By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
1. The likely phenological responses of plants to climate warming can be measured through experimental manipulation of field sites, but results are rarely validated against year-to-year changes in climate. Here, we describe the response of 1–5 years of experimental warming on phenology (budding, flowering and seed maturation) of six common subalpine plant species in the Australian Alps using the International Tundra Experiment (ITEX) protocol.
2. Phenological changes in some species (particularly the forb Craspedia jamesii) were detected in experimental plots within a year of warming, whereas changes in most other species (the forb Erigeron bellidioides, the shrub Asterolasia trymalioides and the graminoids Carex breviculmis and Poa hiemata) did not develop until after 2–4 years; thus, there appears to be a cumulative effect of warming for some species across multiple years.
3. There was evidence of changes in the length of the period between flowering and seed maturity in one species (P. hiemata) that led to a similar timing of seed maturation, suggesting compensation.
4. Year-to-year variation in phenology was greater than variation between warmed and control plots and could be related to differences in thawing degree days (particularly, for E. bellidioides) due to earlier timing of budding and other events under warmer conditions. However, in Carex breviculmis, there was no association between phenology and temperature changes across years.
5. These findings indicate that, although phenological changes occurred earlier in response to warming in all six species, some species showed buffered rather than immediate responses.
6.Synthesis. Warming in ITEX open-top chambers in the Australian Alps produced earlier budding, flowering and seed set in several alpine species. Species also altered the timing of these events, particularly budding, in response to year-to-year temperature variation. Some species responded immediately, whereas in others the cumulative effects of warming across several years were required before a response was detected.
Nevertheless, plants may compensate for climate-induced changes in individual phenological events. Earlier budding or flowering may lead to earlier seed set or be compensated for by later development and maturation of seed, so that the timing of seed production and release is unaltered (Post et al. 2008). Apart from these plastic responses, there is also the possibility that evolution will change individual phenological events. For instance in Brassica rapa, flowering time has shifted in response to drier conditions during development (Franks & Weis 2008). To predict the long-term impact of phenological changes on plant fitness, the effects of compensation and evolution need to be considered and assessed (Post et al. 2008).
The effects of global warming on phenology might be particularly evident in alpine and arctic plants where temperature variation can have a large impact on phenological events (Parmesan & Yohe 2003; Root et al. 2003). Potential changes in these environments have been investigated as part of the International Tundra Experiment (ITEX) (Henry & Molau 1997), which involves passive warming in open-top chambers (OTCs) that cover plots of c. 1–2 m2. In general, plants in warmed plots show earlier budding and flowering (Arft et al. 1999), although species also differ in their responses and may even show later flowering (Hollister, Webber & Bay 2005). There is variation in the speed at which plants shift phenology after experimental warming is started. Some perennial species shift rapidly, whereas in others it appears that cumulative effects of warming are required across multiple growth years (Arft et al. 1999). This suggests that plastic changes in phenology are complex.
Much of the research on ITEX sites has so far focused on testing for general effects of warming on phenology and species composition, but there is also a growing opportunity to link variation detected in experimental plots to year-to-year changes seen in control plots across monitoring periods. For instance, changes in the composition of vegetation in warmed plots involving an increase in woody vegetation match the increase in this type of vegetation seen in control plots as conditions in the arctic have warmed (Wahren, Walker & Bret-Harte 2005). Year-to-year variation can also be used to assess the importance of factors that are not manipulated in ITEX plots, such as changes in seasonal rainfall patterns and snow melt known to influence phenological events in alpine and subalpine plants (Walker, Ingersoll & Webber 1995; Wagner & Reichegger 1997).
Here, we use replicated ITEX plots and year-to-year variation in temperature conditions to investigate phenological changes in six diverse plant species from a subalpine area close to the tree line in south-eastern Australia. Conditions in this region have become warmer with a rise of c. 0.2 °C per decade since the 1970s (Hennessy et al. 2003). There has also been a decrease in rainfall (http://www.bom.gov.au), snow depth and snow cover (Osbourne, Davis & Green 1998; Hennessy et al. 2003; Nicholls 2005). Herbarium records suggest potential changes in flowering time in some species from this region (Gallagher, Hughes & Leishman 2009). In an earlier study on these ITEX experiments based on data from 3 years (Jarrad et al. 2008), there was suggestive evidence that warming resulted in earlier occurrence of key phenological events in some species.
We consider the following questions. Is there evidence for divergence in phenology in six common subalpine species after 5 years of experimental manipulation? Is there evidence of compensation across phenological events? How do the detected effects of ITEX manipulations compare with year-to-year variation in phenology in control sites? To what extent can this variation be predicted by models based on degree-day (DD) accumulation? Do species that show the most rapid responses to warming also track year-to-year variation more closely? We also discuss the implications of these results on the selection of species used to monitor the effects of global warming near the tree line.
Materials and methods
The OTCs used in ITEX experiments are inexpensive and robust and provide protection from wind, but they are only suitable for warming a small area of 1–2 m in diameter. Establishment of the OTCs is described in Jarrad et al. (2008) and only a brief overview is provided here. The OTCs (1.7 m in diameter; 110 cm at the top, 168 cm at the base) were set up at c. 1750 m a.s.l. in November 2003 on the Bogong High Plains, Victoria, Australia, according to the ITEX protocol (Molau & Molgaard 1996). The OTCs were placed over the plots throughout the snow-free period, from snow melt, when 50% of the ground was free of snow (October or November), until 50% of the snow cover returned (June).
This study focuses on results from the two unburnt sites located 0.5 km apart described in Jarrad et al. (2008). There were 13 OTCs and 13 controls (CTL) at each of these sites. The sites had a similar subalpine open heathland community – community 26 in McDougall (1982). The tussock grass Poa hiemata and forb Celmisia pugioniformis predominated in this community at both sites, but there were also various other grasses, daisies, rushes, sedges, forbs and shrubs. Air temperature was recorded at 5 cm above the surface with Onset Hobo data loggers in eight plots (four OTC; four CTL) per site. The OTCs warmed the air by 0.9 °C during the day and 1.0 °C at night based on 2004/2006 data. The OTCs also warmed soil by 1.2–1.4 °C at the surface and 0.9–1.0 °C at 5 cm depth (Jarrad et al. 2008).
We focused on six common subalpine species for monitoring that included the main families and growth forms consisting of grasses (Poaceae), sedges (Cyperaceae), shrubs (Rutaceae) and forbs (Asteraceae, Ranunculaceae). We measured the phenology of plants in the central 1-m2 region of the plots following the ITEX protocol. The species occur in the alpine and high subalpine zones. Plant communities in these zones on the Bogong High Plains are the same floristically and not geographically separated, being part of the same dissected plateau. Celmisia pugioniformis, which occurred in all plots, was not included because it does not flower every year and there were insufficient data points to test for patterns. Plots were visited every 2–4 days during the snow-free period from November 2003 to March 2008. The dates of first flower bud, opening of flower and seed maturation as defined by Jarrad et al. (2008) were recorded when the first plant exhibited the trait for each plot following the ITEX protocol (http://www.geog.ubc.ca/itex/). These pheno-phase dates were expressed in days since 1 June to be more consistent with northern-hemisphere studies.
ITEX comparison To test if phenology has shifted as a consequence of experimental warming, repeated-measures anovas are often carried out on ITEX phenology data from the same plots across years. However, if this analysis is applied to our data from the first and last years, many plots are excluded, because often the same plots did not flower in both of these years. This results in few degrees of freedom for the error term and low power for detecting warming effects.
To overcome this problem, we considered two approaches. First, we used anovas to analyse the effect of treatment on phenological traits for each sampling year (i.e. data were not combined across years). Site was also included in the anova but not presented because the focus of this study is on treatment effects and because site differences were mostly not significant. When presenting the data graphically, site effects were removed by adding or removing the signed difference between site means and the overall mean to individual data points. We analysed traits separately and expected some species–trait–year combinations to differ between treatments by chance, and warming effects are therefore interpreted cautiously except where significance was maintained following Bonferroni correction of probabilities for the number of comparisons per species. For some of the species–trait–year combinations (budding in Erigeron bellidioides 2006/2007, 2007/2008 and in Ranunculus victoriensis 2004/2005; flowering in Craspedia jamesii 2005/2006, E. bellidioides 2007/2008, R. victoriensis 2006/2007, Carex breviculmis 2003/2004; seed maturation in Carex breviculmis 2003/2004), data were not normally distributed by Kolmogorov–Smirnov tests or graphically via probability plots (Quinn & Keough 2002), so non-parametric Mann–Whitney U-tests were used instead. All statistical tests were run with spss for Windows (SPSS, Chicago, IL, USA) or systat 12 (Crane, Bangalore, India).
Second, we combined adjacent early years (to represent the early period) and adjacent late years (to represent the late period) to increase the number of plots with data suitable for a repeated-measures anova. To be included in the analysis, plots had to exhibit the trait in at least one of the years in each of the periods. When combining data from the adjacent early growth years (2003/2004 and 2004/2005, except for R. victoriensis where 2004/2005 and 2005/2006 data were combined because only a few plants flowered in 2003/2004), we initially corrected for differences between the overall mean value for adjacent years. For instance, if flowering time in P. hiemata had a mean of 220 days in 2003/2004 across all plots that flowered and a mean of 230 days in 2004/2005, we added 5 days to all plot values where flowering occurred in 2003/2004 and took away 5 days from all plot values where flowering occurred in 2004/2005. We combined data from the later growing seasons (2006/2007 and 2007/2008) in the same way. anovas were then run with treatment as a fixed factor and period (early or late) as a random factor. There were insufficient data points for budding and seed maturation in the shrub Asterolasia trymalioides for these analyses. Site was also included as a factor in the anova but site (and interaction terms involving site) was only significant for P. hiemata, and these analyses are not presented. Data for the repeated-measures anovas were normally distributed based on Kolmogorov–Smirnov tests and also using probability plots, and variances were similar between treatments based on Levene tests.
Fruit development period Plants might alter seed maturation time to compensate for warming effects on flowering time, to ensure that they produce seed at a similar period within the growing season. To test this, we computed the time of seed maturation minus the time of flowering in each plot where both seed maturation and flowering time were available for a particular plot in a given year. This measure (referred to as the fruit development period) might reflect the ability of plants to compensate for differences in flowering time by changing the rate at which fruit matures. Two-way anovas were used to test for treatment and site effects on fruit developmental period for each species–year combination. We also ran repeated-measures anovas (corrected for site as described above) with year and treatment as factors to test for changes in fruit development period over time using the 2003/2004 and the 2007/2008 data. However, for two species (R. victoriensis and Carex breviculmis), there were insufficient early data in a single year and we therefore combined plots that flowered in either the 2003/2004 or 2004/2005 periods after adjusting for year effects as described before. These data were all normally distributed by Kolmogorov–Smirnov tests and graphical assessments.
Year-to-year variation To test if year-to-year variation in temperature affected phenological traits (budding, flowering, seed maturation) in the different species, a thawing DD model was used. Development of a species towards a phenological event was assumed to depend on air temperature (a threshold of 0 °C was adopted to represent thawing – although higher values did not improve prediction) and degrees accumulated across days. Temperature data were obtained from the air conditions recorded at the sites 5 cm above the ground inside and outside OTCs. However, there were occasional gaps in the data set where temperatures were not recorded because of malfunctioning sensors. In this case, we substituted data from a local weather station (Falls Creek, 36.86°S, 147.28°E) (http://www.bom.gov.au); air temperatures at plots within the growing season were predicted through linear regression from records collected at the weather station (R2 > 0.95).
We initially computed the mean time of an event for each species and trait (averaged across all plots). We then computed the cumulative DD for this mean separately for each combination between plot, site, year and treatment. The cumulative DD was obtained by adding up the number of degrees in a day where the maximum temperature exceeded zero, this was then accumulated across days. DDs for flowering time were computed based on the combined DDs to budding and between budding and flowering, whereas DDs for seed maturation time also encompassed the flowering to seed maturation period. If phenological events depend on temperature, we predicted that there would be a negative relationship between days to the phenological event and cumulative DD for each site–treatment–year combination; i.e. earlier occurrence of phenological event with increasing DD values. These relationships were tested through linear regression analyses of cumulative DDs onto days to the phenological event. We also included time of snow melt in these regression analyses as a separate variable, but found that inclusion of this variable did not improve the fit because this factor is already contributing to the cumulative DD values. We also included treatment (OTC vs. control) in the analyses, but this factor did not significantly influence the regression slope (P >0.05) in any instance, again probably because the OTC difference was already accounted for in the DD values.
Combining probabilities and effect sizes
To examine the consistency of trends across the different species, we followed the Z method as outlined in Whitlock (2005) to combine probabilities from comparisons across individual seasons or single species, initially treating the probabilities as one-tailed but then testing the combined value as a two-tailed test. We also computed effect size by year following the approach of Gurevitch & Hedges (2001). This allowed us to directly compare the effects detected to those from other similar studies.
Phenology across years
Changes in phenology across the 5 years measured are presented Figs 1–3 along with P-values for anovas testing if there were shifts in phenology between warmed control plots each year. These results indicate that, in some species, a number of years were required for the warming treatment effect to be expressed (to produce significant differences between the OTCs and control plots), whereas warming effects were expressed more quickly in other species. The results also indicate that year-to-year variation in phenology was generally larger than the treatment effects.
For the shrub A. trymalioides, flowering time differed significantly (even after Bonferroni correction) between treatments in 2006/2007 and 2007/2008 when the shrub flowered earlier in the OTCs by c. 10 days (Fig. 2). In the repeated-measures anova testing for a shift in phenology between the early and late monitoring period, there was a significant effect of treatment and a Treatment × Year interaction (Table 1), indicating that treatment differences have developed since 2003/2004.
Table 1. Results from repeated-measures anova of phenological events in six subalpine species
Treatment (d.f. = 1)
Between subjects error (d.f.)
Year (d.f. = 1)
Treatment × Year (d.f. = 1)
Within subjects error (d.f.)
Comparisons were between the average of the first two (2003/2004 and 2004/2005) and the last two (2006/2007 and 2007/2008) growth years unless otherwise indicated. Days from 1 June until first appearance of phenological trait was used.
*P <0.05, **P <0.001.
†Insufficient data collected on other traits
‡First 2 years based on 2004/2005 and 2005/2006 data
For the forb Craspedia jamesii, buds, flowers and seeds emerged on average 2–10 days earlier in the OTCs, and the effects of treatments on each of the traits was significant in at least 1 year (Figs 1–3). In the repeated-measures anovas, the effects of treatments on each of the traits was significant and there were no interactions (Table 1), suggesting no difference in response between early and late years. The forb E. bellidioides showed consistently earlier flowering and seed maturation in the OTCs after 2003/2004, and the effects of treatments on each of the traits was significant in 2007/2008 (even after Bonferroni correction) when plants in the OTCs flowered and set seed on average 7–8 days earlier than in the control plots. In the repeated-measures anovas on flowering, both the main effect of treatment and Treatment × Year interaction were significant, whereas for budding only the Treatment × Year interaction was significant (Table 1). In contrast, in the forb R. victoriensis, the effect of treatments was only significant in 2007/2008 due to earlier flowering in the OTCs by an average of 3 days. Treatment effects were not detected in the repeated-measures anovas for this species (Table 1).
In the sedge Carex breviculmis, budding and flowering occurred earlier in the OTCs in later years (2006/2007 and 2007/2008) by 5–10 days on average (Figs 1 and 2). Patterns for seed maturation were inconsistent, but maturation was significantly earlier by 7 days in the OTCs in the final phenological season (Fig. 3). In repeated-measures anovas, there was a significant effect of treatment for budding and significant interactions for flowering and budding (Table 1). Budding and flowering in the grass P. hiemata occurred significantly earlier in the OTCs by on average 4–12 days in 2005/2006 to 2007/2008 (Figs 1 and 2). In the repeated-measures anovas, treatment and Treatment × Year interactions were significant for flowering but not for budding and seed maturation (Table 1).
Comparing patterns across traits and species, events in 2007/2008 occurred on average earlier in the OTCs than in the controls for all 16 trait–species comparisons (with nine of these being significant and a combined probability of P <0.001 based on the Z method; Whitlock 2005). Flowering, seed maturation and budding all took place significantly earlier in the warmed plots (P <0.001, except for budding where P =0.012) when compared across all species. In contrast, in 2004/2005 only 9 of the 16 species–trait comparisons among treatments were on average earlier in the warmed plots (with only one significant in the expected direction, and a combined P-value > 0.05).
For the effect size analyses, results for days to flowering (Fig. 4) indicate that where differences emerge in later years (CIs of effect size do not cross 0), the effect size is c. 1 in some species (e.g. P. hiemata) but greater in others (e.g. A. trymalioides). For budding and seed maturation, effect sizes were also c. 1 when differences among treatments were significant.
Fruit development period
The period between the seed maturation and flowering phases was compared to test whether it had altered in the OTCs due to warming. Treatment differences were significant in several instances (Fig. 5). For Craspedia jamesii, the period between seed maturation and flowering was greater in the controls than in the OTCs in 2003/2004 but this switched in later seasons (significantly so in 2005/2006), although treatment and interaction effects were not significant in the repeated-measures anova (across 2003/2004 and 2007/2008) (Table 2). For E. bellidioides and Carex breviculmis, differences between treatments were consistently small (Fig. 5) and not significant in the repeated-measures anovas (Table 2), although in Carex breviculmis compensation increased across years in both OTCs and controls. Warming treatment effects were significant in the repeated-measures anovas for R. victoriensis and particularly for P. hiemata (Table 2). In these two species, plants from OTCs had extended fruit maturation periods in all years (significant in two years for P. hiemata) indicative of compensation (Fig. 5).
Table 2. Results from repeated-measures anova of compensation of phenological events (first seed maturation minus first flowering time) in six subalpine species
Treatment (d.f. = 1)
Between subjects error (d.f.)
Year (d.f. = 1)
Treatment × Year (d.f. = 1)
Within subjects error (d.f.)
Comparisons were between the average of the first two (2003/2004 and 2004/2005) and the last two (2006/2007 and 2007/2008) growth years unless otherwise indicated.
*P <0.05, **P <0.001.
†First 2 years based on 2004/2005 and 2005/2006 data.
Regression analyses were undertaken to test if year-to-year variation in phenology could be predicted by differences in DDs. For the shrub A. trymalioides, these analyses indicated that budding and flowering were significantly earlier in years with higher cumulative DDs, and patterns were consistent for the OTCs and controls (Fig. 6). No pattern was evident for seed maturation although only a few data points were available.
For the forbs Craspedia jamesii, R. victoriensis and E. bellidioides, there were negative relationships between phenology and DDs for most traits (Fig. 6). R2 values were particularly high for E. bellidioides. For the graminoid Carex breviculmis, changes in phenology were not associated with year-to-year variation in DDs (Fig. 6). However, for the graminoid P. hiemata, phenology and DD were negatively associated and the regression coefficient for budding was significant.
Overall across the species, timing of budding was negatively associated with DDs across years and treatments in five of six species (all five relationships significant, with a combined probability based on the Z method < 0.001). When OTC and control treatments were considered separately, there was a negative relationship in 10 of the 12 comparisons. For flowering time, associations were negative rather than positive for the same number of comparisons as for budding (5 of 6 combined, or 10 of 12 with treatments separated), and the combined P-value across species was < 0.001. For seed maturation, the relationship between timing and DDs was negative in five of six comparisons (9 of 12 if treatments are considered separately), but the combined P-value across species was not significant (> 0.05).
The results complement the early findings on these ITEX plots that suggested phenological responses to warming among Australian subalpine species (Jarrad et al. 2008). However, the use of additional monitoring years has produced much more consistent effects across species compared with idiosyncratic changes noted in the earlier work. Effect sizes are similar in magnitude to those observed for phenological events in other ITEX experiments (e.g. Arft et al. 1999). Moreover, temperature effects were evident from year-to-year variation in phenology being linked to thawing DD predictions.
Species tested differ markedly in the speed at which they have responded. The delayed response of many species is consistent with observations from other ITEX sites (Arft et al. 1999; Molau 2001; Hollister, Webber & Bay 2005). Responses in reproductive traits in some species may take several years to emerge, presumably because the formation of flower buds occurs one to several seasons prior to flowering (Billings 1974). In our study, one of the later species to flower (Craspedia jamesii) was the fastest to respond, whereas one of the earlier-flowering species (R. victoriensis) did not respond until the end of the monitoring period. Another member of the Ranunculus genus exhibits little response to warming in ITEX chambers over 3 years at an alpine site in Norway (Totland & Alatalo 2002). Perhaps early flowering species respond less rapidly to warming because development starts while plants are still covered with snow. Early flowering species might be more vulnerable to climate change if an inflexible flowering time in response to temperature entails fitness costs, such as through a mis-timing of flowering and the availability of pollinators (Kameyama & Kudo 2009).
The present findings indicate that subalpine plants from the Australian Alps respond to shifts in temperature. One other detailed study on this flora suggested that temperature might not be important; Venn & Morgan (2007) found that in three species from the Bogong High Plains (Celmisia pugioniformis, Luzula acutifolia, Poa fawcettiae), flowering time and seed maturation were independent of the time snow melted and might instead be driven by photoperiod rather than temperature. However, Gallagher, Hughes & Leishman (2009), using herbarium records, noted long-term changes in flowering time associated with temperature shifts over the last 50 years in several alpine species. This is consistent with our year-to-year records, which indicate that the phenology of most species from this region is likely to be altered by changes in thermal conditions. Large-scale surveys from other regions have also shown that average temperatures within a year prior to flowering influence flowering time of a range of plants (e.g. Menzel et al. 2006; Miller-Rushing & Primack 2008).
Although budding and flowering time were influenced by temperature, we also found that the period between flowering time and seed maturation could be extended when flowering was early, particularly in one grass (P. hiemata). Due to changes in the fruit development period, there were only minor differences in the timing of seed maturation in this species between the OTC and control plots in 2007/2008, compared with much larger differences in flowering and budding time. Our data on fruit development are consistent with the results of Post et al. (2008) who found that the alpine chickweed Cerastium alpinum could compensate at later stages of development for an effect of experimental warming on emergence, although these authors also found no evidence of such changes in two other species. Poa hiemata is a dominant species in the alpine and subalpine community, and this species may be able to respond to warmer climatic conditions through delaying fruit maturation. In previous work involving transplants of P. hiemata across an environmental gradient, this species showed substantial adaptive capacity both through plastic changes and genetic variation (Byars, Papst & Hoffmann 2007). It remains to be seen if an extended fruit maturation period in the OTCs increases the fitness of P. hiemata.
The changes we observed in the OTCs were smaller than phenological changes occurring from year to year, which were also related to temperature. We found that the timing of budding in particular was linked to cumulative DDs. This approach provides an opportunity to test the importance of the immediate within-season effects of post-thawing temperature on plant development. If sufficiently long time series were available on across-season variation, it might even be possible to test if there is an influence of temperature (and rainfall) in past growth seasons on phenology, as suggested by the delayed phenological responses of several plants in our current ITEX experiments.
Based on herbarium records, Gallagher, Hughes & Leishman (2009) suggested that flowering time in the alpine groundsel, Senecio pectinatus var. major, may be an indicator of climate change in the Australian Alps. Our results indicate that some species respond rapidly to experimental warming, others respond more slowly requiring cumulative effects across multiple years, while there are also sharp differences in responses to year-to-year variation in thermal conditions. Given these diverse patterns, what makes a good indicator species? An ideal species is likely to be one that responds to the cumulative effects of warming over time, is fairly common (unlike S. pectinatus var. major), flowers consistently and does not alter its flowering time dramatically in response to year-to-year variation. Such a species might then show a gradual shift in phenology but a reduced immediate response to thermal conditions within the season where flowering is being measured. Its phenology is more likely to reflect the cumulative effects of several years. In terms of the species considered here, we suspect that two good candidates are flowering time in P. hiemata and flowering time as well as budding in Carex breviculmis. Flowering time in E. bellidioides and A. trymalioides might also be suitable because of the large effect sizes we detected for these species in later years, although phenology in both species did respond to year-to-year variation in temperature. Craspedia jamesii probably responds too quickly and might even be in the process of acclimating to warming given the apparent reduction in effect size in later years (Fig. 4). A related species (Craspedia lamicola) shows a high level of plasticity in its morphological growth form when transplanted across altitudinal gradients (Byars & Hoffmann 2009).
We have so far only considered the effects of temperature on phenological changes, whereas the timing of plant development may also depend on changes in water availability, fire and perhaps CO2 (e.g. Franks & Weis 2008; Hovenden et al. 2008). A reduction in wind exposure might influence phenology in OTCs, although only the central area of the OTCs was monitored in the current experiments. Predictions about phenological changes in alpine and subalpine environments will ultimately require an understanding of how changes in temperature, rainfall and CO2 interact to influence plant development. This will require more complicated models that include variables other than DDs. Ideally, these models will also include predictions about fitness effects, so that the ecological consequences of changes in phenology in perennial plants can be considered. Changes in phenological traits might influence the amount and dispersal of seed in plants due to altered interactions with pollinators and seed dispersers (Kudo & Hirao 2006; Kameyama & Kudo 2009). Finally, although we have considered the timing of first budding, flowering and seed maturation as in other ITEX experiments, it is not clear whether these traits accurately reflect the distribution of phenological events across a season. The estimation of first flowering time will also depend on the size of the flowering plant population (Miller-Rushing, Inouye & Primack 2008). In our plots, where the number of plants of different species in the ITEX chambers was similar from year to year, there was nevertheless variation in the number of plants that flowered.
To conclude, we have found that experimental warming using the ITEX protocol has led to phenological changes in plants in the Australian Alps, particularly in flowering time. Effect sizes were similar to those observed in other ITEX sites in the Northern Hemisphere. Species differ in whether they express these changes soon after warming started or after a number of years, and species also differ in their ability to compensate for shifts in flowering time by the time seed has matured. Most species respond to year-to-year variation in thermal conditions at the budding stage but this source of variation is less apparent at the time of seed maturation. Phenological responses of Australian subalpine plants may therefore involve immediate responses to temperature conditions within a growth season as well as cumulative effects across years.
This research was funded through Australian Research Council Linkage Grants, partnered through the Department of Sustainability and Environment, Parks Victoria, Commonwealth Scientific and Industrial Research Organisation (CSIRO), ESLink Services Pty Ltd, and Mt Hotham Resort Management. A.A.H. was a recipient of a Federation Fellowship while completing this work. Assistance with phenological measurements was provided by Carolyn Blomley, Cherie Campbell, Deborah Cargill, Seraphena Cutler, Brad Farmilo, Katherine Giljohann, Sharon Honicke, Lauren Kiem, Annie Leschen, Danielle Ryan, Paul Smart, Clare Warren and Emma Warren. We are grateful to two anonymous referees for comments that improved this study.