1. The reliability of species distribution models (SDM) to predict the probable response of alien plants to climate change rests on the assumption that plant performance in relation to temperature in the introduced range will be similar to that observed in the native range. Yet, alien plants may exhibit enhanced performance or different environment-distribution relationships following their introduction into a new area. Empirical data are therefore essential to test whether the responsiveness of species to climate is equivalent in the native and introduced ranges.
2. This study tests the assumption that phenological responses of plants to temperature are similar in both their native and introduced range. First flowering date (FFD) is widely used to assess the responsiveness of plants to recent warming arising from global change. For 19 species native to Europe, FFD observed in both the UK and USA between 1970 and 2000 was examined in relation to interannual variation in local temperatures. General trends, variability and responsiveness of FFD to warming were examined for consistency in the contrasting climate of Oxfordshire and Washington DC.
3. Mean FFD in Oxfordshire was a powerful predictor of the same variable in Washington DC, although summer flowering plants in Oxfordshire tended to flower earlier in the season in Washington DC. FFD varied considerably over 30 years, but across all species, the range in FFD revealed a similar trend in both regions with larger ranges observed for earlier flowering species.
4. Comparable trends were found between Oxfordshire and Washington DC in the degree to which flowering advanced or regressed per unit temperature increase. In response to warming, the majority of species flowered earlier in both countries and the degree to which FFD responded to increasing temperature was greatest for species flowering earlier in the year.
5. These equivalent phenological responses to temperature across continents imply prediction of the performance of alien plants under climate change may be derived from a species’ behaviour in its native range. While these findings support the use of SDM, they also indicate that these models could be significantly improved through the integration of phenological relationships parameterized from data in the native range.
Species distribution models (SDM), including bioclimatic envelope and ecological niche models, represent the most frequently applied tools for the prediction of future risks posed by alien species under climate change (Jeschke & Strayer 2008). The use of environment-distribution relationships in the native range to project future distributions in a new region assumes species respond similarly to climate in both the native and introduced range (Wiens et al. 2009). Europe represents the most frequent source of alien plants established in North America (Fridley 2008), and SDM predict that the North American ranges of many introduced European plants will increase significantly under future global change (Peterson et al. 2008). Yet, many European natives have escaped their natural enemies following introduction into North America (Blumenthal et al. 2009) with the result that several exhibit enhanced performance in the introduced range (Bossdorf et al. 2005; Blumenthal & Hufbauer 2007). Furthermore, correlative evidence suggests European plants may exhibit different environment-distribution relationships following their introduction into North America (Broennimann et al. 2007). These observations challenge the underlying SDM assumption that alien plant performance in relation to climate is likely to be similar in the introduced range and hence, question the robustness of future range projections using such models. Nevertheless, such contrary evidence is indirect and observed anomalies in range projections could arise from violation of many other assumptions inherent in SDM (Wiens et al. 2009). Thus, empirical data are essential to test for differences in species responsiveness to climate in their native and introduced ranges. While similar comparisons have been undertaken on other aspect of alien species performance (Van Kleunen et al. 2010; Buswell, Moles & Hartley 2011), none have examined whether responses to temperature might differ.
First flowering date (FFD) is widely used to assess the responsiveness of plants to recent warming arising from global change (Cleland et al. 2007) and is particularly suitable for assessing species responses to climate in their native and alien ranges. Earlier flowering may facilitate increases in local plant abundance as well as geographical range expansion because it is believed that by governing reproductive success, growth and survivorship, phenology will also ultimately determine the probability of species occurrence under particular climatic conditions (Chuine 2010). This view is supported by the finding that longer (and earlier) flowering seasons are important correlates of alien plant establishment (Lloret et al. 2005; Cadotte, Murray & Lovett-Doust 2006) which implies a strong link between plant phenology and invasion. In addition, the flowering phenology of alien plants is known to exhibit rapid adaptive responses to novel environments following introduction (Dlugosch & Parker 2008; Montague, Barrett & Eckert 2008; Keller et al. 2009). Thus, there is significant scope for flowering phenology to adjust to the new environments experienced during plant invasions which could subsequently facilitate improved performance and further spread. Natural enemies may impose strong selection pressures on the timing and duration of flowering, and their absence or scarcity in the introduced range may impact upon plant phenology (Elzinga et al. 2007). Consistent with these assumptions is the finding that where FFD has been compared between native and alien species, results indicate greater responsiveness of the latter taxa to warming (Willis et al. 2010; Hulme 2011).
Comparison of species performance in their native and introduced range are fraught with difficulty as it is often impossible to identify a precise reference population in the native range or account for climatic differences among sites (Colautti, Maron & Barrett 2009; Moloney et al. 2009). Under such constraints, multispecies comparisons provide a more robust means to assess the differences between native and introduced ranges where the focus is on the uniformity of responses across all taxa rather than the absolute differences exhibited by an individual species (e.g. Buswell, Moles & Hartley 2011). To test empirically whether introduced plants exhibit similar responses to climate in their native and introduced ranges, the predictability of FFD and its responsiveness to temperature increase was assessed between North America and Europe. If the underlying assumptions of SDM are robust then, across a range of different species, both the FFD and its responsiveness to temperature in the introduced range should be predictable from observations in the native range. In contrast, if these assumptions of SDM are not supported, species phenology would be expected to behave idiosyncratically in the introduced range reflecting species-specific selection pressures (e.g. different degrees of enemy release and genetic bottlenecks). As a result, phenological patterns across species would not be predictable from observations in the native range (Broennimann et al. 2007; Gallagher et al. 2010; Buswell, Moles & Hartley 2011).
In what is probably a unique opportunity to test these assumptions, FFD observed annually between 1970 and 2000 in the environs of Washington DC (USA) and Oxfordshire (UK) was used to assess whether the flowering responses to temperature of 19 European species introduced into North America were predictable based on observation in their native Europe. Such an approach is equivalent to treating FFD as a measure of plant performance and, similar to many other studies, asking whether performance differs between the native and introduced range (Maron, Elmendorf & Vila 2007; Brodersen, Lavergne & Molofsky 2008; Ebeling, Hensen & Auge 2008; Williams 2009; Herrera, Carruthers & Mills 2011). However, given that even within the native range, FFD will vary in relation to latitude and can often be explained using data on winter temperatures and photoperiod (White 1995; Aasa et al. 2004) differences between Washington DC (USA) and Oxfordshire (UK) are to be expected. Thus, this study aims not to determine whether differences between the introduced and native ranges exist but rather ask whether the flowering responses to temperature in the introduced range are predictable from observations in the native range. Although only one region is compared in each of the introduced and native ranges, the design has the advantage that (i) most of the species examined in this study (with the exception of Melilotus officinalis) were likely introduced into north-eastern USA from England (Mack 2003), (ii) the regions represent two distinct Köppen climates (Peel, Finlayson & McMahon 2007) and daylength regimes (Fig. 1a,b) that should maximize the opportunity for differences in plant phenology, (iii) data are drawn from long-term records across multiple populations of 19 species in each region and (iv) both regions have experienced warming trends since 1970 (Fig. 1c).
Materials and methods
Published FFD observations in the environs of Washington DC (Abu-Asab et al. 2001) and Oxfordshire (Fitter & Fitter 2002) were used to assess whether species common to both datasets exhibit similar or idiosyncratic flowering response to temperature in the two regions. The approach to the surveys differed in the two regions, although in both cases FFD refers to the date at which a mono- or diclinous flower began anthesis or was receptive to pollen. In Washington DC, surveys on 100 taxa were undertaken within an c. 60 km radius from the centre of Washington DC (76°24′0W 38°24′0N) with many different observers (>125) involved in the recording, such that FFD is the earliest observed date of flowering and not necessarily the absolute earliest date on which a species flowered (Abu-Asab et al. 2001). In contrast, in Oxfordshire, FFD for 347 taxa were recorded by a single observer on a more or less daily basis within a few kilometres of Chinnor, Oxfordshire (0°42′9W 51°42′9N) and is much more likely to be the earliest date. Although Fitter & Fitter (2002) provide FFD from 1954 to 2000, data from Abu-Asab et al. (2001) only cover the period between 1970 and 1999. Additional data for the year 2000 in Washington DC were therefore extracted from the Smithsonian spring flowering database (http://botany.si.edu/DCFlora/springflowers) to maximize the comparison between localities over the period 1970–2000. A total of 19 European native species are shared between the two datasets and recorded on at least half of the years between 1970 and 2000 (Table 1). In both regions, multiple populations of each species would have been visited each year though the precise number is unknown. However, the number of years for which FFD was recorded for a species was significantly higher in Oxfordshire than Washington DC (paired t-test: 25·53 ± 0·96 vs. 23·53 ± 0·56 years, t = 2·10, d.f. 18, P = 0·045). While the alien status of two species in North America is unclear, potentially comprising both native and alien populations (Achillea millefolium, Galium aparine), the remainder are unambiguously alien in the latter region. One European species is alien in Washington and Oxfordshire (M. officinalis). Annual temperature data were drawn from College Park, Maryland (from the Maryland State climatologist) and the Central England Temperature Record (UK Meteorological Office 2006). Mean minimum monthly temperatures for the period December to May were calculated for each year. Minimum temperatures were chosen as they better estimate the risk of frost damage to flower buds as this is a major constraint to earlier flowering (Inouye 2008). While FFD ranged from as early as February to as late as July across the 19 species, temperature cues for winter/spring flowering species occur 1–2 months prior to flowering, while for summer flowering species cues may occur as much as 4 months earlier (Fitter et al. 1995). The mean minimum temperature across the months of December–May was therefore used as it covers the period of these environmental cues and has been shown to be a good predictor of flowering response in both Washington DC (Abu-Asab et al. 2001) and Oxfordshire (Hulme 2011). Ordinary least-squared linear regressions between FFD and temperature were undertaken for each species, and the slope used to summarize species responsiveness to warming in terms of the number of days flowering either advanced or regressed per 1 °C increase in temperature. Tests for equivalence among regions in mean FFD and flowering response were undertaken using reduced major axis regression to account for the error associated with each mean value (Bohonak 2004). All other analyses were undertaken in spss (Norusis 2006).
Table 1. Species examined with details of their life-form, photoperiod, alien and weed status in North America, European biome and climate region
The climate in Washington DC experiences significantly lower mean minimum temperatures (1·93 ± 0·17 °C) than Oxfordshire (3·14 ± 0·14 °C) between December and May (one-way anovaF1,59 = 30·97, P < 0·001; Fig 1a). Both regions however, reveal a similar trend between 1970 and 2000 of increasing temperatures (Fig. 1c) which was significant in Washington DC (R2 =0·14, F1,29 = 4·75, P = 0·037) but not Oxfordshire (R2 =0·11, F1,29 = 3·52, P = 0·071). However, neither the slopes (OX 0·03 ± 0·02 vs. WA 0·04 ± 0·02, t = 0·43, d.f. 58, P = 0·671) nor intercepts (OX −52·79 ± 29·82 vs. WA −73·54 ± 34·62, t = 0·45, d.f. 58, P = 0·652) of this relationship differed significantly between the two regions.
Mean FFD in Oxfordshire was a powerful predictor of the same variable in Washington, explaining over 90% of variation (R2 =0·92, F1,17 = 192·73, P < 0·001, Fig. 2). Both the slope (0·67 ± 0·05) and intercept (25·49 ± 5·91) of the relationship were significantly different from 1 (t = 6·60, d.f. 34, P < 0·001) and 0 (t = 4·31, d.f. 34, P < 0·001), respectively. Plants that flowered in summer in Oxfordshire tended to flower earlier in spring in Washington DC. As a consequence, over half of all species in Washington first flowered in April, compared with only four species in Oxfordshire. For any one species, FFD varied considerably over 30 years but the range in observed FFD was significantly correlated between the two regions (r = 0·685, d.f. 17, P = 0·001) and revealed a similar trend of larger ranges for earlier flowering species (OX: R2 = 0·40, F1,17 = 12·76, P = 0·002; WA: R2 = 0·32, F1,17 = 9·59, P < 0·006; Fig. 3a). The range appeared to be a species attribute as neither slopes (OX −0·40 ± 0·11 vs. WA −0·51 ± 0·17, t = 0·58, d.f. 34, P = 0·567) nor intercepts (OX 101·83 ± 14·32 vs. WA 96·99 ± 18·34, t = 0·21, d.f. 34, P = 0·836) of its relationship with FFD differed significantly between the two regions.
Similarly to the range, the degree to which FFD responded to increasing temperature was also a significant function of mean FFD and was greatest for species flowering earlier in the year (Fig. 3b). The best fit lines (OX: R2 = 0·75, F1,17 = 50·25, P < 0·001; WA: R2 = 0·34, F1,17 = 8·83, P = 0·009) revealed neither slopes (OX −0·15 ± 0·02 vs. WA −0·20 ± 0·04, t = 1·16, d.f. 34, P = 0·255) nor intercepts (OX 25·63 ± 2·31 vs. WA 26·17 ± 4·29, t = 0·11, d.f. 34, P = 0·913) to differ significantly between the two regions. As a result, a significant positive association (R2 = 0·58, F1,17 = 23·26, P < 0·001, with slope = 0·89 ± 0·14 and intercept = −1·73 ± 1·28, Fig. 4) between both regions existed in the degree to which flowering responded to increasing temperature. This relationship did not differ from a best fit line with a slope of unity (t-test t = 0·78, d.f. 17, P = 0·443) and intercept of 0 (t = 1·35, d.f. 17, P = 0·194). Approximately 58% of the variation in flowering response to temperature in one region could be explained by observations in the other and reflects the strong and similar relationship in mean FFD between the two regions. The lower amount of variation explained for flowering response compared to mean FFD undoubtedly reflects sampling errors attributable to using a single temperature value across the relative large study regions (especially Washington DC) over which warming is likely to be quite heterogeneous.
Flower initiation is controlled by a diversity of intrinsic and extrinsic mechanisms including developmental clocks, vernalization, photoperiod and thermal time (Kim et al. 2009; Korner & Basler, 2010; Forrest & Miller-Rushing 2010). The finding that the flowering phenology of the majority of plants examined in this study was sensitive to minimum temperatures between December and May lends further support for the role of thermal time in flower initiation (Cleland et al. 2007). However, species generally flowered earlier in Washington DC than Oxfordshire, even though temperatures between December and May were much colder. Photoperiod did not appear to explain this intercontinental difference in mean FFD as in April over twice as many species flowered in Washington DC than Oxfordshire irrespective of the shorter daylength. On the other hand, vernalization is likely to be important for winter annuals/biennials or long-day perennials that comprise the majority of the species examined in this study (Table 1). Because the thermal time required for flowering generally declines as the vernalization period lengthens (Kim et al. 2009), the earlier mean FFD in Washington DC is consistent with plants experiencing colder winters than in Oxfordshire.
While shifts in phenology may be the result of local adaptation (Dlugosch & Parker 2008; Montague, Barrett & Eckert 2008), such shifts could also arise through phenotypic plasticity. The latter scenario is supported by this study where even in a single region FFD can vary considerably in response to annual variations in temperature. Over the 30-year period, FFD for an individual species varied by between 20 and 100 days indicating considerable plasticity in response to interannual temperature variation. There was no indication that species in Washington DC were any less plastic in their response to temperature, as might be expected if alien species experienced reduced genetic variation following introduction (Doi, Takahashi & Katano 2010). This may reflect that the species examined in this study are weeds that have mostly been accidentally transported across continents multiple times potentially limiting any genetic bottlenecks (Mack 2003). The differences in the absolute values for mean FFD or flowering response, while statistically significant could easily have been different had other regions in each continent been chosen for comparison. All study species are widely distributed in North America and Europe (Table 1), and FFD certainly varies across the wide latitudes and longitudes encompassed in the native and introduced ranges (White 1995; Sparks, Jeffree & Jeffree 2000). For this reason, it is the similarity in the relative performance across a range of species that is a better indicator of equivalent phenological responses in native and introduced ranges.
The equivalent phenological responses of species in both Europe and North America are strongly indicative that the mechanisms of flower initiation are conserved across large geographical distances. There was no indication that the origin of a species influenced phenological response and neither those species doubtfully alien in North America (A. millefolium and G. aparine) or alien in both regions (M. officinalis) were outliers in the overall trends. Similarly, while the performance of Alliaria petiolata and Barbarea vulgaris has been observed to differ between European and North American populations (Bossdorf et al. 2005), their phenological patterns in response to temperature were consistent with the trend observed across all species in this study. Support for equivalent biogeographic patterns in flowering phenology is found in the parallel latitudinal clines in phenology across both continents that exist for some of the study species, e.g. Tussilago farfara (Aasa et al. 2004; Lavoie & Lachance 2006) indicating that FFD will be analogous where environments are similar. Furthermore, where the flowering phenology of European species introduced to North America has been shown to adapt rapidly to local environmental conditions, this is generally in a manner consistent with patterns observed in the native range (Dlugosch & Parker 2008; Montague, Barrett & Eckert 2008; Keller et al. 2009).
Although this study examined the flowering phenology of many species that comprised multiple populations observed across different landscapes, the study regions reflect only part of the native and introduced range. Thus, criticism may be levelled as to the generality of the findings. However, the equivalent long-term patterns seen in FFD, its range and response to temperature across many species observed in two climatically distinct regions using two different sampling approaches are indicative of reasonably robust relationships. In addition, FFD may also be sensitive to variation in plant population size (Miller-Rushing, Inouye & Primack 2008) that, while likely to have occurred over the 30-year observation period, would not be expected to be correlated between the two regions. Given these considerable sources of noise in the two data sets that might have masked any relationship, the comparability of intercontinental patterns in flowering phenology is all the more impressive. Moreover, the patterns observed in Oxfordshire and Washington DC are consistent with those seen across wider regions in each continent (White 1995; Sparks, Jeffree & Jeffree 2000). Alternative approaches that might involve reciprocal garden experiments would need considerable replication in each continent and be required to run for at least a decade to provide similar insights into the phenology–temperature relationships (Colautti, Maron & Barrett 2009; Moloney et al. 2009). Thus, the data presented here represent the best available information on whether changes in flowering phenology may occur following species introduction to new regions.
A key insight from this work is that the future risks posed by alien plants in North America under climate change may be predictable from an understanding of their phenology in the native range. Increasing evidence points to the utility of phenology-based models in predicting the fate of species under climate-change scenarios but as yet such tools have not been applied to forecast range changes of alien species (Chuine 2010). Earlier flowering should allow plant distributions to increase and could facilitate invasion if alien plants exhibit a more marked phenological response than natives (Willis et al. 2010; Hulme 2011). While the intercontinental comparisons suggest aliens do not exhibit improved phenological performance following introduction, if alien species represent a biased subset of species that flower earlier or originate from warmer climates this may pre-adapt taxa to respond more rapidly to climate warming. Such biases may be common in species selected as ornamentals (Korner & Basler 2010) and these represent a major component of alien floras world-wide (Hulme et al. 2008). The results presented here suggest using such phenological insights into species responses to temperature in their native range to address risks of plant invasion under climate change elsewhere may represent a significant improvement upon current attempts based solely on bioclimate matching (Broennimann et al. 2007; Gallagher et al. 2010).
I thank Richard Duncan for providing constructive comments on the manuscript and Katie Collins, Assistant Maryland State Climatologist, for climate data.