Climate warming prolongs the time interval between leaf-out and flowering in temperate trees: Effects of chilling, forcing and photoperiod

Abstract

1. Leaf-out and flowering are two key phenological events of plants, denoting the respective onsets of visible vegetative growth and reproduction during the year. For each species, the schedule of vegetative growth and reproduction is crucial to the maximization of its fitness. Warming-induced advances of leaf-out and flowering have been reported frequently; however, it is unclear whether the responses of the two events are equal for any given species.
2. Using long-term phenological records in Europe, we examined simultaneously the responses of both leaf-out and flowering of four common temperate tree species to climate warming and further examined the effects of winter chilling, spring forcing and photoperiod on the responses of the two events.
3. We found that regardless whether flowering or leaf-out occurred first, the first event advanced more than the second during 1950–2013, resulting in a prolonged time interval between the two events. The temporal changes were also supported by a similar geographical trend that the time interval between the two events increased from cold to warm sites. Due to the warming-induced reduction in chilling, the spring forcing accumulated until the second event was increased more than the forcing accumulated until the first event, and that reduced the temperature sensitivity of the second event. In addition to the effect of chilling, the shorter photoperiod, associated with the advanced spring phenology, was also likely to substantially increase the spring forcing accumulated until the second event, which thus slowed down its advance, compared to the advance of the first event. The relative contributions of chilling and photoperiod to the increased forcing varied between species and events, with chilling mostly outweighing photoperiod.
4. Synthesis. This study provides the large-scale empirical evidence of prolonged time interval between leaf-out and flowering with climate warming. The unequal advances of the two events may alter the partition of resources between vegetative growth and reproduction and cause different changes of spring frost damage to vegetative and reproductive tissues, which may alter species fitness and further affect ecosystem structure and function.

1 INTRODUCTION

Warming-induced shifts in the timing of phenological events have been frequently reported in recent decades (Cleland et al., 2007; Fitter & Fitter, 2002; Parmesan & Yohe, 2003; Root et al., 2003; Sherry et al., 2007), which affect nearly all aspects of ecology and evolution in the ecosystems (e.g. species interactions, population dynamics; Forrest & Miller-Rushing, 2010). Many studies have reported species-specific phenological shifts in response to climate change (Forrest & Miller-Rushing, 2010; Renner & Zohner, 2018) and found that phenological timing in early-season species was more responsive to warming than that in late-season species (Menzel et al., 2006; Miller-Rushing & Primack, 2008). Efforts have also been devoted to studies addressing responses of different phenological phases to the warming, such as the widely reported advances in spring leaf unfolding and delays in autumn leaf coloration (Chen et al., 2019; Menzel & Fabian, 1999), and the unequal advances of flowering and fruit ripening (Haggerty & Galloway, 2011; Post et al., 2008). However, little attention has been devoted to examining whether leaf-out and flowering of a given species respond equally or differently to climate warming.

Leaf-out and flowering are two key phenological events of plants, denoting the respective onsets of visible vegetative and reproductive development during the year. Warming-induced changes in the timing of these two phenological events affect the ecosystem functioning and ecosystem services in different ways (Sherry et al., 2007). Shifts in the vegetative phenology have considerable consequences on processes of ecosystem functioning, such as carbon uptake (Piao et al., 2017), whereas changes in reproductive phenology generally alter population and community dynamics in the future generations, thus affecting evolutionary processes (Fitter & Fitter, 2002; Sherry et al., 2007). For example, changes in flowering time may disrupt plant–pollinator interactions, particularly when the pollinators are seasonal (e.g. insects), and reduce seed production of plants and food resources to the pollinators, thereby influencing the survival and success of both species (Fitter & Fitter, 2002; Memmott et al., 2007). Differences in the shifts of flowering time may affect gene flow between species, and alter competitive interactions during reproduction, potentially altering species composition of the community (Sherry et al., 2007).

Leaf-out and flowering of any given species have evolved to occur in a predetermined sequence (Hall, 2000). In many species, leaf-out occurs before flowering, whereas in others, especially in many wind-pollinated broad-leaved trees, flowering occurs before the vegetative development, as the developing leaves would prevent the dispersion of pollen (Loescher et al., 1990). The two sequences reflect different strategies in partitioning the limited resources, particularly the stored resources acquired in the previous growing season, to vegetative and reproductive development, as the physiologically active organs (e.g. leaves, flowers) compete for resources (Gougherty & Gougherty, 2018; Lieberman, 1982). Flower-before-leaf species rely on the stored resources to support the reproductive development before the leaves develop, whereas leaf-before-flower species can utilize both stored and newly synthesized resources simultaneously and thus may have greater access to resources for reproduction following the onset of photosynthesis in leaves (Gough et al., 2010). Although recent studies suggest that the association between the timing of leaf-out and flowering may be weak (Ettinger et al., 2018; Mulder & Spellman, 2019), changes in the time lag between the two events may affect the degree of separation of resources for vegetative growth and reproduction within trees (Singh & Kushwaha, 2005). For example, Maust et al. (1999) found that, in some cultivated fruit trees, the accelerated development of flower buds may delay and suppress the emergence of leaves and canopy establishment. As the species have optimized the allocation of resources towards the maximization of fitness under historical climates (Cody, 1966; Hegazy et al., 2005), the possibly altered resources allocation may affect the species fitness and further affect ecosystem structure and function (Doust, 1989; King & Roughgarden, 1982; Kozłowski, 1992).

Warming-induced shifts in the timing of leaf-out and flowering also alter the exposure of vulnerable leaves and flowers to late spring frosts (Cannell, 1985; Hänninen, 1991, 2006; Vitasse et al., 2014), respectively, affecting the growth and reproduction of plants (Inouye, 2000). It has been reported that the temporal changes in the risk of frost damage vary between species: the species whose phenology is especially responsive to warming are more prone to increased risk of spring frost damage than the less responsive ones (Ma et al., 2019). Similarly, the possibly unequal responses of leaf-out and flowering for a given species may cause different changes in the risk of frost damage to leaves and flowers, which affect the vegetative growth and reproduction to different degrees.

Temperature is the primary environmental factor affecting spring phenology in temperate plants (Menzel et al., 2006). Plants often require a certain exposure to chilling temperatures during winter to release endodormancy before the vegetative bud burst, or flowering, can be triggered by a sufficient exposure to forcing temperatures (i.e. the sum of temperatures above a specific threshold required to cause ontogenetic development towards the visible phenological events, such as bud burst or flowering, henceforth ‘forcing’; Coville, 1920; Hänninen, 2016). Additionally, the reduction of chilling in a warmer climate may cause a nonlinear increase in the forcing requirement, and thus slow down, or even reverse, the warming-induced advance of spring phenology (Ford et al., 2016; Fu, Zhao, et al., 2015; Murray et al., 1989). As climate warming affects the accumulation of both chilling and forcing, their interactive effects on the timing of leaf-out and flowering may be unequal and thus cause different responses of the two phenological events (Flynn & Wolkovich, 2018; Hänninen & Tanino, 2011).

In addition to temperature, photoperiod (daylength) may also limit the ability of temperate plants to respond to climate warming (Körner & Basler, 2010; Marchin et al., 2015; Way & Montgomery, 2015; Zohner et al., 2016; Zohner & Renner, 2015). Although photoperiod does not change with warming, plants experience shorter photoperiod prevailing at the earlier occurrence of the spring phenological events (Flynn & Wolkovich, 2018). The associated shorter photoperiod, by substantially increasing the forcing requirement, may reduce the temperature sensitivity of spring phenology (Flynn & Wolkovich, 2018; Fu, Piao, et al., 2019; Fu, Zhang, et al., 2019). By comparing warming-induced spring phenological differences among temperate trees, Geng et al. (2020) recently found that late-season species have a higher sensitivity to photoperiod shortage and a greater increase in forcing requirement than early-season species, leading to a relatively smaller advance in late-season species and increased interspecific time differences in spring phenology. This pattern found between early- and late-season species may be extended to early and late phenological events of a given species. That is, given the predetermined sequence of leaf-out and flowering in each species, we expect a greater photoperiod limitation in the late event than in the early event. In addition, the shorter photoperiod due to advanced phenology may have a larger influence on the forcing requirement of the late event, probably leading to unequal advances of the two events. Given these complicated and interactive effects of temperature (winter chilling and spring forcing) and photoperiod, we hypothesize that leaf-out and flowering of a given temperate tree species respond differently to a warming climate, which consequently alters the time interval between the two spring phenological events.

Using observations of first leaf date (FLD) and first flower date (FFD) of four common temperate tree species in Europe during 1950–2013, we aimed (a) to detect and quantify the temporal changes in FLD and FFD and the time interval between them and (b) to further elucidate the explanations behind the observed trends. For the second purpose, we compared the changes of winter chilling, spring forcing and photoperiod between the two phenological events, and further examined the effect of reduced chilling and shorter photoperiod on the spring forcing accumulated until the two events. To support the temporal changes observed during 1950–2013, we also examined in a similar way differences of the timing of FLD and FFD and climatic indices among sites representing a gradient from cold to warm sites.

2 MATERIALS AND METHODS

2.1 In situ phenology dataset and climate data

Phenology dataset was obtained from the Pan European Phenology (PEP) network (www.pep725.eu; Templ et al., 2018), which provides a free access to in situ phenology records of a variety of plants across Europe. As we studied simultaneously the responses of leaf-out and flowering to climate warming, we could select only species where both events were recorded. Only for four tree species in the PEP dataset, sufficiently long-term records over a large geographical region are available. Out of the four species, three are deciduous (Alnus glutinosa, Fraxinus excelsior and Aesculus hippocastanum) and one is evergreen (Pinus sylvestris). We used both FLD and FFD of these four temperate tree species during 1950–2013. For the three deciduous species, FLD was determined as the date when first leaves unfolded (denoted as BBCH code 11 in the PEP dataset) and for the evergreen P. sylvestris, the date when first leaves separated (BBCH code 10). FLD indicates the start of the visible vegetative development in each species. For all four species, FFD was determined as the beginning of flowering, which is denoted as BBCH code 60 in the PEP dataset.

For each species, the following screening criteria were applied: (a) For each site, we used observations only in those years when both FLD and FFD were recorded and (b) We kept in the dataset sites with more than 20 years of observations during the period 1950–2013. In most sites, observations were available for many more years. Alternative time thresholds of 30 and 40 years were also used to test the robustness of the result. The geographical distribution of the in situ phenological observation sites for each species is provided in Figure S1. For each species, the time interval between FLD and FFD per year at each site was calculated by subtracting the date of the first event from the date of the second event. Based on the records across all years and all sites, in A. hippocastanum and P. sylvestris, leaf-out was taken as the first and flowering as the second event so that the time interval was computed as FFD-FLD, while in A. glutinosa and F. excelsior, flowering was taken as the first and leaf-out as the second event so that the time interval was calculated is FLD-FFD. The mean time interval between FLD and FFD for each species was computed as the average of the time intervals between FLD and FFD across all years and all sites. To show the results of four species with different mean time intervals between FLD and FFD, the relative time interval was calculated for each species as the time interval between the two events divided by the mean time interval.

Daily temperature data from a gridded climate dataset E-OBS with a spatial resolution of 0.25° was used in this study (Cornes et al., 2018). Data in E-OBS are based on EU-FP6 project ENSEMBLES (http://ensembles-eu.metoffice.com) and the ECA&D project (http://www.ecad.eu; Haylock et al., 2008). To better capture temperature variation in each site, in particular in the mountain areas, temperature data were adjusted using elevation difference between the site and the climate grid cell, based on a temperature lapse rate of 6.4°C/km (Olsson & Jönsson, 2014).

2.2 Temporal changes in FLD, FFD and the time interval between FLD and FFD

For each species, linear mixed effects models (LMMs), including phenological observation site as a random variable, were employed to compute long-term shifting rates of FLD, FFD and the time interval between FLD and FFD (days/decade):

(1)

where y_ij is the FLD, FFD or the time interval between FLD and FFD in year i at the site j, x_ij is the year i at the site j, α is the intercept with the y-axis, β is the long-term shifting rate, a_j is the random effect of site j and ε_ij is the error term. To detect any nonlinear trends, data-driven and flexible generalized additive models (GAMs) were further applied to examine the long-term temporal changes of the relative time interval between FLD and FFD over time.

2.3 Temporal changes in winter chilling, spring forcing and photoperiod for FLD and FFD

Winter chilling, that is, the chilling units accumulated until the phenological event to occur, was calculated as the number of days, from 1st November in the previous year to FLD or FFD, when the daily mean air temperature was below 5°C (Cannell & Smith, 1983; Murray et al., 1989; Weinberger, 1950). Spring forcing, that is, the forcing units accumulated until the phenological event to occur, was computed as follows:

(2)

where t₀ is the starting date for forcing accumulation, here assumed to be 1st January (Cannell & Smith, 1983; Fu, Piao, et al., 2015; Murray et al., 1989), t_d is the FLD or FFD, T_t is the daily mean air temperature and T_th the threshold temperature for forcing accumulation. Temperatures above 5°C usually contribute to forcing in temperate regions, so 5°C was used as the value of T_th (Cannell & Smith, 1983; Murray et al., 1989). An alternative calculation of forcing using a sigmoidal function of the daily mean air temperature (Fu, Piao, et al., 2015; Hänninen, 1990) generated very similar results and thus we only reported results obtained using Equation (2). Photoperiod is the duration of day light time on the FLD or FFD at a given phenological observation site. Photoperiod on the FLD or FFD was used as a proxy of photoperiod condition for the two phenological events (Fu, Zhang, et al., 2019). It was calculated using r package insol (http://www.meteoexploration.com/R/insol/).

For each species, LMMs (Equation 1), including phenological site as a random effect, were used to compute long-term changing rates of winter chilling, spring forcing and photoperiod for FLD and FFD. Alternatively, linear regressions were also used to examine the temporal changes of yearly mean values of three climatic indices across all studied sites and very similar results were obtained.

2.4 The influence of winter chilling and photoperiod on spring forcing

To examine the influence of the warming-induced reduction of chilling on the spring forcing, linear regression between yearly mean values of winter chilling and spring forcing was performed separately for FLD and FFD in each species. For FLD and FFD separately in each species, we also divided the data into three subsets, based on the photoperiod conditions prevailing at the corresponding phenological date: short (<1st quantile of photoperiod), medium (between 1st and 3rd quantiles) and long (>3rd quantile) photoperiod, and further checked the influence of chilling on spring forcing under three different photoperiod conditions. An analysis analogous to the one described above was carried out to check the effect of shorter photoperiod on the spring forcing, both for all data and for three subsets of different chilling conditions: low (<1st quantile of chilling), medium (between 1st and 3rd quantiles) and high (>3rd quantile) chilling. In addition, the relative influence of winter chilling and photoperiod on the spring forcing was further calculated to assess their relative contributions to the changes of spring forcing accumulated until FLD and FFD in each species (Grömping, 2006).

2.5 Differences of the phenological and climatic indices among sites representing a gradient from cold to warm sites

An analysis analogous to the one described above for the temporal changes was carried out geographically by comparing sites forming a temperature gradient from cold to warm sites. For each species, LMMs (Equation 1), including phenological site as a random effect, were used to compute differences of FLD and FFD, and of the winter chilling, spring forcing and photoperiod among sites with different site temperatures. Site temperature was calculated for each phenological observation site as the average of the mean annual temperatures across all years providing phenological observations. It summarizes the local climate of the sites and roughly captures the thermal conditions occurring during phenological development, thus representing spatial variation of climate among sites (Ma et al., 2018; Rossi et al., 2016). For the analysis in relation to site temperature, sites with extremely high or low temperatures (out of 99% quartile range) were excluded due to the small number of such sites.

All data analyses were conducted with R software (R Development Core Team, 2018). LMMs and GAMs were conducted using lme4 (Bates et al., 2014) and mgcv (Wood, 2001) packages.

3 RESULTS

3.1 Temporal changes of FLD, FFD and time interval between FLD and FFD

Both FLD and FFD significantly advanced during 1950–2013, with the shifting rate ranging from −1.4 to −3.8 days/decade (Table 1). In the two flower-before-leaf species (A. glutinosa and F. excelsior), FFD advanced more than FLD, while in the two leaf-before-flower species (A. hippocastanum and P. sylvestris), FLD was more responsive than FFD (Table 1; Figure S2). Thus, in all species, regardless whether flowering or leaf-out occurred first, the first event advanced more than the second event. Consequently, the time interval between the two events prolonged during 1950–2013 (Figure 1), with the rate varying from 0.6 to 1.3 days/decades (Table 1). The above results were based on phenological records at sites with time series longer than 20 years. We also carried out a robust analysis using smaller datasets with alterative time thresholds of 30 and 40 years. Regardless of the time threshold used, the prolonged time interval between leaf-out and flowering was always observed (Figure S3).

TABLE 1. First leaf date (FLD), first flower date (FFD) and the time interval between them in four temperate tree species in Europe during 1950–2013. For each variable, Mean indicates the mean value averaged over all years and observation sites; Rate (days/decade) derived from LMM indicates the rate of change during 1950–2013 and Response (days/°C) derived from LMM represents the slope of the dependence of the variable on site temperature

Species	#obs	#sites	First leaf date (FLD)			First flower date (FFD)			Time interval of FLD and FFD (days)
			Mean DOY	Rate (days/decade)	Response (days/°C)	Mean DOY	Rate (days/decade)	Response (days/°C)	Mean	Rate (days/decade)	Response (days/°C)
Aesculus hippocastanum	105,002	3,023	111 ± 11	−2.03 ± 0.02*	−5.29 ± 0.09*	130 ± 10	−1.44 ± 0.02*	−4.95 ± 0.07*	19 ± 9	0.58 ± 0.02*	0.34 ± 0.07*
Pinus sylvestris	39,262	1,331	131 ± 11	−2.38 ± 0.04*	−3.61 ± 0.16*	138 ± 10	−1.47 ± 0.03*	−2.88 ± 0.12*	7 ± 10	0.90 ± 0.04*	0.73 ± 0.15*
Alnus glutinosa	44,642	1,457	110 ± 13	−2.42 ± 0.04*	−4.74 ± 0.17*	74 ± 21	−3.75 ± 0.07*	−7.40 ± 0.27*	36 ± 19	1.33 ± 0.06*	2.65 ± 0.28*
Fraxinus excelsior	42,583	1,400	129 ± 10	−1.89 ± 0.03*	−4.24 ± 0.13*	118 ± 13	−2.98 ± 0.04*	−5.07 ± 0.18*	11 ± 11	1.09 ± 0.04*	0.83 ± 0.16*

• * Denotes significance at p = 0.01. SD (standard deviation) of the mean value and SE (standard error) of the rate or the response value derived from LMMs are provided.

FIGURE 1

Relative time interval between first leaf date (FLD) and first flower date (FFD) in four tree species in Europe during 1950–2013. The relative time interval was calculated for each species as the actual time interval divided by the mean time interval of the tree species so that the horizontal line at the value of 1 represents the mean time interval between FLD and FFD for the species. Blue solid line indicates the general trend of the relative time interval obtained with generalized additive model. Coloured dots represent for the corresponding species the mean values averaged over the different phenological observation sites during 1 year

3.2 Temporal changes of winter chilling, spring forcing and photoperiod for FLD and FFD

In all species, the winter chilling accumulated for FLD and FFD reduced almost at the same rate during 1950–2013 (Figure 2a; Table S1). In the two leaf-before-flower species (A. hippocastanum and P. sylvestris), increased forcing was observed for both FLD and FFD, and the rate of increase was much higher in the second than in the first event (Figure 2b; Table S1). In the two flower-before-leaf species (A. glutinosa and F. excelsior), the second event FLD had increased forcing, but the first event FFD even had a slight decrease in forcing during the examined time period (Figure 2b; Table S1). Thus, in all four species, the increase in forcing was greater in the second event than in the first event during 1950–2013. With the advances of FLD and FFD, all species experienced the reduction in photoperiod for both events during 1950–2013 (Figure 2c). Moreover, in all species, due to the unequal advances of the two phenological events, the reduction of photoperiod in the first event was larger than that in second event (Figure 2c; Table S1).

FIGURE 2

Winter chilling (a), spring forcing (b) and photoperiod (c) of first flower date (FFD) and first leaf date (FLD) in four tree species in Europe during 1950–2013. Solid lines indicate the trends obtained with LMMs. Slopes are provided in Table S1. The dots represent the mean values averaged over the different phenological observation sites for 1 year. CU and FU denote chilling and forcing units, respectively

3.3 The influence of winter chilling and photoperiod on spring forcing

In all species, as winter chilling reduced under climate warming, the spring forcing accumulated until both FFD and FLD increased, but the increasing rate of spring forcing was higher in the second event than in the first event (Figure 3a; Table S2). Similar results were observed for three subsets of the data with short, medium and long photoperiod conditions (Figure S4).

FIGURE 3

Spring forcing accumulated until first leaf date (FLD) and first flower date (FFD) in relation to the corresponding winter chilling (a) and photoperiod (b) in four temperate tree species in Europe during 1950–2013. Solid lines indicate the trends obtained with linear models. Slopes are provided in Table S2. The dots represent the mean values averaged over the different phenological observation sites for 1 year. CU and FU denote chilling and forcing units, respectively

A negative correlation between photoperiod and spring forcing was found for both FFD and FLD of all species (Figure 3b). However, with the reduction of photoperiod, the increase in spring forcing was higher in the second than in the first event in most species, except A. hippocastanum (Figure 3b; Table S2). Similar results were observed for three subsets of the data with low, medium and high chilling conditions (Figure S5).

The relative influence of chilling and photoperiod to spring forcing differed among the two events and among the studied species. In most cases, chilling had a larger contribution to increased spring forcing than photoperiod, whereas in FFD of P. sylvestris and A. glutinosa, photoperiod played a greater role than chilling (Figure 4).

FIGURE 4

Relative influence of chilling and photoperiod on the spring forcing accumulated until first flower date (FFD) and first leaf date (FLD), in four temperate tree species in Europe during 1950–2013

3.4 Differences of the phenological and climatic indices among sites representing a gradient from cold to warm sites

During 1950–2013, site temperature across the whole study area varied from 6.2 to 11.0°C, with an average of 8.9°C (Figure 5). Both FLD and FFD significantly advanced with increasing site temperature, with a response varying from −2.9 to −7.4 days/°C (Table 1; Figure S6). In all species, the first event responded to the site temperature more than the second event. Accordingly, the time interval between the two events lengthened with increasing site temperature at a response varying from 0.3 to 2.7 days/°C (Figure 5; Table 1).

FIGURE 5

Relative time interval between first leaf date (FLD) and first flower date (FFD) in four tree species in Europe during 1950–2013 in relation to site temperature. The relative time interval was calculated as the actual time interval divided by the mean time interval of the tree species so that the horizontal line at the value of 1 represents the mean time interval between FLD and FFD for the species. The site temperature was calculated for each phenological observation site as the average of the mean annual temperature values across all years providing phenological observations. Blue solid line indicates the general trend obtained with generalized additive model. Coloured dots represent for the corresponding species the mean values averaged over the phenological observation sites having their site temperatures within 0.1°C

With the increasing site temperatures, the winter chilling of both FFD and FLD decreased, the response to site temperature being almost the same for the two events (Figure S7a; Table S3). The spring forcing of FFD and FLD of all four species increased with site temperature (Figure S7b; Table S3). However, the increase in forcing was in all species higher in the second than in the first event, regardless whether FLD or FFD occurred first. The reduction of photoperiod in the first event was larger than that in the second event from cold to warm sites (Figure S7c; Table S3).

4 DISCUSSION

Although advances of both leaf-out and flowering of plants have been extensively reported (Cleland et al., 2007; Fitter & Fitter, 2002; Parmesan & Yohe, 2003; Root et al., 2003; Sherry et al., 2007), it is unclear whether the advances of the two events of a given species are equal. By simultaneously analysing the responses of both leaf-out and flowering timing of four tree species over time and space, we found that although both events advanced with elevated temperature, their advances were not equal. Rather, the first event was more responsive to warming climate than the second one, regardless whether flowering or leaf-out occurred first. Consequently, the time interval between leaf-out and flowering prolonged during 1950–2013, at a rate of 0.6–1.3 days/decades. Geographically, the time interval between the two phenological events enlarged from cold to warm sites, the response being 0.3–2.7 days/°C. The findings of the geographical analysis thus further support our results concerning the temporal changes. Additionally, the prolonged time interval was also accompanied by the reduced variability (increased synchrony) among populations of both leaf-out and flowering in the four examined species (Ma et al., 2018), further supporting the separation in time of the two events with climate warming. Due to limited data availability as described in the Section 2, only four species were examined in the present study. However, these four species included both flower-before-leaf and leaf-before-flower species, both deciduous and evergreen species, and both early- and late-season species, which could therefore reflect the responses of a variety of temperate tree species.

The schedule of vegetative growth and reproductive development is crucial to the maximization of species fitness (Doust, 1989; Kozłowski, 1992). Under historical climates, each species has evolved an optimal timing of both vegetative and reproductive phenology to enhance its fitness (Richardson et al., 2018). The prolonged time interval between leaf-out and flowering indicates an increasing time lag between the onset of vegetative growth and reproductive development with elevated temperature. The rates of time lag increase were in the present study moderate (0.58–1.33 days/decades). However, the time lag tends to increase further with continued warming, which probably affects the transfer and allocation of resources (e.g. carbohydrates, nutrients and water) between vegetative growth and reproduction (Kozłowski, 1992). In flower-before-leaf species, more stored resources might be devoted to support the flowering and subsequent reproductive development before the leaves develop (Gougherty & Gougherty, 2018). This may accelerate the development of fruit sets and delay vegetative development, resulting in low leaf/fruit ratios, as reported in cultivated fruit trees that flowering occurs before leaf-out (Maust et al., 1999). However, Bolmgren and Cowan (2008) found no relationship between flowering time and seed mass or plant height in woody species, and suggested that the stored resources may allow the decoupling of vegetative growth and reproduction. These inconsistent findings are probably because plants’ ‘decisions’ on resources storage and expenditure are complex and often include substantial time lags. In particular, perennial species allocate resources to growth and reproduction not only within the growing season but also between growing seasons (Johansson et al., 2013). This complicates the effect of shifted phenology on resources allocation within plants. In the case of leaf-before-flower species, more resources can be accumulated during the vegetative growth to better support the subsequent reproduction, thereby benefiting size and quality of fruits and seeds (Williamson et al., 2002). However, due to the time lags between resource availability and investment in reproduction and growth, it is also possible that resources are more completely devoted to vegetative growth (e.g. shoot, root) rather than reproduction. While this study focused on the response of timing of leaf-out and flowering and the underlying mechanisms, we call for further studies that simultaneously monitor the timing of phenological events and the allocation of resources within plants to better evaluate the consequences of altered phenology under climate warming.

The advanced leaf-out and flowering implies the possibly increased susceptibility of both newly developed leaves and flowers to spring frost damage (Augspurger, 2013; Liu et al., 2018; Ma et al., 2019). However, the greater advancing rates of the first event observed in all species suggest that the first event is more likely to have increased risk of frost damage than the second one (Ma et al., 2019). In flower-before-leaf species, warming increases the risk of frost damages more in reproductive than in vegetative tissues. These species would suffer losses of seed production, but not that much direct loss of carbon gain. In leaf-before-flower species, the situation would be reversed. It is possible that seed production is also reduced due to a severe loss of carbon uptake. These changes may affect the growth, survival and fitness of species, which may ultimately alter the structure and function of terrestrial ecosystems (Inouye, 2000; King & Roughgarden, 1982; Kozłowski, 1992).

The unequal advances of the two spring phenological events are partly attributable to warming-induced changes in winter chilling and spring forcing. Consistent with previous studies (Cannell & Smith, 1983; Fu, Piao, et al., 2015), we observed that under the reduction of winter chilling, spring forcing accumulated until the two events increased in all studied species. However, in all species, the increase in forcing was much greater in the second than in the first event (Figure 2b), indicating that the forcing needed for the second event was more sensitive to the reduced chilling than that for the first event (Figure 3a; Figure S4). This is also consistent with the nonlinear response of spring phenology to increased temperature, as reported in previous studies (Ford et al., 2016; Fu, Zhao, et al., 2015; Ma et al., 2018; Pope et al., 2013). The temporal changes were further supported by the similar geographical trend that the increase in forcing was higher in the second than in the first event as the chilling reduced from cold to warm sites (Figure S7; Table S3). Our results suggest that the second event has a higher chilling requirement, and thus the forcing was largely elevated when the chilling became inadequate. Consequently, a larger increase in forcing in the second event greatly counteracted the accelerating effect of warming so that the acceleration of second event caused by warming was less than that of the first event.

In addition to the effect of winter chilling, photoperiod is likely another environmental factor causing the unequal increases in spring forcing in two events. In agreement with previous studies (Flynn & Wolkovich, 2018; Fu, Piao, et al., 2019; Fu, Zhang, et al., 2019), the shorter photoperiod due to earlier spring phenology tended to increase the spring forcing of both leaf-out and flowering (Figure 3b). Except A. hippocastanum, the first event experienced a larger reduction in photoperiod but a lower increase in forcing, whereas the second event had a smaller reduction in photoperiod but a greater increase in forcing (Figure 3b; Table S2). Thus, in addition to the effect of reduced chilling, the influence of shorter photoperiod on the forcing accumulation was also larger in the second event, and therefore its advance with warming was reduced, as compared with the advance of the first event.

Although both reduced chilling and shorter photoperiod tended to increase the forcing of the two spring phenological events, their relative contributions to the increased forcing varied between species and phenological events (Figure 4). Generally, chilling outweighs photoperiod in increasing forcing and thus preventing too early development of two phenological events (Laube et al., 2014). However, the relative contributions of chilling and photoperiod to forcing had no consistent tendency between the first and the second events (Figure 4). As all climatic indices co-varied in the observational data, further temperature- and photoperiod-manipulation experiments are needed to better clarify and evaluate the independent and interactive effects of different climatic factors on the uneven advances of the two phenological events observed in the present study (Hänninen et al., 2019).

In this study, winter chilling and spring forcing were used to explain the different responses of leaf-out and flowering. Unfortunately, the experimental evidence for the details of chilling and forcing, such as temperature responses of chilling and forcing, and the time windows for chilling and forcing, remains extremely scarce for any given tree species and provenance (Hänninen, 2016; Hänninen et al., 2019). In the present study, the calculation of winter chilling and spring forcing followed Weinberger (1950), Cannell and Smith (1983), Murray et al. (1989), Hänninen (1990) and Fu, Zhao, et al. (2015); see Section 2). Similar to earlier studies examining long-term observational phenological records (Chen et al., 2019; Fu, Piao, et al., 2015; Fu, Zhao, et al., 2015; Ma et al., 2018), chilling and forcing in the present study should be considered as rough climatic indices approximating the climatic effects on the underlying ecophysiological phenomena, and correlating with the phenological timing in spring; rather than strict ecophysiological model variables (Chuine et al., 2013; Hänninen & Kramer, 2007) addressing the phenomena explicitly.

In all species, the spring forcing for leaf-out increased over time, in agreement with Fu, Piao, et al. (2015), whereas divergent changes in forcing for flowering were observed. The forcing needed for flowering increased in those species where flowering is the second event and decreased slightly in species where flowering is the first event (Figure 2b). In the first case, the changes of forcing required counteracted the accelerating effect of warming, whereas in the second case, the changes of forcing caused a further acceleration. Our results suggested that unlike leaf-out, the forcing required for flowering might not increase with climate warming in temperate woody species. However, further studies on more species are needed to examine whether this pattern generally exists in temperate tree species. Our results also implied the complexity of the responses of flowering phenology to warming temperatures. Previous studies (Chen et al., 2019; Fu, Piao, et al., 2015; Fu, Zhao, et al., 2015; Ma et al., 2018), as well as this one, typically focused on the effect of temperature in months prior to flowering (winter chilling and spring forcing) on the timing of flowering. Diggle and Mulder (2019) reported that temperature could affect flowering by altering development of buds in the year(s) prior to flowering. This suggested that temperature over the entire multi-year course of flower development needs to be examined to improve our understanding of the responses of flowering to warming temperatures.

A better understanding of long-term phenological responses of trees to climate warming is a prerequisite for assessing and predicting the impacts of climate warming on forest ecosystems (Forrest & Miller-Rushing, 2010). Our finding of unequal responses of leaf-out and flowering and the prolonged time interval between the two events deepens our understanding of how climate change impacts species-level phenology. Our findings may also be one manifestation of the mechanisms how climate warming causes profound changes in species fitness, distribution and co-evolution in temperate terrestrial ecosystems.

ACKNOWLEDGEMENTS

This project was funded by National Natural Science Foundation of China (31971499, 31600392, 41661144007 and 41861124001), Natural Science Foundation of Guangdong Province (2016A030310013, 2019B121202007) and the International Collaborative Key Project of the CAS (GJHZ1752). The authors acknowledge all members of the PEP725 network for collecting and providing the phenological data. The authors also acknowledge the E-OBS dataset from the EU-FP6 project ENSEMBLES (http://ensembles-eu.metoffice.com) and the data providers in the ECA&D project (http://www.ecad.eu).

AUTHORS' CONTRIBUTIONS

J.-G.H., Q.M. and H.H. designed the study; Q.M. performed the analysis and wrote the manuscript with assistance from J.-G.H., H.H., X.L. and F.B. All authors discussed and commented on the manuscript.

Open Research

PEER REVIEW

The peer review history for this article is available at https://publons.com/publon/10.1111/1365-2745.13558.

DATA AVAILABILITY STATEMENT

Data available from the Dryad Digital Repository https://doi.org/10.5061/dryad.jdfn2z391 (Ma et al., 2020).