Author for correspondence: G. R. Edwards Tel: +44 (0)207 5942 782 Fax: +44 (0)207 5942 919 Email:email@example.com
• The effects of elevated CO2 on seedling recruitment are presented for the pasture species Anthoxanthum odoratum, Cerastium glomeratum, Leontodon saxatilis, Lolium perenne, Poa pratensis and Trifolium repens.
• Seeds collected from pastures maintained at ambient (360 µl l−1) and elevated (475 µl l−1) CO2 were measured for germination and seedling growth in a reciprocal design under ambient and elevated CO2.
• Seedlings of A. odoratum and L. perenne from elevated-CO2-developed seeds were heavier than those from ambient-CO2-developed seeds but only when the seedlings were grown at elevated CO2. Elevated-CO2-developed seeds had higher germination and seedling mass in T. repens and lower germination and seedling mass in L. saxatilis, irrespective of the CO2 concentration after sowing. There were more L. saxatilis and T. repens seedlings in pastures maintained at elevated than ambient CO2; a seed addition experiment showed that this was due mainly to the increased seed production of these species under elevated CO2. Seedling recruitment of C. glomeratum and P. pratensis was unaffected by elevated CO2.
• Species exhibited variable responses to seed development under elevated CO2 that could potentially alter community composition by influencing seedling recruitment.
Changes in seedling recruitment might also arise from direct or indirect effects of elevated CO2 after seed dispersal on germination and seedling growth and survival. Germination, for instance, has been reported to be both increased (Ziska & Bunce, 1993) and decreased (Popay & Roberts, 1970) under elevated CO2. Furthermore, elevated CO2 may alter the size, growth rate and identity of competing plants (Bazzaz et al., 1992; Potvin & Vasseur, 1997), which in turn may affect seedling emergence and survival (Edwards & Crawley, 1999). This could potentially modify the effects of CO2-induced changes in seed mass, quality and number on seedling recruitment. For example, an increase in seed mass under elevated CO2 might be accompanied by an increase in the size of existing plants, so leading to increased competition, with little net effect on seedling recruitment.
The purpose of this study was to examine the effects of elevated CO2 on seedling recruitment of one annual and five perennial species in a sheep-grazed permanent pasture on dry, sandy soil in New Zealand. Our goal was to assess the effects of elevated CO2 during the development of seeds on the parent plant (e.g. seed quality and seed mass) and the direct and indirect effects of elevated CO2 after seed dispersal. In one series of experiments, seeds collected from parent plants grown at ambient and elevated CO2 in a free air carbon dioxide enrichment (Face) experiment were measured for germination and seedling growth in a reciprocal design at ambient and elevated CO2. As we were interested in how plant–plant interactions may affect how plants respond to elevated CO2 (Bazzaz & McConnaughay, 1992), the experiment was conducted for seedlings grown individually in compost in a pot experiment and for seedlings transplanted into undisturbed pastures. In a further experiment to assess the importance of possible changes in the number of seeds produced under elevated CO2, the emergence and survival of seedlings was measured after a fixed number of seeds collected from parent plants grown at ambient and elevated CO2 were sown into pastures maintained at ambient and elevated CO2. The implications of the changes in seedling recruitment to the species composition changes that were taking place in the pasture under elevated CO2 are discussed.
Materials and Methods
The study was conducted during 1999 in a free air carbon dioxide enrichment (Face) experiment located in a pasture at Bulls, Manawatu, New Zealand (40°814′S, 175°816′E). The pasture had been under permanent grass since the 1940s during which time it had been grazed by sheep, cattle and goats, with occasional hay cuts taken. An inspection of the vegetation at the site in spring 1996 found 25 vascular plant species, with the C3 grasses, Agrostis capillaris Sibth., Anthoxanthum odoratum L. and Lolium perenne L., the C4 grass, Paspalum dilatatum Poir., the legumes, Trifolium repens L. and Trifolium subterraneum L. and the herbs, Leontodon saxatilis Lam syn. L. taraxacoides and Hypochaeris radicata L., being predominant. All vascular plant species at the site were exotic. The soil at the site was a Pukepuke black sand (Cowie & Hall, 1965) with a 0.25-m black loamy, fine sand top horizon underlain by greyish-brown, fine sand-textured horizons. A total of 60 soil cores (0.025 m diameter × 0.075 m deep) were collected across the whole site in March 1997, air dried and tested for pH in water (Lee et al., 1991), extractable K (Davies, 1952), plant available phosphate (Grigg, 1975) and sulphate S (Watkinson & Kear, 1991). Mean values were: pH = 5.8, K = 0.15 cM(+) kg soil−1, P (phosphate) = 20 µg ml soil−1 and S (sulphate) = 7 µg ml soil−1. The site was fertilized during the experiment with 30 g m−2 of superphosphate in spring and 5 g m−2 of potassium sulphate in spring, autumn and winter. The site is characterized by periods of summer drought which result in low standing biomass, death of perennial vegetation and a high proportion of bare ground (3–5% of ground area) at the end of summer (Edwards et al., 2001). During the experimental period January to December 1999, rainfall was below the long-term 50-year average in summer (January–March: average = 185 mm, 1999 = 141 mm), similar to the long-term average in autumn (April–June: average = 231 mm, 1999 = 255 mm) and winter (July–September: average = 224 mm, 1999 = 215 mm), and greater than the long-term average in spring (October–December: average = 229, 1999 = 270 mm). The Face rings were not irrigated during the study.
Six rings, each 12 m in diameter, were marked out at least 20 m apart at the site in January 1997. Rings with similar vascular plant species composition and soil characteristics were paired to give three blocks. One ring in each block was randomly allocated to be fumigated with CO2 and the other was left as an ambient CO2 control. Each ring was ring-fenced to preclude sheep located in the surrounding field from grazing the rings when not desired.
Fumigation was by free air carbon dioxide enrichment (Face) technology (Lewin et al., 1992). The Face system was installed in the pasture during the first nine months of 1997, with enrichment beginning on 1 October 1997. The target enrichment concentration was 475 µl l−1 CO2 during the photoperiod. Fumigation was through 24 standpipes, each 0.05 m in diameter, located at equi-distance intervals on the perimeter of each ring. Pure CO2 was piped to each elevated CO2 ring, mixed with ambient air in the stand-pipes and blown across the ring from the upwind side. The CO2 concentration (daily mean of 1 min averages) of the elevated CO2 rings for the experimental period January 1998 to December 1999, measured in the centre of each ring during the photoperiod at 0.3 m above the ground surface, was 474.6 µl l−1 with a standard deviation of 12.6. The 1 min average values were within ±10% of the set point of 475 µl l−1 for 83% of the time. The CO2 concentration of the ambient rings was not continuously measured during the same period. Measurements taken every 3 min during the photoperiod at 0.3 m above the ground surface in ambient CO2 rings over a 3-day period in October 1999 showed a mean of 365.7 µl l−1 with a standard deviation of 10.9. The lack of night-time enrichment raises the possibility of potential effects of elevated CO2 on mitochondrial respiration (Drake et al., 1999), and therefore on plant carbon balance and ultimately on seed quantity and quality. However, as recent evidence suggests only very small effects of elevated CO2 on respiration (Amthor 2000) it seems unlikely that the lack of night time enrichment led to any marked differences in plant carbon balance.
Adult sheep intermittently grazed all rings. Grazing commenced when the total above-ground biomass, averaged across all rings (determined by a pasture meter), reached 1800–2000 kg d. wt ha−1 and continued until the above-ground biomass was reduced to approximately 500–700 kg d. wt ha−1. Both ambient and elevated CO2 rings were grazed at the same time and for the same duration (1–1.5 d). Rings were grazed by opening the gates into each ring and allowing a flock of approximately 200 sheep that had been grazing in the surrounding field for 3–4 d free access to all of the rings. Up to 15 sheep were observed grazing in one ring at any one time. The dates of grazing during the experiments reported here were 6 July, 15 October and 1 December 1999. The grazing procedure raises the possibility of sheep transferring seeds between ambient and elevated CO2 rings in dung (Ozer, 1979). However, as few seeds are excreted in dung within 1 day of ingestion (Ozer, 1979), most seeds excreted during the period rings were grazed would have originated from the field surrounding the rings. Thus, we might expect all rings to receive a constant input of ambient CO2 seeds in dung, and there to be little transfer of seeds from elevated to ambient CO2 rings.
Study species and seed collection
The six species studied were A. odoratum (perennial grass) Cerastium glomeratum Thuill (annual herb), L. saxatilis (perennial herb), L. perenne (perennial grass), Poa. pratensis L. (perennial grass) and T. repens (perennial legume). All were C3 species. Inflorescences of the six study species were collected from early November 1998 (early spring) to late February 1999 (late summer). At least 30 inflorescences were collected per species from each ambient and elevated CO2 ring. All inflorescences were collected before seed dispersal. At the time of seed collection, CO2 enrichment had been occurring for at least 13 months. Fully formed seeds were extracted from each inflorescence and counted. The bulked sample of seeds from each inflorescence was weighed and divided by the number of seeds per inflorescence to determine mean seed mass. A 0.3–0.4 g sample of seeds of each species except C. glomeratum from each ring was analysed for carbon and nitrogen using a carbon-hydrogen-nitrogen analyser (Leco Corporation, St. Joseph, MI, USA). The remaining seeds from inflorescences collected from all dates were then bulked to give one sample for ambient CO2 rings and one for elevated CO2 rings.
A total of 12 seed trays, each measuring 0.3 m wide × 0.4 m long × 0.06 m deep, were filled with sterile compost and laid out in the centre of each ring. The bottom of the trays were raised 0.2 m above the ground surface (0.26 m to soil surface in tray) on top of small wooden pegs. Seeds of the six study species developed on parent plants in ambient and elevated CO2 rings were randomly assigned to the trays (6 species × 2 seed development sources = 12 trays per ring). A total of 200 seeds, randomly selected from the bulked sample from ambient and elevated CO2 rings, were sown on the soil surface of each tray on 23 April 1999 and covered with a 2-mm layer of compost. The mass of the seeds sown into each tray was recorded before sowing as a covariate. Each tray received 1 l of water every fourth day. The trays were checked every 2 d for 32 d with leaf emergence taken as an indication of germination. Seedlings were removed after emergence to prevent double counting. The percentage of seeds germinated after 16 and 32 d was calculated as a measure of germination. Both days were calculated in order to examine treatment effects on the speed of germination.
Seedling growth experiment
Two metal frames each containing 48 pots were laid out in the centre of each ring (i.e. 96 pots per ring). The pots, each measuring 0.04-m square × 0.15-m deep, fitted tightly against neighbouring pots within the metal frame, and the bottom of the pots were raised 0.2 m above the ground surface (0.35 m to soil surface in tray) on top of small wooden pegs. Each pot was filled with sterile compost and watered with half strength Hoagland’s nutrient solution (Hoagland & Snyder, 1933) before seed planting and every 7 d during the experiment. The trays were treated with molluscicide pellets to prevent slug damage to the emerging seedlings. Eight replicates of seeds of the six study species from parent plants grown in ambient and elevated CO2 rings were randomly assigned to the 96 pots in each ring (6 species × 2 CO2 seed development sources × 8 replicates = 96 pots per ring). Two seeds were sown in each pot on 28 April 1999. Each seed was weighed before sowing. Seeds were checked daily for leaf appearance as an indication of germination. The first seedling that germinated was treated as the study seedling and the second seedling was removed as it emerged. Seedlings were harvested and aboveground and belowground structures separated 40 d after leaf emergence in A. odoratum, P. pratensis and T. repens and 32 d after leaf appearance in C. glomeratum, L. perenne and L. saxatilis. The percentage of total biomass made up of root at final harvest was calculated as a measure of biomass allocation.
In the germination and seedling growth experiments, the canopy height of the pasture in the Face rings was typically < 0.1 m. This means the seedlings in the trays would have been at least 0.16 m above the canopy. Thus, there should not have been any adverse shading consequences for germination and seedling growth. In addition, the rings were not grazed by sheep during the germination and seedling growth experiments. Thus, there was no need to protect trays with a cage. No CO2 measurements were taken at the exact height above ground level of the soil surface of the trays in the germination (0.26 m) and seedling growth (0.35 m) experiments. The CO2 measurements made at 0.3 m above ground level (see Face experiment in the Materials and Methods section) are used to represent the CO2 concentration of the seedlings in the trays.
Seedling transplant experiment
Seeds of the six study species from parent plants grown in ambient and elevated CO2 rings were germinated on soil in the glasshouse in autumn 1999. Ten seedlings of similar size from each species and parental CO2 seed source were transplanted into the pasture in each ring about 25 d after leaf emergence. Transplanting was on 23 May 1999 for L. perenne, C. glomeratum and L. saxatilis and on 10 June 1999 for A. odoratum, P. pratensis and T. repens. At transplanting, most grass seedlings had three leaves while the dicotyledonous seedlings had one true leaf. Seedlings were transplanted in a random order at 0.05-m intervals along a 6-m line transect marked out in each ring (6 species × 2 CO2 seed development sources × 10 replicate seedlings = 120 seedlings per transect). Seedlings were transplanted into 0.02-m diameter × 0.03-m deep holes dug in the soil, and firmed into place. While some roots of the vegetation surrounding the transplanted seedlings were damaged, no damage was done to the aboveground vegetation. To give an indication of the conditions experienced by the seedlings at transplanting, the height of the vegetation and the cover of bare ground were measured. The tallest point of the canopy above the soil surface was measured at a location 0.02 m to the west of where each seedling was transplanted. Bare ground was recorded by dropping a 0.0025-m diameter pin at 0.05 m intervals down a transect running parallel (a distance 0.1 m away) to that where seedlings were transplanted and recording hits on canopy (either live or dead vegetation) and bare ground. ANOVA showed no significant effect of elevated CO2 on vegetation height or bare ground cover at either transplanting date (mean vegetation height [cm] ± SE: 23 May, ambient = 5.3 ± 0.8, elevated = 5.9 ± 1.6, 10 June, ambient = 5.7 ± 0.6, elevated = 5.8 ± 0.5; percentage bare ground ± SE: 23 May, ambient = 3.9 ± 0.2, elevated = 4.1 ± 0.4, 10 June, ambient = 3.0 ± 0.2, elevated = 3.2 ± 0.1). The aboveground biomass of surviving seedlings was harvested on 14 October 1999, oven dried and weighed. Seedling harvests occurred on the day before sheep grazed the pastures and over three months since the last grazing had occurred.
Seed sowing experiment
The question of whether or not potential differences in the number of seeds produced under elevated CO2 would influence seedling numbers was tested by a seed addition experiment carried out in the intact vegetation of the Face rings. Four 1.5 m × 1.5 m squares were marked out in each of the 3 ambient and 3 elevated rings in early March 1999. There was one square in each compass quarter of each ring. Each square was further divided into 18 contiguous 0.25 m × 0.5 m quadrats arranged in 6 columns of 3 quadrats. Six of the 18 quadrats in each square were randomly allocated to have no seeds sown in them. These six quadrats acted as unsown controls where seedling recruitment from the natural seed rain and seed bank was measured. One of the unsown quadrats in each square was allocated to each of the six study species for statistical analysis. Seeds of the six study species developed on parent plants grown in ambient and elevated CO2 (12 species-seed development combinations) were randomly allocated to be sown in the remaining 12 quadrats in each square. Only one species-seed development combination was sown in each quadrat. A total of 30 seeds of the appropriate species-seed development combination was sown into each quadrat (equivalent to 240 extra seeds m−2) on 24 March 1999. Seeds were sown by hand onto the canopy and no attempt was made to force seeds to the soil surface. The mass of the seeds sown into each quadrat was recorded before sowing as a covariate. Seedling emergence and survival were measured at approx. monthly intervals from 9 April 1999, with final measurements made on 11 December 1999. At each date, all new seedlings were identified and counted, and the position of the seedling was marked on a map of the quadrat. At subsequent dates, all seedlings were recorded as live or dead, with missing seedlings assigned to the dead category. A seedling herbarium made by the authors that including all species found at the site was used as an aid in seedling identification.
The mass and C : N ratio of seeds were analysed by ANOVA with block and CO2 effects in the model. Analysis for seed mass was based on the mean seed mass averaged across the inflorescences collected from each ambient and elevated CO2 ring. Percentage germination in the germination experiment, leaf and root mass and allocation to root in the seedling growth experiment, and the aboveground mass of seedlings in the transplant experiment were analysed by ANOVA using a split plot design. The CO2 concentration that the seeds or seedlings were planted in was the whole plot factor (ambient or elevated) and the CO2 concentration of the parent plants that the seeds developed on (ambient or elevated) was the subplot factor. A block factor was included in the ANOVA in the transplant experiment where seedlings were transplanted into the existing vegetation but not in the germination or seedling growth experiments where seedlings grew on compost in raised pots in the centre of each ring. The percentage germination data and percentage allocation to root data were arcsine square root transformed before analysis. Where significant interactions where found, post hoc testing of differences between treatment means was carried out using Tukey’s multiple comparison test.
The seed sowing experiment was analysed using a split plot design with the CO2 concentration of the rings that the seeds were sown in as the whole plot factor (ambient or elevated) and seed sowing as the subplot factor (unsown, ambient CO2 seeds sown, elevated CO2 seeds sown). Analysis of seedling numbers and survival was based on seedling counts totalled across the four replicate 0.25 m × 0.5 m quadrats in each ring. The number of seedlings surviving at the final recording on 11 December (end of spring) was analysed for each species using Poisson errors with a log link function. The survival of seedlings that emerged in autumn (April and May) until 11 December was analysed using binomial errors with a logit link function.
The mass of T. repens seeds was greater for seeds collected from plants growing at elevated than ambient CO2 (F1,2 = 27.9, P < 0.05) (Table 1). Seed mass of the remaining five species was unaffected by elevated CO2 during seed development (Table 1). There was no significant of elevated CO2 on the C : N ratio of seeds (Table 1).
Table 1. The mass and C : N ratio of seeds collected from parent plants growing in ambient (360 µl l−1) and elevated (475 µl l−1) CO2 rings for the six study species
Seed mass (mg)
C : N ratio
Seed mass was measured on seeds removed from at least 30 inflorescences per species in each of the three ambient and three elevated CO2 rings and values for each ring were averaged for statistical analyses. C : N ratios were measured on 0.3–0.4 g samples taken from the bulked samples of seeds collected for each ring. Asterisks on elevated value indicate significant effect of elevated CO2 from ANOVA (F1,2) at P < 0.05 (*) level. Where no asterisks are shown, values were not significantly different. nd, not determined. Values are means ± SE (n = 3).
The percentage germination of T. repens seeds was greater for seeds developed at elevated than ambient CO2 after both 16 (F1,4 = 19.7, P < 0.05) and 32 d (F1,4 = 20.6, P < 0.05) (Table 2). When the mass of seeds sown into each tray was incorporated in the ANOVA as a covariate, the effect of elevated CO2 during seed development disappeared and a significant seed mass effect was revealed (16 d: F1,3 = 12.2; 32 d F1,3 = 11.6; both P < 0.05). The correlation of seed mass and germination was significant after both 16 (r = 0.87, n = 12, P < 0.01) and 32 (r = 0.74, n = 12, P < 0.01) d for T. repens. The percentage germination of L. saxatilis seeds was lower for seeds developed at elevated than ambient CO2 after 32 d (F1,4 = 22.7, P < 0.01). When the mass of seeds sown into each tray was incorporated in the ANOVA as a covariate, the effect of elevated CO2 during seed development remained. The correlation between seed mass and germination was not significant (r = 0.23, n = 12, P < 0.1). There was no significant effect of elevated CO2 during seed development on germination in A. odoratum, C. glomeratum, L. perenne and P. pratensis (Table 2). There were no significant effects of elevated CO2 after planting or the interaction of elevated CO2 during seed development and after planting on germination in any species (Table 2).
Table 2. The effects on the percentage germination after 16 and 32 days of the CO2 concentration (ambient or elevated) (a) during the development of seeds on the parent plant and (b) after the seeds were planted. The experiment was a split plot design with the CO2 concentration after seed planting the whole plot factor and the CO2 concentration during seed development the sub plot factor
Period CO2 treatment applied
Data are untransformed values. All tests of significance were conducted on the arcsine square root scale. Asterisks on the elevated value indicate significant main effects of the CO2 concentration during seed development or after seed planting from split-plot ANOVA (F1,4 for each treatment) at P < 0.05 (*) and P < 0.01 (**) level. None of the interactions between CO2 concentration during seed development and after seed planting were significant. Values are means ± SE (n = 3 for after planting, n = 6 for seed development). Ambient = 360 µl l−1 CO2; elevated = 475 µl l−1 CO2.
The root and leaf mass of T. repens seedlings was greater when they were grown from seeds developed at elevated than ambient CO2 (root: F1,4 = 18.8, P < 0.05; leaf: F1,4 = 11.7, P < 0.05) and when they were grown in elevated than ambient CO2 (root: F1,4 = 19.3, P < 0.05; leaf: F1,4 = 12.7, P < 0.05) (Table 3). The interaction was not siwnificant. When the mass of each seed planted was incorporated in the ANOVA as a covariate, the effect of elevated CO2 during seed development disappeared and a significant seed mass effect was revealed (root mass: F1,3 = 14.5; leaf mass F1,3 = 15.1 both P < 0.05). The effect of elevated CO2 during seed development remained. The correlation of seed mass and d. wt was significant for both root (r = 0.63, n = 96, P < 0.05) and leaf mass (r = 0.66, n = 96, P < 0.01) for T. repens. The root and leaf mass of L. saxatilis seedlings was lower when they were grown from seeds developed at elevated than ambient CO2 (root: F1,4 = 94.1, P < 0.05; leaf: F1,4 = 45.1, P < 0.05) but was unaffected by elevated CO2 during seedling growth or the interaction (Table 3). The significant effect of elevated CO2 during seed development remained when seed mass was fitted as a covariate, with the effect of seed mass not significant. There were no significant effects of elevated CO2 during seed development and seedling growth or the interaction on root and leaf mass of C. glomeratum and P. pratensis seedlings.
Table 3. The effects on the root and leaf mass (mg d. wt) of the CO2 concentration (ambient or elevated) (a) during the development of seeds on the parent plant and (b) during the growth of seedlings. Seedlings were grown as individual plants on compost. The experiment was a split plot design with the CO2 concentration during seedling growth the whole plot factor and the CO2 concentration during seed development the sub plot factor
Period CO2 treatment applied
Cerastium glomeratum and Leontodon saxatilis were harvested 32 d after leaf emergence, while Poa pratensis and Trifolium repens were harvested 40 days after leaf emergence. Asterisks on the elevated value indicate significant main effects of the CO2 concentration during seed development or during seedling growth from split-plot ANOVA (F1,4 for each treatment) at P < 0.05 (*) and P < 0.01 (**) level. None of the interactions between the CO2 concentration during seed development and during seedling growth were significant for these four species. Values are means ± SE (n = 3 for seedling growth, n = 6 for seed development). Ambient = 360 µl l−1 CO2; elevated = 475 µl l−1 CO2.
There was a significant effect of the interaction between elevated CO2 during seed development and during seedling growth on the root and leaf mass of A. odoratum (root: F1,4 = 8.2, P < 0.05; leaf: F1,4 = 10.4, P < 0.05) and L. perenne (root: F1,4 = 49.2, P < 0.01; leaf: F1,4 = 11.7, P < 0.05) seedlings (Fig. 1). For both species, elevated CO2 during seedling growth increased the root and leaf mass of seedlings. Elevated CO2 during seed development resulted in increased root and leaf mass only when seedlings were grown at elevated CO2. This effect remained when seed mass was fitted as a covariate, with the effect of seed mass not significant.
The percentage of biomass at final harvest allocated to roots was greater for seedlings grown at ambient than elevated CO2 for A. odoratum (ambient = 32.0%, elevated = 39.6%, F1,4 = 8.7, P < 0.05) and P. pratensis (ambient = 31.3%, elevated = 33.4%, F1,4 = 13.4, P < 0.05) but unaffected by elevated CO2 during seedling growth for the other species (C. glomeratum: ambient = 23.9%, elevated = 24.2%; L. saxatilis: ambient = 21.7%, elevated = 22.6%, L. perenne: ambient = 33.6%, elevated = 34.9%; T. repens: ambient = 38.4%, elevated = 39.5%). There was no significant effect of elevated CO2 during seed development or the interaction of elevated CO2 during seed development and seedling growth on the percentage of biomass allocated to root in any species.
There were no significant effects of elevated CO2 during seed development or during seedling growth, or the interaction, on the aboveground biomass of seedlings of any species transplanted into the pasture (mean aboveground biomass [mg d. wt per plant] averaged across all treatments: A. odoratum= 17.7, C. glomeratum= 36.4, L. saxatilis= 30.1, L. perenne= 20.1, P. pratensis= 18.3, T. repens= 106.3).
Seed sowing experiment
A total of 2147 new seedlings of the six species were recorded with 1987 (92%) of these emerging in April and May (autumn). There were more seedlings in both sown and unsown plots at the final recording on 11 December (end of spring) in elevated than ambient CO2 rings for T. repens (χ2 = 24.7, df = 1, P < 0.01) and L. saxatilis (χ2 = 21.9, df = 1, P < 0.01) (Fig. 2). There was no significant effect of the CO2 concentration of the ring on seedling numbers for the other four species (Fig. 2). There was a significant effect on the number of seedlings at the final recording of the seed sowing treatment for A. odoratum, L. saxatilis, L. perenne and T. repens but not for C. glomeratum and P. pratensis (Fig. 2). For the four species where there was a significant effect of seed sowing, neither the ambient or elevated CO2 levels of the seed sowing factor could be combined with the unsown level without a significant change in deviance, indicating sowing both ambient and elevated CO2 seeds increased seedling numbers. For A. odoratum, L. saxatilis and L. perenne, combining the ambient and elevated CO2 levels of the seed sowing factor into a single level did not result in a significant change in deviance, thus indicating no difference in seedling numbers between plots sown with ambient and elevated CO2 seeds. For T. repens, however, there were more seedlings where seeds developed at elevated CO2 were sown as combining the ambient and elevated CO2 levels of the seed sowing factor did result in a significant change in deviance (χ2 = 3.9, df = 1, P < 0.05). There were no significant interactions between seed sowing and the CO2 concentration of the rings that the seeds were sown into on the number of seedlings at the final recording for any of the species.
Further analyses of the difference in seedling numbers between sown and unsown plots on 11 December (Poisson errors and log link) showed that this was greater for T. repens seeds developed at elevated than ambient CO2 (χ2 = 9.1, df = 1, P < 0.01) but was unaffected by seed source for the other species (Fig. 2). When the mass of seeds sown was incorporated into the analysis as a covariate for T. repens, the effect of seed source disappeared and a significant seed mass effect was revealed (χ2 = 9.6, df = 1, P < 0.01). The correlation of seed mass and seedling numbers was significant for T. repens (r = 0.59, n = 12, P < 0.05). The size of the difference between sown and unsown plots did not differ between ambient and elevated CO2 rings for any of the species.
Seedling survival of all species from emergence in autumn until end of the spring was unaffected by the addition of seeds, the CO2 concentration of the rings that the seeds were sown into, or the interaction of these effects (mean percentage of seedlings surviving averaged across all treatments: A. odoratum= 66%, C. glomeratum= 68%, L. saxatilis= 47%, L. perenne= 71%, P. pratensis= 60%, T. repens= 55%).
This study was designed to evaluate the consequences of elevated atmospheric CO2 during the development of seeds on the parent plant and after seed dispersal to seedling performance for a range of pasture species. One of the most striking results was that seeds of T. repens harvested from plants grown at elevated CO2 had higher germination, produced seedlings which had greater mass when grown as individuals in compost and resulted in more seedlings when a constant number of seeds were sown into the pasture than seeds harvested from plants grown at ambient CO2. Consistent with previous studies (Scheidegger & Nösberger, 1984; Nijs et al., 1989), the mass of T. repens seedlings grown as individuals on compost was also higher when seedlings were grown at elevated CO2. The seed development effects are probably attributable to the fact that T. repens seeds developed at elevated CO2 had greater mass. When seed mass was fitted as a covariate in each experiment, the effect of elevated CO2 during seed development was replaced by a strong effect of seed mass. Previous studies have often shown that seed mass is positively correlated with germination, early seedling growth and the ability of seedlings to establish in dense vegetation (Crawley & Nachapong, 1985; Parrish & Bazzaz, 1985; Bewley & Black, 1994).
In contrast to T. repens, seeds of L. saxatilis harvested from plants grown at elevated CO2 had reduced germination and produced seedlings which had lower mass when grown as individuals on compost than seeds harvested from plants grown at ambient CO2. The reduced seedling performance of elevated CO2-developed seeds is consistent with recent studies with Arabidopsis thaliana Heynh. (Andalo et al., 1996, 1998) and Bromus rebens L. (Huxman et al., 1998). The result did not appear to reflect differences in seed mass as this was unaffected by elevated CO2 during seed development in L. saxatilis. A possible explanation is that the result reflects the effects of elevated CO2 on the nutrient content of the seeds. Although not significantly different, the C : N ratio of L. saxatilis seeds developed at elevated CO2 was 12% higher than those developed at ambient CO2. Previous studies have noted that seeds with higher N concentrations can have improved seedling performance (Parrish & Bazzaz, 1985; Huxman et al., 1998).
The root and leaf mass of A. odoratum and L. perenne seedlings grown as individual plants on compost were greater for seeds developed at elevated CO2 but only when the seedlings were grown in the elevated CO2 atmosphere. A similar interaction was found by Bezemer et al. (1998), who reported plants of Poa annua L. exhibited a much higher responsiveness to elevated CO2 when the seeds from which they were grown originated from plants reared at elevated than ambient CO2. The mechanism underlying the interaction in our study is unclear at this stage. CO2 stimulation of seedling growth has been reported to increase with increasing seed mass at the intraspecific (e.g. Miao, 1995; Steinger et al., 2000) and interspecific (Bazzaz & Miao, 1993) level. However, interactions with seed mass or seed quality are unlikely to be the case in this study as seed mass and seed C : N ratios of A. odoratum and L. perenne were unaffected by elevated CO2 during seed development.
A notable feature of the results was that none of the effects of elevated CO2 during seed development or during seedling growth detected for A. odoratum,L. perenne, L. saxatilis and T. repens seedlings grown as individual plants on compost were found when seedlings were transplanted into the pasture. The contrasting response may reflect that other factors affecting growth (e.g. herbivory, competition for light and soil nutrients) were more limiting to plant growth for seedlings grown in the pasture than for seedlings grown as individual plants with ample nutrients and water (Woodward et al., 1991; Bazzaz & McConnaughay, 1992; Leishman et al., 1999). An alternative explanation for the failure of effects of elevated CO2 during seedling growth to be evident for transplanted seedlings may be that seedlings at the base of the canopy in both ambient and elevated CO2 rings experienced high CO2 concentrations (Fuller, 1948; Bazzaz & Williams, 1991), and so were less likely to experience competition for CO2 as a resource (Bazzaz & McConnaughay, 1992). However, measurements of the CO2 concentration taken in the pasture canopy (mean pasture height = 0.08 m) at 0.015 m above ground level during the photoperiod over a 5-d period at the conclusion of our experiment in late spring gave mean values for ambient (380 µl l−1) and elevated (482 µl l−1) CO2 rings that were both similar to those found at 0.3 m above ground level (i.e. where the germination and seedling growth experiments were conducted; P.C.D. Newton, unpublished). This finding suggests it is unlikely that high CO2 concentrations at the base of the canopy for transplanted seedlings in both ambient and elevated CO2 rings contributed to the difference between experiments in our study. Regardless of the exact mechanism leading to the differing responses between experiments, the results highlight further that the response of individually grown plants to elevated CO2 can differ substantially from those within monospecific or multispecific stands (Bazzaz & McConnaughay, 1992) and stresses the need for caution in extrapolating from studies with individual plants to community responses (Navas, 1998).
There were about 20% more seedlings of L. saxatilis and T. repens in elevated than ambient CO2 rings where no seeds were sown. This result is likely to partly reflect increased seed production for these two species under elevated CO2. A previous study in the same grassland showed that L. saxatilis and T. repens both showed increased seed production under elevated CO2 (Edwards et al., 2001). Moreover, both species appeared to be seed-limited as they showed increased seedling numbers when extra seeds were sown in this study. For T. repens, the result may also be partly attributable to effects of elevated CO2 during seed development: the number of extra seedlings produced by sowing a constant number of seeds was greater when the sown seeds were developed at elevated than ambient CO2. An alternative explanation for the increased seedling recruitment, that is, differences in the extent and quality of establishment microsite conditions between CO2 treatments leading to elevated CO2 rings being more conducive to seedling recruitment, is unlikely because the number of extra seedlings that arose from the sowing of a constant number of seeds was similar between ambient and elevated CO2 rings.
Seedling survival in the pasture from the start of autumn until the end of spring was unaffected by elevated CO2 during seedling growth. Reduced seedling survival in elevated CO2 rings may have been expected had the biomass of the existing vegetation or the rate of canopy closure been enhanced under elevated CO2 (Bazzaz et al., 1992). Previous studies in permanent grasslands like the one studied here have shown that experimental treatments that cause increased above-ground biomass (e.g. herbivore exclusion, fertilizer application) often result in reduced seedling survival due to increased competition for light, water and nutrients (Howe & Snaydon, 1986; Edwards & Crawley, 1999). However, there was little effect of elevated CO2 on the biomass of the above-ground vegetation or the rate of growth between grazing events for the duration of this experiment (Edwards et al., 2001).
It is unclear at this stage whether the differences in seedling numbers of L. saxatilis and T. repens would persist in the long term and lead to differences in species composition. Seedling mortality is often high in pastures and differences in seedling numbers, even like those reported here which are for seedlings of 7–8 m of age, often disappear due to density-dependent seedling mortality or effects of competing vegetation (e.g. Edwards & Crawley, 1999). However, it is interesting to note that both L. saxatilis and T. repens showed increased abundance in the biomass in elevated CO2 rings in the first 2 yr of the experiment (Edwards et al., 2001; Newton et al., 2001). This suggests that increased seedling recruitment may be an important mechanism leading to changes in species composition, although how important they are relatively to changes in vegetative growth under elevated CO2 (e.g. tillering and branching, mortality of adult plants, see Clark et al., 1997) is unclear. However, the importance of seedling recruitment as a mechanism causing the increased abundance of L. saxatilis is highlighted by the finding that the basal diameter of mature rosettes of this species did not differ between ambient and elevated CO2 rings (Edwards et al., 2001).
This study used seeds collected from a pasture that had been exposed to elevated CO2 for at least 13 m. Despite this time period, some of the plants that seeds were collected from in the CO2-enriched rings may have been exposed to ambient CO2 in the pretreatment period; all study species except C. glomeratum were perennial. This raises the possibility of differences in storage (e.g. in roots, stolons) laid down in the pretreatment period interacting with elevated CO2 in the treatment period to influence seed quality, mass and number, and subsequent seedling performance (see Field et al., 1996). It is not clear from this study, however, whether variations in the amount of storage between species played any role in determining the differing responses between perennial species (e.g. response in T. repens but not P. pratensis). Continued monitoring of the perennial plants in this grassland with reference to the length of time that they have been enriched with CO2 will be required to ascertain the importance of these effects.
In summary, the effect of elevated CO2 during seed development on seedling performance (germination and growth) of individual spaced plants of the six study species was highly variable, ranging from no impact in C. glomeratum and P. pratensis to reduced performance in L. saxatilis and increased performance in A. odoratum, L. perenne and T. repens. Such species-specific effects on seedling performance, in addition to changes in the number of seeds produced under elevated CO2 (Edwards et al., 2001), could potentially lead to changes in species composition. However, our study also highlighted that the effects on seedling performance due to elevated CO2 during seed development were much smaller when measured within intact pasture communities. Future work aimed at predicting the effect of elevated CO2 on plant communities must not only consider whether elevated CO2 alters reproductive parameters like seed mass and quality, but also under what conditions changes in these factors will lead to changes in germination, seedling growth and seedling numbers.
The Face experiment was funded by research grant C10632 from the New Zealand Foundation for Research, Science and Technology. Grant Edwards was funded by a New Zealand Science and Technology Post-Doctoral Fellowship. We thank Peter Curtis and two anonymous referees for helpful comments on the manuscript that greatly improved the clarity of the paper. We thank Elaine Glasgow for help with field work.