Introduction

Phenology is one of the most sensitive responses of the natural world to a changing climate. Of all the evidence considered by the IPCC in its most recent report, the bulk of evidence of changes to the natural world concerned phenological change (Rosenzweig et al. 2008). Indeed, phenological change may be seen as a vanguard for wider change in the environment (Cleland et al. 2007).

Plant phenology has been shown to be very responsive to temperatures, and a number of multi-species studies have reported shifting phenology, particularly to earlier leafing and flowering in spring (Bradley et al. 1999; Menzel and Fabian 1999; Abu-Asab et al. 2001; Fitter and Fitter 2002; Peñuelas et al. 2002; Menzel et al. 2006). However, many of these papers covered data only up to the end of the twentieth century and had a focus on spring events. Reports on autumn phenology are much less common (but see Menzel and Fabian 1999), and there have been few multi-species papers covering the first few years of the twenty-first century which, because of their exceptional warmth, would be expected to be associated with exceptionally early spring phenology (e.g. White et al. 2009).

Studies that report many species have distinct advantages. They allow a comparison between species without the confounding environmental differences associated with different studies. Studies of records from a limited geographical area have further benefits in that the genetic diversity of the recorded material is likely to be smaller and the studied environment likely more homogenous (e.g. Hepper 2003). In this respect, botanical gardens may be particularly valuable in recording phenology. They are often long-established with their own meteorological station and continuity of personnel. Careful observation of various phases of the same species may well be possible without the need to search the environment to locate a particular species.

There have been few studies that have looked at several phases of the same species and the relationships between successive phases. Furthermore, there have been few multi-species studies from Eastern Europe (see maps in: Rosenzweig et al. 2008).

In this paper, we examine a 31-year record (1977–2007) of 66 phenophases from the Poznań Botanical Garden (Poland). The studied species include a number of iconic and more obscure ones. Of the former, Horse Chestnut Aesculus hippocastanum has been widely planted in Europe, has very obvious phenological phases, and has been widely reported in the phenological literature, for example the two-century record of first leafing from Geneva, Switzerland (Defila and Clot 2001). The purpose of our paper is to (1) identify trends in plant phenology from early spring to late autumn, (2) estimate the responsiveness of species to mean monthly air temperatures, (3) investigate the interphase intervals of the same species, and (4) look at the consistency of changes at the beginning and end of the growing season.

Materials and methods

A large number of plant phenophases have been monitored at the Adam Mickiewicz University Botanical Garden in Poznań, Poland (www.ogrod.edu.pl/info_eng.php) (52º25′N 16º53′E). The Botanical Garden was founded in 1925, occupies an area of 22 ha at an altitude of 89 m asl and contains nearly 7,000 Special Collections. For this paper, we abstracted data on the dates of 66 phenophases for the period 1977–2007 incorporating 42 species (see Table 1 for list). All observations were made within the Garden and on a daily basis. The definitions of the phenophases used are as follows:First shoot – first shoots appearing above the ground; First leaf – first fully open leaf; First flower – first open flower; Pollen – first pollen shed; End of flowering – last flower; Earing – inflorescence of cereals emerges; Seeding – first ripe seeds produced; First senescence – first evidence of above ground portion of plants dying; Leaf colouring – first colour change; Die back – above ground portions of plant fully dead; Bare – all leaves fallen.

Table 1 A summary of the examined phenophases and the regressions of phenophases on year and on mean temperature
Full size table

All dates were converted prior to analysis into days after 31 December, hereafter day of the year (DOY) where 1 =  1 January,etc. Eighteen intervals between successive phenophases of the same species were calculated where considered biologically meaningful (see Table 2 for list).

Table 2 Trends in 18 phase intervals, the correlation between the two phases and the significance of the earlier phases in a regression model after fitting 3-month temperature
Full size table

Mean monthly air temperatures, collected to standard WMO guidelines, were obtained from the meteorological station situated within the Botanical Garden.

Trends through time were estimated using linear regression of phenophases on year. Temperature responses were estimated by regression of phenophases on the mean temperature for the three calendar months ending in the month in which the mean of the phenophase occurred; thus, for example, an event whose mean date was in May would be compared to the mean temperature from March to May. This is a rather broadbrush approach but has been shown to be usually sufficient, particularly for spring events (Estrella et al. 2007).

The 18 phenophase intervals were subjected to regression on year to check for trends over time. A correlation was calculated between the two phases from which the interval was derived. Finally the end phases were regressed on the first phases after fitting the 3-month mean temperature mentioned above. This was in order to see if the first phases influenced the later ones after temperature effects were removed.

To assess variability changes within early spring events we calculated the standard deviation annually among all six phenophases occurring, on average, before the end of March. These are the first six events in Table 1. A similar exercise to look at autumn variability was based on the standard deviation of the eight “bare” phenophases listed towards the end of Table 1. Trends in these variability measures were assessed by correlation with year.

Results

Trends through time

Table 1 summarises the examined phenophases, their trends through time and their response to the mean temperature of the three calendar months leading up to and including the mean date of that phase. Significant changes in timing were detected in 22 of the 66 phenophases; 14 significant advances and 8 significant delays. There was a strong association between timing of the phase and the trend through time (r 64  = 0.624, P <0.001; Fig. 1) with spring events tending to get earlier and autumn events later.

Fig. 1
figure 1

Trends through time (days/year) for 66 phenophases recorded at the Poznań Botanical Garden in the period 1977-2007 plotted against the mean date (day of the year) of the phenophase. A dotted reference line has been added; phases below the line got earlier, above the line later

Full size image

Response to temperature

In comparison with 3-monthly mean temperatures, 44 of the 66 phenophases showed a significant response to temperature (Table 1). Of these, 39 indicated earlier events with warmer temperatures. All events with mean dates before DOY 193 (July 12) had a significant negative relationship (warmer = earlier) with temperature. The five significantly positive relationships all occurred in events with mean dates post DOY 274 (October 1). Overall, a strong correlation between temperature response and mean date was also apparent (r 64  = 0.825, P <0.001; Fig. 2).

Fig. 2
figure 2

Temperature responses (days/°C) for 66 phenophases recorded at the Poznań Botanical Garden in the period 1977-2007 plotted against the mean date (day of the year) of the phenophase. A dotted reference line has been added; phases below the line getting earlier with warmer temperatures, above the line later

Full size image

A highly significant correlation existed between the regression estimates of phenophase on year and the regression estimates of phenophase on temperature, i.e. columns 7 and 11 in Table 1 (r 64  = 0.741, P <0.001).

Trends and temperature responses in interphase intervals

Table 2 lists the 18 considered intervals. Seven of these had changed significantly during the study period; all of them getting longer. Fifteen of the 18 intervals had positive trends through time suggesting extended phase intervals for the majority of species, some of which can be interpreted loosely as the length of the growing season. For 7 of these intervals significant positive correlations existed between the two phases used to derive the interval (Table 2). However, the correlations were not typically large suggesting that the intervals are rarely predetermined, but rather influenced by annual climatic conditions. In fact only 3 of the earlier phases were significant in modelling the later phase once temperature effects had been accounted for (Table 2).

Inter-species variability in spring and autumn

The standard deviation between the six early phenophases plotted against year is shown in Fig. 3. This has significantly increased over time (r 29  = 0.520, P = 0.003) and is greater in early springs (correlation with the mean date of the six phenophases r 29  = –0.502, P = 0.004). The standard deviation between the eight “bare” phenophases plotted against year is shown in Fig. 4. This has also significantly increased over time (r 29  = 0.553, P = 0.001) and is greater in late autumns (correlation with the mean date of the eight phenophases r 29  = 0.720, P < 0.001).

Fig. 3
figure 3

The standard deviation between the dates of six early phenophases recorded at the Poznań Botanical Garden for the period 1977–2007

Full size image
Fig. 4
figure 4

The standard deviation between the dates of eight “bare” phenophases recorded at the Poznań Botanical Garden for the period 1977–2007

Full size image

Discussion

Botanical gardens can offer many advantages in studies of phenology and climate impacts (Donaldson 2009; Primack and Miller-Rushing 2009). They are typically long established, with professional staff and good archives. There is typically a stability in both staffing and methods that results in continuity of recording protocols, and a longevity that can rarely be achieved when records are made by individuals (but see Fitter and Fitter 2002). Phenological recording may often involve the same specimen in a relatively small area and thus eliminate some of the noise associated with phenological records made in the wild over large areas. Their compact area also makes interspecies comparisons more valid since environmental conditions will be much more similar. Many botanical gardens, such as that in Poznań, also have their own meteorological station enhancing the value of the plant records that have been made. Botanical garden archives offer additional possibilities (Miller-Rushing and Primack 2008; Donaldson 2009; Primack and Miller-Rushing 2009). Sadly, we are not aware of any other contemporary species-rich data sources within Poland with which to compare our data.

The results reported here confirm the responsiveness of plant phenology, particularly of spring events, to temperature. These help to confirm the value of plant phenology as a climate change indicator. Autumn events have typically been equivocal in their response to temperature but our results suggest a delay in autumn events associated with rising temperatures. Further work is needed to tease apart the relative importance of mean temperature and other autumn drivers such as wind, sunshine and frost on leaf fall. We investigated 18 interphase intervals. Seven of these had become significantly longer which broadly suggests a lengthening of the growing season. In all but three cases the earlier phase timing was not significant after temperature had been accounted for. Thus it appears that the later phases are far more influenced by prevailing temperature than the timing of preceding phases. However, for Aesculus hippocastanum, leaf colouring appeared positively correlated with earlier phases suggesting that early leafing resulted in an earlier end of season. However, this pattern was not apparent in Betula pendula or Corylus avellana. Further investigation of whether leaves have a limited lifetime (“shelf life”), thus associating early springs with early autumns, may be justified (see also Cleland et al. 2007).

We calculated a measure of spring synchrony based on the standard deviation between the dates of the six early spring events, and a similar one for autumn based on the standard deviation between the bare dates for eight tree species. Both of these measures increased significantly over time indicating reduced synchrony in both seasons. Whilst we are limited in the choice of phenophases to assess synchrony, examination of their coefficients for trend and temperature response in Table 1 do not suggest that synchrony measures are overly influenced by a single phenophase. We believe that most people associate the seasons, particularly spring and autumn, with biological events. These results suggest that the sharply defined spring at this mid-continent location at the beginning of our study period (low standard deviation in Fig. 3) has become less consistent over time. A similar change has occurred in autumn. Thus the perceived boundaries between winter and spring, and between autumn and winter, have become increasingly blurred. The consequences of this phenomenon to wildlife, particularly those with specific dependent links in food webs, remains to be seen.