Population Ecology

, Volume 48, Issue 1, pp 49–58

Habitat-specific responses in the flowering phenology and seed set of alpine plants to climate variation: implications for global-change impacts


    • Graduate School of Environmental Earth ScienceHokkaido University
  • Akira S. Hirao
    • Graduate School of Environmental Earth ScienceHokkaido University
Original Article Special feature: Global climate change and the dynamics of biological communities

DOI: 10.1007/s10144-005-0242-z

Cite this article as:
Kudo, G. & Hirao, A.S. Popul Ecol (2006) 48: 49. doi:10.1007/s10144-005-0242-z


The timing of the snowmelt is a crucial factor in determining the phenological schedule of alpine plants. A long-term monitoring of snowmelt regimes in a Japanese alpine area revealed that the onset of the snowmelt season has been accelerated during the last 17 years in early snowmelt sites but that such a trend has not been detected in late snowmelt sites. This indicates that the global warming effect on the snowmelt pattern may be site-specific. The flowering phenology of fellfield plants in an exposed wind-blown habitat was consistent between an unusually warm year (1998) and a normal year (2001). In contrast, the flowering occurrence of snowbed plants varied greatly between the years depending on the snowmelt time. There was a large number of flowering species in the fellfield community from mid- to late to late June and from mid- to late July. The flowering peak of an early-melt snowbed plant community was in the middle of the flowering season and that of a late-melt snowbed community was in the early flowering season. These habitat-specific phenological patterns were consistent between 1998 and 2001. The effects of the variation in flowering timing on seed-set success were evaluated for an entomophilous snowbed herb, Peucedanum multivittatum, along the snowmelt gradient during a 5-year period. When flowering occurred prior to early August, mean temperature during the flowering season positively influenced the seed set. When flowering occurred later than early August, however, the plants enjoyed high seed-set success irrespective of temperature conditions if frost damage was absent. These observations are probably explained based on the availability of pollinators, which depends not only on ambient temperature but also on seasonal progress. These results suggest that the effects of climate change on biological interaction may vary depending on the specific habitat in the alpine ecosystem in which diverse snowmelt patterns create complicated seasonality for plants within a very localized area.


Alpine plantsFlowering phenologyGlobal warmingSeed productionSnowmelt time


The timing of flowering in plants is temperature-sensitive, and evidence is accumulating that global warming is accelerating this important determinant of reproductive success in plants (Fitter and Fitter 2002). Global warming is expected to have serious impacts on arctic and alpine ecosystems (IPCC 2001), in which the length of the growing season is seriously restricted. The timing of the flowering event is one of the more important factors determining the reproductive success of alpine plants because late flowering often results in the failure of seeds to mature as a result of the season being too short and early flowering may result in restrictions in pollen availability due to low pollinator activity (Kudo 1991, 1993; Molau 1993). Therefore, temporal changes in the flowering schedule should affect reproductive success of arctic and alpine plants. Two major factors determine the flowering phenology of alpine plants: ambient temperature and the timing of the snowmelt. The latter is an important ecological factor because snow distribution in alpine regions is heterogeneous even within a local area, and this heterogeneous snow distribution is a major factor creating diverse plant communities in the alpine ecosystem (Billings and Bliss 1959; Kudo and Ito 1992). Although the spatial pattern of snow dynamics is consistent among years, the actual timing of the snowmelt varies greatly from year to year, thereby reflecting the inter-annual variation in climate conditions. Consequently, the long-term monitoring of snowmelt patterns and flowering phenology is important in order to be able to detect the effects of global change on alpine ecosystems. To date, very few long-term studies of flowering phenology have been conducted on subarctic tundra communities (Thórhalsdóttir 1998; Molau et al. 2005), and phenological studies on the middle-latitude alpine regions have been reported for only a few restricted species (e.g. Walker et al. 1995; Saavedra et al. 2003). Alpine flora of middle-latitude mountains are isolated and scattered around the southernmost edges of their respective geographic distributions and, as a result, the middle-latitude alpine ecosystem must be more sensitive to the warming effect than the high-latitude tundra ecosystem.

The seasonal transition of ambient air temperature in the middle-latitude alpine region is generally steeper than that in the high-latitude tundra (Kudo and Suzuki 2002). A steep temperature gradient from the beginning to the middle of the growth season in the middle-latitude alpine causes a more distinct seasonal trend of pollinator activity than that found in the subarctic alpine, with the result that the pollination success of insect-pollinated plants obviously increases as the season progress (Kudo and Suzuki 2002). Therefore, any modification of the flowering schedule due to global climate change may cause serious impacts on the pollination success of middle-latitude alpine plants if plants and pollinators respond to the warming differently. However, the effects of climate change on the flowering phenology of alpine plants may not uniform – even within a very localized area – due to the heterogeneous distribution of snow reflecting the undulating topographic features in mountain regions. The alpine ecosystem consists of a fellfield habitat and a snowbed habitat. The fellfield habitat is situated on wind-blown ridges and plateaus where snow disappears early in the growing season and plants are exposed to cold weather during the winter and spring. In contrast, the snowbed habitat is covered with thick snow until the middle of the growing season and the initiation of plant growth is determined by the time of snowmelt. As the snowmelt time is influenced not only by ambient temperature but also by irradiation, precipitation, and the amount of snowfall during the winter, the effect of temperature on the phenological events of snowbed plants may be indirect. Thus, the impact of global warming on the flowering phenology may differ between the fellfield and snowbed habitats.

In this article, we compare inter-annual variations in the flowering patterns of fellfield and snowbed communities in the alpine region of northern Japan between an extreme warm and early-snowmelt year and a normal year. Based on data obtained from long-term monitoring of the snowmelt pattern along a snowmelt gradient, we predict the habitat-specific warming effects on flowering phenology.

We also evaluate the effects of temporal variations in the flowering season, as determined along a snowmelt gradient over a 5-year period, on the reproductive success of an entomophilous snowbed herb, Peucedanum multivittatum Maxim. (Apiaceae). Global climate change may impact on the reproductive state of alpine plants in two important ways: by means of modifying the thermal conditions during the flowering season (as a direct warming effect) and by modifying the flowering schedule caused by the temporal variation in snowmelt timing (as an indirect warming effect). An increase in ambient temperature often accelerates the activity of pollinator insects in arctic and alpine regions, resulting in higher pollination success (Inouye and Pyke 1988; Totland 1994; Bergman et al. 1996). Seasonal changes in the flowering schedule may change the availability of pollinators if phenological responses to global warming differ between plants and pollinators (Kudo et al. 2004; Inouye et al. 2000). Peucedanum multivittatum is an ideal plants to assess the relationship between flowering conditions and pollination success because (1) this species is self-compatible but pollinator visitation is still necessary for considerable seed set due to the well-organized dichogamous flowering within individuals (Kudo 1997); (2) the distribution range is wide, extending from early to late snowmelt locations, with the result that the flowering season is highly variable along a snowmelt gradient (Kudo 1997). In this study, we focus on the discriminating between the direct warming effect and the effect of flowering timing on the reproductive success of alpine plants.

In a final step, we attempt to determine the heterogeneous responses of alpine plants to global climate change based on the unique structure of the alpine ecosystem in which the steep snowmelt gradient creates diverse seasonality within a localized area.

Materials and methods

Study site

This study was conducted in the central part of the Taisetsu Mountains of Hokkaido in northern Japan (43°33′N, 142°52′E). In 1988, we established five 20×20 m2 plots (denoted plots A–E) on a southeast-facing slope near Lake Hisago (Fig. 1). These five plots were arranged by snowmelt order along a snowmelt gradient (1,790–1,900 m a.s.l.) and were all within 700 m of each other. The snowbed vegetation consisted primarily of dwarf shrubs and forbs (see Kudo and Ito 1992 for species composition). The snowmelt pattern of these plots has been monitored since 1988. In addition, one fellfield plot, denoted plot O (about 50×50 m2), was established on a plateau at 1,910 m.a.s.l. (Fig. 1), at a location where snow-cover is thin during the winter due to strong winds blowing the snow away and where the soil surface is usually exposed from snow-cover by early- to mid-May. Ericaceous dwarf shrubs and lichens are common components of the fellfield (Kudo and Suzuki 2002).
Fig. 1

Location of survey plots in the study site. Each snowbed plot (A –E) is about 20×20 m2 size, and the fellfeild plot (O) is about 50×50 m2 size. The fellfield plot is usually exposed from snow-cover by mid-May, prior to the initiation of plant growth. Snowmelt of the snowbed plots progresses in the order of A (earliest) to E (latest) (see Fig. 2)

The Taisetsu Mountains are characterized by a cold, snowy winter and a mild, wet summer. Over the study period, the mean air temperature measured at 40 cm above the ground near Lake Hisago (1,700 m a.s.l.; Fig. 1) was 4.1°C in May, 9.0°C in June, 13.2°C in July, 12.9°C in August, and 7.7°C in September (Table 1). The mean precipitation during the summer (June–August) was 767 mm, ranging from 494 to 1,010 mm (over a 7-year period; data supplied by G. Kudo). Taken over a 10-year period, the summers of 1998 and 2000 were very warm, and the summers of 1995 and 2002 were cool.
Table 1

Mean monthly temperaturea (°C) at 1,700 m a.s.l. near the research site








No data






No data
















































No data











aMeasurements were made at 1-h intervals at the height of 40 cm from the ground

Phenological survey

The flowering sequences of major entomophilous plant species were recorded throughout the growth season in plots O, B and D in 1998 and 2001. The growth season in 1998 was the warmest one of the decade, and the timing of the snowmelt in this year was the earliest it had been for the last 17 years (see Results). However, the temperature condition in 2001 was normal. The flowering stage of individual species was classified by visual observation as follows: rank A: <25% of the flower buds have opened; B: 25–50% of flower buds have opened; C: 50–75% of the flower buds have opened; D: 75–100% of the flower buds have opened; E: no flower buds remain and about 50% of flowering has finished; F: no flower buds remain and about 75% of flowering has finished. In order to quantify the flower sequence, we scored ranks A and F with a “1”, ranks B and E with a “2”, rank C with a “3”, and rank D with a “4”. Observation intervals were 3–10 days. The number of species observed was 14 in plot O, 19 in plot B, and 12 in plot D.

Seed-set success

Peucedanum multivittatum is a Japanese endemic alpine herb that commonly grows in a snowbed habitat. This species is andromonoecious in that a terminal umbel contains both hermaphrodite and male flowers, and lateral umbels usually have only male flowers. Each hermaphrodite flower has two pistils, each of which contains one ovule. The flowering of lateral umbels occurs following anthesis of the terminal umbel (Kudo 1997). Major pollinating insects are syrphid and other flies, and sometimes bumblebees. The flowering season and seed-set success of P. multivittatum were monitored in plots A, C, D and E in 1995, 1997, 1998, 2000 and 2001 (but monitoring in plot E was performed only in 1995, 1997 and 1998). At the beginning of the flowering season, 40–100 individual plants with flower buds were randomly selected in each plot, and the number of hermaphrodite flowers was counted. During the fruiting season, mature fruits were harvested, the number of developed seeds was counted, and the seed-set percentage was calculated. The peak of the flowering season in individual plots was measured as mentioned above. The mean daily temperature during the major flowering season in each plot was calculated on the basis of hourly measurements recorded 40 cm above the soil surface at 1,700 m.a.s.l. (Fig. 1). We did not compensate for elevation differences in temperature for individual plots because spatial altitudinal differences among plots was small.

To evaluate the effects of temporal variations in the onset of flowering and of ambient temperature of the flowering season on seed-set success, we applied a generalized linear mixed model of a penalized quasi-likelihood method (PQL) (Venables and Ripley 2002) using a glmmPQL function in the “MASS” library of an open source system, R ver. 2.0.1 (http://www.r-project.org/). In this analysis, “plot (P)”, “peak flowering season (S)” and “mean temperature during the flowering season (T)” were considered to be fixed factors and “year” to be a random factor. The peak flowering season of each plot was expressed as the sequence at 5-day intervals from July 1; i.e. July 1–5 = 1, July 6–10 = 2, and so on. Data of 1995, 1997 and 1998 were used for this analysis because data from plot E were lacking for 2000 and 2001. The logistic regression model used in this analysis is as follows:

$$ \log _{{\text{e}}} {\left( {\frac{p} {{1 - p}}} \right)} = \beta _{{\text{0}}} + \beta _{{{P}}} P + \beta _{{{S}}} S + \beta _{{{T}}} T + \beta _{{{{ST}}}} {{ST}} + \varepsilon , $$
where p is the probability of ovules setting seeds in individual plants, β0, βP, βS, βT and βST are regression coefficients, and ε is an error term of the random factor (year).


Snowmelt conditions

The time of snowmelt varied greatly among years within each snowbed plot, but snowmelt progressed sequentially from plot A (earliest) to plot E (latest) every year (Fig. 2). On average, snowmelt occurred in early June in plot A, late June in plot B, early July in plot C, late July in plot D, and early August in plot E. The mean difference in the onset of snowmelt for all plots between the earliest (1998) and latest (1993) year was approximately 2 months.
Fig. 2

Inter-annual variations in snowmelt time of the individual plots during 1988–2004. Day number means calendar days from January 1. Snowmelt always progressed from plot A (earliest) to plot E (latest). The time of snowmelt has become earlier during the past 17 years in plots A and B (P<0.05), but significant trends were not been detected in the other plots

The time of snowmelt has become earlier during the past 17 years in early snowmelt plots A and B (r = –0.52, P=0.042 and = –0.52, P=0.043, respectively; Spearman’s rank correlation test), but no significant trends have been detected in other plots (P>0.1). In the fellfield plot, snow usually disappeared prior to mid-May, although occasional spring snow remained on the ground for short periods from mid- to late May. In 1998, the snow disappeared in late April.

Flowering phenology

At the community level, the number of flowering species in the fellfield (plot O) was large from mid- to late June and from mid- to late July (Fig. 3), with the later peak being larger than the earlier peak. The length of the complete flowering season in plot O was 85 days in 1998 and 80 days in 2001. In the early-snowmelt snowbed, plot B, peak flowering occurred in the middle of flowering season: late June in 1998 and mid-July in 2001 (Fig. 3). The length of complete flowering season in plot B was 85 days in 1998 and 75 days in 2001. In the late-snowmelt snowbed, plot D, peak flowering occurred in the early flowering season within the plot: mid-July in 1998 and mid-August in 2001 (Fig. 3). The length of the complete flowering season in plot D was 55 days in both years.
Fig. 3

Flowering patterns at the community level in the fellfield (plot O), early-melt snowbed (plot B), and late-melt snowbed (plot D) in 1998 (solid gray lines) and 2001 (dashed dark lines). The number of species observed was 14 in plot O, 19 in plot B and 12 in plot D. Flowering duration in plot O was 85 days in 1998 and 80 days in 2001; in plot B, 85 and 75 days; in plot D, 55 and 55 days

At the fellfeild plot, flowering time and duration of individual species were similar between 1998 and 2001, irrespective of the great differences in temperature conditions (Fig. 4). The mean flowering period of individual species was 18.2 days (range: 15–30 days) in 1998 and 19.6 days (range: 15–30 days) in 2001. The peak flowering season of Arctous alpinus was 10 days earlier in 1998 than in 2001, but it almost the same between these years for other species. Many flower buds of Diapensia lapponica did not open in 1998 due to frost damage, which resulted in low floral density during the peak season.
Fig. 4

Flowering patterns of major fellfield species in plot O in the extremely warm year (1998, solid gray lines) and a normal year (2001, broken dark lines). See text for flowering ranks

The flowering sequences of 12 species found in plot D are shown in Fig. 5 as being representative of snowbed plots. In contrast with the patterns in the fellfield plot, the flowering timing of snowbed plants was strongly influenced by the snowmelt time, and flowering in 2001 occurred 15–30 days later than it did in 1998. However, the duration of flowering by individual species was similar between the years. The mean flowering period was 17.1 days (range: 10–25 days) in 1998 and 17.5 days (range: 10–30 days) in 2001.
Fig. 5

Flowering patterns of major snowbed species in plot D in the extremely early snowmelt year (1998, solid gray lines) and normal year (2001, broken dark lines). Snow disappeared on July1 in 1998 and July 23 in 2001. See text for flowering ranks

Variation in seed-set success

Variations in seed-set percentages of P. multivittatum over a 5-year period (3 years for plot E) are shown in Fig. 6. Mean seed-set values throughout the plots and years was 30.8%. The seed-set success tended to be lower in the early-melt plots and higher in the late-melt plots, although there were large variations among years within plots. In plot A, the seed-set percentage in a warm year (1997, when the mean temperature during the flowering season was 19.4°C) was higher than that in other years (Fig. 7). In plot C, the seed-set success was positively correlated on the air temperature during the flowering season. When the daily mean air temperature was lower than 15°C, seed set was lower than 25% in these plots. In plots D and E, however, seed set was large even under cool conditions (11–13°C) in which flowering usually occurred later than early August. Only in 1997 did plants growing in plot D show a relatively low seed set (23%, shown as a gray square in Fig. 7), which was due to many flowers being damaged by frost in late August.
Fig. 6

Seed-set percentages of Peucedanum multivittatum under natural conditions in each plot (A, C, D and E) and year (1995, 1997, 1998, 2000 and 2001). The horizontal axis shows the time of peak flowering in each year. The box plot represents the 75th, 50th and 25th percentile, the top whisker ranges from the 75th to the 90th percentile and the bottom whisker from the 25th to the 10th percentile. Mean seed set throughout the plots and years was 30.8% (shown as broken lines)

Fig. 7

Relationship between the daily mean temperature during the flowering season and the mean value of seed set of Peucedanum multivittatum in each plot and year. In 1997, many plants in plot D were damaged by frost in late August, resulting in low seed set (shown as a filled square)

The results of the logistic regression indicated that flowering season significantly influenced the seed-set success (Table 2). The effect of temperature during the flowering season was not significant, but the interaction between flowering season and temperature was significant (Table 2). This means that the effect of temperature on pollination success may vary as the season progress. When flowering occurred early in the season, seed-set success was influenced by the temperature of the flowering season, in which warmer conditions resulted in higher seed set. In contrast, when flowering occurred in relatively later in the season, temperature had little effect on seed-set success if frost damage was absent.
Table 2

Results of logistic regressiona for seed-set success of Peucedanum multivittatum






Intercept (Plot A)





Plot B





Plot D





Plot E





Flowering season (S)





Mean temperature (T)





S × T





aCoefficients of intercept, plots, flowering season, mean temperature during flowering, and the interaction between flowering timing and temperature are shown with standard error (SE)


Phenological variation

Responses of flowering phenology to the extremely warm year were completely different between the fellfield and snowbed communities. Because snow usually disappears prior to the initiation of growth in the fellfield, variation in snowmelt time may have few effects on phenological events. In contrast, the timing of snowmelt strongly influences the initiation of growth in the snowbed because ambient temperature at this time is usually enough warm for plant growth (Holway and Ward 1965; Kudo 1991; Kudo and Suzuki 1999). There was a 23-day difference in the timing of the snowmelt between 1998 and 2001 in both plots B and D, but the phenological difference between the years was smaller in plot B than in plot D (Fig. 3). This means that the variation in snowmelt time may not always correspond precisely to the variation in flowering time even in snowbed plants when snowmelt occurs in very early in the season (e.g. mid-May 1998 in plot B).

Surprisingly, flowering phenology of the fellfield community was not accelerated in 1998 even though the daily mean temperature in May and June was respectively 2.7 and 2.5°C warmer in 1998 than in 2001. In contrast to the daily mean temperature, the daily minimum temperature in June 1998 (0.9°C) was lower than that in 2001 (3.4°C), and the number of days that the minimum temperature fell below 0°C was 14 days in 1998 and 7 days in 2001. Thus, low nighttime temperatures may have decelerated the development of floral buds in 1998 despite the warm daytime temperature. In their warming experiment using the open-top-chamber (OTC) in the fellfield of the Taisetsu Mountains, Suzuki and Kudo (2000) found that the flowering phenology of the dwarf shrub, Vaccinium uliginosum, did not change relative to that of the control when, as a result of the warming treatment, the daily mean temperature was 2.5–2.9°C higher but the daily minimum temperature was 0.3°C lower than the control situation. A similar result was reported in alpine Norway in which flower development of Ranunculus glacialis was conservative when subjected to the warming treatment by the OTC (Totland and Alatalo 2002). These results indicate the importance of the minimum temperature for the onset of flowering.

Flowering phenology of plants is a temperature-dependent phenomenon in many cases (Rathcke and Lacey 1985), and the cumulative temperature above a certain threshold value from a certain starting date (i.e., effective cumulative temperature) has often been used as an indicator of flowering traits (Rathcke and Lacey 1985; Diekmann 1996; Molau et al. 2005). The flowering pattern of a local community is determined by the interaction between seasonal transition of temperature and the thermal-demand for the onset of flowering of individual species (Kudo and Suzuki 1999). When a community is first exposed from the snow cover early in the growing season, the effective cumulative temperature will initially increase very gradually, but the rate of temperature increase will accelerate after that as the season progresses because the air temperature increases rapidly from mid-June to mid-July. In such a case, realized flowering will be distorted towards later within the flowering season (negative skewness). When a community is exposed from snow cover in the mid-summer, when the ambient temperature is at its maximum level, the effective cumulative temperature will increase constantly at first, but the increase rate decelerates later due to the decrease in air temperature as autumn approaches. In this case, realized flowering will be distorted towards earlier within the flowering season (positive skewness). The study of Kudo and Suzuki (1999) revealed that the distribution of the temperature-demand of the fellfield species for flowering is bimodal with a wide range, while that of the snowbed species is unimodal with a narrow range that results in simultaneous flowering. Therefore, the flowering pattern of individual communities is potentially habitat-specific and conservative against inter-annual climatic variations if changes in temperature regime are not drastic.

Seed-set success

Our 5-year census of seed-set percentages in P. multivittatum along the snowmelt gradient demonstrates that the temperature effect on seed-set success varies as the season progresses. The positive effect of ambient temperature of the flowering season was obvious only in plots A and C, where flowering commonly occurred in July. Such a seasonal variation may be related to seasonal changes in the activity and availability of pollinating insects. Flowers of P. multivittatum are commonly visited by dipteran insects (syrphid and other flies) throughout the season, but worker bumblebees sometimes visit the flowers in August to collect pollen (G. Kudo, personal observation). Because worker bumblebees usually appear in late July and are at their most active phase in August in the Taisetsu Mountains, only late-flowering plants can utilize worker bumblebees as pollinators. Dipteran insects are the most common pollinators in arctic and alpine regions, and their activity highly depends on ambient temperature (Hocking 1968; Kevan 1972; Levesque and Burger 1982; Inouye and Pyke 1988; Totland 1994). Thus, the positive correlation between ambient temperature and seed set in early-melt plots may reflect the temperature-dependent activity of the pollinators. As bumblebees are endothermic social insects, they are effective pollinators in arctic and alpine regions even under cool conditions (Fægri and van der Pijl 1979; Heinrich 1979; Bergman et al. 1996). This may result in a high success rate with respect to seed set that is independent of temperature conditions in the late-melt plots.

An increase in the seed-set success as the season progresses has been reported at this site for both intra-specific (Kudo 1993 for Rhododendron aureum) and inter-specific comparisons (Kudo and Suzuki 2002), primarily due to both a seasonal increase in ambient temperature and the seasonality of pollinator availability. In contrast, several studies have shown that seed-set success is higher in early-flowering individuals of species endemic to the Scandinavian Mountains (Stenström and Molau 1992; Totland 1994, 1997a, b) and the USA Rocky Mountains (Galen and Stanton 1991). Thus, the seasonal trend of seed-set pattern may be a region-specific phenomenon (Kudo and Suzuki 2002).

On the other hand, a delayed flowering has been found to occasionally result in the failure of the seed of P. multivittatum to mature due to the lack of a seed-developing period (onset of winter or frost damage in late summer) (Kudo 1997 and result of this study). P. multivittatum is a relatively late-flowering species among snowbed plants and is sensitive to harsh weather conditions in the late summer. In 2002, for example, most plants in the late-melt plots failed to set seeds due to the physical damage caused by an early snowfall in late August (A.S. Hirao, personal observation). It is predicted that late-flowering species will bloom earlier, reproduce more, and grow larger under conditions of global warming (Totland 1997b; Molau et al. 2005). However, this prediction will only actually be realized if and when both flowering phenology and insect seasonality make the same forward temporal shift. The extent of phenological synchronization between flowering season and pollinator activity is a crucial issue for the prediction of climate change impacts on the alpine ecosystem (Inouye et al. 2000).

Implications for climate change

There are many predictions and reports that ongoing climate change may reduce the snow-pack and accelerate the time of snowmelt in arctic and alpine regions (e.g., CSIRO 1994; Whetton et al. 1996; IPCC 2001). However, the average date of snowmelt in the Colorado Rocky Mountains has not changed during the past 25 years despite a trend for warmer spring temperatures, probably due to the increase in winter snowfall (Inouye et al. 2000). In the Taisetsu Mountains, an acceleration of the snowmelt time has been detected only in the early-melt places during the last 17 years. Therefore, warming effects on snow regimes seem to vary between regions and between sites within a region. In the Taisetsu Mountains, topographic features may determine the maximum snow depth around the bottom of the snowbed because wind-blown snow accumulates on depression locations. In contrast, the depth of snow around the marginal snowbed may depend on the snowfall, and the timing of snowmelt may be sensitive to spring weather conditions due to a relatively shallow accumulation of snow. Thus, plants growing in the early-melt habitat may be more sensitive than those growing in the late-melt habitat with respect to their response to climate change.

An acceleration of the snowmelt time often decreases the survival, growth, and reproductive activity of alpine plants in the early-melt habitat when early exposure from the protection of the snow-cover enhances the risk of frost damage and water stress (Walker et al. 1995; Inouye et al. 2002; Saavedra et al. 2003). Based on a 10-year study carried out in the subalpine region of northern Sweden, Molau (1996) reported that during the last 3 years of the study the biomass of an evergreen cushion plant, Diapensia lapponica, abruptly decreased by 22% and its flower production decreased by 55% as a result of frost damage caused by an extremely early snowmelt in the early-melt habitat; conversely, plants in the late-melt habitat showed little or no damage. In our study also, floral buds of D. lapponica were damaged by frost in the extremely warm year (1998). Although the phenological responses of fellfield plants to the warm condition were small, early exposure from snow-cover may enhance the risk of frost damage because reproductive organs of arctic and alpine plants are sensitive to freezing temperatures (Billings 1974). Thus, the loss of snow insulation in the early season may cause a serious frost impact on the survival and reproduction of plants in the fellfield and the early-melt snowbed habitats under global warming (Inouye 2000).

The existence of snowmelt gradient is a crucial factor in maintaining the biodiversity of the alpine ecosystem in that creates various habitat types and seasonally diverse phenological features within a local area. Our most recent study revealed the existence of genetic structure among neighboring populations having different snowmelt conditions, probably because segregated flowering seasons isolate the gene flow via pollination process (Hirao and Kudo 2004). Modification of the flowering phenology along the snowmelt gradient may change the spatial genetic structure among local populations if climate change increases the flowering overlap among populations.


We thank S. Suzuki, T. Kasagi and Y. Shimono for their assistance in the field work. This study was partly supported by a grant-in-aid from the Ministry of Environment for the Global Environment Research Fund (F-052).

Copyright information

© The Society of Population Ecology and Springer-Verlag Tokyo 2005