- Top of page
- Materials and Methods
- Supporting Information
Species shifts in latitude and altitude provide strong evidence of the impact of recent climate warming on biodiversity (Kelly & Goulden, 2008; Lenoir et al., 2008; Rosenzweig et al., 2008). Yet climate impacts may be underestimated where limited dispersal retards population expansion (Parmesan & Yohe, 2003) or where land-use constrains species ranges (Davis & Shaw, 2001). Alternatively, climate impacts may be overestimated where changes in species distributions are driven by human transport that facilitates the spread of alien species (Hulme et al., 2009). Thus differences in dispersal and/or land-use change could obscure distribution changes in responses to climate. Instead, measures of plant performance such as phenological shifts, particularly in first flowering date (FFD), may be better indicators of the potential impact of climate change (Peñuelas & Filella, 2001; Root et al., 2005; Parmesan, 2007).
Plant phenology is particularly sensitive to climate and a key indicator of environmental change (Badeck et al., 2004; Estrella et al., 2007; Peñuelas et al., 2009; Yang & Rudolf, 2010), but while earlier flowering should allow plant distributions to increase (Hegland et al., 2009; Miller-Rushing & Weltzin, 2009; Steltzer & Post, 2009), a link between FFD and changes in species distributions has not been observed. Such a link would enable phenological changes to be more rigorously used as early warning systems of potential climate impacts on species distributions (Menzel et al., 2006; Cleland et al., 2007; Crimmins et al., 2009). Process-based models predict such a link (Chuine & Beaubien, 2001; Morin et al., 2007) under the assumption that phenology governs reproductive success, growth and survivorship which ultimately determines the probability of species occurrence under particular climatic conditions (Cleland et al., 2007; Steltzer & Post, 2009). Finding such a link is a significant challenge as few long-term studies have examined temperature-related changes in FFD for more than a handful of species in natural population s (Abu-Asab et al., 2001; Fitter & Fitter, 2002; Peñuelas et al., 2002; Willis et al., 2008; Gordo & Sanz, 2009) and thus any test requires correspondence between several scarce data sets: long-term records of FFD, climate and distribution change.
Under what circumstances might we expect the response of flowering phenology and species distributions to climate to be linked? A widely held belief is that species from warmer climates are pre-adapted to respond rapidly to increasing temperatures and this could be facilitated by flowering phenology tracking climate change (Walther et al., 2009). Alternatively, the flowering phenology of life-forms with short generation times, such as annual weeds and opportunistic woody species, may be less constrained by photoperiod and more responsive to temperature (Körner & Basler, 2010), resulting in rapid spread. Warmer regions of species origin and opportunistic life-histories are likely to coincide in the case of many alien plant species (Hulme, 2009) and such taxa might reveal a stronger link between flowering phenology and increases in distribution than natives. However, against these synergistic effects of flowering phenology, life-history and climate change, species distributions may be constrained by edaphic conditions less immediately influenced by variation in temperature such as soil fertility, moisture, pH or disturbance. Furthermore, these drivers and constraints are only likely to act where species are already at equilibrium with the current climate. The distribution of recently introduced alien plant species appears to reflect the length of time a species has had to expand its range since it was first introduced (Williamson et al., 2009) and potentially may not be at equilibrium with climate.
Clearly an assessment of a link between flowering phenology and species distributions is much needed as it will be crucial to gauge the potential impact of global warming on plants and address whether alien species pose a greater threat under climate change (Walther et al., 2002). To test for a link between flowering phenology and species distributions, the FFDs of 347 terrestrial plant species recorded from 1970 to 2000 in Chinnor, south-central England (Fitter & Fitter, 2002) were analysed in relation to their change in distribution in Britain over the same period. Variation in the number of days FFD occurred earlier per 1°C warming and the Change Index (CI), a measure of the relative change in species distribution between 1969 and 1999 (Preston et al., 2002a; Hill et al., 2004), was examined in relation to a set of plant attributes that might strengthen or weaken any link between flowering phenology and species distributions.
To assess whether alien species and/or opportunistic life-forms were more likely to exhibit a link between flowering response and distribution, analyses examined the importance of life-form (annual, herbaceous or woody perennial) and species status (native, archaeophyte (alien introduced before 1500 ad) or neophyte (alien introduced after 1500 ad)) for flowering response and distribution change. Within each status group, more detailed analyses examined whether species from warmer climes showed the most marked response in phenology and distribution change using a measure of a species’ current bioclimatic niche in the UK. These analyses also accounted for measures that reflected a species response to less temperature-dependent factors (important environmental gradients of light, soil moisture, pH and fertility) in constraining phenology or distribution change. Finally, to assess whether recent changes in distribution simply reflect progressive range expansion rather than a response to the environment, for neophytes the date of introduction to the British Isles was included in these models. The results indicate that, at least for native species, a link between FFD and the CI exists, with species exhibiting delayed flowering undergoing relative distribution declines. Yet earlier flowering did not guarantee range increase and no such relationship was found for neophytes.
- Top of page
- Materials and Methods
- Supporting Information
A flowering response was most evident for earlier flowering species in all status groups (Table 1, Fig. 1). Independently of flowering date, neophytes flowered on average mean ± SE = 2.20 ± 0.87 d earlier than natives per 1°C warming (by analysis of covariance (ANCOVA); F2,337 = 3.52, P = 0.031; Fig. 2a), with no significant difference between archaeophytes and natives. The magnitude of flowering response was also greater for annuals than for other life-forms (F2,337 = 8.98, P < 0.0001; Fig. 2b) and this was consistent for all status groups (two-way interaction; F4,337 = 2.25, P = 0.064). Significant MAMs (all P < 0.001) were found for natives (R2 = 0.424), archaeophytes (R2 = 0.163), and neophytes (R2 = 0.306), and there was good correspondence between variables in the MAM and those most commonly found in the ‘best’ subset models using AICc (Table 1). Mean first flowering date was the single variable with greatest explanatory power in the MAM for natives, archaeophytes and neophytes and in all the ‘best’ subset AICc models. For natives, flowering response was greater for shade-tolerant species (e.g. with low Ellenberg light scores), reflecting the response of the spring-flowering woodland flora, whereas short-statured archaeophytes of more fertile sites, mostly arable weeds, exhibited a greater advance (Table 1). Species from warmer climates were expected to respond more markedly to climate change. No bioclimatic variables were retained in the MAM but Tjan was included in the minimum AICc model for both natives and archaeophytes. Nevertheless, Tjan has less support than the other variables in the minimum AICc model as it was not consistently found in the AICc models identified in the ‘best’ subset. Furthermore, while native species encountering warmer winter temperatures exhibited the greatest advance, the opposite was true for archaeophytes.
Table 1. Parameters in the full models from separate general linear model (GLM) analyses for native (N = 271), archaeophyte (N = 42) and neophyte (N = 34) taxa summarizing the strength (beta) and statistical significance (P) of plant height, bioclimatic profile (Tjan and Tjul) and response to environmental gradients (Ellenberg scores for light, moisture, pH and fertility)
|R2|| ||0.440|| ||0.395|| ||0.557|
|F|| ||25.722|| ||2.040|| ||5.195|
|P|| ||0.000|| ||0.083|| ||0.000|
|(b) Change Index|
|R2|| ||0.090|| ||0.127|| ||0.247|
|F|| ||3.236|| ||0.456|| ||1.355|
|P|| ||0.002|| ||0.875|| ||0.252|
Figure 1. Negative relationship between flowering response (positive scores reflect earlier flowering) and the mean date of first flowering for native, archaeophyte and neophyte species. Neither the slopes nor intercepts derived from linear regressions differed significantly among status groups: native (closed circle, dashed line) y = −0.073x + 16.18, adjusted R2 = 0.39, P < 0.001; archaeophyte (open triangle, continuous line) y = −0.074x + 17.86, R2 = 0.29, P < 0.001; neophyte (open square, dotted line) y = −0.094x + 20.42, R2 = 0.28, P < 0.001.
Download figure to PowerPoint
Figure 2. Variation in flowering response (shaded bars) and the Change Index (open bars) for (a) species status and (b) life-form. The significance of the post hoc comparisons refer to Bonferroni-type simultaneous confidence intervals based on Student’s t distribution and bars labelled with the same letter are not statistically different from each other at P = 0.05.
Download figure to PowerPoint
Although the FFD of natives and archaeophytes occurred c. 6 d earlier for every 1°C rise in temperature, overall there was only limited evidence for any significant increase or decrease in the CI since 1970; in contrast, neophytes exhibited a relative increase in their distribution over the same period (F2,338 = 27.30, P < 0.0001; Fig. 2a). Differences were found among life-forms (F2,338 = 4.71, P = 0.010; Fig. 2b), where woody perennials revealed a relative increase in distribution, whereas the CI for annuals and herbaceous perennials was not significantly different from zero. This trend among life-forms was similar across all status groups (two-way interaction; F2,338 = 1.50, P = 0.203) but was the opposite to that found for flowering response.
There was only partial evidence that changes in flowering phenology fed back to changes in species distribution. For archaeophytes, soil fertility was the only variable retained in the MAM (R2 = 0.130, P = 0.019), and was included in all ‘best’ AICc subset models (Table 1). A further variable, Tjul, was also included in most of the ‘best’ subset AICc models, implying that species of fertile and warm environments exhibited the greatest increase in CI. The CI for neophytes was not significantly related to flowering response or any other environmental covariate. However, there was a positive relationship between the date of naturalization in the wild and relative change in distribution since 1970 (Fig. 3a, full model beta = 0.483, P = 0.004). Only date of naturalization was retained in the MAM (R2 = 0.210, P = 0.007) and was the only variable consistently found in all the ‘best’ subset AICc models.
Figure 3. Positive relationships between the Change Index and (a) date of first naturalization in the wild for neophyte species (y = 0.007x − 11.329, adjusted R2 = 0.213, P = 0.004); (b) flowering response for native species (y = 0.021x – 0.159, adjusted R2 = 0.023, P = 0.007).
Download figure to PowerPoint
In contrast to the alien taxa, flowering response was related to the relative changes in distribution of natives. Native species that showed the greatest flowering response, persisted in shaded habitats and benefited from fertile soils exhibited increases in relative distribution (Table 1, Fig. 3b). These three variables were found in the MAM (R2 = 0.061, P < 0.001) and all equivalent ‘best’ AICc subsets. Two other variables, height and pH, were included in all bar one of the ‘best’ subset models but not the MAM. Taller species and those of more acid soils exhibited greater increases in CI. There was no support for the view that species from warmer regions of the UK might exhibit the greatest relative change in distribution, with neither bioclimatic variable included in the ‘best’ subset AICc models or the MAM. Of those native species where FFD did not initiate earlier with warming, all but one declined in distribution, whereas species whose distributions showed a relative increase had a greater response. However, an earlier flowering response was not by itself a guarantee of a relative increase in distribution. Almost two-thirds of native species exhibiting earlier flowering in response to warming also showed relative declines in distribution, probably in response to other environmental pressures related to variation in light and soil fertility (Table 1).
Could these national scale results translate to patterns at the local scale? There are no data on the local abundance of species in the vicinity of Chinnor but it might be assumed that species increasing in the local area may be more likely to be recorded in any one year. Thus a proxy for abundance may be the frequency with which a species was recorded in the phenological time series. At this local scale, a positive relationship did exist between flowering response and the frequency with which a species was recorded, but again this was only for native species (beta = 0.301, P < 0.001). Native species whose FFD failed to respond significantly to warming tended to be under-recorded in the Chinnor data set (Fig. 4). As at the national scale, no relationship was found for either archaeophytes (beta = −0.077, P = 0.630) or neophytes (beta = 0.287, P = 0.099). These results are consistent with the interpretation that changes in flowering phenology subsequently impact upon local abundance.
Figure 4. Positive relationship between the frequency with which phenological records were obtained for species between 1970 and 2000 in the vicinity of Chinnor and the flowering response over the same period (y = 0.296x + 22.451, adjusted R2 = 0.087, P < 0.001).
Download figure to PowerPoint
- Top of page
- Materials and Methods
- Supporting Information
The primary causes of change in the distribution of the British flora over the last 30 yr have been land-use transformation (Preston et al., 2002a,b) and eutrophication (Smart et al., 2005) rather than climate change (Haines-Young et al., 2003; Hulme, 2009). The results of the present study suggest that, while many species have shown relative declines in response to these pressures, there is still a detectable climate signal. The response in FFD explained only a small proportion of the variation in native CI and thus this result should not be over-interpreted. The explanatory power of the analysis is limited by changes in FFD being only one component of plant phenology that influences plant demography (Morin et al., 2007). Even under experimental warming, plants respond in idiosyncratic and unpredictable ways as a result of different combinations of changes in organ size and physiological rates (Lambrecht et al., 2007). In addition, while flower phenology at a single site may parallel patterns at a national scale, such correlations are not perfect (Sparks et al., 2000; Menzel et al., 2001), although phenological models fitted locally can predict regional phenology (Chuine et al., 2000). Finally, additional environmental drivers impose further constraints on distribution change (Hulme, 2009). Under such circumstances the explanatory power of the responsiveness of FFD to warming may be of less interest than how its effect ranks against other explanatory variables (Freckleton, 2009). Thus the significance of the relationship may be better reflected by the variation explained by flowering response being similar in magnitude to the light and soil fertility covariates, which are likely indicators of the response of the British flora to land-use change (Preston et al., 2002b) and eutrophication (Smart et al., 2005).
Native plants whose phenology did not respond significantly to climate warming revealed a relative decline in distribution (e.g. Stachys officinalis and Scabiosa columbaria). Later flowering may impair plant performance and reproductive output (Suttle et al., 2007), especially if plant species that respond to temperature change may better maintain interactions with pollinators (Willis et al., 2008; Walther et al., 2009). Later flowering may also result in a shorter growing season and potentially lower productivity either in absolute terms (Steltzer & Post, 2009) or relative to competitors that flower earlier. A combination of a shorter growing season, decreased productivity, reduced competitive ability, and lower fecundity could result in reduced local abundance and distribution range as well as making species less able to deal with other environmental pressures such as disturbance, habitat fragmentation and eutrophication. Partial support for such a hypothesis is found in observed changes in local abundance at Chinnor that appeared to mirror trends at the national scale.
However, it is important to distinguish cause and effect because declines in population size may result in sampling later FFD, irrespective of any role of climate (Miller-Rushing et al., 2008). Are these changes in local abundance a result of differences in flowering phenology translating into changes in plant demography or simply a sampling artefact arising from a higher likelihood of picking up earlier flowering dates in larger populations? While changes in plant population size may have occurred in Chinnor, this seems an unlikely determinant of the changes in FFD. First, if the relationship is a sampling artefact it would not generate a strong relationship between first flowering date and the magnitude of the response to warming, as this would assume that the magnitude of the sampling artefact was strongly correlated with the mean date of flowering. In contrast, there is a sound physiological basis to expect spring-flowering species to respond more strongly to temperature changes (Lapointe, 2001). Secondly, a sampling artefact should be observed for all species irrespective of status, rather than only natives. If the magnitude of advance is an artefact of changes in abundance, then a strong relationship between the two variables should have been observed for neophytes as these species exhibited the most marked increase in both variables. Thirdly, the phenological records reveal high interannual variation in FFD (mean CVs: native, 9.36; archaeophyte, 13.09; neophyte, 13.61) that is more consistent with changes in flowering responses to annual temperature differences than dramatic changes in plant population size. Thus, while it is important to recognize the potential effect of a sampling artefact, the evidence above suggests that such an artefact may play only a minor role in the patterns observed. This interpretation is consistent with the conclusions drawn regarding the role of delayed phenological response and the declining local abundance of woodland species in Massachusetts (Willis et al., 2008).
Even improved plant performance as a result of earlier flowering may be no guarantee of population persistence in the face of these environmental pressures. Archaeophytes of fertile soils, mainly arable weeds (e.g. Viola arvensis and Lamium album), exhibited earlier flowering but they have also shown the strongest relative declines in their distributions in the British Isles as a result of the intensification of agriculture (Preston et al., 2002b). Similarly, the FFD of native species typical of shaded habitats (with Ellenberg light scores < 5; e.g. Mercurialis perennis and Viola hirta) showed the strongest response to warming, yet native species best suited to open conditions (with Ellenberg light scores > 5; e.g. Plantago lanceolata and Rumex acetosa), such as might result from disturbance, showed greater relative increases in distribution (Table 1).
Given their marked earlier flowering and increase in distribution, the absence of a significant association between flowering response and CI for neophytes is in stark contrast to native species. However, the populations and hence ranges of invasive alien species may be expected to increase irrespective of changes in climate and most models of their spread have not required a specific climate driver (Hastings et al., 2005). Under these circumstances, change would be greatest for more recently introduced species which are potentially furthest away from reaching any limits imposed by climate. For example, the distribution of Acer pseudoplatanus introduced in 1635 has shown little relative change in the UK between 1969 and 1999, whereas the relative distribution of Cerastium tomentosum, first recorded in the wild in 1915, has increased fivefold over the same period. This study has shown that time since naturalization, even after > 100 yr, appears to have a strong effect on the relative change in neophyte distributions since the 1970s, more so than recent environmental pressures such as climate-induced early flowering or eutrophication. This suggests that many alien species distributions have yet to reach equilibrium with the environment (Williamson et al., 2009). Nevertheless, the higher proportion of alien species that exhibited earlier flowering in response to warming compared with native species indicates that these taxa may be better adapted to higher temperatures and is consistent with their bioclimate profiles being significantly warmer and drier than those of natives (Hulme, 2009).
This study highlights that changes in flowering phenology are not only a sign of climate change (Peñuelas & Filella, 2001; Root et al., 2005; Peñuelas et al., 2009; Steltzer & Post, 2009; Walther et al., 2009) but, at least for native species, could be viewed as an indicator of the potential impacts of global warming on plant species distributions. While species from warmer climes (at least natives) and opportunistic species (e.g. annuals) exhibited a more marked flowering response, this did not appear to be directly translated in greater relative distribution change. It appears that environmental constraints, particularly relating to soil fertility and pH, mediate the magnitude of distribution change for species whose ranges are most likely to be in equilibrium with current climate. For species whose ranges are not in equilibrium with current climate (e.g. neophytes), there is no evidence of a relationship between flowering response and distribution change.
Three important caveats emerge from this study. First, most emphasis in phenological studies has been placed on identifying species showing a positive response to climate change rather than those that have shown delayed or no change in phenology (Parmesan & Yohe, 2003; Root et al., 2005; Parmesan, 2006; Rosenzweig et al., 2008). Yet it is the species that fail to track climate change that are of particular concern as they may decline as a result of reduced productivity, shorter growing seasons and/or phenological mismatch (Parmesan, 2006; Suttle et al., 2007; Willis et al., 2008; Steltzer & Post, 2009; Walther et al., 2009). Over 20% of species across Europe may be failing to track climate change (Menzel et al., 2006). Secondly, many phenological networks utilize alien ornamental species as a basis for their observations (Sparks et al., 2000; Menzel et al., 2006; Cleland et al., 2007; Morisette et al., 2009) and, while their responsiveness to temperature makes them suitable indicators, aliens may not be representative of phenological changes in native communities, particularly given their tendency for earlier FFD (Miller-Rushing & Primack, 2008; Peñuelas et al., 2009). Thirdly, current distributions of alien species and also certain native species (Davis & Shaw, 2001; Svenning & Skov, 2004) may reflect dispersal limitation rather than climatic limits and may exhibit only limited tracking of future climate. Thus, while a link between phenological response and distribution change may exist, this study warns against the uncritical extrapolation of such data to predictions of future changes in species distributions (Walther et al., 2009).