Responses to elevated temperature and CO2 in the perennial grass Agrostis curtisii in relation to population origin



1. Evolutionary responses to climate change will depend on the presence of heritable variation within species populations for traits that increase fitness under the changing conditions. Patterns of ecotypic differentiation in relation to latitude in some species suggest that such variation exists in relation to temperature responses. Response to elevated CO2, whether heritable or not, is not expected to be related to latitudinal or climatic differences within temperate regions.

2. To test these ideas, seeds were collected from 10 populations of the outbreeding perennial grass Agrostis curtisii across its range in Europe from south Wales to Portugal. Plants were grown under ambient and elevated temperature and CO2 conditions, in a factorial design, in solardomes; two half sibs from each population were planted in separate pots in each of the two replicate domes with each combination of treatments. One half sib was harvested at the end of the first summer, the second at the end of the second summer.

3. Survival was uniformly high and flowering uniformly low across treatments and populations.

4. Responses to temperature and CO2 treatments varied over time for almost all populations. Treatment effects were not significant on plants harvested in year 1, although there was a trend towards higher shoot biomass under the elevated temperature and CO2 treatment. In year 2 shoot biomass was significantly higher under the elevated temperature treatment across all populations and there was a strong trend towards decreased biomass under elevated CO2.

5. There were no significant correlations of plant response to either CO2 or temperature with climate at origin.

6. These results warn of the dangers of extrapolating evolutionary plant responses to CO2 from short-term experiments.


Increasing levels of atmospheric CO2, resultant warming and changes in water availability predicted over the next century (Houghton, Callander & Varney 1992; Houghton et al. 1995) will inevitably result in changes to the processes that determine the population sizes of plant species at any location. In some cases at least, there will be strong selection pressures which may lead to evolutionary responses to the changing conditions. Populations of plants containing heritable variation in traits, which result in increased fitness under such pressures, are in theory capable of responding to climate change. Although highly confounded by other environmental variation, the ecotypic differentiation observed in many geographically wide-ranging species often has a clear relationship with climatic differences (e.g. Turesson 1925; Clausen, Keck & Hiesey 1940; Bocher 1949). Whereas such correlations suggest that evolutionary responses to climate change may occur, they do not always do so. In the case of CO2 concentrations, Beerling & Woodward (1996) observed that the reduced leaf stomatal densities found in early evolving plant groups, as a result of high palaeo-CO2 concentrations, have been retained by living representatives of the same plant groups over extensive periods of time and atmospheric changes. In contrast, the stomatal densities of the more recently evolved angiosperms have shown marked changes in response to long-term atmospheric CO2 changes (Beerling & Kelly 1997).

Recent investigations of differentiation in wide-ranging plant species have lent general support to the classic genecological studies cited above (Gray 1996). For example, a trend common to several species is an increase in the northern part of the species’ range in the length of the prereproductive period. This occurs in Daucus carota (Lacey 1988), Melilotus alba (Smith 1927) and Verbascum thapsus (Reinartz 1984) in North America, and Aster tripolium (Gray 1987, 1997), Beta maritima (Boudry et al. 1994), Holcus lanatus (Bocher & Larsen 1958) and Pulicaria dysenterica (R. E. Daniels and I. L. Moy, personal communication) in Europe. In Aster, Beta and Verbascum it results in populations of annuals in the south and typically often long lived perennials in the north.

Such patterns of variation can be interpreted as being caused by differences in climate. Reliable, warm growing seasons alternating with periods of drought are likely to favour annual life histories, while in cooler, more variable conditions, garnering resources by incremental growth and postponing reproduction until a critical size is reached is likely to be a better strategy (Gray 1997). However, it is important to emphasize that differences in climate, experienced by populations at different latitudes, and in the species’ response to climate differences, are confounded by variation in other key variables—most notably variation in day-length and in the timing and amounts of seasonal rainfall. For this reason, responses to differences in temperature, when grown in a common daylength, may not accurately reflect differences in the climate from which they originated.

Responses of plants to variation in CO2 concentrations (in the short term) are poorly understood. A number of experiments have revealed contrasting responses within species to artificially enhanced CO2. Germination among five families of Plantago lanceolata varied under enhanced CO2 (Wulff & Alexander 1985) as did lifetime fecundity in Raphanus raphanistrum (Curtis et al. 1994), grain production in Oryza sativa (Ziska & Teramura 1992) and total biomass production in Arabidopsis thaliana (Norton, Firbank & Watkinson 1995). A study conducted by Bazzaz et al. (1995) on populations of both Abutilon theophrasti and Betula alleghaniensis, which revealed genetic differentiation within populations to CO2, showed that those individuals which responded favourably to elevated CO2 in the absence of competition did not have the highest fitness in competitive stands. A number of other studies of grassland species have however, revealed no genetic variability in the responses of species to enhanced CO2 (Leadley & Stocklin 1996; Luscher et al. 1996; Luscher, Hendrey & Nosberger 1998).

In this paper, we investigate differences between geographically dispersed populations of a perennial grass, Agrostis curtisii Kerguelen to see if there is a relationship between growth under contrasting temperature and CO2 regimes and climate of origin. Agrostis curtisii is an appropriate species for such a study because it has a relatively narrow, and well-defined climate envelope (Carey et al. 1995), being confined to an oceanic zone extending from south Wales to south Portugal—a so-called Lusitanian distribution (Fig. 1a). There is variation in rainfall from west to east, but the major axis of variation from north to south is in daylength and temperature (Table 1). The potential distribution of A. curtisii according to the IS92a scenario of a climate possible for the year 2050 (Houghton, Callender & Varney 1992) which forecasts a temperature rise of between 1 and 1·5 °C across Britain is given in 1Fig. 1b. Agrostis curtisii occurs mainly on acid, well-drained soils and is particularly common on recently burnt heathland. It is self-incompatible, and isozyme studies have revealed clinal patterns of genetic variation among British populations, with the typically low levels of between-population genetic diversity for supposedly neutral markers displayed by perennial outbreeders (Gray 1996).

Figure 1.

. (a) The distribution of Agrostis curtisii in Europe based on presence/absence data on a 50 km grid; (b) the forecasted distribution of A. curtisii in Europe modelled on forecasted European climate data under climate change scenario IS92a produced by using logistic regression to relate climate to the probability of the occurrence of A. curtisii in each 50 km square (from Carey et al. 1995).

Table 1.  . Thirty year climate averages (1962–1992) for the sites of origin of the populations of Agrostis curtisii used in the experiment, with the ambient temperature conditions in the 2 years of the solardome experiment for comparison Thumbnail image of

Therefore, we predict that, in relation to the climate change scenarios described at the beginning, differences between populations in their response to elevated temperature may be related to differences in their origins, because temperature is a major variable to which local populations could have made an adaptive response. A separate study has estimated the heritability of variation in several traits to establish whether such a response is actually possible (I. L. Moy & A. J. Gray, unpublished data). As variation in CO2 across the the species’ geographical range is likely to have been small in the past, and potentially random with respect to position, it is possible that a genotypic response to elevated CO2 may not necessarily be linked to population origin within the range investigated here. To test these hypotheses, we grew A. curtisii plants originally sampled from a wide range of environments in controlled conditions under elevated temperature and CO2.

Materials and methods

The experiment was undertaken using the solardome facility at the Institute of Terrestrial Ecology, Bangor, UK. This system provides both tracking of ambient temperature and CO2 and elevated temperature levels at 3 °C above ambient, and 340 p.p.m. above ambient CO2, thereby mimicking the atmospheric changes expected towards the end of the next century (Houghton et al. 1990). The facility comprises eight solardomes in a 2 × 2 factorial design with two temperature and two CO2 treatments and two replicates per treatment. Technical details are provided by Rafarel & Ashenden (1992) and Rafarel, Ashenden & Roberts (1995).

In the summers of 1992 and 1993, 10 populations of A. curtisii were located throughout the species’ range (Fig. 2), and seeds were collected from a subsample of four individuals from 50 originally sampled from each population. The experimental design involved growing two plants (which were at least half sibs) in individual pots from each of the four parent plants from each population in each solardome. One half sib was harvested after one growing season, and the remaining half sib after the second season. The treatments were therefore four plants × two harvests × 10 populations in each replicate CO2 and temperature treatment, 80 plants in each solardome, and 640 plants in total.

Figure 2.

. Map of Europe showing the sites from which the 10 A. curtisii populations originated: site 1, Surrey; 2, Wales; 3, Devon; 4, Dorset; 5, Cornwall; 6, Brittany; 7, Galicia; 8, Basque; 9, Arcachon; 10, Portugal.

The seeds were sown on 3 December 1993 in a 50:50 John Innes No. 1 compost/sand mix (supplied by Joseph Arnold and Sons, Leighton Buzzard, UK) in a warm glasshouse. The majority of seeds had germinated after 7 days, and seedlings were allowed to establish before being potted on, with two seedlings per pot for each of four pairs of half sibs from each population (except for the Surrey population, where germination was sufficient to allow only three pairs), into pots 9 cm × 9 cm × 10 cm deep. These were transferred into the solardomes on 14 January 1994, positioned at random within the group of pots within each dome, and thinned to one seedling per pot on 9 March. The pots were watered as required, watering being applied equally to all domes.

One of each pair of half sibs was selected at random and harvested at soil level on 26–27 July 1994. Immediately afterwards, the remaining 39 plants per solardome were transplanted into pots of c. 9 cm diameter and 15 cm deep. Nutrients (full strength Phostrogen) were applied in April, May and June 1995, and the pots were re-randomized every few months throughout the experiment. The remaining plants were harvested on 19–20 July 1995 after flowering of some of the plants, but before seed set. At both harvests, the numbers of tillers and flowering spikelets (where present) were counted, and plant biomass was recorded after drying for 72 h at 70 °C.

Climate data for the points of origin were taken from 30 years average data from nearby weather stations for the period 1962–1992, provided by the Climatic Research Unit at the University of East Anglia.


January minimum, July maximum and annual precipitation were used to summarize the weather data (Table 1), following Carey et al. (1995). Shoot biomass per plant at each harvest was analysed using a split plot analysis of variance, in which the main plot effects were the result of solardome replication and solardome treatments (temperature and CO2) and the subplot treatments were plant origin and the interactions between plant origin and solardome treatments. The error term for the main plot effects was the solardome treatment × solardome replicate interaction. The use of four unrelated individuals from each population in each half of the experiment ensured that the experiment included some measure of the potential variability between families within each population.

The relationships between tiller number and plant size were investigated using correlation and stepwise regression to test for main treatment effects. The degree of consistency of rank order of plant size by population between the different treatments was explored using Spearman Rank correlation. The interaction of climate at the source of populations with shoot biomass in year 2 was analysed first by simply looking at mean biomass, and then at the response to the treatment, defined as the absolute mean difference between biomass at ambient and at elevated temperature or CO2 using log transformed values. Spearman Rank correlation was used to investigate relationships between the three climate variables at each site and variables for both temperature response and CO2 response.



All seedlings survived the first year, and 96% of plants survived until harvest at the end of the second summer. The 4% of the plants that did not survive were spread over seven of the 10 populations; there were no significant differences in survivorship between populations.


The analysis of variance showed no significant effect of solardome treatment on log transformed shoot biomass per plant (Table 2), although there was an indication of greater biomass under the treatment which involved both elevated temperature + elevated CO2 than under the other treatments (Fig. 3a). There were, however, significant differences between populations (Table 2), with the greatest biomass achieved by the Basque population (Fig. 4). The same pattern was observed in tiller number per plant, because of the high correlation with shoot biomass (Fig. 5a). The ANOVA revealed a significant population × temperature interaction (Table 2), and the rankings of the biomass per plant of the different populations were fairly consistent across treatments, with populations from Wales, Dorset and Galicia at the bottom and Basque, Pyrenees and Cornwall at the top (Fig. 4a,b). Spearman Rank correlations of the mean shoot biomass per plant for each population in each of the four solardome treatments revealed significant correlations with the exception of correlations between the temperature × CO2 treatment and ambient temperature treatments (Table 3). The responses to temperature varied inconsistently between populations, hence the population × temperature interaction (Fig. 4a), but responses to CO2 were more uniform, with most populations showing increased growth in the elevated CO2 treatment (Fig. 4b). No plants flowered in this first year.

Table 2.  . Analyses of variance of shoot biomass (log transformed) of A. curtisii from 10 sources of origin across its distribution grown under contrasting temperature and CO2 treatments in two blocks of solardomes (see text for details). Data are given for plants harvested at the end of the first and the second season's growth. The error term for the treatments (CO2 and temperature) is the treatment × block interaction. *P < 0·05; **P < 0·01; ***P < 0·001 Thumbnail image of
Figure 3.

. The effect of CO2 and temperature in the factorial solardome treatments (tc, ambient temperature, ambient CO2; tC, ambient temperature, elevated CO2; Tc, elevated temperature, ambient CO2; TC, elevated temperature, elevated CO2) on the shoot biomass (g) ± SE for each treatment averaged over all populations in years (a) 1 and (b) 2.

Figure 4.

. Shoot biomass ± SE of populations of A. curtisii from 10 areas of origin, averaged over ambient (open) and elevated (black) (a, c) temperature and (b, d) CO2 regimes, harvested in the first (a, b) and second (c, d) years.

Figure 5.

. Relationship between shoot biomass and number of tillers in (a) year 1 (r = 0·72, n = 304, P <0·001) and (b) year 2 (r = 0·85, n = 304, P <0·001). Note the log scale.

Table 3.  . Summary of Spearman Rank correlations on the mean biomass per plant for each of the 10 populations under each of the four solardome treatments in years 1 and 2: tc, ambient temperature, ambient CO2; tC, ambient temperature, elevated CO2; Tc, elevated temperature, ambient CO2; TC, elevated temperature, elevated CO2. The significances of rs are: NS, not significant; *P < 0·05; **P < 0·01; ***P < 0·001 Thumbnail image of


By the end of the second year, shoot biomass was higher under the elevated temperature regime (Fig. 3b, Table 2). Although ANOVA revealed no significant effect of CO2 on plant performance it is notable that all of the populations showed a negative response to elevated CO2 (Fig. 4d) with the single exception of plants from Galicia which showed a slight positive response. Again, tiller number was correlated with biomass across the treatments (Fig. 5b), and showed no evidence of treatment effects. As in year 1, there were significant differences between populations in terms of biomass production, with the Basque population again achieving the greatest biomass. Spearman Rank correlations of the mean shoot biomass per plant for each population in each of the four solardome treatments revealed fewer significant correlations than in year 1 (Table 3). Only 15 out of c. 300 surviving plants flowered; these plants were not obviously associated with particular populations or treatments.


January minimum and July maximum temperatures were negatively intercorrelated (r = – 0·73, n = 10, P < 0·05), but there were no correlations of climate at origin with plant performance in terms of either mean biomass, or the response to the treatment (as defined above).


The results indicate that the growth (both in biomass production and tiller number) of A. curtisii was positively influenced by elevated temperature over the two seasons. This response has been shown in another solardome experiment using the annual grass Vulpia ciliata (Firbank et al. 1995) and is probably owing to an enhanced capacity for photosynthesis with increasing temperatures, up to an optimal temperature beyond which negative effects ensue (Atkinson & Porter 1996). Nil to marginal increases in growth across the populations in response to elevated CO2 in the first year were replaced by negative responses across almost all populations in the second. Although the negative response was not significant, the lack of significance probably has more to do with the enforced lack of replication and poor statistical power, than with any lack of treatment effect.

The tendency for plant biomass to be lower under enhanced CO2 levels after the second year's growth is an unusual finding, as the majority of studies (many of them short-term) show positive effects of CO2 on plant shoot growth (e.g. Drake & Leadley 1991; Hunt et al. 1991, 1993; Lawlor & Mitchell 1991; Idso & Idso 1994; Curtis & Wang 1998). Whether this reduction in shoot biomass reflects a reduction in total biomass is, however, unclear. It may reflect a CO2 induced alteration in the pattern of biomass allocation. Fitter et al. (1996) for example, found that shoot biomass in monoliths of two contrasting vegetation types grown in solardomes, was unaltered, but root biomass was increased by between 40 and 50%. In contrast, other studies have found little impact of elevated CO2 on root mass ratios (e.g. Schenk et al. 1995; Hebeisen et al. 1997), although that is not to say that there are no effects on root distribution, water use and nutrient uptake. Clearly, where possible, a measure of root biomass as well as shoot biomass allows for a more complete picture of CO2 effects on plants.

There is clear evidence for population differentiation within A. curtisii, with for example, plants from the Basque population consistently producing the highest biomass per plant whilst the biomass from other populations (notably Wales, Dorset and Galicia) was consistently low. Spearman Rank correlations on plant biomass showed that in general populations performed similarly regardless of treatment in year 1, although by year 2 there were only two significant correlations between treatments. As the ANOVA revealed no significant origin × treatment interactions this was most probably owing to greater variability in plant biomass in all populations after 2 years’ growth.

Genetically based variation in plant size between populations is well documented along latitudinal and altitudinal gradients (e.g. Clausen, Keck & Hiesey 1940, 1948; Bradshaw 1960; Reinartz 1984), but the only correlation between biomass at harvest and variables describing climate of origin was in this case between annual precipitation and biomass under elevated CO2. There was no evidence of life history variation between the populations in terms of phenology. It might have been expected that plants from southerly latitudes would have had a shorter pre-reproductive period (Reinartz 1984; Gray 1987, 1997; Lacey 1988; Boudry et al. 1994), but only a few plants flowered during the course of the experiment and there was no pattern in relation to the origin of the population.

In relation to the predicted population responses to environmental change, these results do not support our first hypothesis that response to differences in temperature in a short-lived perennial should relate to differences in the climate of origin, reflecting past natural selection to a key climatic factor. Populations did not show responses to temperature in relation to their origin. This may reflect the overlapping variability of climate within the range looked at, or may possibly be owing to complications arising from the location of the experimental site in relation to the sites of population origin (although ambient temperatures were high in both 1994 and 1995, see Table 1).

The overall differences in the effect of elevated CO2 in the two seasons of growth, although non-significant, were consistent with the findings of Luscher & Nosberger (1997), who found differences between the first and second years in CO2 effects on biomass in grasses and legumes (increase in year 1, less increase in year 2), and Farnsworth & Bazzaz 1995), who showed that the enhanced early vegetative growth in high CO2 of nine herbaceous annuals could not be related to final reproductive output, and hence fitness. The differential response to CO2 between years observed in the solardomes casts further doubt on the attempt to relate short-term responses to possible adaptive responses in the field.

However, there is evidence to support our second hypothesis, that response to elevated CO2 will be unrelated to climate of origin, as there were no relationships between either mean biomass under elevated CO2 or the reduction in plant biomass at final harvest under elevated CO2 compared to ambient, and variables describing the weather at the site of origin. This contrasts with the results of several studies which have demonstrated genetic variation in growth responses to elevated CO2 in herbaceous plants both between populations [e.g. in Arabidopsis thaliana (Norton et al. 1995) and Oryza sativa (Ziska et al. 1996)] and within populations [e.g. Raphanus raphanastrum (Curtis et al. 1994), Abutilon theophrasti and Betula alleghaniensis (Bazzaz et al. 1995)].


We thank Roger Daniels, Isabel Moy, Rebecca Mogg and Alan Raybould for helping to sample seeds, Trevor Ashenden and Laurence Jones for maintaining the solardome facility, and Martine Van der Poll and Kamal Ibrahim for help with the analysis. Climate data were provided by the Climatic Research Unit of the University of East Anglia. This work forms part of the NERC TIGER (Terrestrial Initiative in Global Environmental Research programme) IV.2a, award numbers T03087b6 (LGF), GST/02/639 (ARW) and T03087F6 (AJG).