Global warming and eutrophication: effects on water chemistry and autotrophic communities in experimental hypertrophic shallow lake mesocosms

Summary

• 1
  Shallow lakes are important components of the biosphere, but they are also highly vulnerable to damage from human activities in their catchments, such as nutrient pollution. They may also be particularly vulnerable to current warming trends.
• 2
  Forty-eight tanks were used to create 3-m³ mesocosms of shallow lake communities, in which the effects of warming by 4 °C and regular nutrient loading at two levels relevant to current degrees of eutrophication were studied in the presence and absence of fish.
• 3
  Warming changed concentrations of soluble phosphate, total nitrogen and conductivity, increased total plant biomass and decreased the amount of phytoplankton through shading by floating plants. Nutrient additions decreased total plant biomass but increased floating plant biomass. Nitrogen increase and warming increased floating plant biomass and decreased plant species richness. The plant community remained intact and did not switch to the turbid-water, phytoplankton-dominated community often predicted to be a consequence of global warming and eutrophication.
• 4
  Synthesis and applications. Likely future temperature increase will exacerbate some, but not all symptoms of eutrophication in shallow lakes. Alone it will not cause a switch from plant-dominated to algal-dominated systems, but may result in nuisance growths of floating lemnids. Currently underplayed, nitrogen loading should be taken more seriously in the management of European freshwaters.

Introduction

Shallow lakes and wetlands dominate inland waters (Downing et al. 2006), and are particularly important in carbon storage (Downing et al. 2008). They are also prone to environmental change because of their large surface-to-volume ratios, and to switch between states of high biodiversity and plant domination to waters made turbid by suspended sediment or phytoplankton (Scheffer et al. 1993). Such switches are influenced by nutrient loading, but driven by deliberate weed cutting, overgrazing by birds or cyprinid fish, herbicide and other biocide run-off, change in water level or salinity (Moss, Madgwick & Phillips 1996). Water level and salinity may be indirectly affected by climate, but there are suggestions that direct effects of increasing temperature may also act as a driver (Mooij et al. 2005).

Global temperatures are rising, bringing with them changes in seasonality and precipitation (IPCC 2007). Current predictions are that temperatures over the coming century may rise up to 10 °C with uncertainty dependent on what mitigation measures are taken and how rapidly these are adopted (IPCC 2007). Current policy assumes that a rise of 2–5 °C is likely (IPCC 2007; Monbiot 2007).

Of several approaches for predicting ecological effects of such change, experimentation at the ecosystem level is the most direct, yet rarest because of its expense. Two such experiments have hitherto been conducted in controlled, heated mesocosms: in the UK (McKee et al. 2000, 2003; Feuchtmayr et al. 2007), and in Denmark (Liboriussen et al. 2005). The UK experiment found that a 3 °C temperature increase accelerated flowering of Potamogeton natans and increased the predominance of Lagarosiphon major, a South African exotic. Macrophyte dominance was maintained even with increased nutrient loading and warming had comparatively little effect on phytoplankton composition or biomass (Moss et al. 2003). The additional nutrient treatment was modest, however, compared with the current state of many shallow lakes. The Danish experiment is still underway.

Meanwhile, comparisons across latitudinal gradients suggest substantial change in shallow lakes as temperatures increase. Warmer lakes have longer growing seasons and greater diversities of highly fecund fish that exert increased predation, reduce zooplankton grazing and lead to greater phytoplankton crops and development of floating plant communities (Gyllstrøm et al. 2005; Jeppesen et al. 2005; Meerhoff et al. 2007). The influences of temperature as such, and biogeographical history, however, cannot easily be separated in such comparisons. Experimental studies in lakes across Europe, using parallel mesocosm systems that eliminated much of the biogeographical effect also showed a stronger tendency for phytoplankton dominance with decreasing latitude (Moss et al. 2004) owing to greater fish predation on zooplankton grazer communities. Nonetheless, there is still much uncertainty as to how future temperatures will influence shallow lake structure and function. The continuation of many other pressures adds further complications.

Eutrophication, for example, remains a major problem for both lakes and society. Increasingly both phosphorus and nitrogen are implicated (Elser et al. 2007), not predominantly just phosphorus, as was widely believed in the past. In summer, abundant phosphorus release from sediment stores, and uptake and denitrification of available nitrogen compounds may lead to nitrogen scarcity. Increased nitrate loading leads to reduced diversity of submerged plant communities (James et al. 2005; Barker et al. 2008). Submerged plants derive much of their nutrient supply from sediments (Barko, Gunnison & Carpenter 1991), but their growth is inhibited by availability of light (Falkowski & Raven 2007) that may be denied by growth of phytoplankton, periphyton and floating plants, all of which must take their nutrients from the water. Floating plant communities appear to be favoured by increased nutrient loading (Scheffer et al. 2003). Adding to the problem of eutrophication. Adding to the problem of eutrophication whose interactions with climate warning are still poorly understood, rising temperatures will influence many aspects of nutrient processing within the systems, as well as change loading rates through its influences on hydrology.

Changes in fish communities could also complicate the direct effects of temperature increase. Fish act through predation on grazer communities of zooplankton and periphyton feeders (Moss et al. 1996) and may be particularly vulnerable to warming both through their limited thermal tolerance and through indirect effects of oxygen concentrations (Alabaster & Lloyd 1982). Warming may result in fish deaths or declines and changes in predation levels. Eutrophication exacerbates such consequences by providing greater biomasses of many organisms that increase respiration rates and oxygen consumption in the vulnerable hours of the night (Bronmark & Weisner 1992). A loss of fish might result in increased grazer invertebrate communities, and hence, lower phytoplankton crops and greater macrophyte biomass (Gyllstrøm et al. 2005; Meerhoff et al. 2007).

The literature on temperature and nutrient effects on shallow lake communities thus suggests several future scenarios. Here we performed an experiment designed to simulate new predictions of likely future temperature in conditions that closely reflect current states of eutrophication, in contrast to our first experiment with lower nutrient concentrations and lower temperature predictions. We used 48 temperature-controlled 3-m³ mesocosms factorially divided into two thermal regimes (ambient and + 4 °C), presence or absence of fish, and three nutrient-loading regimes, in hypertrophic systems. Our hypotheses were that (i) warming would increase phytoplankton or floating plant growth, particularly with increased nitrogen loading, with an alternative that warming would disfavour phytoplankton growth through increased grazing following detrimental effects on fish populations, and (ii) warming would increase or maintain submerged plant growth where loadings were not increased.

Methods

mesocosm establishment

Our tank system was at Ness Gardens in north-western England (53°16′ N, 3°03′ W) and included 48 tanks, each 1 m deep and 2 m in diameter, sunk in the ground to about 20 cm of the rim (see McKee et al. 2000, 2003). Heating and control treatments were applied in a randomized block design with four blocks. Plastic-pipe heating elements were placed on the sediment either as a ‘dummy’ in 24 unheated tanks or connected to the heating system in 24 heated tanks. In the latter, water at 45 °C was pumped through submerged pipes to increase the tank temperature, which was measured every 15 min throughout the experiment by sensors at depths of 45 cm, whilst a control system ensured a 4 °C higher water temperature in heated tanks compared with the nearest unheated tank (McKee et al. 2000). Heated tanks thus followed the same daily and seasonal temperature cycles of the ambient temperature tanks, plus 4 °C.

In October 2005 each tank received a 20-cm deep sediment mixture containing (by volume) 50% garden loam and 50% organic material (47·5% chopped organic oat straw and 2·5% rotted organic cow manure) to simulate the sediment of a eutrophic pond. Three cubic metres of local borehole water were then added. Heating began in October 2005 and three nutrient levels were imposed in a randomized block design from January 2006, with a standard load every 2 weeks. Sixteen tanks received 2500 µg L⁻¹ N and 50 µg L⁻¹ P (high addition, N2), 16 tanks received 250 µg L⁻¹ N and 50 µg L⁻¹ P (low addition, N1) using NaNO₃ and KH₂PO₄, and 16 tanks did not receive nutrient additions (N0).

Algal and animal communities were established from well-mixed inocula from five local lakes and ponds, added to the mesocosms before January 2006. Two and three weeks after inoculation, organisms were cross-mixed between mesocosms via standardized sweep-net samples to ensure similar starting conditions. As part of the randomized block design, four sticklebacks Gasterosteus aculeatus L. of both genders were added to half of the mesocosms in May 2006. These populations were supplemented by additional fish to an average biomass of 13·45 g per mesocosm in July 2006.

Aquatic plants established naturally during early summer of 2006 from seeds and fragments introduced with the animal and algal inocula. Communities were kept from full growth by cutting until July 2006 when checks were made that all species were present in each tank. Omissions were rectified by deliberate additions and the communities then allowed to establish unhindered. Elodea nuttallii, Lemna trisulca, Ceratophyllum demersum, Lemna minor, Lemna minuta, Potamogeton berchtoldii, P. crispus and P. natans were present. Spirodela polyrhiza appeared during 2007. Nomenclature follows Stace (1997).

All mesocosms were allowed to settle until January 2007 when sampling and monitoring of the macrophyte community began. During this period, evaporation losses from the mesocosms were replaced with de-ionized water when not compensated for by rainfall.

monitoring

Water samples were taken for analysis prior to nutrient dosing every 2 weeks, using a tube that integrated the water column. We used Johnes & Heathwaite (1992) for total nitrogen (TN) and total phosphorus (TP) and Mackereth, Heron & Talling (1989) for soluble reactive phosphorus (SRP), nitrate (NO₃-N), and ammonium (NH₄-N). Conductivity and pH were measured electronically. Chlorophyll-a, an estimate of phytoplankton biomass, was measured by a submersible fluorometer (Cyclops-7, Turner Designs). Oxygen was measured at mid-depth (35 cm) with a YSI Model 85 probe between 10·30 h and 12·30 h. The phytoplankton community was qualitatively examined to genus level.

Fish were monitored monthly by standardized trapping, using non-return plastic bottle traps over 24 hr. Fish were weighed fresh and returned to the mesocosms. If fish had died, new additions were made to maintain the fish treatments until summer 2006. No further additions were made during the period described in this study but numbers and biomass were monitored. Submerged and floating plants and filamentous algae were recorded every 2–3 weeks during 2007 as proportion of the water volume infested (PVI). This monitoring was performed by the same person to ensure comparability and a submersible glass-bottomed box was used to help visibility.

During the final harvest in September 2007, plants and filamentous algae were manually removed, hand-sorted, placed in mesh bags, and spun-dry at the highest setting in a domestic spin drier for a standard time of 5 min. Fresh weight was then determined. Sub-samples were retained for dry weight determination following drying in a laboratory oven at 80 °C. Dry weight contents of different species were sufficiently close (±5%), however, for fresh weights to be used in subsequent calculations. After water had been pumped out of the mesocosms, three random samples of surface sediment were taken for drying at 80 °C: loss on ignition was determined by burning at 350 °C and TN and TP were determined by acid digestion and methods as above.

statistics

Results were tested for normality (Kolmogorov–Smirnov test) and transformed to log(n + 1) where necessary, They were analysed by repeated measures analysis of variance (anova), done in two ways for each variable: first with block as a covariate and Bonferroni tests applied, and secondly, with block removed and Tukey tests used for between-subjects factors. Where these differed, the most conservative result has been taken. Proportional values were arcsine-transformed. Repeated measures anova requires that data have homogeneity of covariance, which was determined by Mauchly's test of sphericity, followed by Greenhouse–Geisser tests if the Mauchly test was not significant. Time was universally significant but block never so. Where the test showed no overall significance of treatment but there was a distinctive seasonal pattern, data were re-analysed on a date-by-date basis using one-way anova. Data obtained from the final harvest were subjected to three-way anova. Linear and non-linear regressions were used to investigate the shape of seasonal trends in species richness and Shannon–Weaver diversity. SPSS 15 was the main statistical package used.

Results

The differential temperature in the experimental mesocosms was maintained as intended (Fig. 1) within ±0·2 °C. Noon temperatures reached 21 °C in unheated mesocosms and 25 °C in heated ones and did not drop below about 3 °C in any mesocosm.

Figure 1. Changes in noon mean water temperature between January and September 2007 in 24 mesocosms warmed by 4 °C above ambient and 24 unheated mesocosms.

Oxygen concentrations measured at a standard time in the morning were 39% lower on average in heated mesocosms (Table 1), which was mainly due to biological effects, (accounting for at least 34%), rather than to physical effects on solubility. Nutrients and fish had no significant effects on oxygen levels, but towards the end of the experiment, oxygen levels in heated and high nutrient mesocosms were very low and fish struggled to survive (Table 1, and R. J. Moran, B. Moss, H. Feuchtmayr, K. Hatton, T. Heyes and D. Atkinson, unpublished data).

Table 1.  Mean effects of heating (H), nutrient (N) and fish treatments (F) on water chemistry and fish biomass in 48 mesocosms between January and September 2007 (N = 16 to 18 sampling dates for water chemistry, N = 9 for fish biomass). Abbreviations used: no addition, N0; low addition N1; high addition, N2; no fish, –F; fish present, +F; not significant, ns. Wherever results were significant, P values are indicated as *P < 0·05, **P < 0·01, ***P < 0·001; P values ≤ 0·08 are also given; n.a., not applicable

	Mean value							Probability
	Unheated	H	−F	+F	N0	N1	N2	H	F	N	HF	HN	NF	HNF
Oxygen concentration (mg L^−1)	6·6	4·0	5·7	4·9	5·9	5·1	4·9	***	ns	ns	ns	ns	ns	ns
Oxygen saturation (%)	59·9	39·4	53·8	45·4	56·1	47·4	45·4	***	0·07	ns	ns	ns	ns	ns
SRP (µg L^−1)	92	158	148	102	94	113	168	*	ns	ns	ns	**	ns	ns
TP (µg L^−1)	352	381	298	435	353	329	418	ns	*	ns	*	ns	ns	ns
Nitrate-N (mg L^−1)	0·31	0·22	0·39	0·14	0·01	0·02	0·77	ns	***	**	0·06	ns	**	*
Ammonium-N (mg L^−1)	0·13	0·09	0·09	0·13	0·04	0·05	0·23	ns	ns	***	*	**	ns	*
TN (mg L^−1)	2·6	2·2	2·3	2·5	1·7	1·8	3·7	*	ns	***	*	**	ns	ns
pH	7·38	7·46	7·44	7·40	7·44	7·35	7·47	**	ns	ns	ns	ns	ns	ns
Conductivity (µS cm^−2)	471	572	519	523	432	459	672	***	ns	***	0·06	ns	ns	ns
Chlorophyll a (µg L^−1)	66·3	20·8	18·7	68·5	50·9	32·1	47·8	***	***	ns	***	ns	0·07	ns
Fish biomass (g per tank)	5·3	2·4	n.a.	n.a.	6·1	2·7	2·8	***	n.a.	***	n.a.	ns	n.a.	n.a.

SRP increased significantly in heated mesocosms (Fig. 2; Table 1). Overall SRP was not affected by fish or nutrient treatments but began to show significant decreases with fish and increases with nutrient level as the summer progressed (Fig. 2: anova for individual dates) although external phosphorus loading had been kept constant. Concentrations of SRP were substantial in summer (overall means from 166 to 204 µg L⁻¹), suggesting major release from the sediments. TP (Fig. 3, Table 1) showed no significant difference on warming, but increased significantly with fish, especially in the heated mesocosms (H × F interaction, Table 1). TP concentrations in the N1 treatment did not differ from those where no nutrients were added but the higher addition (N2) showed significant increases (anova for individual dates) from around 100 µg L⁻¹ in January to 500–600 µg L⁻¹ in late summer.

Figure 2. Changes in mean SRP concentrations between January and September 2007 in 48 mesocosms subjected to different temperature (ambient + 4 °C, ambient), nutrient addition (N0, N1, N2, see Methods) and fish (without, with) treatments. Asterisks (*) indicate significantly different results (P < 0·05 from anovas for individual dates) for warming and predation treatments. For nutrient additions, significant differences are denoted by 02 between N0 and N2; 01 between N0 and N1; and 12 between N1 and N2.

Figure 3. Changes in mean TP concentrations between January and September 2007 in 48 mesocosms subjected to different temperature (ambient + 4 °C, ambient), nutrient addition (N0, N1, N2, see Methods) and fish (without, with) treatments. Abbreviations are as in Fig. 2.

Nitrate additions (Fig. 4; Table 1) were rapidly metabolized with only trace concentrations detectable 2 weeks after addition (Fig. 4; Table 1). Fish decreased nitrate concentrations in winter and in spring. Ammonium concentrations also increased in the N2 treatment as did TN concentrations (Fig. 5, Table 1), which were seasonally much less variable than those of ammonium and nitrate. Fish alone had no effect on ammonium and TN concentrations, and effects of heating depended mainly on nutrient and fish treatments (H × N, H × F, and H × N × F interactions, Table 1). Effects of treatments on mean pH were slight (Table 1), but conductivity decreased with time, aided by rainwater dilution, except in N2 mesocosms where it was greater than in N1 and N0, because of the increased salt addition. The significantly higher mean conductivity in heated mesocosms (Table 1) was not due to evaporative concentration because water levels were maintained by adding de-ionized water. Conductivity (Table 1) was typical of waters drawn from sedimentary rocks, as expected from the borehole source used.

Figure 4. Changes in mean NO₃-N concentrations between January and September 2007 in 48 mesocosms subjected to different temperature (ambient + 4 °C, ambient), nutrient addition (N0, N1, N2, see Methods) and fish (without, with) treatments. Abbreviations as in Fig. 2.

Figure 5. Changes in mean TN and NH₄-N concentrations between January and September 2007 in 48 mesocosms subjected to different nutrient addition (N0, N1, N2, see Methods) treatments. Abbreviations are as in Fig. 2.

Treatments caused no significant differences in sediment depth, sedimentary loss on ignition, or TN and TP concentrations at final harvest (means ± SD and anova tests; for sediment depth 17·63 ± 2·15 cm, P > 0·29; for loss on ignition 8·0 ± 1·75%, P > 0·14; for TN 0·83 ± 0·12 mg g⁻¹, P > 0·08; for TP 0·21 ± 0·04 mg g⁻¹, P > 0·19). Depth had decreased by 2–3 cm from the depth initially established. Loss on ignition was small (7–8%). TN and TP concentrations were well within natural ranges.

Phytoplankton biomass (Table 1; Fig. 6) was low in heated mesocosms but increased significantly in unheated ones. It was also much higher in mesocosms with fish, especially in those that were heated, but showed no systematic pattern in relation to nutrient additions. The phytoplankton communities were largely of cryptomonads and small Chlorococcales. Cyanobacteria were scarce.

Figure 6. Changes in phytoplankton biomass (mean chlorophyll a) between January and September 2007 in 48 mesocosms subjected to different temperature (ambient + 4 °C, ambient), nutrient addition (N0, N1, N2, see Methods) and fish (without, with) treatments. Abbreviations are as in Fig. 2.

Heating was associated with a trend towards increased total PVI (significant at the 6% level) (Table 2; Fig. 7), particularly through an increase in surface lemnids (predominantly Lemna minor), Ceratophyllum demersum, and filamentous algae (Spirogyra, Ulothrix, Cladophora). Some Lemna minuta was also present but could not readily be distinguished from L. minor. Heating increased the PVI of Spirodela polyrhiza, but although the effects were very clear in mesocosms where this species occurred, the overall effect was not statistically significant since the plant came in as a late casual and did not appear in all replicates. Heating decreased the PVI of Lemna trisulca. Fish had relatively small effects, increasing filamentous algae and C. demersum, but decreasing L. trisulca and Potamogeton crispus. Nutrients had greater effects, decreasing total PVI from N0 to N2 through an effect on filamentous algae and P. crispus, while increasing the PVI of surface lemnids (L. minor/minuta).

Table 2.  Effects of heating, fish and nutrient treatments on the regularly measured PVI of aquatic macrophytes in 48 mesocosms between January and September 2007. Mean values of PVI are given as fractions. There were no significant three-way (H × N × F) effects or block effects. For abbreviations used, see Table 1

	Mean PVI (as proportion)							Probability
	Unheated	H	−F	+ F	N0	N1	N2	H	F	N	HF	HN	NF
Total PVI	0·35	0·45	0·39	0·40	0·50	0·43	0·26	0·06	ns	**	ns	ns	ns
Filamentous algae	0·02	0·05	0·01	0·06	0·07	0·03	0·01	*	**	*	ns	ns	**
Ceratophyllum demersum	0·05	0·17	0·08	0·13	0·1	0·13	0·08	***	*	ns	ns	ns	ns
Elodea nuttalli	0·18	0·14	0·18	0·13	0·23	0·15	0·09	ns	ns	ns	ns	ns	ns
Potamogeton crispus	0·05	0·06	0·08	0·03	0·08	0·07	0·01	ns	*	*	ns	ns	ns
P. natans	0·002	<0·0001	0·002	<0·001	<0·001	<0·001	0·003	ns	ns	ns	ns	ns	ns
P. berchtoldii	0·002	0·001	0·003	<0·0001	0·004	0·001	<0·0001	ns	ns	ns	ns	ns	ns
Lemna trisulca	0·04	0·01	0·03	0·02	0·02	0·03	0·02	***	*	ns	ns	0·07	ns
L. minor/minuta	0·01	0·03	0·02	0·02	0·01	0·02	0·03	***	ns	*	ns	ns	ns
Spirodela polyrhiza	<0·0001	0·001	0·001	<0·00001	<0·0001	<0·0001	0·001	ns	ns	ns	ns	ns	ns
Lemna spp. cover (%)	14	47	27	34	18	28	45	***	ns	**	ns	ns	ns

Figure 7. Changes in mean total macrophyte PVI between January and September 2007 in 48 mesocosms subjected to different temperature (ambient + 4 °C, ambient), nutrient addition (N0, N1, N2, see Methods) and fish (without, with) treatments. Abbreviations are as in Fig. 2.

Results from PVI were confirmed by those obtained from final biomass harvesting (Fig. 8; Table 3). Plant crops were substantial with individual mesocosms producing up to 4 kg fresh weight m⁻². Dry weight averaged 11·1 ± 1·7% of fresh weight. Heating increased total biomass slightly through its greater than eightfold increase in floating lemnids (Fig. 8), and more than doubling of Ceratophyllum demersum. However, among the submerged vascular plants, this increase in Ceratophyllum demersum contrasted with a nearly sixfold decrease in L. trisulca. Fish increased the abundance of floating lemnids approximately threefold but were otherwise of small influence, whilst nutrients decreased total biomass slightly despite a more-than-threefold increase in floating biomass.

Figure 8. Effects of warming, fish and nutrient addition on the final fresh macrophyte biomass (mean ±95% confidence limit) harvested from a set of mesocosms in September 2007. Asterisks indicate significant results (P < 0·05).

Table 3.  Effects of heating, fish and nutrient treatments on the final fresh-weight harvest of aquatic macrophytes (g fresh weight m⁻²) in 48 mesocosms between January and September 2007. There were no significant block effects. For abbreviations used, see Table 1

	Mean biomass (g fresh weight m^−2)							Probability
	Unheated	H	−F	+F	N0	N1	N2	H	F	N	HF	HN	NF	HNF
Total biomass	2083	2430	2260	2260	2480	2450	1835	***	ns	**	ns	ns	ns	ns
Submerged plants	2028	1880	2040	1875	2351	2163	1355	ns	ns	0·08	ns	ns	ns	ns
Floating plants	56	547	220	382	132	291	480	***	ns	*	ns	ns	*	*
Filamentous algae	3·3	25·8	25·7	3·5	21·1	8·7	13·9	ns	ns	ns	*	ns	ns	ns
Ceratophyllum demersum	630	1390	849	1170	1120	1257	652	*	ns	ns	ns	ns	ns	ns
Elodea nuttalli	735	357	706	386	881	435	321	ns	ns	ns	ns	ns	ns	ns
Potamogeton crispus	7·1	0·1	7·0	0·2	5·7	4·9	0·2	ns	ns	ns	ns	ns	ns	ns
P. natans	1·3	0	1·3	0	0	0	0	ns	ns	ns	ns	ns	ns	ns
P. berchtoldii	0·1	0	0·1	0	0·2	0	0	ns	ns	0·06	ns	0·06	0·06	0·06
Lemna trisulca	650	114	450	314	323	457	367	***	ns	ns	ns	**	ns	ns
L. minor/minuta	55	449	122	382	131	291	332	***	*	ns	*	ns	ns	ns
Spirodela polyrhiza	0·1	99	99	0·2	0·2	0·1	148	ns	ns	0·07	ns	0·07	0·07	0·07

Heating significantly increased Shannon–Weaver diversity and the Simpson index of diversity of the macrophyte community, but decreased species richness (Table 4). Fish affected none of these. Nutrients increased species richness and Shannon–Weaver diversity between N0 and N1, then substantially decreased them between N1 and N2 (Table 4). There were also seasonal patterns in the change in species richness (Table 5) and Shannon–Weaver diversity (not shown). Heating, increase in nutrients and presence of fish all converted seasonal patterns in species richness from curves with mid-season peaks to linear declines with time (Table 5). Trends for Shannon–Weaver diversity were for heating to change a pattern of slight increase to one of decrease with time, and for nutrients to change an increase at N0 to a mid-peak relationship at N1 and then to a flat relationship at N2. Fish changed a seasonal increase in their absence to a flat or mid-peak relationship.

Table 4.  Mean effects of heating (H), nutrient (N) and fish treatments (F) on aquatic macrophyte species richness, Shannon-Weaver index and Simpson index of diversity in 48 mesocosms between January and September 2007 (N = 13). For abbreviations used, see Table 1

	Mean value							Probability
	Unheated	H	−F	+F	N0	N1	N2	H	F	N	HF	HN	NF	HNF
Species richness	3·78	3·57	3·78	3·59	3·87	3·99	3·19	*	ns	***	ns	ns	ns	**
Shannon–Weaver index	0·60	0·70	0·64	0·65	0·63	0·73	0·58	***	ns	***	ns	**	ns	**
Simpson index of diversity 1-D	0·34	0·42	0·4	0·37	0·36	0·43	0·35	***	ns	ns	ns	ns	ns	*

Table 5.  Regression relationships with time for species richness of aquatic macrophytes in 48 mesocosms between January and September 2007. For abbreviations used, see Table 1

Treatment	Relationship	Constant	Function (x)	Function (x^2)	R^2	P	Trend
Unheated	Linear	3·70	0·011		0·01	ns
Unheated	Polynomial	2·92	0·324	−0·022	0·57	**	Mid-peak
H	Linear	4·17	−0·086		0·80	***
H	Polynomial	3·98	0·012	−0·005	0·84	***	Decline
−F	Linear	3·82	−0·006		0·01	ns
−F	Polynomial	3·29	0·21	−0·015	0·62	***	Mid-peak
+F	Linear	4·07	−0·07		0·49	***	Decline
+F	Polynomial	3·65	0·10	−0·012	0·65	***	Decline
N0	Linear	3·97	−0·013		0·02	ns
N0	Polynomial	3·29	0·26	−0·019	0·49	***	Mid-peak
N1	Linear	4·4	−0·058		0·26	ns
N1	Polynomial	3·48	0·307	−0·026	0·83	***	Mid-peak
N2	Linear	3·49	−0·043		0·43	**	Decline
N2	Polynomial	3·61	−0·092	0·0035	0·46	***	Decline

Discussion

Various and complex significant effects resulted from warming these hypertrophic systems by 4 °C above ambient temperatures. A synthesis is that warming and nutrients increased the growth of plants, but favoured floating species, best able to exploit the combined high availability of light at the water surface and nutrients. Growth of these floating plants may have been enhanced by warming directly, and possibly by warming-induced SRP release from the sediment. These plants then appeared to have suppressed species just below the surface (L. trisulca) and probably phytoplankton. Shade-tolerant species such as C. demersum, which can persist under Lemna covers (Preston & Croft 1997), fared better under warming. Warming also reduced fish biomass (Table 1, and R. J. Moran, B. Moss, H. Feuchtmayr, K. Hatton, T. Heyes and D. Atkinson, unpublished data), probably through oxygen stress. As zooplankton biomass and the zooplankton : chlorophyll-a ratio were not significantly higher with warming (H. Feuchtmayr, B. Moss, I. Harvey, R. J. Moran, K. Hatton, L. Connor, D. Spiller and D. Atkinson, unpublished data), it seems unlikely that zooplankton were responsible for the lower phytoplankton biomass. However, the anticipated warming-induced increase in zooplankton per capita grazing rates cannot yet be excluded as an influence on phytoplankton abundance.

Several aspects of the results deserve emphasis. These are the contrasts with a previous experiment in these mesocosms; their links with current understanding of alternative states in shallow lakes; the key role of nitrogen; and their implications for predicting ecosystem state in a warming world.

Our former experiment used deliberately designed plant communities, including Lagarosiphon major, Elodea nuttallii, and Potamogeton natans, a 3 °C rise in temperature above ambient, rather infertile inorganic sediments, and a fertilization regime in which maximum concentrations given were 0·5 mg N L⁻¹ and 0·05 mg P L⁻¹. The N1 treatment in the present experiment used a lesser and the N2 a greater addition, more in line with current loadings in countries with intensive agriculture (James et al. 2005). The present sediments were also richer and the plant community was self-determined. In the current experiment, a 4 °C rise over an ambient temperature that was about the same as that in 2000 (McKee et al. 2003) (maximum ambient temperature in 1999, 23·7 °C; in 2000, 24·8 °C; in 2007, 24·9 °C; summer mean temperature in 1999, 16·4 °C; in 2000, 15·5 °C; in 2007, 15·5 °C), thus gave slightly warmer conditions.

In the previous experiment, warming did not change phytoplankton biomass, even with increased nutrients and fish. Macrophyte communities remained dominant (McKee et al. 2003), as they did in the present experiment, but warming in 2007 led to lower phytoplankton crops, almost certainly because of shading by floating plants. Warming previously increased phosphorus concentrations and conductivity, decreased pH and oxygen saturation and increased the frequency of severe de-oxygenation. Currently, it had similar but more extreme effects; warming increased daytime pH, most probably due to increased photosynthetic withdrawal of CO₂. The rise in conductivity with warming probably reflects increased mineral weathering in the sediments. In the previous experiment, total plant abundance remained high and was unaffected by warming but the proportion of a warm-water exotic, Lagarosiphon major, increased. Warming had no influence on E. nuttallii, as in the present experiment. Effects of nutrient addition and the presence of fish were formerly independent of warming and tended to increase and decrease macrophyte abundance, respectively. This was not the case in the current experiment where warming and nutrients together increased floating lemnid growth, which had been negligible in the previous experiment although Lemna minor was present. Comparison of the two experiments thus suggests that symptoms of eutrophication are exacerbated by warming. Collectively, these include de-oxygenation, increased phosphorus owing to release from the sediments, reduced fish biomass, reduced plant species richness and increase in warm-water exotics. One common symptom of eutrophication, increased phytoplankton biomass, was not shown, owing to the consequences of other symptoms, particularly increased shading by floating plants. Reduction or disappearance of submerged plants, seen by many workers (Hartmann 1977) as per se a symptom of advanced eutrophication, did not occur.

Over a gradient of nutrient concentrations, shallow lakes can have a clear-water state dominated by macrophytes, and an alternative undesirable turbid state dominated by phytoplankton. There is much interest in the mechanisms by which these states switch and the role of nutrients in them (Moss et al. 1996). Scheffer et al. (2003) suggested the establishment of floating plant dominance as a third, alternative stable state associated with increased nutrient loading. However, our current experiment has demonstrated that substantial submerged growth is still possible under a lemnid cover.

Many advocates of alternative states in shallow lakes assume that the switch to a phytoplankton-dominated state is driven by increased nutrient loading (see Scheffer 1998). Others note experimental evidence (Balls, Moss & Irvine 1989; Irvine, Moss & Balls 1989) that nutrients alone do not promote such a shift and suggest that an external driver is necessary. The evidence of both our former and current mesocosm experiments is that nutrients alone are insufficient to displace a plant-dominated state. The external driver view allows that the thresholds for operation of drivers may be determined by nutrient availability, but clearly a high nutrient loading in the current experiment was insufficient to displace the plants.

Changes in fish communities such that piscivores decrease and zooplanktivores increase, thus exerting more predation pressure on grazer zooplankton and plant-associated invertebrates, are common features of switches from macrophyte dominance. They lead to increased phytoplankton or periphyton growth (Phillips, Eminson & Moss 1978). Jones & Sayer (2003) found a link between fish biomass and macrophyte decline through loss of plant-associated invertebrates that graze on periphyton. We recorded filamentous algae, the major component of macrophyte periphyton in the mesocosms. However, filamentous algal biomass was very low, probably due to high abundances of snails: average abundance per surface area of 500 m⁻² at final harvest (R. J. Moran, B. Moss, H. Feuchtmayr, K. Hatton, T. Heyes and D. Atkinson, unpublished data). Fish had no significant impact on gastropods or filamentous algae by the end of our experiment. This was probably due to an increase in lemnid cover causing a decrease in oxygen levels which had a negative effect on fish biomass (R. J. Moran, B. Moss, H. Feuchtmayr, K. Hatton, T. Heyes and D. Atkinson, unpublished data). Diurnal studies in June/July 2007 showed near anaerobiosis especially in heated tanks by dawn.

Temperature increase has been suggested as a driver for loss of plants and increase in turbid conditions in a review of how global warming will affect shallow lakes (Mooij et al. 2005). We concur that there will be severe changes with temperature increase and we are aware that single experiments may give only one of several possible outcomes, dependent on the many combinations of starting conditions. However, for the moment, the experimental data do not suggest a loss of macrophyte dominance through temperature increase alone, even at high nutrient loadings, although a move towards floating plant abundance seems likely in small temperate hypertrophic water bodies, which parallels the preponderance of floating plants in warm temperate and tropical lakes (Meerhoff et al. 2007).

Temperature increase and nitrate loading reduced species richness. Currently, there is a reluctance by Government Agencies to apply standards for nitrogen in schemes for assessing ecological quality under the Water Framework Directive, on the grounds of too limited evidence (UKTAG 2007). James et al. (2005) produced correlative evidence that nitrate concentrations above about 1 mg NO₃-N L⁻¹ led to reduced species richness of aquatic plants and Barker et al. (2008) confirmed this experimentally. Our current experiment adds more evidence. Official reluctance to declare nitrate standards is misplaced.

Acknowledgements

We are grateful to Jan Hatton, Tom Heyes, Richard Hewitt, Becky Ford and John Segar for help with data collection and measurements. Support was provided by EU Contract no. GOCE-CT-2003-505540 (Euro-limpacs). The experimental design involving fish was approved by the Home Office Inspectorate as not requiring a licence.