Introduction

To predict the impact of climate change on population dynamics of plant species, it is crucial to understand how plant growth and reproduction respond to environmental variations (Hedhly et al. 2008). In high latitude ecosystems with clear seasonality, temperature is one of the most important environmental factors influencing plant growth. Generally, enzyme activity, cell division, and photosynthetic activity decrease in cool temperatures in many plants (Fitter and Hay 1987; Tardieu et al. 2000). However, previous studies reported that spring ephemerals, mostly perennial herbs inhabiting the floor of cool-temperate deciduous forests, grow better at cool temperatures (Lapointe 2001; Lapointe and Lerat 2006; Badri et al. 2007; Gandin et al. 2011; Bernatchez and Lapointe 2012). For example, Erythronium americanum (Lapointe and Lerat 2006; Gandin et al. 2011) and Crocus vernus (Badri et al. 2007; Lundmark et al. 2009) developed bigger storage organs under cool conditions. The enhanced growth under cool conditions is correlated with extended leaf longevity and/or continuous sink intensity of bulbs or corms in terms of carbon partitioning compared with warmer temperatures (Lundmark et al. 2009; Gandin et al. 2011).

Initiation of growth in spring ephemerals depends mostly on the time of snowmelt and subsequent temperature (Schemske et al. 1978; Fitter et al. 1995; Whigham 2004). The significant variations in temperature encountered so far have been mainly recorded at the end of winter and the beginning of spring (Easterling et al. 1997; Sparks et al. 2000). Actually, the timing of extreme temperatures could have a significant impact on the existence of spring ephemerals that emerge very early in the spring (Whigham 2004; Kudo et al. 2008). Furthermore, increased spring temperature commonly advances the flushing of canopy trees (Menzel 2000, 2002; Vitasse et al. 2009). Earlier closure of the canopy could also reduce carbon assimilation by limiting the favorable light period on the forest floor and hence reduce photosynthetic carbon accumulation (Niesenbaum 1993; Rothstein and Zak 2001; Ida and Kudo 2008). Thus, the impact of climate change on spring ephemerals should be evaluated in terms of direct warming effects and indirect light conditions.

Previous studies have looked at the extent of growth of spring ephemerals in different growth temperature regimes but only for a limited range of species. In addition, probably because of the simple whole-plant morphology (i.e., one source versus one sink), most of the work has concentrated on non-reproductive individuals (Gandin et al. 2011; Gutjahr and Lapointe 2008; Lapointe and Lerat 2006; Bernatchez and Lapointe 2012). To predict the response of spring ephemerals to climate change, however, the responses of reproductive plants should be clarified because the timing and frequency of extreme temperature events could be important also for sexual reproductive phases and final reproductive output (Kudo et al. 2004). In addition, sensitivity to thermal environment may vary between reproductive and non-reproductive plants, reflecting the specific carbon allocation strategy (Sunmonu and Kudo 2014).

Gagea lutea is an ideal model plant to quantify the responses of reproduction to warming climate because reproductive individuals have two sources (leaf and bract) versus two sink functions (fruit and bulb) in terms of carbon assimilation during a growth period (see illustration in Sunmonu et al. 2013). Leaves and bracts act as specialized source organs for bulb growth and current seed production, respectively, but photosynthetic products from bracts could be flexibly used for bulb growth when plants fail to set fruits (Sunmonu et al. 2013). Therefore, by monitoring reproductive individuals of G. lutea in warm conditions, we could clarify whether irrespective of reproductive status their growth is also limited at warm temperatures, as found in non-reproductive counterparts of other spring ephemerals.

In this study, we explored the hypothesis that the extent of reproductive output and bulb growth in warm conditions would depend on the responses to source organs for each sink function (i.e., bract for seed production and leaf for bulb growth). Seed production and bulb growth may not be sensitive to climate change if the lifespan and carbon assimilation of the leaf and bract are not restricted under warm conditions. Apart from the source–sink balance for resource allocation, warmer temperature may directly influence the pre-zygote process even under conditions of good pollination success, i.e., fertilization success, such as pollen viability, pollen tube growth, stigma receptivity, and ovule viability, which may also decrease seed production (Hedhly et al. 2008). Because pollen activity is generally sensitive to temperature (Hedhly et al. 2005; Kakani et al. 2005), the thermal influence on pollen performance may also influence reproductive success in a warm climate.

By linking data for environmental factors, physiological and phenological responses of leaves and bracts (source functions), reproductive activities and bulb growth (sink functions) among forest, open, and greenhouse (GH; warming manipulation in the open site) conditions, we investigated phenological and physiological responses of photosynthetic activity, reproductive performance, and vegetative growth to earlier and warmer spring in G. lutea. In this experiment, we intended to predict if warm spring conditions impact on spring ephemerals by separating temperature effects (GH vs. open habitat) from light effects (open vs. forest habitat). We expected that responses in terms of reproductive (fruit and seed production) and vegetative (bulb growth) performance to growth conditions would vary based on the responses of individual source organs. The objectives of this study were to clarify the effects of early and warmer spring on (a) leaf and bract characteristics, (b) reproductive output, and (c) bulb growth in G. lutea, a typical spring ephemeral species.

Materials and methods

Study species and experimental design

Gagea lutea Ker-Gawl. (Liliaceae) is a polycarpic perennial herbaceous species inhabiting northern temperate forests. This species has a typical spring ephemeral lifecycle; flowering starts immediately after snowmelt concurrently with leaf expansion (in mid- to late April) and fruits mature about 2 weeks after anthesis. It produces 1–10 flowers and 24–39 ovules per plant on average (Nishikawa 1998). It is pollinated by insects, and cross-pollination is more effective for seed production (Kudo et al. 2004). Aboveground shoots usually die at the same time as seed dispersal, at the time of canopy closure in late May. Thus, the short period between snowmelt and canopy closure is when this species accumulates resources in the underground bulb. Non-reproductive individuals produce only one leaf, while one basal leaf and a pair of long and short leaf-like bracts on the scape are produced in reproductive plants (Sunmonu et al. 2013).

Bulbs of G. lutea were collected (n = 203) from a secondary deciduous forest within the campus of Hokkaido University, Sapporo, Japan (43º04′57″N, 141º20′22″E) in late autumn of 2012. This forest is usually covered with snow from early December to early April, and common canopy trees in this fragmented forest include Ulmus davidiana var. japonica, Cercidiphyllum japonica, Betula platyphylla var. japonica, and Populus maximowiczii. The bulbs were immediately taken to the laboratory and their volumes were measured. Width (W) and length (L) of individual bulbs were measured using a digital caliper, and the volume was estimated as π × W 2 × L/6 based on the shape of the bulb. In this study, we used volume as an index of bulb size because of the non-destructive measure. In our preliminary measurement, however, we confirmed a strong correlation between bulb volume and dry mass (r 2 = 0.837, n = 15). Individual bulbs were then planted in pots with numbered tags for identification, and the pots were randomly transferred to three plots: forest (n = 67) and two open plots outside the forest (n = 68 in each plot). In March of 2013, advancement of snowmelt timing was performed at one of the open plots by manually removing snow twice (14th and 21st March). Then we set a GH over the plot to facilitate rapid natural snowmelt of the remaining snow and increase the temperature. The GH was 2 m in width, 3 m in length, 2 m in height, and covered with clear plastic sheet. Hence, we established three plots in this study; forest (natural habitat), open (continuously bright but same snowmelt time as the forest plot), and GH plot (continuously bright and warm with early snowmelt). In the first year (2013), plants were made to acclimate to the environmental conditions. In late autumn of 2013 before snow cover, bulb sizes from all plots were measured again and replanted in preparation for the next spring. The same exercise was repeated in early spring 2014 (snow removal on 20th and 26th March), but to generate similar snowmelt dates between forest and open plots, we added 50–70 cm of snow to the forest plot because snow depth was deeper at the open plot.

Growth conditions

Preliminary growth conditions were characterized by monitoring soil temperature among plots throughout the growth season in 2013. For this, six automatic data loggers, two per plot (HOBO, UA-002, Onset Computer Corporation, Bourne, MA), were randomly set in pots to measure soil temperature at a depth of 10 cm at 1-h intervals from December 2012 to June 2013. The data obtained by two loggers in each plot were averaged. In 2014, air temperature (in every plot) and photosynthetically active radiation (PAR; in the forest and open plots) were recorded during the experimental period at 1-h intervals using a combined data logger with a solar radiation monitor and thermometer (HOBO weather station, Onset Co., MA, USA) from 7 March to 5 June. Averages of 24 measurements within 1 day were stored as daily means. We did not measure soil temperature in 2014 because of technical difficulties.

Physiological measurements of leaves and bracts

After shoot emergence in spring 2014, all reproductive plants producing floral buds were monitored. To investigate the physiological responses of leaves and bracts to environmental manipulations among plots, leaf and bract maximum photosynthetic rates (P max) at saturation irradiance (1500 μmol m−2 s−1) and dark respiration rates were measured using a portable LI-6400 photosynthesis system (Li-Cor, Lincoln, NE, USA). Three of the experimental plants were selected per plot at each of three growth stages: 7th April at floral bud stage, 16th April at flowering stage, and 8th May at early fruiting stage in the GH plot; 18th April, 12th May, and 23rd May in the forest plot; and 18th April, 10th May, and 24th May in the open plot. Respiration rate was measured after leaving the leaf for 5–7 min in the dark (0 μmol m−2 s−1 irradiance). Leaf temperature in the chamber was controlled at 20 °C, and the concentration of CO2 in ambient air entering the leaf chamber was maintained at 380 μmol mol−1. This temperature corresponded to the typical daytime temperature on clear days in the growth period of G. lutea (see results) in which photosynthetic activity is maximal (Sunmonu et al. 2013). Leaf-to-air vapor pressure deficit (VPD) was controlled to be less than or equal to 1.1 kPa. Because the width of leaves and bracts was smaller than the chamber size (2 × 3 cm), we corrected photosynthetic parameters by the replacement of chamber area by actual leaf area that was included in the chamber. This protocol of photosynthetic measurement was determined based on our previous studies in G. lutea (Sunmonu et al. 2013; Sunmonu and Kudo 2014).

To assess the seasonal changes in photosynthetic area, the leaf and bract sizes (length, L and width, W) of all plants were measured using a digital caliper on a weekly basis, but as soon as senescence started, monitoring and measurement changed to every other day till the end of the growth period. This measurement involved only the green area that was photosynthetically active. Leaf and bract area (A) was estimated as A = 0.83 × L × W (r 2 = 0.968, n = 5). After removing damaged plants, 45, 49, and 40 plants were present at the forest, open, and GH plots, respectively.

Reproductive output

To evaluate the reproductive output in various environmental conditions, the number of floral buds was recorded for each plant in every plot during the flowering period. Then artificial outcrossing by hand-pollination was conducted for every flower to eliminate pollen limitation for seed production. Soon before seed dispersal, all infructescences were harvested and taken to the laboratory, where individual fruits were opened carefully and the numbers of mature seeds and undeveloped ovules in each fruit were counted. Fruit-set ratio was expressed as matured fruit number divided by original flower number, and seed production was taken as the ratio of mature seed number to original ovule number produced per plant. Duration of the flowering period was also recorded for all plots to clarify any environmental effects on flowering phenology.

Bulb growth

Annual bulb growth was measured to clarify the response of perennial organs (i.e., vegetative growth) to environmental variations among plots. Initial bulb sizes were measured in November 2013 (see above). Final bulb sizes were measured again in late June 2014 after the growth period. Bulb growth was taken as the final bulb volume after one growth season in response to plot differences and initial bulb size.

Pollen germination

To test the effect of temperature on fertilization success, pollen germination ability was measured at five different temperatures in the laboratory. First, agar-based media with a suitable sucrose concentration (10 %) was prepared in a test tube. Next, flowers with fresh and dehisced anthers were collected from a nearby G. lutea population, placed in plastic bags, and brought immediately to the laboratory. Two drops of media were placed separately on each of 15 glass slides kept on petri dishes lined with moist filter paper. Uniform pollen grain samples were dispersed vertically on each medium field, and three slides (six media) each were immediately incubated at temperatures of 10, 15, 20, 25, or 30 °C at approximately 80 % humidity for 24 h. Pollen germination was determined by direct microscopic observation (BX43, Olympus Corporation, Tokyo, Japan). Pollen was considered germinated when the length of the germinated pollen tube exceeded the grain’s diameter. For each temperature treatment, the numbers of germinated and shrunken pollen grains were counted using a microscope, and the germination percentage was evaluated by dividing the number of germinated pollen grains per field of view by the total number of pollen per field of view. Furthermore, the length of pollen tubes of randomly selected 40–60 pollen tubes in each temperature was measured using digital photographs.

Statistical analysis

Leaf and bract P max and dark respiration rates were analyzed using generalized linear models (GLMs) with a postulated gamma error distribution with log-link function, in which plot (forest, open, GH) and shoot age (day from emergence) were set as explanatory variables. Leaf and bract survival rates were compared among plots using Cox proportional hazards regression models. Maximum leaf and bract sizes were compared among plots by analysis of covariance (ANCOVA), in which initial bulb size (2013) was included as a covariate after log-transformation. Tukey’s HSD test was used for post hoc multiple comparisons. Flower and fruit production per plant were analyzed using a GLM with a postulated Poisson error distribution with log-link function, in which plot and initial bulb size in 2013 (after log-transformation) were set as explanatory variables. Fruit-set success (fruit/flower ratio) and seed-set success (seed/ovule ratio) were compared using a GLM with a postulated binomial error distribution, in which the plot was set as an explanatory variable. Final bulb size (June 2014) was compared using a GLM with a postulated gamma error distribution with log-link function, in which plot, flower number, and fruit number were set as explanatory variables and initial bulb size (2013) was included as an offset variable after log-transformation. Pollen germination rate and pollen tube length were compared using GLMs with a postulated binomial error distribution with logit link function and a gamma error distribution with log-link function, respectively, in which temperature (10, 15, 20, 25, and 30 °C) was set as an explanatory variable. All statistical analyses were conducted using an open source system, R version 3.0.1 (R Development Core Team 2013).

Results

Growth conditions and phenology

In the preliminary measurement of soil temperature in 2013, soil conditions were constantly maintained around 0–1 °C during the winter (December to February) in every plot, indicating a lack of soil frost. Mean soil temperature during the growth period (April and May) was 7.3 °C (ranging from 0.1 to 20.2 °C) in the forest plot. Daily mean soil temperature was 4.3 °C warmer in the GH plot and 0.6 °C warmer in the open plot in comparison with the forest plot (see Appendix Fig. 6).

Air temperature during the experimental period in 2014 showed a similar trend to soil temperature in 2013. Mean air temperature throughout the growth period (April and May) was 10.2 °C (ranging from −1.2 °C to 28.0 °C) in the forest plot. Daily mean temperatures in the GH and open plots were 3.5 and 0.8 °C, respectively, warmer than the forest plot (Fig. 1a). Daily maximum temperatures in the GH and open plots were 15.3 and 3.1 °C, respectively, warmer than in the forest plot. PAR in the open plot was two times larger than that in the forest plot (Fig. 1b). As the season progressed, the difference in PAR between the open and forest habitats became larger owing to developing canopy closure in the forest.

Fig. 1
figure 1

Fluctuations in a air temperature and b photosynthetically active radiation (PAR) during the growth season of Gagea lutea in each plot. PAR was measured only in the forest and open plots. c Green leaf period and flowering period in each plot in 2014

Full size image

Acceleration of snowmelt in the GH plot advanced both shoot growth initiation and reproductive phenology (Fig. 1c). Although the duration of flowering period was similar among plots (14, 15, and 17 days in the open, forest, and GH plot, respectively), flowering started earlier in the GH plot (8 days after shoot emergence) compared with the forest and open plots (16 days after shoot emergence).

Physiological traits of leaves and bracts

Leaf P max did not differ among plots (P > 0.10) but decreased with leaf age (P < 0.001; Table 1a). A significant interaction existed between the GH plot and age (P = 0.001) owing to a rapid decrease in P max in the GH plot (Fig. 2a). This indicated that the period of high photosynthesis was short in warm conditions. Leaf respiration rates tended to decrease with leaf age in the open and forest plots but remained high in the GH plot, especially at the fruiting stage, leading to a significant interaction between GH and age (P < 0.001; Table 1c, Fig. 2c). Retention of high respiration rate during the fruiting period indicated the high respiratory loss in the GH plot compared with the open and forest conditions.

Table 1 Results of generalized linear models for P max and dark respiration of leaves and bracts among plots (open, forest, and greenhouse [GH]) and shoot age
Full size table
Fig. 2
figure 2

Seasonal transitions in leaf P max (a), bract P max (b), leaf dark respiration rate (c), and bract dark respiration rate (d) in the open, forest, and greenhouse (GH) plots. The age of plants is expressed as days since shoot emergence. Measurements were conducted at pre-flowering, flowering, and fruit developing stages in each plot. Mean ± SE. See Table 1 for statistical results

Full size image

Bract P max was slightly lower in the forest plot in comparison with the open and GH plots, corresponding to lower irradiance under the canopy (Table 1b, Fig. 2b). Similar to leaf P max, bract P max decreased with age (P < 0.001) and showed an interaction between GH and age (P < 0.001) because of a rapid decrease in photosynthetic activity at fruiting stage in warm conditions. Although bract respiration rates decreased with age in the open and forest plots, plants in the GH plot retained high respiration rates at fruiting stage, leading to a significant interaction between GH and age (P = 0.006; Table 1d, Fig. 2d). Thus, warmer temperature caused greater respiratory loss also for bracts.

Leaf lifespan as well as bract lifespan varied significantly among plots (Fig. 3). Initiation of leaf senescence began 25, 34, and 34 days after shoot emergence in the GH, open, and forest plots, respectively. Mean (±SE) leaf longevity was 43.3 ± 0.5, 49.4 ± 0.5, and 50.3 ± 0.4 days in the GH, open, and forest plots, respectively, whereas mean bract longevity was 42.1 ± 0.5, 48.8 ± 0.5, and 52.6 ± 0.4 days, respectively. Leaf longevity was significantly shortened in the GH plot (P < 0.001) but no difference between the open and forest plots (P = 0.36). Bract longevity was shortened in the open and GH plots compared with the forest plot, as shown by the Cox proportional hazard regression (P < 0.001, Fig. 3).

Fig. 3
figure 3

Survival curves of leaves (a) and bracts (b) in the open, forest, and greenhouse (GH) plots in 2014. Results of a generalized linear model are indicated (*** P < 0.001)

Full size image

Maximum leaf and bract sizes increased with initial bulb size (P = 0.015 and P < 0.001, respectively) and varied among plots (P < 0.001; Table 2). Plants in the GH plot produced significantly smaller leaves (5.0 ± 0.3 cm2) and bracts (1.7 ± 0.1 cm2) compared with plants in the open plot (11.5 ± 0.5 cm2 and 2.8 ± 0.2 cm2, respectively) and the forest plot (11.2 ± 0.5 cm2 and 2.7 ± 0.1 cm2, respectively; P < 0.05 by Tukey’s test).

Table 2 Results of ANCOVA for maximum leaf and bract size among plots (open, forest, and greenhouse [GH]) according to bulb size (2013) as representative of plant size
Full size table

Reproductive activity

Flower and fruit production in 2014 depended highly on bulb size in 2013 (P < 0.001; Appendix Table 5). There was no significant difference in flower number per plant among plots when size effect was considered using a GLM (P > 0.05; Table 3, Appendix Table 5). In contrast, fruit number per plant, fruit-set rate, and seed-set rate were significantly smaller in the GH plot (P = 0.002, P = 0.001 and P < 0.001, respectively), whereas they were similar between the forest and open plots (P > 0.10). These results indicated that reproductive success was restricted in the GH plot even when hand-pollination was conducted.

Table 3 Flower, fruit, and seed production in the forest, open, and greenhouse [GH] plots. Mean ± SE
Full size table

Bulb growth

Final bulb volume in June 2014 was significantly smaller in the GH plot (P = 0.028) but larger in the forest plot (P < 0.001) in comparison with the open plot (Table 4, Fig. 4). Flower production negatively influenced bulb size (P = 0.048), whereas fruit production did not (P = 0.78). Bulb size increased by 26 % in the forest plot during the growth period, but showed a 14 % decrease in the open plot and a 33 % decrease in the GH plot. Therefore, vegetative growth was greatest in the original habitat under the canopy, but it was restricted in warm and constantly bright conditions.

Table 4 Result of generalized linear model for final bulb size (2014), in which plot (open, forest, and greenhouse [GH]), flower production, and fruit production were set as explanatory variables
Full size table
Fig. 4
figure 4

a Final bulb size in volume (cm3) and b change in bulb volume during a growth period in the open, forest, and greenhouse (GH) plots. Results of a generalized linear model are indicated (* P < 0.05, *** P < 0.0001). Box plots indicate 25, 50, and 75 percentiles, and whiskers indicate 10 and 90 percentiles of the data distribution. See Table 4 for details for the statistical results

Full size image

Pollen germination activity

Mean pollen germination rates were 18.9, 23.5, 30.5, 17.5, and 7.9 % at 10, 15, 20, 25, and 30 °C, respectively. The highest germination rate was recorded at the intermediate temperature (Fig. 5, Appendix Table 6). The highest temperature, which mimicked the GH conditions, significantly inhibited pollen germination (P < 0.001). Hence, high temperatures during the flowering period might have serious deleterious effects on fertilization success and subsequent seed production in G. lutea.

Fig. 5
figure 5

Pollen germination rates under various temperatures. Results of a generalized linear model are indicated (ns P > 0.05, * P < 0.05, *** P < 0.0001) in which 10 °C is set as an interception. Box plots indicate 25, 50, and 75 percentiles, and whiskers indicate 10 and 90 percentiles of the data distribution. See Appendix Table 6 for details of the statistical results

Full size image

Mean values of pollen tube length were 0.88 ± 0.04 mm, 0.98 ± 0.04 mm, 1.08 ± 0.04 mm, 1.12 ± 0.04 mm, and 0.98 ± 0.05 mm at 10, 15, 20, 25, and 30 °C, respectively. The highest pollen tube growth was recorded at the intermediate temperature (P < 0.001; Appendix Table 6).

Discussion

Responses of vegetative growth: leaf–bulb relationship

Our experiment revealed that the physiological and phenological responses of leaves are influenced more strongly by temperature than by light conditions, because their photosynthetic activity, dark respiration rate, size, and longevity were similar between the open and forest plots but clearly different in the GH plot. A rapid decrease in photosynthetic activity, large respiration loss, small leaf area, and short lifespan in the GH plot clearly indicated that the overall performance of assimilative ability was negatively affected by warm conditions. These findings supported the previous studies on other spring ephemeral plants (Lapointe and Lerat 2006; Badri et al. 2007; Lundmark et al. 2009; Gandin et al. 2011).

Leaf P max decreased quickly with time, reflecting the short lifespan of spring ephemerals (Constable et al. 2007), but the decreasing rate was accelerated in the GH plot (Fig. 2a), indicating earlier physiological aging in warm conditions. Dark respiration rates in the GH plot stayed at a high level even during the fruiting period (Fig. 2c), suggesting that most carbon fixed in this period might be exhausted by respiration loss. High respiratory loss has been reported in other spring ephemerals grown in warm conditions (Gandin et al. 2011; Bernatchez and Lapointe 2012). Shorter leaf longevity only in the GH plot (Fig. 3a) indicated that leaf lifespan of spring ephemerals was more strongly determined by temperature rather than light conditions. Yoshie (2008) reported an extended longevity for G. lutea in cool growth temperatures in contrast to a summer-green forb Maianthemum dilatatum in which cool temperatures shortened leaf lifespan. Despite a relatively high P max in the GH plot early in the season, reduced leaf longevity and sizes, in addition to high respiration loss, should restrict carbon assimilation in warm conditions, resulting in weaker source function. Especially, leaf size of plants in the GH was less than half of plants in the open and forest plots. This might be caused by heat stress in the GH, where daily maximum temperature often exceeded 30 °C (Fig. 1a).

Several studies have demonstrated that photosynthetic activity is regulated by the sink intensity of storage organs (Sawada et al. 2003; Lundmark et al. 2009; Gandin et al. 2011). In the spring geophyte Erythronium americanum, leaf senescence was induced by a reduction in carbohydrate sink demand once the bulb was filled with carbohydrates (Lapointe 2001). At a warm temperature, faster starch accumulation causes smaller sink capacity of bulbs and leads to shorter leaf lifespan, resulting in smaller bulb size (Gandin et al. 2011). A similar physiological mechanism is expected in G. lutea. Plants grown in the forest plot developed the largest final bulb volume, whereas plants in the GH plot accumulated least volume (Fig. 4a). Therefore, vegetative growth of G. lutea was negatively influenced by warm condition as reported in other spring ephemerals such as E. americanum (Lapointe 2001; Gandin et al. 2011), Allium tricoccum (Bernatchez and Lapointe 2012), and C. vernus (Badri et al. 2007; Lundmark et al. 2009). On the other hand, careful consideration is necessary to evaluate the climate change impacts on spring ephemerals from this study, because warming effect by the GH was available only during the daytime on clear days, while night-time temperatures were similar to the forest plot in our experiment (Fig. 1a). Warm night conditions may accelerate the respiration loss of plants, resulting in further decrease in bulb growth under warm conditions.

Bulb volume was also smaller in the open plot compared with the forest plot (Fig. 4b). This trend was in contrast to the previous report on this species (Sunmonu and Kudo 2014), where the bulb size was maintained when plants were transferred to an open habitat during the second half of the growth season. Longer exposure to direct sunlight in open conditions in the present experiment might have led to high transpiration and even heat and water stress, resulting in the restriction of bulb growth.

Responses of reproductive characteristics: bract–fruit relationship

Onset of flowering occurred 1 week earlier from the initiation of growth in the GH plot compared with the forest and open plots (Fig. 1c). This indicated that flowering initiation was temperature dependent for plants growing in cool conditions (Fitter et al. 1995). However, flowering duration and flower number were less affected by warming in our experiment. The absence of changes in flowering period may be partly because hand-pollination was performed soon after flowering for every flower. If floral longevity was determined by the timing of pollen deposition on the stigmas (Primack 1985; Ishii and Sakai 2000), this is a reasonable response. The previous warming experiment in a subalpine meadow also documented earlier onset of flowering but no change in flowering duration in warm conditions (Price and Waser 1998). The construction of aboveground shoots in spring ephemeral herbs commonly depends on resources stored during the preceding year, i.e., the size of storage organs (Muller 1978; Routhier and Lapointe 2002). Flower production of G. lutea also depends on initial bulb size (Schnittler et al. 2009; Sunmonu et al. 2013). Therefore, we predict that flower production in continuous warm conditions may decrease from year to year in response to decreasing bulb size.

Although bract P max in the forest plot was relatively small in comparison with the open and GH plots during the early to middle growth period, the decline of P max progressed gradually toward fruiting stage in the forest plot, whereas P max in the GH plot was intensively suppressed at the fruiting stage (Fig. 2b). Similar to leaf responses, bracts in the GH plot showed the shortest lifespan and smallest size, whereas longevity was largest in the forest plot (Fig. 3b). Longer bract lifespan under the canopy may compensate for the smaller P max of bracts in terms of assimilative function (Constable et al. 2007), resulting in similar fruit and seed production to the open plot. Short lifespan, rapid decrease in photosynthetic activity, high respiration rate even in the fruiting stage, and small size of bracts in the GH plot indicated low photosynthetic assimilation in warm conditions, as detected in the leaves. Because photosynthesis by bracts is completely responsible for seed production in this species (Sunmonu et al. 2013; Sunmonu and Kudo 2014), reduced bract assimilation in the GH plot might be responsible for the lower fruit and seed production.

Similar to the sink–source balance between bulb growth and leaf photosynthesis, as mentioned before, the sink intensity of fruits may also affect the photosynthetic activity and transportation of photosynthetic products to reproductive organs (Iglesias et al. 2002; Ida et al. 2013). When the sink intensity of the reproductive function decreased, the activity of photosynthetic function supporting the reproductive sink may be decreased, resulting in earlier senescence and shorter longevity (Iglesias et al. 2002). Low fertilization success causes the low sink intensity of developing fruits. The pollen germination rate of G. lutea was restricted at high temperature (Fig. 5), and it might have resulted in low fertilization success in the GH plot. Inhibitions of pollen vigor and stigma receptivity by heat stress were reported in some crop plants (Devasirvatham et al. 2012; Kaushal et al. 2013). If low reproductive output in the GH plot was caused by the failure of ovule fertilization, the small sink intensity of developing fruits might accelerate the senescence of bracts, which could lead to feedback restriction of fruit development. Our previous study revealed that photosynthetic products from bracts were flexibly used for bulb growth when all flowers were removed (Sunmonu and Kudo 2014). Because bulb growth seemed to be stopped at the fruiting stage in the GH plot, the small sink intensity of both bulb and fruits might accelerate the senescence of bracts at the same time as leaf senescence.

Contrary to our previous study (Sunmonu and Kudo 2014), seed-set rates did not differ between the open and forest plots (Table 3), indicating no advantage of an extended bright period for seed production in this year. A previous study reported higher seed-set success in plants inhabiting a forest-edge habitat compared with plants growing beneath a forest canopy (Nishikawa 2009). These contradictory results indicate that the advantage of longer bright conditions may vary from year to year. Growth initiation of G. lutea in natural conditions in 2014 occurred 1 week earlier than an average year owing to earlier snowmelt (Fig. 1c). Thus, most reproductive plants might have completed seed production by the time of canopy closure, thereby making light resource limitation for seed production undetectable in an early snowmelt year.

In conclusion, cool temperature early in the spring in a forest is beneficial for spring ephemerals, and future climate warming is predicted to be detrimental to the growth and reproduction of spring ephemerals. On the other hand, the effects of early canopy closure on the fitness of spring ephemerals are not consistent depending on whether the bright period from snowmelt to canopy closure was longer or shorter than the period for the completion of bulb growth and seed production. Reproductive mode, i.e., the balance between sexual reproduction and vegetative reproduction, is crucial to predict the fate of population dynamics of perennial herbs under conditions of global climate change. G. lutea can reproduce by both seed production (sexual reproduction) at large size and bulbils (vegetative reproduction) at small size (Schnittler et al. 2009). Because we focused on the responses of only reproductive plants with a large size in the present study, it is unclear how climate warming affects bulbil production in this species. Therefore, long-term monitoring of the population dynamics is needed to evaluate the exact trend because changes in growth rate and seed-set success should affect the population dynamics of perennial plant species.