How important is climate? Effects of warming, nutrient addition and fish on phytoplankton in shallow lake microcosms

Authors


Brian Moss, School of Biological Sciences, Derby Building, University of Liverpool, Liverpool L69 3GS, UK (fax 0151 7945017; e-mail brmoss@liverpool.ac.uk).

Summary

  • 1Climate is changing. Predictions are for at least a 3 °C rise in mean temperature in northern Europe over the next century. Existing severe impacts of nutrients and inappropriate fish stocking in freshwater systems remain.
  • 2Effects of warming by 3 °C above ambient, nutrient addition and the presence or absence of sticklebacks Gasterosteus aculeatus were studied in experimental microcosms dominated by submerged plants, mimicking shallow lake ecosystems.
  • 3Warming had considerably smaller effects on the phytoplankton community than did fish and nutrients. It had very minor effects on chlorophyll a and total phytoplankton biovolume. However, it significantly decreased the biovolumes of Cryptophyceae (a major component in the controls) and Dinophyceae. Contrary to expectation, warming did not increase the abundance of blue-green algae (cyanophytes). Warming decreased the abundances of Cryptomonas erosa (Cryptophyceae) and Oocystis pusilla (Chlorophycota) and increased those of two other green algae, Tetraedron minimum and Micractinium pusillum. It had no effect on a further 17 species that were predominant in a community of about 90 species.
  • 4Fish and nutrients, either together or separately, generally increased the crops of most of the 21 abundant species and of the algal groups. Exceptions were for diatoms and chrysophytes, which were very minor components of the communities. Fish, but neither nutrients nor warming, increased the number of species of phytoplankton detected. This was probably through removal of zooplankton grazers, and parallels terrestrial studies where the presence of top predators, by controlling herbivores, leads to increased plant diversity.
  • 5There was no particular pattern in the taxonomy or biological characteristics of those species affected by the treatments. In particular, there was no link between organism size (a surrogate for many important biological features of phytoplankton species) and the effects of warming, nutrient addition or presence or absence of fish. However, all species were relatively small and potentially vulnerable to grazing.
  • 6Synthesis and applications. The results suggest that fears of an increasing abundance of cyanophytes with current projections of global warming may be unrealized, at least in shallow unstratified lakes still dominated by macrophytes. However, they emphasize that eutrophication and fish manipulations remain very important impact factors that determine the abundance of phytoplankton and subsequent problems caused by large growths.

Introduction

Earth is steadily warming and predictions for the UK and much of Europe suggest a rise in temperature of 3–5 °C in the next half century (Houghton et al. 2001). Temperature increase, although only one of several related climate changes, including rainfall intensity and timing and evaporation, is the key variable of change and one that pervades almost all biological processes. All ecosystems will therefore be affected to some extent by global warming and there is much interest in predicting what its effects might be.

Because of their importance in natural economies, shallow wetland systems (Dugan 1994) are particularly important and already vulnerable. Shallow lakes, once dominated by submerged plants (macrophytes) in clear water, have often been converted to turbid, algal-dominated lakes by a variety of food web-linked mechanisms in an environment of increased nutrient loading (Moss, Madgwick & Phillips 1996) despite an array of mechanisms that to some extent can stabilize the plant-dominated state (Moss 1999). These mechanisms include nutrient competition between plants and algae and provision of refuges for invertebrate grazers on periphyton or phytoplankton.

A variety of switch mechanisms (Moss, Madgwick & Phillips 1996; Moss 1999), either directly damaging the plants (vertebrate grazing, mechanical damage, herbicide run-off or treatment) or indirectly influencing the effectiveness of invertebrate grazers (changes in fish community, toxins, increased salinity), may allow algal populations to suppress plant growth and become dominant. The thresholds at which these switches may act have not been precisely quantified but appear to be influenced by nutrient availability (Jeppesen et al. 1997). The greater the nutrient load, the more likely it is that a switch to phytoplankton dominance will take effect. Because of the pervasive influence of temperature on biological processes, it seems possible that warming might add to the list of switches and increase the risk of destabilization of submerged plant communities. This was our first hypothesis.

Phytoplankton ecology per se provides a valuable model for investigation of changing environments, with results possibly applicable to communities of larger organisms that are experimentally not so tractable over short periods (Reynolds 1997). Over just a few weeks phytoplankters undergo many generations, giving ample opportunity for the selection of forms more appropriately fitted to changed environments. Annually there is a periodicity of species that has reasonable predictability, at least in terms of overall groups of algae or life strategies (Hutchinson 1967b). Diatoms (Bacillariophyceae), for example, tend to be abundant in the well-mixed periods of spring and autumn in temperate lakes (Lund 1964, 1965); chrysophytes often follow in the short period, when nutrients depleted by diatoms have not yet been replaced by internal regeneration or new supplies from the catchment. Small ‘weedy’ species, often small green flagellates (Chlorophyta) or cryptomonads (Cryptophyceae), have irregular cycles controlled by zooplankton grazing, itself linked to fish predation. In turn fish recruitment depends crucially on the weather during egg development and zooplankton abundance during maturation of the fry and young-of-the year. In mid-summer, when previous intense biological activity has increased the chemical complexity of the water, a mixed community of green algae and dinoflagellates (Dinophyceae), often including mixotrophic species, tends to occur. Then, in late summer, larger forms of blue-green algae (Cyanophyta) and dinoflagellates may predominate, perhaps favoured by their abilities to obtain scarce nutrients from the bottom waters through vertical migration. They may be helped by removal of competition from smaller forms kept scarce by grazers. Along with late-appearing green algae, they may also simply divide less frequently and hence take more time to attain prominence (Lund 1971).

There is always a much larger pool of species available than the array that grows most abundantly in a given year (Hutchinson 1967a; Chorus & Schlag 1993). Weather can alter water temperature greatly in different years and different species have different temperature ranges and optima for growth (Moss 1973b; Seip, Sneek & Snipen 1994; Seip & Reynolds 1995), as they have for other environmental factors (Moss 1973a; Reynolds, Irish & Elliott 2001, Reynolds et al. 2002). Substitution of alternative species could thus maintain overall phytoplankton community function despite much warming. Because phytoplankton crops also appear to be controlled extensively by nutrient availability and grazer impact, we hypothesize that there might be no effect of warming on overall community size but some change in species composition.

Phytoplankton communities that develop when submerged plant communities are displaced from shallow lakes vary greatly, depending on the balance of nitrogen and phosphorus availability, the retention time of the water mass and the degree of mixing. Diatoms tend to be favoured by greater mixing (Moss & Balls 1989; Reynolds 1997), green algae by very high nutrient concentrations (Jensen et al. 1994) and cyanophytes by moderately high phosphorus levels, a scarcity of free carbon dioxide, stability of the water mass and long retention time (Shapiro 1990). However, there are two general trends in community composition that might be expected with warming. First, environmental impacts should favour pioneer or weedy species over those of more stable habitats. Very small species tend to have irregular annual cycles. They grow rapidly but are consumed easily by grazers. They are the equivalents of r-selected pioneer species with high birth and death rates, and are opportunistic and generalist in their habitat requirements. Larger species tend to have more specific needs and a more complexly defined habitat. Therefore, we predict that warming trends along with other experimental treatments will have fewer influences on smaller species and that there will be more complex effects on larger ones. The nature of the community is, of course, important in the provision of domestic drinking water supplies. Small species pose greater filtration problems.

Secondly, we expect a greater representation of cyanophytes with warming. Cyanophytes evolved in the Precambrian period when temperatures were much higher than at present (Falkowski 2002) and, although widespread in the oceans and freshwaters, even in Antarctica, they show particular association with warm habitats. They tend to be most abundant in late summer in temperate lakes (Lund 1965; Hutchinson 1967b; Moss & Balls 1989) and frequently dominate tropical lakes (Talling 1986), desert soil crusts and hot springs (Brock 1967, 1978). Their temperature optima for growth tend to be higher than those of green algae and much higher than those of diatoms (Brock 1967; Reynolds, Irish & Elliott 2001).

In freshwaters, planktonic cyanophytes are associated with toxin problems and in recent years many freshwaters have been closed to public recreation because of the risk of cyanophyte blooms (Codd 1995). Multiple deaths occurred in a dialysis unit in Brazil as a result of use of water in which cyanophytes had grown (Pouria et al. 1998), despite their having been filtered out. In drinking water supplies they are also problematic because of odour and filtration problems (Uhlmann 1979). Global warming may therefore bring a greater incidence of cyanophytes and their associated problems in freshwaters. In shallow lakes where plants have been lost, circumstances leading to cyanophyte dominance also make restoration of the plant communities difficult because cyanophytes are often uningestible, indigestible or nutritionally poor for invertebrate grazers (Arnold 1971; de Bernardi & Giussano 1990).

Our overall prediction therefore is that increased warming will bring a greater risk of collapse of submerged plant communities in shallow lakes and that there will be a switch towards phytoplankton communities increasingly dominated by cyanophytes; that overall production will be unchanged by warming trends; that there will be some change in community species composition; and that there will be fewer effects on small species than on larger ones.

Although several lines of observational and modelling evidence provide pointers to eventual warming effects (Carpenter et al. 1992; Anderson, Robertson & Magnusson 1996; Melack et al. 1997; Hostetler & Small 1999; George 2002), long-term experiments at the whole-ecosystem scale perhaps provide the most reliable predictors (Schindler 1998). Such experiments have considerable logistic problems and replication of treatments may not be possible. A compromise, therefore, is the use of replicated microcosms, each containing a simplified, although still structurally complex, version of the whole ecosystem. Although microcosms cannot contain a complete food web (Carpenter 1996), they offer substantial increases in reality compared with laboratory systems (Petchey et al. 1999; Fox & Morin 2001). Experiments at an ecosystem scale to examine the effects of warming on freshwater communities are as yet very rare.

To test our predictions we created 48 experimental, open-air microcosms (McKee et al. 2000) that could be heated to specified temperatures above ambient. We stocked them with a standard community resembling those of shallow lakes, with sediment, submerged plants, associated invertebrates and a plankton community. Because warming trends are occurring against a background of other major impacts, including eutrophication (Moss 1996) and changes in fish communities (Carvalho & Moss 1995; Williams, Moss & Eaton 2002; Williams & Moss 2003), we incorporated nutrient and fish treatments combined with warming treatments. We then examined the microcosm phytoplankton communities that resulted. Effects on water chemistry, zooplankton and plants have been reported elsewhere (McKee et al. 2002a,b, 2003).

Methods

microcosm construction

The microcosms were circular fibre-glass tanks, 1·02 m deep and 2 m in diameter, containing 3200 l of water and sediment. For insulation they were sunk into the ground on land in the University of Liverpool Botanic Gardens at Ness on the Wirral Peninsula, UK (53°16′N, 3°02′W). Each of the 48 tanks contained either a dummy heating element or a functioning element placed in its bottom. Hot water was circulated through the functional elements from a boiler at rates determined by a control sensor and system positioned in each tank. The control system allowed the tank water to be maintained at a predetermined temperature above the ambient sensed in the nearest unheated control tank. The control system operated with a sensitivity of +0·25 °C and functioned without any interruption or breakdown for 24 months. Evaporative losses in summer were replaced with deionized water generated on site. Full details of the system, specifications for the components and evidence of the precision of temperature control are given in McKee et al. (2000).

Before the experiment began, each tank was provided with a standard sediment (5 cm depth, 7 : 1 sand : loam) and filled with water from a bore hole installed on site (pH 7·3, alkalinity 3·4 mequiv l−1, Mg2+ 38 mg l−1, Ca 103 mg l−1). Plankton was introduced as a net-concentrated inoculum from a local canal. Macrophytes were introduced as standard inocula of equal weights of plants from local sources held in weighted netting ‘sandwiches’ placed randomly on the bottom. Potamogeton natans L., Elodea nuttallii Planch. H. St John and Lagarosiphon major Ridl. Moss were used, with a deliberate choice of one native species, one naturalized species and one recent invader, respectively. Invertebrates were inevitably introduced with the plants and many others flew in. Invertebrate and plankton communities were regularly cross-mixed during the summer of 1998, before the experiment began in September 1998, to ensure uniform presence. By the beginning of the experiment visually uniform, well-established plant and invertebrate communities were present in each tank.

experimental design

Three temperature treatments, two nutrient treatments and two fish treatments (presence or absence) were used in a randomized block design with quadruple replication of each of the 12 treatment combinations from September 1998 until September 2000. Temperature treatments were: control (C) (no heating); heating to 3 °C above ambient year around (H); heating to 3 °C above ambient during summer, defined as April to September inclusive (h). These regimes were based on scenarios predicted for north-west Europe for the mid-twenty-first century by the International Panel on Climate Change (Houghton et al. 1996). Nutrient treatments included a control in which no additional nutrients were added and a treatment in which sodium nitrate and sodium phosphate were added every 3 weeks in winter and every 2 weeks in summer, to give instantaneous additional concentrations of 500 mg l−1 N and 50 mg l−1 P in winter, and 170 mg l−1 N and 17 mg l−1 P in summer. These values were determined following inspection of loading rates of local eutrophicated shallow lakes. Fish treatments included absence of fish and addition at the beginning of the experiment of 21 adult sticklebacks Gasterosteus aculeatus L. of both sexes. The fish freely bred and the final mean biomass per relevant microcosm was 27 + 18·2 g m−2.

phytoplankton methods

Phytoplankton samples were taken with a tube from the entire water column samples from each microcosm every 3 weeks in winter and every 2 weeks in summer, mixed and preserved with Lugol's iodine solution. A total of 2064 samples was subsequently examined and all the component species counted following sedimentation in counting chambers and examination with a Leitz inverted microscope at ×400 magnification. Biovolumes were obtained by measurement of a random sample of 50 individuals of each organism and, for calculation, approximation of the species’ shape to the most appropriate geometric solid. Chlorophyll a was determined spectrophotometrically after filtration through glass-fibre filters and extraction with 90% acetone. For statistical analysis mean values were calculated, log-transformed where appropriate and submitted to repeated-measures analysis of variance, with time as the repeated-measures variable and warming, fish and nutrients as factors. For some comparisons, where seasonality was examined, means were calculated for the two winters (W1 and W2) and the two summers (S1 and S2) of the experiment. Significant results from the analyses of variance were subjected to Tukey honest significant difference tests to reveal the specific effects of treatments.

Results

overall community

Eighty-nine photosynthetic planktonic taxa were recorded in the samples, together with a further 24 diatom species sporadically displaced from the surfaces of the plants by sampling or wind disturbance (Table 1). Thirty-six of the 89 taxa were flagellates; many of the remainder were small coccoid or colonial forms. There were few large colonies, all of them flagellate or, in the cases of Botryococcus braunii Kutz. and Gomphosphaeria sp., buoyant through oil storage and gas vesicles, respectively. Overall the community was that expected from weedy, shallow waters where turbulence is minimal and motility and the low sinking rates of small forms have selective advantage. Planktonic diatoms, which have relatively high sinking rates, were very infrequent.

Table 1.  Composition of the phytoplankton community in a set of experimental microcosms
DivisionNumber (%) of taxaMajor genera
Chromophyta21 (23·6) 
Cryptophyceae 4 (4·5)Cryptomonas, Rhodomonas
Chrysophyceae 8 (9·0)Mallomonas, Chrysococcus, Synura, Chrysochromulina, Dinobryon
Bacillariophyceae 6 (6·7)Aulacoseira, Nitzschia, Cyclotella, Asterionella, Stephanodiscus
Dinophyceae 3 (3·4)Peredinium, Gymnodinium
Chlorophyta46 (51·7) 
Conjugatophyceae 6 (6·7)Cosmarium, Staurastrum, Closterium
 
Chlorophyceae (Volvocales)11 (12·4)Chlamydomonas, Gonium, Haematococcus, Pyramimonas, Pandorina, Eudorina, Chlorogonium, Carteria
Chlorophyceae (Chlorococcales)29 (32·6)Scenedesmus, Crucigenia, Oocystis, Tetraedron, Schroederia, Ankyra, Micractinium, Pediastrum, Sphaerocystis, Golenkinia, Monoraphidium, Ankistrodesmus, Closteriopsis, Dictyosphaerium, Botryococcus, Asterococcus, Lagerheimia, Pseudococcomyxa
Euglenophyta10 (11·2) 
Euglenophyceae10 (11·2)Euglena, Phacus, Lepocinclis
Cyanophyta10 (11·2) 
Chroococcaceae 4 (4·5)Dactylococcopsis, Chroococcus, Merismopaedia, Aphanothece
Oscillatoriaceae 4 (4·5)Oscillatoria, Phormidium
Nostocaceae 2 (2·2)Anabaena, Gomphosphaeria
Anoxyphotobacteria 2 (2·2)Lampropedia

The phytoplankton community developed in abundance during the first year (Table 2). Most algal groups and species showed changes between seasons with increases in biovolume, particularly during the second summer. This did not reflect availability because all abundant species were recorded within the first year. It may rather reflect accumulation of nutrients in the system with continued nutrient loading. Cryptophyceae were predominant in the first winter and remained a major component in the second winter. Chlorophyta became predominant in the first summer and remained predominant thereafter. Other than Euglenophyta in the second summer, other algal groups, including Cyanophyta, accounted for very small relative (and absolute) contributions to the total biovolume.

Table 2.  Changes in mean seasonal biovolume of the major algal groups (µm3 ml−1 × 10−3). The standard deviation (SD) and percentage composition by season (%) are also given
Algal groupFirst winterFirst summerSecond winterSecond summer
MenSD%MeanSD%MeanSD%MeanSD%
Bacillariophyceae8·79·5 0·4169362 2·118·427 0·2760·688·6 0·5
Chlorophyta30167613·2748013 20092·34800920069·5865011 40064·3
Chrysophyceae19·530·4 0·90·551·6 0·012636 0·4120111 0·9
Cryptophyceae1950365085·5418567 5·220001630292970232022·1
Cyanophyta0·931·9 0·041·231·62 0·021·944·1 0·03103238 0·8
Dinophyceae00 026·443 0·335190 0·7255298 1·9
Euglenophyta0·030·1 0·0011236·8 0·158·726 0·113002390 9·7

chlorophyllaand biovolume of major groups

Phytoplankton chlorophyll a concentrations were generally low (overall mean per microcosm = 14 µg l−1, SD = 25, n= 192). There were significant increases in chlorophyll a and total algal biovolume with either fish presence or nutrient addition, but almost no significant effects of warming (Table 3). The major interaction effect was between fish and nutrients, with the combined effects of the two treatments significantly increasing chlorophyll a or total biovolume over the values obtained with the individual treatments.

Table 3.  Statistical significance of treatment effects as elucidated by repeated-measures analysis of variance of chlorophyll a concentrations, total phytoplankton biovolume and biovolumes of the major component groups of the phytoplankton. For each variable, the analyses used log-transformed mean values calculated for seasons when these groups were present (W, winter; S, summer; 1, first year; 2, second year). Significance is indicated as: NS, not significant, *P < 0·05, **P < 0·01, ***P < 0·001
TreatmentChlorophyll aTotal biovolumeBacillariophyceaeChlorophytaChrysophyceaeCryptophyceaeCyanophytaDinophyceaeEuglenophyta
Warming (W)NSNSNSNSNS*NS*NS
Fish (F)******NS***NS***********
Nutrients (N)******NS**NS*******NS
Season (S)******NS***NS*************
W × FNSNSNSNSNSNSNSNSNS
W × NNSNSNSNSNSNSNSNSNS
F × N******NS**NS***NSNSNS
W × SNSNSNSNSNSNSNSNSNS
F × S*NSNSNSNS**NSNSNS
N × S***NSNSNSNSNSNSNS
W × F × NNSNSNSNSNSNSNS**NS
W × F × S*NSNSNSNSNSNSNSNS
W × N × SNSNSNSNSNSNSNSNSNS
F × N × SNSNSNSNSNS*NSNSNS
W × F × N × SNSNSNSNSNSNS**NS
 W1, S1, W2, S2W1, S1, W2, S2W1, S1, W2, S2W1, S1, W2, S2W1, W2, S2W1, S1,W2, S2W1, S1, W2, S2S1, W2, S2S1, W2, S2

There was a significant warming × fish treatment × season interaction in the repeated-measures anova suggesting statistically significant variation overall in chlorophyll concentrations dependent on warming, although the effect was ecologically only minor. It was conditioned particularly by results from the second winter, when chlorophyll concentrations in the continuously warmed microcosms containing fish were unusually high. Within-seasons analyses yielded significant warming treatment effects only during this second winter (McKee et al. 2003). Warming significantly reduced (Table 4) the biovolume of the Cryptophyceae, a group that was proportionately more abundant in the winters than in the summers. The biovolume of Dinophyceae was also reduced by warming although the absolute change was very small. Warming had no effect on the biovolume of the predominant Chlorophyta or on any of the remaining groups (Tables 3 and 4).

Table 4.  Mean seasonal values for biovolume (µm3 ml−1 × 10−6) of the algal groups in the phytoplankton in a microcosm experiment investigating the effects of warming (C, ambient temperature; h, summer warming; H, all-year warming), fish (F) and nutrient (N) treatments. W, winter; S, summer; 1, first year; 2, second year. Values in parentheses are the appropriate mean square error estimates (from the repeated-measures analyses of variance, data log (x + 1) transformed) for attaching back-calculated error intervals to each mean. A dash indicates that the group was not detected in that season
  CFNCFCNChFNhFhNhHFNHFHNH
Bacillariophyceae(5·102)W1211·21·32·313·23·8313·80·512·73·8330·10·67
S17·15·738·24209065·530·6126032·768·2043·6
W215·62·386·70·09448·560·18805·060·7519·835·2
S258·82921·252·8116896078·3412·698·19·946·25
Chlorophyta(0·741)W166·226·3354·944941957167·40239825538·425·5
S112 200364151778046 100451552126516 50024536121657
W21280102213090·328 000544508035919 80057626·9180
S210 70028603587133521 1004750922066540 00058406653190
Chrysophyceae(6·580)W10·071020·193330·03235·60·35714·50·01741·44·971·15
S1
W256·618·54·0126019·312·135·8022·5017·4
S217033364·430·36·2512727·4186023218·5239
Cryptophyceae(0·228)W16390156154175416065·110478·011 600149120248
S11689118197430207108387123152077·674·692·0
W23730724232217305190616146557646609421262824
S2681098549722467701904313113541760109017402900
Cyanophyta(1·705)W10·30·0220·310·0120·7710·6716·820·0140·1481·750·3680·022
S10·590·580·060·1361·822124·1504·600·450·0070·292
W20·850·690·032·494·260·15300·4214·30·1270·0130
S215·414047·66·1284776·54·225·938·93·023·7723·8
Dinophyceae(2·937)W1
S1012700·245080·80·245072·932·702·2
W22·032541·353·389·9637·802·3622944·710·617·6
S21808952·255·6338657318·910·15014471·1346·8
EuglenophytaW1
(2·602)S11·530·29003·770·573002·461295·730·82
W200·9600·162·10·78800·3990·99·0800
S268378334·211·8185060247060·52690842012·333·0

Nutrient addition significantly increased the biovolumes of Chlorophyta, Cryptophyceae and Cyanophyta and significantly reduced that of the dinoflagellates. It had no significant effect on the scarce diatoms, chrysophytes and euglenophytes (Tables 3 and 4). Presence of fish had more widespread effects, with increases in all groups showing significant change. There were no effects on diatoms or chrysophytes. Interaction effects (Table 3) were mostly associated with the mutually supportive combination of fish and nutrients in increasing the biovolumes.

effects on individual species

The mean number of species recorded from each microcosm over the entire course of the experiment was 32·6 (SD = 4·3, n= 48, range 22–48). To investigate diversity further, we calculated Hill's numbers (Ludwig & Reynolds 1988): number of species recorded (N0), number of abundant species (N1) and number of very abundant species (N2). These values (untransformed) were then subjected to analysis of variance. Only the fish treatment significantly changed N0, more species being recorded where fish were present (Table 5). There were no two- or three-factor interaction effects on N0. Nor were there any main or two-factor effects on N1 and N2, although there were marginally significant three-factor interaction effects on both (Table 5). The probable cause of this was the unusually high N1 and N2 values obtained from the summer-only warmed microcosms containing fish but not receiving additional nutrient input. However, the effects were not significant according to subsequent Tukey honest significant difference testing.

Table 5.  Effects of warming, fish and nutrient treatments on mean phytoplankton diversity. The first part of the table shows Hill's numbers: number of species recorded (N0), number of abundant species (N1) and number of very abundant species (N2). The second part of the table shows results of analysis of variance of these indices. C, ambient temperature; h, summer warming; H, all-year warming. Mean number of species in each treatment was calculated for each microcosm so that there were four replicates per value calculated. anovas (n = 16 (warming) or 24) were carried out without transformation. Probabilities below 0·05 are given in bold type
(a) Mean species richness per treatment
WarmingFishNutrientsN0N1N2
MeanSDMeanSDMeanSD
C++32·85·16·31·14·41·3
C+36·51·05·32·63·31·8
C+28·85·94·70·73·60·7
C28·56·26·20·84·80·6
h++41·55·84·40·32·80·5
h+34·52·58·72·95·92·2
h+27·53·15·91·74·41·0
h28·33·64·91·23·61·3
H++39·84·25·11·23·41·1
H+36·04·25·53·23·72·1
H+26·33·34·51·23·30·9
H31·53·44·70·93·51·0
(b) Results of anova
EffectN0N1N2
FPFPFP
Warming (W) 0·72  0·501·350·281·330·28
Fish (F)45·7< 0·0012·150·150·020·89
Nutrients (N) 0·03  0·872·190·151·770·19
W × F 0·99  0·380·200·820·270·77
W × N 1·44  0·250·810·450·850·43
F × N 2·90  0·100·900·340·500·48
W × F × N 2·79  0·075·090·015·900·006

Many of the 89 species recorded were patchy in distribution either in time or microcosm, as is normal for phytoplankton communities. However, a suite of 21 species that occurred more consistently was identified and for which statistical analysis was likely to be informative. The criterion was that a species should have constituted at least 10% of the numbers or biovolume in at least one treatment in at least one season. The list included one diatom, 15 chlorophytes, two cryptophytes and one each of chrysophytes, dinoflagellates and euglenophytes (Table 6). No cyanophyte was sufficiently abundant to be listed. Further analysis of treatment effects on these species was carried out.

Table 6.  Significance of main and interaction effects of warming (W), fish (F), nutrients (N) and season (S) on the most abundant (by number or biovolume) phytoplankton species in an experiment carried out in experimental microcosms. Species codes are: a, Chlamydomonas grovei G.S. West; b, Asterococcus limneticus G.M. Smith; c, Rhodomonas minuta var nannoplanktica (Skuja) Javornicky; d, Nitzschia sp.; e, Chrysochromulina parva Lackey; f, Chlamydomonas sp.; g, Tetraedron minimum (A. Braun); h, Scenedesmus dimorphus (Turpin) Kutzing; i, Monoraphidium contortum (Thuret) Komarkova-Legnerova; j, Lagerheimia ciliata (Lagerheim) Chodat; k, Micractinium pusillum Fresenius; l, Cosmarium regnellii Wille; m, Pseudococcomyxa simplex (Mainx) Fott; n, Haematococcus pluvialis Flotow emend. Wille; o, Peredinium cinctum (O.F. Muller); p, Sphaerocystis planktonica (Korshikov) Bourrelly; q, Oocystis pusilla Hangsgirg; r, Cryptomonas erosa Ehrenberg; s, Trachelomonas volvocina Ehrenberg; t, Pandorina morum (O.F. Muller); u, Cosmarium botrytis Meneghini ex Ralfs. Significance testing was carried out only for those seasons (W, winter; S, summer) when the species was detected. Presence of either NS or a significance value (*P < 0·05, **P < 0·01, ***P < 0·001) indicates that the species was present. Dashes indicate that it was not detected in all seasons. Sig. inter. is the total number of significant interactions (main, two-way, three-way and four-way) recorded for the species with all instances involving season ignored to preserve uniformity of comparison. The maximal number would be 7
Species Vol. (µm3)abcdefghijklmnopqrstu
10157070100108150240260260270l 37073017001800200037004000420017 50027 000
WNSNSNSNSNSNS***NSNSNS*NSNSNSNSNS****NSNSNS
FNS*****NS****************************NSNSNS*********
NNS*********NSNSNSNS********NS***NS************
S**********************************************
W × FNSNSNSNSNSNS***NSNSNSNSNSNS*NSNSNSNS*NSNS
W × N*NSNSNSNSNSNSNSNSNSNSNSNSNSNSNSNSNSNSNSNS
F × NNS*NSNSNSNS**NS******NSNS***NS*******NS
W × SNSNSNSNSNSNS*NSNSNSNSNSNSNSNSNS
F × S***NSNSNS*****************************
N × S***NSNSNS*NS*****NSNSNS**NSNS*
W × F × NNSNSNSNSNSNSNSNSNSNSNSNSNSNSNSNSNSNSNSNSNS
W × F × SNSNSNSNSNSNSNSNSNS*NSNSNSNS*NS
W × N × SNSNSNSNSNSNSNSNSNSNSNSNSNSNSNSNS
F × N × S***NSNSNS*****NSNSNSNSNSNSNSNS
W × F × N × SNSNSNSNSNSNSNSNSNSNSNSNSNSNSNSNS 
Sig. inter.133111421343223123432

Data are shown in Table 6, with species arranged in order of the biovolume of the organism. Two species, Tetraedron minimum and Micractinium pusillum (both Chlorophyta), showed significant increases in their populations with warming, while Oocystis pusilla (Chlorophyta) and Cryptomonas erosa (Cryptophyceae) showed a significant decrease. The other 17 species were unaffected by warming. In the four cases of significant change, increases with warming were proportionately much greater than decreases (Table 7). Nutrient addition, perhaps surprisingly, was associated with about equal numbers of increases (six species), decreases (eight species) and lack of effect (seven species). In contrast, the presence of fish increased the populations of the majority of this suite of species, was associated with decreases in four and had no effect on only two. There was no particular pattern of main effects associated with organism size, nor any pattern of summed effects (main plus all interactions) (Table 6). This was true even after application of all conventional transformations.

Table 7.  Changes in population size (organisms per ml) of those abundant species (Table 6) showing significant changes in response to warming in a microcosm experiment investigating the effects of warming (C, ambient temperature; h, summer warming; H, all-year warming), fish (F) and nutrient (N) treatments. Seasonal means are shown (W, winter; S, summer; 1, first year; 2, second year). Values in parentheses are the appropriate mean square error estimates (from the repeated-measures analyses of variance, data log (x + 1) transformed) for attaching back-calculated error intervals to each mean
  CFNCFCNChFNhFhNhHFNHFHNH
TetraedronW1    –  –   –  –    –   –  –  –    –   –  –  –
minimumS1    –  –   –  –    –   –  –  –    –   –  –  –
(0·322)W2    2  6   1  1   34  38  5  2 3 970 681  1  0
S2   46557  11 13  8941960 25  9 6 1207600 11 12
MicractiniumW1   71  2   1  4 1 660   5  5  2 8 110   2134  1
pusillumS110 200  3   0 4678 800 190  1 15 4 300 105240295
(0·179)W2  929  0   0  126 500   0 17  219 340   0 25  1
S2  364  1   1  1 2 670 529 42  344 200   7 12 63
OocystisW1    –  –   –  –    –   –  –  –    –   –  –  –
pusillaS1   10 56   0 78    3   9  2  2   12  10  0 12
(0·171)W2    1  4  24 19    0   1  1  1    0   6  0  1
S2    5 20  36211    0  10 46 59    3  11  0149
CryptomonasW1 1 570 20  25 36 1 020   6 10 10 2 870  21 19 55
erosaS1  706 48 295241  396  38210 32  769  35 96 37
(0·287)W2  931176 580431 1 300 156359138 1 160 229307202
 S2 1 7002421240549 1 930 224777331  438 271421706

Discussion

We had predicted that increased warming would bring a greater risk of collapse of submerged plant communities, that there would be a switch towards phytoplankton communities increasingly dominated by cyanophytes and with some changes in species composition but with overall production unchanged by warming trends, and that there would be fewer effects on small species than on larger ones.

The plant community proved resilient (McKee et al. 2002b, 2003). There was no switch to phytoplankton dominance, even at the highest nutrient levels in the presence of fish. The stabilizing mechanisms that preserve submerged plant communities in the face of increased nutrient loading have thus been shown to be powerful. The phytoplankton community, measured either by chlorophyll a or biovolume, did not change significantly with warming either throughout the year or in summer only. Both increased with nutrient addition and in the presence of fish. The mean increase in chlorophyll a was about 10-fold to 35 mg l−1 (from a control value of 3·3 mg l−1) when nutrients were added and fish were also present, but this is not a very large value compared with turbid phytoplankton-dominated systems where sometimes 300 mg l−1 is achieved.

The fish predation intensity, judged from the biomass and its impacts on both zooplankton and benthic invertebrate communities (McKee et al. 2003), was comparatively high. This should have increased chlorophyll concentrations to a greater extent than might be expected in whole-lake systems with piscivores present. However, chlorophyll concentrations remained moderate, underlining the resilience of the plant-dominated system. It seems unlikely that the predicted warming trend alone will act as a switch in these systems, and even with the support of nutrients and fish predation on zooplankton a switch was not achieved. The effects of already known switches are thus likely to remain more important than the effects of warming.

Warming did change the phytoplankton community but only by modestly decreasing the cryptophyte and dinoflagellate components, particularly the former, which was most abundant in the controls in winter. Neither of these changes is likely to be ecologically very important. The anticipated increase in blue-green algae did not occur, despite the presence of several cyanophyte species. Cyanophytes are important components of turbid, shallow lake communities and it may be that warming will increase the proportion of these algae in lakes where aquatic plants have been lost and phytoplankton dominance has been established. It may also increase their extent in stratified lakes if the hypolimnion becomes increasingly deoxygenated at higher temperatures. But for plant-dominated lakes our experiment suggests that the risk is small. None of the blue-green algae even entered our list of relatively abundant species.

The phytoplankton community that established was as diverse as those of natural lakes and the assemblage was typical of shallow lakes. Of those phytoplankton species that were sufficiently widespread within the treatments for statistical analysis to be used reliably, only a minority was influenced by warming. Two chlorophyte species increased with warming, one declined. Within groups, different species have different temperature optima and most grow very well over wide ranges. The range and optima of the three species have not been established in culture but what has been observed seems most likely to be a simple species substitution with warming, such as is common in year-to-year variations in phytoplankton communities dependent on weather. In a series of microcosm experiments carried out in 1998 and 1999 over a European transect from Finland to southern Spain, temperature differences in mid-summer between years averaged 2·9 ± 1·6 °C, with a range from 0·3 to 4·3 (Stephen et al. 2004). The predicted temperature change from global warming during the twenty-first century is about the same as this inter-year difference. Not surprisingly, the change in phytoplankton community in the experiments with warming was slight. The fourth species that was significantly affected by warming was a cryptophyte, more abundant in general in winter in the controls than in the summer. Warmer winters are predicted for the future in north-west Europe. Changes in the winter flora of lakes might thus be more prominent than changes in the summer flora. However, no particular algal problems are associated with winter and it is unlikely that this effect will have great practical relevance. The extreme dispersability of algae (most genera and many species are ubiquitous) also means that the difficulties of dispersion, in colonizing new geographical ranges created by warming, that might be faced by larger organisms with longer and more specialized life histories are unlikely to be important.

Shallow lakes will inevitably support a phytoplankton community fitted to their particular conditions. Water associated with plant beds is relatively calm, lacking the vigorous mixing of the pelagial zone of larger, deeper lakes. Life forms fitted to such conditions include small and motile species, which have low net sinking rates. Our communities epitomized this. There is no reason to think that such communities will be any more or less susceptible to the effects of warming than those of deep lakes, but they do differ in their susceptibilities to factors such as grazing. Effects of fish (and thus indirectly of the grazers on which the fish feed) were generally more prominent than those of nutrients or warming. Fish allowed an increase in diversity as well as increased biovolumes of all the prominent algal groups. This is consistent with terrestrial experience where the loss of top predators tends to lead to an increase in grazing and a progressive loss of plant species (Terborgh 1992; Schmitz 2003). Analogies must not be drawn too far, however. In shallow lake communities diversity includes not only micro-organisms but also plants and macroinvertebrates. In such communities, the general principle still applies, predatory fish are associated with high diversity, clear water conditions in which cascading effects lead to an abundance of grazing zooplankters. These, by decimating the phytoplankton, lead to conditions where the macrophyte community and its associated invertebrates can flourish at high diversity.

Do these results suggest that there is little cause for concern as temperatures continue to rise in the next few decades from the approximately 0·6 °C rise (compared with the recent historic mean) at present to a predicted 3+ °C rise later this century? This is a difficult question. Microcosms inevitably do not replicate all the features of natural lakes. They cannot reflect extreme events of hydrology, for example. But, especially if sited in the open air and subject to natural rainfall, they do mimic quite well the conditions in small, shallow lakes and ponds, in which the macrophyte architecture strongly modifies many facets of the external environment (deadening of water movements, modification of water chemistry, for example). Macrophyte-dominated shallow lake communities are complex with many inherent stabilizers. Deep lake phytoplankton communities might be more vulnerable to temperature change, although their ability through high residual diversity to cope with between-year temperature variations as great as those predicted for global warming suggests that this is unlikely. An increase in cyanophytes, because of their preference for the microaerophilic conditions associated with hypolimnia, which could become more severely deoxygenated with warming, is possible. However, the most severe effects of global climate change might come, not from temperature rise per se, but from the associated increasing incidence of extreme weather events and rising sea levels. Extreme flooding will remodel river systems and rising sea level will extend estuarine conditions at the expense of freshwater ones. Whether these changes will constitute boon or disaster from a conservation or ecological point of view is certainly not clear, but there can be no relent regarding the tackling of existing, perhaps more significant, problems of eutrophication (diminished diversity, reduced drinking water quality, toxic blooms, changed fish communities) and inappropriate fish stocking (removal of predators, introduction of favoured angling species such as rainbow trout Onchorhyncus mykiss and common carp Cyprinus carpio), irrespective of warming trends.

Ancillary