (F), defined for each functional group considering the phytoplankton assemblage likely to occur in a pristine lake of the corresponding type (Padisák et al., 2006). A later version of this index was extended for the evaluation of river phytoplankton (Q_R, Borics et al., 2007). Since the F numbers are (lake) type specific and can be adjusted for different kinds of human impacts, both Q and Q_R indices are conceptually different from most other metrics proposed for assessing the ecological quality of lakes, particularly in response to nutrient enrichment, the most widespread pressure affecting lakes (Thackeray et al., 2013). The Q index provides results coherent with other phytoplankton‐based quality indices (e.g. Becker, Huszar & Crossetti, 2009; Belkinova et al., 2014; Molina‐Navarro et al., 2014) and, specifically, with the German PSI (Mischke et al., 2008), the Polish PMPL index (Pasztaleniec & Poniewozik, 2010) and the Algal Group Index (AGI; Teneva et al., 2014).

Morabito & Carvalho (2012), Lyche‐Solheim et al. (2013) and Thackeray et al. (2013) evaluated different phytoplankton metrics to assess the ecological quality of lakes in response to eutrophication (expressed as total phosphorus, TP). The Size Phytoplankton Index (SPI) and the Morpho‐Functional Group Index (MFGI) were derived from the TTSS and the MFG, respectively. Both indices showed a significant (P < 0.01) relationship with TP, but with different results and also non‐significant relationships in different European regions. A combination of SPI and MFGI in a unique index (Functional Traits Index, FTI) improved the correlation with TP (Morabito & Carvalho, 2012).

Synoptic view and critical evaluations

Comparative analyses of functional classifications

Several authors have pointed out the strengths and weaknesses of FG, MFG and MBFG both in lakes and rivers (see references in Table 4). In general, these studies showed that both FG coda and MFG were suitable tools for explaining changes in phytoplankton assemblages in relation to major environmental drivers. These two approaches often produce similar (overlapping) results. However, since FG coda is associated with well‐described environmental templates, they are generally acknowledged as being more helpful in explaining phytoplankton variability in relation to environmental factors. Classification of MBFG, based on seven groups, is closer to the diatom ecological guild approach (three groups; Passy, 2007) than to either FG or MFG. Morphologically Based Functional Groups can explain large‐scale variations (Abonyi et al., 2014) and therefore are suitable for analysing large, ecoregional data sets (Izaguirre et al., 2012; Hu, Han & Naselli‐Flores, 2013; Žutinić et al., 2014). At finer regional and temporal scales, functional groupings (either FG or MFG) apparently perform better (Abonyi et al., 2014).

Table 4. Summary of analyses comparing different phytoplankton functional classifications

Origin/site	Data set	Functional classifications compared	Statistical methods	Reference
Large floodplain rivers: (Mura, Drava, Danube and Sava) (Croatia)	Spatial and temporal, 24 samples	FG, MBFG	Canonical Correspondence Analysis (CCA), Self‐organising Maps (SOM)	Stanković et al. (2012)
87 Andalusian lakes and ponds (S‐Spain)	Spatial, 87 samples	FG, MBFG	Pearson correlations, Generalised Linear Models (GLM)	Gallego et al. (2012)
Three Pampa lakes, three sites per lake (Argentina)	Spatial and temporal, 72 samples	FG, MFG, MBFG	Redundancy Analysis (RDA), Detrended Correspondence Analysis (DCA)	Izaguirre et al. (2012)
Three small reservoirs (S‐China)	Spatial and temporal, 18 samples	FG, MFG, MBFG	CCA	Hu et al. (2013)
River Loire (France)	Spatial and temporal, 170 samples	FG, MFG, MBFG	SOM	Abonyi et al. (2014)
A lateral channel of the Upper Paraná floodplain (Brazil)	Temporal, 49 samples	FG, MBFG	PCA, CCA, Indicator Value Analysis (IndVal)	Bortolini et al. (2014)
Two deep karstic lakes (Plitvice NP, Croatia)	Temporal, 384 samples	FG, MFG, MBFG	Principal Components Analysis (PCA), CCA	Žutinić et al. (2014)

A critical evaluation of functional approaches

Classifications founded on the concept of life history traits (r/K and CSR) have limited applicability in the study of phytoplankton. As stated by Roff (1992), ‘attempts to transfer the concept to actual populations without regard to the realities of the complexities in life history have probably been detrimental rather than helpful’ (see also Ricklefs and Miller, 2000). The r/K concept set the stage for the definition of successive approaches, such as the CSR classification. Nevertheless, as in the r/K approach, the CSR classification is more of conceptual value, highlighting the importance of the strong link between size and shape, and functional properties. The classification of phytoplankton into three CSR classes does provide only a very limited set of attributes to study phytoplankton life history traits.

In the morphologically based classifications (BSS/NBS, TTSS, PGS and MBFG; Table 2), the presence of similar structures and/or sizes/shapes in phylogenetically distantly related species can be interpreted as a set of common analogous traits under strong natural selection. Although morphology and structure have implicitly functional roles, most (MBFG) or all (BSS/NBS, TTSS, PGS) physiological complexities are not taken into account. Characters such as (among others) pigment composition and photosynthetic efficiency are vital characteristics that cannot be predicted and modelled by size and shape. With these approaches, the common possession of silica walls in the large Aulacoseira and small (<5 μm) Cyclotella can identify the two genera as functionally equivalent, although they differ in their sinking rate in stable water columns (Winder, Reuter & Schladow, 2009). Similarly, large mucilaginous colonies share many related characters, such as the resistance to grazing and reduced susceptibility to sinking. However, no one can deny the differences between the large Microcystis colonies, which can move upwards by several metres per day, and the large colonial, non‐motile Chlorococcales s.l. In a few occasions, attempts were made to investigate the reliability of size‐based classifications in the trophic evaluation of waterbodies. However, in the case of trophic indices based on size classes (SPI and MFGI), better relationships with TP were obtained using other classical metrics, based on chlorophyll‐a, the biovolume of cyanobacteria and species composition (Thackeray et al., 2013). More generally, the low discriminatory power of classifications based on a limited number of groups has been highlighted by Izaguirre et al. (2012), Stanković et al. (2012) and Žutinić et al. (2014).

Functional Groups and MFG make explicit use of functional properties in the delineation of groups of species. There are advantages in this, due to the recognition of specific ecological capabilities otherwise not distinguishable on a structural basis (e.g. mixotrophy, light optimum requirements); nevertheless, there is a high level of subjectivity in the approach.

An advantage of FG is that ecological features are linked with the trophic state or habitat preferences (Fig. 2; Table 2). Unlike any of the other classifications, species with very similar morphological characteristics but distinct environmental tolerances, such as Planktothrix agardhii and P. rubescens, are clearly separated into two functional groups, namely S1 and R, respectively. Similar considerations apply to other groups of species, for example small Cyclotella spp. or Aulacoseira spp.. On the other hand, the low number of representative species in each of the different FG forces investigators to ‘guess’ the inclusion of new species not yet assigned into a well‐defined group. This issue was addressed by Padisák et al. (2009), where a number of misplacements were identified. A serious risk for FG is the blurring of differences in ecological tolerance between the groups, due to the addition of further species to the existing coda. This classification must not be used when the ecological preferences of species are insufficiently known. The clear ecological delineation of FG coda was a prerequisite for the derivation of the assemblage Q and Q_R indices used in the evaluation of ecological status. However, the large degree of subjectivity in the choice of the factor number F poses serious limits to the possibility to generalising this approach, with applications limited to a case‐by‐case evaluation.

Contrary to the FG classification, MFG does not have a clear habitat characterisation, and investigations in this direction have begun only recently (e.g. Izaguirre et al., 2012; Salmaso, Naselli‐Flores & Padisák, 2012; Gallina et al., 2013; Hu et al., 2013; Mihaljević et al., 2013; Thackeray et al., 2013). Since species can be accommodated in several functional groups, MFG classifications can be efficiently used to overcome the problems related to differences in taxonomic accuracy and species identification in different ecosystems (e.g. Tolotti et al., 2010).

The relationship between the functional groups is summarised in Fig. 3. The specificity of classifications purely based on size or shapes is apparent in the separation of PGS and BSS/NBS/TTSS. At a lower dissimilarity, r/K and CSR are closely connected, forming a separate group. The use of diversified structural attributes put the MBFG nearer to the FG and MFG.

Figure 3

Relationships between the functional groups based on the distinctive binary characteristics reported in Table 2, namely dimensions, shape, structural characters, functions, ecological attributes, habitat and taxonomy. Distances are based on the Jaccard dissimilarity, whereas the cluster analysis was performed with the Ward's method (R Core Team, 2014).

Potentials and weaknesses of the functional approach: perspectives for the progress in phytoplankton ecology

Functional classifications allow comparisons between ecosystems around the world. Distant lakes appear different based on species composition, but they could share a phytoplankton with similar functional characteristics. The identification of common traits under a range of environmental conditions (e.g. differing in the proportion of land use, nutrient inputs, grazing and climate) can improve our ability to generalise results, finding common patterns useful also in predicting shifts in the phytoplankton under climate‐ and environmental‐change scenarios. Nevertheless, functional groups are not meant to be a substitute for the whole extent of information that can be gathered from species. The knowledge of which species dominate a functional group is of primary importance when information on conservation, trophic roles, toxicity or other characters pertaining to populations or strains are essential in addressing particular ecological questions or environmental issues.

The definition of functional groups requires categorising similarities in biological and ecological traits. At a broader scale (e.g. the ecosystem), such an approach has fostered important conceptualisations regarding the functioning of trophic webs and ecosystem energetics (Weisse et al., 1990). At a finer scale, the numerous criteria and approaches that have been proposed to group adaptive phytoplankton traits may also reflect the lack of unifying concepts. Environmental drivers act at every level of the biological complexity, from single traits to communities. Therefore, the excessive reduction of traits can affect negatively the sensitivity and efficiency of classifications to explain the observed species distributions and changes.

The misuse of functional classifications outside a specific range of applications can have serious consequences in the interpretation of ecological processes. When using functional groups, we should take into account the limits and uncertainty implicit in the conceptualisation of ecological redundancy (Naeem, 1998). Isbell et al. (2011) argued that even more species would be needed to maintain ecosystem processes and services than suggested by previous studies. If only one or a few processes are considered, many species appear redundant within a specific set of environmental constrains.

Future progress will necessarily be founded on the delineation of functional traits defined with a broader and stronger theoretical framework. It is highly unlikely that the continuous application of functional classifications or selected traits would open the way to strong generalisations, that is in a ‘let's apply and see what happens’ approach. Adopting a deductive method (Ritchie, 2010), the robustness of functional traits and classifications should be tested experimentally, based on clearly defined hypotheses addressing the power of traits and their mutual relationships. Simple examples include testing the increase of slow‐sinking functional groups (or of selected traits, e.g. the fraction of gas‐vacuolated cyanobacteria) in experimental setups and warming lakes (cf. Winder & Sommer, 2012), and testing whether physiological traits of phytoplankton can explain how species respond to environmental gradients (e.g. light and phosphorus; Edwards, Litchman & Klausmeier, 2013). Quantification of the links between species traits and environment should make use of robust statistical analyses, including methods specifically devised to test the species traits–environment relationships (e.g. the ‘fourth‐corner’ method; Dray & Dufour, 2007; Dray & Legendre, 2008). On the other hand, following an inductive approach, refinement of functional classifications will require the study of the relationships between functional traits and environmental drivers. Important outcomes should include the identification of common patterns of change along environmental gradients.

Functional approaches have been based on discernible biological traits, integrated with phenology, ecology and taxonomy (Table 2). In this respect, more accurate phylogenetic analyses should be assessed for their potential to contribute synergistically to trait‐based approaches (see Westoby, 2006; Kraft et al., 2007; Cavender‐Bares et al., 2009; Litchman et al., 2010; Vamosi, 2014). At the finer taxonomic levels, and considering the rapid progress in both molecular genetics and ecological metabolomics, future directions should also take into account ‘cryptic adaptive traits’. Examples include the ability to produce a variety of toxins or to withstand hydrostatic pressure gradients through the synthesis of gas vesicles of different strengths in different strains of cyanobacteria (D'Alelio et al., 2011; Kurmayer et al., 2011; Salmaso et al., 2013); the ability to exploit various light intensities and nitrogen compounds in different genotypes of Prochlorococcus (Moore et al., 2002); the increasing presence of mycosporine‐like amino acids (MAAs) and enhanced absorption of ultraviolet radiation of phytoplankton in high‐altitude lakes (Ficek, Dera & Woźniak, 2013). As a cautionary note, the existence of ecotypes with different physiological adaptations (see also Rohrlack et al., 2008; Zapomělová et al., 2010; Üveges et al., 2012; D'Alelio, Salmaso & Gandolfi, 2013) should be taken into account in the evaluation of the limits implicit in the use and interpretation of functional approaches based on easily measurable traits.

In conclusion, the various functional classifications available represent a first step towards the use of tools integrating phytoplankton ‘functions’. Besides classical traits, an important future challenge will be to integrate, together with the advances in molecular genetics, metabolomics and physiology, our growing knowledge of phytoplankton taxa in the definition of functional classifications.

Acknowledgments

The first ideas of this manuscript were presented at the 8th SEFS (Symposium for European Freshwater Sciences), Münster (Germany), July 1–5, 2013. Partial support was provided by the Hungarian National Science Foundation (OTKA‐K75552) and by the University of Palermo (2012‐ATE‐0148). We acknowledge Prof. Alan Hildrew and three reviewers for their constructive advice to improve the manuscript.