Global warming reduces plant reproductive output for temperate multi-inflorescence species on the Tibetan plateau


Author for correspondence:
Shucun Sun
Tel: +86 25 83686670


  • Temperature is projected to increase more during the winter than during the summer in cold regions. The effects of winter warming on reproductive effort have not been examined for temperate plant species.
  • Here, we report the results of experimentally induced seasonal winter warming (0.4 and 2.4°C increases in growing and nongrowing seasons, respectively, using warmed and ambient open-top chambers in a Tibetan Plateau alpine meadow) for nine indeterminate-growing species producing multiple (single-flowered or multi-flowered) inflorescences and three determinate-growing species producing single inflorescences after a 3-yr period of warming.
  • Warming reduced significantly flower number and seed production per plant for all nine multi-inflorescence species, but not for the three single-inflorescence species. Warming had an insignificant effect on the fruit to flower number ratio, seed size and seed number per fruit among species. The reduction in seed production was largely attributable to the decline in flower number per plant. The flowering onset time was unaffected for nine of the 12 species. Therefore, the decline in flower production and seed production in response to winter warming probably reflects a physiological response (e.g. metabolic changes associated with flower production).
  • Collectively, the data indicate that global warming may reduce flower and seed production for temperate herbaceous species and will probably have a differential effect on single- vs multi-inflorescence species.


The Earth’s mean global surface temperature is projected to increase in the foreseeable future (Solomon et al., 2007). Models and recent trends indicate that temperatures are expected to increase to a significantly greater extent in nongrowing (‘winter’) seasons, particularly at high latitudes and altitudes (Bonsal et al., 2001; Shabbar & Bonsal, 2003; Solomon et al., 2007). Thus, there is a need for research focusing on the effects of winter warming on communities and ecosystems that are expected to experience the greatest climatic changes during the winter months, for example alpine and Arctic plant communities (Campbell et al., 2005; Bokhorst et al., 2008).

One potential effect of winter warming is a change in seed production as a consequence of a change in flowering phenology and/or a direct decrease in flower production. Artificial warming has frequently been reported to either advance or delay the time of flowering (Sherry et al., 2007; Hoffmann et al., 2010). The advance of flowering may lead to smaller flowering plants that may be attended by a reduction in the resources to produce seeds (resulting in a decrease in seed number per fruit, for example). A delay in the time to flowering may reduce the amount of time during the growing season for plants to fully develop fruits and seeds (resulting in small seeds, for example). It is also possible that warming per se might negatively affect floral formation and development by disrupting a vernalization requirement for flower production. Many plants adapted to temperate climates require vernalization and must experience a period of low winter temperature to initiate or accelerate flower production (Chouard, 1960). Plants growing at lower temperatures often have more flowers than their counterparts growing in warmed conditions (Walch & van Hasslet, 1991), presumably because low temperatures accelerate the floral transition by restraining somatic growth (Tromp, 1984). For example, Dianthus species flower more abundantly following vernalization (Chouard, 1960), as do Brunonia, Calandrinia and Lavandula angustifolia (Niu et al., 2002). By contrast, higher than normal temperatures can delay flowering and reduce flower number in many species (Warner & Erwin, 2006), because they may depress flower formation by diminishing or eliminating the effect of vernalization (Hideyuki & Takashi, 2003). Warming has been proven to restrict vernalization and to result in a reduction in flower or seed production (Saure, 1985; Hennessy & Claytongreene, 1995).

However, few studies have successfully explored the effect of winter warming on plant seed production of temperate species, in part because of the logistical problems of simulating warmer than normal temperatures in winter under field conditions. An additional limitation to the available information is that many experiments are designed to set temperatures consistently higher than controls, such that seasonal variations in temperature are not considered (Conant et al., 2008; Byrne et al., 2009). In the field, winter warming is hard to realize because of snow cover (Kennedy, 1995; Alatalo & Totland, 1997; Hobbie & Chapin, 1998). In particular, open-top chamber (OTC) warming facilities work in a passive way; relative to the control, air temperature within OTCs is often increased to a greater extent in summer than in winter (Mitchell, 1990). Alatalo & Totland (1997) have suggested that winter temperatures above the snow cover do not affect plant growth for species that typically grow in snow-covered sites. This perhaps reflects the fact that, although many studies have recorded a decline in seed production in response to warming (e.g. Bokhorst et al., 2008; Hovenden et al., 2008), few have considered ‘vernalization’ as a viable explanation for the decline in seed production under experimental warming.

Here, we report the results of a long-term winter warming experiment in the Tibetan Plateau, which is one of the regions that has experienced the most rapid temperature increase in the world and which has been shown to exhibit a greater increase in the nongrowing season than in the growing season (Lu & Liu, 2010). A recent remote sensing analysis has revealed that the onset of spring greening has been delayed, resulting in a shorter growing season, and the delay has been attributed to a reduction in winter vernalization requirements for the onset of spring growth (Yu et al., 2010). This observation highlights the importance of winter warming to plant growth, and points to a need for the further study of reproduction. In order to test whether this winter warming negatively affects seed production by means of weakening vernalization, we conducted an artificial warming experiment using OTCs that achieved significant seasonal-biased warming in the nongrowing season to mimic model simulations (Zhao et al., 2002; Mearns et al., 2003). In the fourth year after warming, we examined a suite of variables, including flowering onset time, total seed mass and number, seed size, seed number per fruit, fruit set ratio (the ratio of the number of fruits to flowers) and flower number per plant, for 12 coexisting herbaceous species in an alpine meadow. This gamut of variables was intentionally selected to cope with the potential effects of using OTCs on wind-pollinated vs biotically pollinated species. For example, pollination efficiency might be reduced in wind-pollinated species, whereas the use of OTCs probably has little differential effect on variables such as flower number or size, or the onset of flowering, as a function of a pollination syndrome of a species. As will be shown, the use of OTCs manifested no obviously consistent effects on wind-pollinated vs biotically pollinated species.

Our goal was to determine whether seed production was changed in these species and to identify the factors that contributed most to any change in seed production. We hypothesized that vernalization would be a viable explanation if flower production accounted for the largest variation in seed production among the species. In addition, because this alpine meadow is cold and wet, and hence plant growth is limited by low temperatures, plant height and aboveground biomass should be higher in warmed compared with unwarmed treatments (see Li et al., 2011). In addition, because herbaceous species often require high light intensities during flowering time, such species need to grow tall to flower in dense communities (Sun & Frelich, 2011). To this end, we investigated community height for both treatments and flowering height (i.e. the height at which plants flower) for each species in order to determine whether warming systematically increased biomass allocation to height growth (i.e. increased flowering height), leading to reduced seed production. These results are discussed in the context of the potential effects of winter warming on plant community structure.

Materials and Methods

Study site

Our study was conducted in the Hongyuan Alpine Meadow Ecosystem Research Station (Chinese Academy of Sciences), located in Sichuan Province in the eastern Qinghai-Tibetan Plateau (32°48′N, 102°33′E). The altitude is 3500 m. The climate is cold, continental and characterized by a short and cool spring, summer and autumn, and a long winter. In the period 1961–2008, as recorded by Hongyuan County Climate Station (located 5 km from the study site), the annual mean temperature was 0.9°C, with maximum and minimum monthly means of 10.9 and −10.3°C in July and January, respectively. The annual mean precipitation was 690 mm (80% of which occurs between May and August).

The meadow in which the study was conducted is dominated by sedges (e.g. Blysmus sinocompressus and Carex enervis ssp. chuanxibeiensis) and grasses (e.g. Deschampsia caespitosa). Forb species, including Anemone trullifolia var. linearis, Potentilla anserina, Haplosphaera himalayensis, Aster alpinus and Gentiana formosa, are also occasionally abundant. The total vegetation coverage in mid-summer is over 90%; the average maximum plant height is c. 30 cm (Wu et al., 2011). The soil is characterized by a high organic content (250 g kg−1) and low total nitrogen (N) (8 g kg−1) and phosphorus (P) (5 mg kg−1).

Experimental warming

In 2007, we fenced a 0.5-ha plot in which > 90% of the angiosperm species in the area could be found in any 25 × 25 cm2 patch (Liu et al., 2011). Twenty OTCs measuring 2 × 2 × 2 m3 (rectangular without hanging roof) were installed in random locations within the plot with at least 3 m distance between the chambers. Half of the chambers were designated as ambient OTCs. These were surrounded with a thin (< 0.1 mm) steel screen with a mesh size of 0.2 × 0.2 mm2. The remaining OTCs were designated warm OTCs. These were covered with polycarbonate film with a light transparency of > 90%. All the chambers were sunk to a depth of 20 cm in the soil. The ambient OTCs were deliberately installed to remove the confounding effect of the physical setting of OTCs on plant reproduction. Honeybees (Apis cerana cerana) and lepidopteran species are abundant, and many are pollinators and hence may be responsible for seed production for some of the herbaceous species. Birds occasionally visit and feed on the seeds of plants in the study site. The ambient OTCs act as the same barrier for the animals as do the warmed OTCs. Moreover, measurements showed that the aboveground plant biomass was indistinguishable between ambient OTCs (N = 5, mean = 345.4 ± 7.4 (SE) g m−2) and field (354.3 ± 8.6 (SE) g m−2) within the fenced plot in the sampling year (t = 0.877, = 0.401). The plant height was visibly comparable between ambient and field conditions.

Three years (2007–2010) of continuous measurements (at 15-min intervals) using thermometers and data-loggers placed in the centers of the OTCs (DS1921G, Maxim Integrated Products, Sunnyvale, CA, USA) showed that the mean annual temperature at the soil surface was, on average, 1.5°C higher in the warmed compared with the ambient chambers over the experimental period (see Supporting Information Table S1).

In our study site, sunlight is intense throughout the year and, on average, nearly 300 d yr−1 are sunny (as recorded by Hongyuan County Climate Station). Because of the high-altitude area, the study site is characterized by large day–night temperature differences; for example, the air temperature may increase above 20°C at midday and drop below freezing at night in the winter. Although snowfall frequently occurs during the nongrowing season, during the duration of our experiments snow cover rarely persisted for > 3 d (Y. Liu, pers obs). Thus, we reasoned that the effects of winter warming, if present, would be apparent (data showed a 2.4°C increase at the soil surface during the nongrowing season and a c. 0.4°C increase during the growing season; see Table S1). Finally, measurements during the growing season showed that the daily average relative humidity exceeded 80%, and that it was slightly higher in the ambient OTCs (86.2%) than in the warmed OTCs (83.4%; Li et al., 2011).

Species selection

The OTCs contained c. 20 C3 perennial herbaceous species, including three grass species, three sedge species and 14 forb species. We tried to sample all the species from the meadow as much as possible. OTCs that contained three or more fruiting plants per species were sampled for each species, and only those species that were sampled in three or more OTCs were included in this study. This resulted in 12 species available for study (nine forb, two sedge and one grass species), nine of which were characterized by indeterminate growth and bore multiple (single-flowered or multi-flowered) inflorescences per plant (‘multi-inflorescence species’; eight eudicot species and one monocot species), and three of which were characterized by determinate growth and bore only one inflorescence per plant (‘single-inflorescence species’; all monocots). In addition, the eight eudicot species are primarily animal pollinated, whereas the four monocot species (two sedge species and two grass species) are mainly wind pollinated (see Table 1 and Fig. S1). These 12 species accounted for > 80% of the total aboveground biomass in both warmed and ambient OTCs (data not shown).

Table 1.   The flower type and pollination syndrome of the sampled species
SpeciesFamilyFlower vs inflorescenceSingle- vs multi-flower/inflorescenceInsect vs wind pollinationN of warmed OTCN of ambient OTC
  1. The numbers of sampled warmed and ambient open-top chamber (OTCs) are also provided for each species.

Anemone trullifolia var. linearisRanunculaceaeFlowerMultipleWind1010
Caltha scaposaRanunculaceaeFlowerMultipleInsect610
Gentiana formosaGentianaceaeFlowerMultipleInsect1010
Ranunculus brotherusiiRanunculaceaeFlowerMultipleInsect510
Trollius farreriRanunculaceaeFlowerMultipleInsect510
Aster alpinusAsteraceaeInflorescenceMultipleInsect610
Chamaesium paradoxumApiaceaeInflorescenceMultipleInsect410
Deschampsia caespitosaPoaceaeInflorescenceMultipleWind1010
Haplosphaera himalayensisApiaceaeInflorescenceMultipleInsect410
Allium sikkimenseLiliaceaeInflorescenceSingleInsect39
Blysmus sinocompressusCyperaceaeInflorescenceSingleWind1010
Carex enervis ssp. chuanxibeiensisCyperaceaeInflorescenceSingleWind1010

Reproductive variables

The reproductive status of the 12 species was investigated from March 2010 to early October 2010. We recorded the time of flowering onset, flower or inflorescence number per plant (here-after referred to as the flower number), fruit number per plant, seed number per fruit and seed size for each species. Because of the different flowering phenologies (from early May to late August), the time each species was observed varied among the 12 species. At least three (and up to 10) plants per species in each OTC were randomly tagged for investigation. At least three OTCs containing 10 plants per species were sampled for each treatment (warmed vs ambient).

For each species, we recorded the flowering height as the maximum height of plant photosynthetic tissue (see also Cornelissen et al., 2003) on the day on which plants flowered, and the plant density within each OTC in mid-August when the plant stem density was expected to be at its peak. We also recorded the number of flowering plants and number of flowers (or inflorescences) per plant every 3–7 d, depending on the species. After flowering, we recorded the number of fruiting inflorescences per plant and calculated the fruit set ratio (i.e. the ratio of fruiting inflorescence number to inflorescence number per plant). We also calculated the proportion of reproductive plants as the ratio of the number of flowered plants to the total plant number per OTC per species. When fruits were mature and were about to abscise, we collected fruits from the mid-region of inflorescences located at mid-plant height for at least three plants per species from at least three OTCs. We subsequently counted the viable and aborted seeds within each fruit; the viable seeds were oven dried at 65°C for 48 h and then weighed to estimate seed size (individual seed dry mass = total seed mass divided by seed number). The seed set ratio, calculated as the ratio of the viable seed number to the total seed number within fruits, was also obtained for each species.

Community traits

The average plant density and height for each species and at the community level were calculated for the ambient and warmed OTCs as follows. For each of the two treatments, the total plant density (at the community level) was calculated as the total plant number in all OTCs sampled in July when the plant density was highest. The relative species density was calculated as the total number of plants per species divided by the total plant density. Subsequently, the relative species density was multiplied by the average height of the plants of each species, which was summed to obtain the community height for each OTC, which was further averaged to obtain the average community height per treatment.

Data analysis

All data were averaged first for each OTC and then for each treatment and each species; the data were then log10 transformed to achieve normality for subsequent statistical analyses. We used one-way ANOVA to examine the effects of experimental warming on seed size, seed mass or seed number per plant, seed number per fruit, flower (or inflorescence) number per plant, fruit set ratio, flowering height and proportion of flowering plants for each species. One-way ANOVA was also used to test the significance level of the differences in community height and community density between warmed and ambient OTCs.

A generalized linear mixed model (GLMM) analysis was performed to determine the extent to which reproductive variables (e.g. seed size, seed number per fruit and flower number) and warming explain the variation in total seed mass and seed number (per plant) among species and OTCs. We selected total seed mass and seed number per plant as dependent variables, warming, seed size, seed number per fruit, flower number and fruit set ratio as independent fixed factors, and species as a random factor. Because fruit number per plant and seed number per plant should be closely associated with flower number per plant, they were not included in the GLMM analysis.

In order to further determine whether the effect of warming on seed production could be explained by the flower number per plant, regression analysis was used to examine the relationships between the ratios of the difference in seed production (total seed number and mass) between the treatments (i.e. the difference in total seed number/total seed mass in ambient vs elevated OTCs) and the difference in flower number. All data analyses were completed using the software Statistica 6.0 (Statsoft Inc., 2001).

In addition, because cross-species (regression) relationships may arise from phylogenetic relationships among species (Felsenstein, 1985; Harvey & Pagel, 1991), a generalized least-squares (GLS) approach was used (Pagel, 1999; Freckleton et al., 2002) to estimate the strength of the phylogenetic effect on the relationship between flower production and seed production (total seed mass and number). A maximum likelihood framework was used to estimate the index of phylogenetic association λ, which ranged from zero (indicating phylogenetic independence) to unity (indicating complete phylogenetic dependence). The likelihood ratio test was also used to compare whether the model with the maximum likelihood value of λ differed from models with λ values of zero or unity. The phylogenetic tree was constructed following the program Phylomatic (Webb et al., 2008) and the ‘Flora of China’ (ECCAS (Editor Committee of the Chinese Academy of Sciences for Flora of China), 1974–1999), with the branch lengths being a constant. Simple regressions in the phylogenetic framework (GLS method) were used to analyze the relationships between flower number per plant and seed production. The GLS analysis was performed in R version 2.14 (R Development Core Team, 2011) using the package APE (Paradis et al., 2004).


Warming effect on seed production

Warming decreased significantly the total seed mass and total seed number in nine species, all of which were multi-inflorescence species, but did not have an observable effect on the three single-inflorescence species (Fig. 1a,b; Table S2). Collectively, the negative effect of warming was statistically significant for the pooled species set (= 0.046; binomial test). Warming decreased total seed mass and number by 56.7% and 60.4% on average, respectively, for the multi-inflorescence species. The decrease in total seed mass ranged between 30.9% (H. himalayensis) and 81.9% (D. caespitosa) among the multi-inflorescence species, and the decrease in total seed number was between 10.2% and 84.1% across all species.

Figure 1.

The effect of warming on total seed mass (a) and total seed number (b) per plant, flower number (c) and fruit number (d) per plant, fruit set ratio (e), seed set ratio (f), seed size (g) and seed number per fruit (h) of the three single- (closed circles) and nine multi-inflorescence (open circles) species from an alpine meadow, as indicated by the mean values (± 1SE) of ambient vs warmed open-top chambers (OTCs). Because the means are close for some species, not all the species are visible in some panels (e.g. for the flower number per plant, fruit number per plant and fruit set ratio). Generally, the values under the dashed line denote a negative effect on warming, and vice versa for those above the line. See text and Supporting Information Table S2 for details of statistical significance.

Warming reduced significantly both the flower number and fruit number per plant for the multi-inflorescence species (see, however, A. trullifolia var. linearis, = 0.09; Fig. 1c,d; Table S2), but had no effect on single-inflorescence species. The reduction in the proportion of seed production was not significantly different between grass and forb species (F = 1.553, = 0.241). In addition, the fruit set ratio was indistinguishable between warmed and ambient chambers for all species (Fig. 1e); the seed set ratio was enhanced significantly by warming in two animal-pollinated forb species (Trollius farreri and G. formosa) and in one wind-pollinated sedge species (C. enervis ssp. chuanxibeiensis), but was unaffected for the other nine species (Fig. 1f; Table S2), indicating an insignificant warming effect on the seed set ratio at the community level (= 0.146; binomial test).

Warming had no consistent effect on the other reproductive variables of interest between the two species groups (P > 0.05 for each variable, binomial test). For example, warming increased significantly the seed size for Ranunculus brotherusii and decreased it for T. farreri (Fig. 1g; Table S2), both of which are multi-inflorescence species. Likewise, the seed number per fruit was decreased significantly by warming in D. caespitosa, A. trullifolia var. linearis, G. formosa and Chamaesium paradoxum, but was increased significantly for T. farreri, all of which are multi-inflorescence species. None of the single-inflorescence species were affected significantly by warming with regard to the seed number per fruit (Fig. 1h; Table S2), and the proportion of flowered plants was not significantly different between warmed and ambient OTCs for each species (Table 2).

Table 2.   Means (± SE) of flowering onset time (FOT), flowering height (FH), length of flowering time (LFT) and proportion of flowered plants (PEP) for 12 species (nine multi-inflorescence species and three single-inflorescence species) in both warmed and ambient open-top chambers (OTCs)
 Ambient OTCsWarmed OTCsP
  1. The significance level of the difference in the parameters between warmed and ambient OTCs is denoted by P values.

Multi-inflorescence species
 Anemone trullifolia var. linearisLFT61 ± 6.8149 ± 6.940.246
FOT131 ± 2.84138 ± 2.810.112
FH30.8 ± 1.0438.9 ± 0.93<0.001
PEP87.00 ± 0.9587.45 ± 0.720.709
 Aster alpinusLFT39 ± 7.3243 ± 5.680.725
FOT161 ± 4.68159 ± 2.560.789
FH35 ± 0.9640.4 ± 1.790.013
PEP95.00 ± 1.6996.15 ± 2.130.671
 Caltha scaposaLFT43 ± 6.8326 ± 5.190.071
FOT126 ± 3.12130 ± 2.050.335
FH16 ± 0.3414.5 ± 0.430.098
PEP59.50 ± 1.5361 ± 2.080.557
 Chamaesium paradoxumLFT23 ± 5.6544 ± 4.070.023
FOT177 ± 3.86168 ± 1.850.124
FH50.3 ± 1.7652.8 ± 1.110.902
PEP35.36 ± 1.5032.56 ± 9.460.750
 Deschampsia caespitosaLFT58 ± 6.3947 ± 3.760.173
FOT164 ± 6.81180 ± 1.660.031
FH99 ± 1.89110 ± 1.75<0.001
PEP97.90 ± 0.5598.95 ± 0.410.136
 Gentiana formosaLFT100 ± 9.9264 ± 7.420.061
FOT218 ± 3.02227 ± 2.810.312
FH18.4 ± 0.7026.8 ± 0.89<0.001
PEP100.00 ± 0.00100 ± 0.00 
 Haplosphaera himalayensisLFT55 ± 7.2471 ± 9.480.177
FOT157 ± 3.37153 ± 4.40.460
FH32.6 ± 1.3241.1 ± 1.01<0.001
PEP96.90 ± 0.3996.75 ± 0.450.803
 Ranunculus brotherusiiLFT83 ± 7.5875 ± 10.470.503
FOT147 ± 2.98155 ± 2.890.050
FH38.4 ± 1.0245.1 ± 1.14<0.001
PEP97.05 ± 0.3996.65 ± 0.440.501
 Trollius farreriLFT43 ± 5.0739 ± 13.820.753
FOT144 ± 2.21143 ± 5.920.879
FH38.4 ± 1.6146.4 ± 1.30.018
PEP76.76 ± 2.1776.59 ± 2.830.961
Single-inflorescence species
 Allium sikkimenseLFT47 ± 6.3758 ± 8.060.322
FOT220 ± 2.57217 ± 3.320.468
FH36.6 ± 1.0140.9 ± 1.270.016
PEP97.45 ± 0.5398.05 ± 0.480.409
 Blysmus sinocompressusLFT16 ± 3.4723 ± 7.050.365
FOT183 ± 2.25176 ± 5.20.175
FH29.6 ± 2.6532.3 ± 3.30.911
PEP19.25 ± 1.6319 ± 1.650.915
 Carex enervis ssp. chuanxibeiensisLFT43 ± 12.8941 ± 4.340.938
FOT137 ± 5.46134 ± 1.970.658
FH29.6 ± 2.6532.3 ± 3.30.911
PEP4.20 ± 0.284.35 ± 0.260.698

Relationships among reproductive variables

Warming and the reproductive variables accounted for 94% of the variation in total seed mass among OTCs and among species (< 0.001). Among the reproductive factors, flower number per plant contributed most to the variation, followed by seed size and seed number per fruit (Table 3). Similarly, the variation in total seed number per plant was largely attributable to the interaction between species and warming, flower number per plant, seed number per fruit and fruit set ratio (r2 = 0.950, < 0.001). Among the reproductive variables, flower number per plant contributed the most to the variation, followed by the seed number per fruit (Table 3).

Table 3.   Results of generalized linear mixed model analysis showing the effects of different parameters (species as a random factor and the others as fixed parameters) on seed production
  1. The data were the averages for each species and for each chamber, with N = 202. FlN, flower number per plant; FSR, fruit set ratio; SeS, seed mass per 1000 seeds (mg); SMP, total seed mass per plant (mg); SNF, seed number per fruit; SNP, total seed number per plant.

Warming × speciesRandom25491351118.840<0.001
Error 2152535175  
Warming × speciesRandom10420401110.319<0.001
Error 1597303174  

Regression analyses consistently revealed that the ratio of the difference in flower number between warmed and ambient treatments explained c. 70% and 80%, respectively, of the variation in the proportional reduction in total seed mass and number (Fig. 2a,b) (both < 0.001).

Figure 2.

Relationships between warming-reduced proportions of flower number and total seed mass (a) and between warming-reduced proportions of flower number and total seed number (b) resulting from cross-species regression analyses. The proportions are calculated as the ratios of the reduced flower number (or seed number and seed mass; flower number per plant in ambient chambers minus that in warmed chambers) to the flower number per plant in ambient chambers. Open circles, multi-inflorescence species; closed circles, single-inflorescence species.

The results of phylogenetic GLS further showed that the warming effect on plant seed production was largely a result of a change in flower number per plant between warmed and unwarmed chambers, and R2 values (0.740 and 0.812 for the relationships between reduced proportion of flower and reduced proportions of total seed mass and number, respectively) were close to those of the cross-species regressions (see Table S3). In addition, the maximum likelihood values of λ were close to zero for both flower number per plant vs total seed number and mass relationships, significantly different from unity but not from zero (at the level of < 0.001), suggesting that there was no significant phylogenetic association between flower production and seed production.

The effect of warming on flowering onset and height

Warming delayed significantly the onset of flowering for T. farreri and A. alpinus, but advanced the onset time of R. brotherusii (Table 2). Among the other nine species, the flowering onset time was not affected significantly (Table 2). Compared with the ambient controls, reproductive plants were significantly taller in warmed OTCs in nine of the 12 species (and marginally significant in the case of Caltha scaposa, Table 2). No significant difference in flowering height was found for C. enervis ssp. chuanxibeiensis, B. sinocompressus and C. paradoxum (Table 2). Community height was also significantly greater in warmed than ambient OTCs, but no significant difference was found in total plant density (Fig. 3).

Figure 3.

Mean community height (a) and total plant density (b) in warmed open-top chambers (OTCs; closed column) and ambient OTCs (open column). The error bars denote 1SE. *, < 0.05.


We have shown that warming (typically winter warming) reduces significantly the number of flowers and seed output per plant, specifically in multi-inflorescence species, and that the reduction in seed output largely results from a reduction in flower production. Importantly, the λ values (indistinguishable from zero) revealed by GLS analyses indicate that the cross-species relationships without phylogenetic correction provide an equally strong relationship between flower production and seed production. Therefore, these relationships are not the result of a phylogenetic bias. The data specifically show that warming reduces significantly both flower and fruit number per plant among the multi-inflorescence species, whereas seed size and (or) seed number per fruit are generally unaffected. Moreover, because the fruit set ratio is not affected by warming, the difference in fruit number per plant must be the result of the difference between the two species groups in the flower number per plant. Consistently, the result of GLMM indicates that the flower number contributes most to the variation in seed output among species and among OTCs, whereas regression analyses also indicate that the change in flower number per plant accounts most for the negative effect of warming on seed production. In addition, the reduction in flower and seed number is not attributable to changes in flowering phenology or pollination limitation, because the effects of warming on the flowering phenology, fruit set ratio and seed set ratio are inconsistent and mostly insignificant among the study species. Therefore, we attribute the decline in flower and seed production to the warming effect on the vernalization of temperate species.

Recent research has focused on the effect of climate change on seed production in order to predict the long-term consequences of short-term effects of global climate change on community composition and ecosystem properties. Many studies have obtained either positive (Alatalo & Totland, 1997; Arft et al., 1999; Kudo & Suzuki, 2003; Aerts et al., 2004; Kudo & Hirao, 2006) or neutral (Totland & Alatalo, 2002; Kudo & Suzuki, 2003; Hovenden et al., 2007) effects of warming on flower and seed production of species from alpine and arctic areas. These studies are often performed without a temperature increase in winter (Alatalo & Totland, 1997; Arft et al., 1999; Kudo & Suzuki, 2003), in part because OTCs may potentially act as snow traps during the winter months (Marion et al., 1997). By contrast, we report a negative effect of warming on flower and seed production for the nine multi-inflorescence species. The reduced seed production resulting from reduced flower production in the warmed OTCs is consistent with the hypothesis that warming may weaken or abate the effects of vernalization (Hideyuki & Takashi, 2003), thereby decreasing plant flower and seed production (Saure, 1985; Hennessy & Claytongreene, 1995). As noted, the climate of the study site is characterized by a long cold period and a relatively short growing season. Therefore, it is reasonable to speculate that the onset of flowering and flower production should be under strong selection for plants endemic to this climate. Studies on the physiology of vernalization (Chouard, 1960) suggest that low temperature may induce flower bud formation, involving complex interactions among many different enzymes and hormones. In this regard, Saavedra et al. (2003) have shown that warming decreases flower production in Delphinium nuttallianum in an Arctic area. These authors attribute this reduction to the negative effect of warming on the length of the snow-cover season, in accordance with the strong positive correlation between snowpack and flower production of the species. Likewise, a recent study has shown that extreme and sudden winter warming events may reduce flower and seed production of Arctic shrub species (Bokhorst et al., 2008), although the warming effect on vernalization was not employed as a viable explanation.

Several additional factors may potentially be employed to explain the reduction in flower and seed production among species. Changes in flowering height and flower phenology are obvious candidates. For example, with increasing plant density, many grassland species must grow sufficiently tall to gain sufficient light to flower (Vile et al., 2006), attract insects (Galen & Cuba, 2001) or disperse pollen (Niklas, 1985). In our study, total plant density was not affected significantly by warming. However, the average community height was significantly greater in the warmed OTCs. In accordance with the increased community height, flowering height increased significantly in eight of the nine multi-inflorescence species (Table 2), which probably provides an advantage in terms of light inception and pollination visitation during flowering (Vile et al., 2006). Taller plants probably required a greater investment in stem growth in the warmed OTCs, leading to the reduction in flower and seed production. However, there was no significant association between the changes in flowering height and flower production and seed output (both > 0.1, data not shown). An alternative explanation is that significant changes in flowering phenology often tend to decrease the potential for flower formation and development. Plants cannot reproduce much earlier than the time at which they have accumulated sufficient material resources; early-flowering plants tend to be shaded by late-flowering plants, which reduces the chances for large seed production, even though the advance in flowering phenology increases and provides plants with more time to develop large seeds (Bolmgren & Cowan, 2008). They also cannot delay significantly reproduction because of a limited ‘window of opportunity’ to produce and fully develop fruits/seeds before the end of the growing season (Bolmgren & Cowan, 2008; Du & Qi, 2010). However, the onset of flowering and the flowering period were not consistently altered by warming for the species in our study. As shown by Yu et al. (2010), spring warming advances spring greening on the Tibetan plateau, whereas winter warming tends to delay the onset of spring growth. The interplay between the spring and winter warming effects might have resulted in the unchanged flowering onset time in this study.

As noted, we are sensitive to the concern that the physical effects of OTCs might have contributed to the differences in seed production observed among our study species. For example, OTCs might interfere with pollinator visitation, thereby decreasing seed production for the species that require cross-fertilization. However, both types of OTC should have similar effects on pollinator visitation rates because the mesh of the screen of the ambient OTCs is too small for all potential pollinators (e.g. flies and bees in the study area). Moreover, the forb species, whose seed production could be pollinator limited, did not show a change in fruit set ratio and seed set ratio, suggesting that the pollination success was largely unaffected for the species studied between warmed and ambient OTCs. Thus, the difference in seed production between ambient and warmed OTCs is not easily attributable to the difference in the OTC effect on pollinator behavior, although it is not clear whether the same levels of reduced seed production would have been observed if pollinators had not been limited by the chambers. Along the same lines, the screen sides of the ambient OTCs might be less of a barrier to wind-borne pollen than the plastic sheets of the warmed OTCs, such that seed production among the wind-pollinated species within warmed OTCs might have been more limited. However, among the three wind-pollinated species, two (C. enervis ssp. chuanxibeiensis and B. sinocompressus) showed no decrease in seed production. Indeed, our data clearly show that the reduction in seed production in the multi-inflorescence species was primarily a result of the decline in flower production, which cannot be interpreted in terms of pollination limitation.

Another concern is the difference in the reproductive response to warming between single- and multi-inflorescence species. Nested ANOVAs (with species nested within single- and multi-inflorescence species groups) showed that the flower type and the interaction between warming and flower type have a significant effect on both the flower number per plant and total seed mass (Table S4), which is consistent with the observation that the reduction in flower and seed production occurred in all the multi-inflorescence species, but in none of the single-inflorescence species. These contrasting reproductive responses to warming cannot simply be attributed to the phylogenetic effect between single- and multi-inflorescence species. Eight of the nine multi-inflorescence species are dicots, whereas all three of the single-inflorescence species are monocots. However, one monocot, the grass D. caespitosa, is a multi-inflorescence species that showed a reduction in its flower and seed production in warmed chambers. One possible explanation is that natural selection favors single-inflorescence species that are insensitive to temperature change under the climatic conditions of sites with snow-free winters. Species that produce a single inflorescence per growing season invest all of their effort in one reproductive opportunity and therefore cannot afford to lose this investment. Another possibility is that warming has a greater effect on species characterized by indeterminate growth, which tend to take a longer time to develop inflorescences.

In summary, warming reduced flower and seed production for the multi-inflorescence species, but not for the single-inflorescence species, of the Tibetan alpine meadow. The reduction was assumed to be primarily a result of the warming effect of vernalization in the species, and less likely to be caused by changes in flowering phenology, flowering height, the physical setting of OTCs and phylogenetic effects. Our results also suggest that, compared with multi-inflorescence species, single-inflorescence species may increase in dominance in plant communities of the future if post-fruiting life-history events remain unchanged and are continuous. However, many herbaceous species can reproduce both vegetatively and sexually, such as in this study; local communities may experience migration or encroachment from plants from warmed communities. It is therefore hard to predict which species or species group would become more dominant in warmed communities. These factors complicate predictions on changes in community composition as a result of global warming.


We thank Lee E. Frelich for helpful comments on an early version of the manuscript and Yibin Yuan, Jian Feng and Xinwei Wu for field assistance. This study was supported by NSFC (31130008), the Fundamental Research Funds for the Central Universities.