In 1735, de Réaumur discovered that plant phenological events (the timing of life cycle phases, such as leafing or flowering) can be linked to temperature sums of the preceding period (de Réaumur, 1735). Several additional factors which influence phenological onset dates have since been identified, such as length of chilling period (Murray et al., 1989; Sogaard et al., 2008; Laube et al., 2014), photoperiod (Heide, 1993; Körner & Basler, 2010; Caffarra & Donnelly, 2011; Basler & Körner, 2012; Laube et al., 2014), temperature of the previous autumn (Heide, 2003), nutrient availability (Jochner et al., 2013b), precipitation (Peñuelas et al., 2002; Estiarte et al., 2011), and light regime (Linkosalo & Lechowicz, 2006). An overwhelming number of phenological studies have confirmed that temperature is the main driver for phenological events (Chuine et al., 2010), while all other factors are supposed to capture some of the remaining, unexplained variance. The underlying physiological processes for this temperature-sensitivity remain unclear, and temperature sensors involved in budburst have not yet been found (Samach & Wigge, 2005; Penfield, 2008).
The basic model to predict spring phenology is the thermal time model, which determines the critical forcing temperature sum above a certain threshold necessary for onset. More sophisticated models additionally include chilling requirements since the lengths of the preceding cold period reduce the forcing requirements of species (Murray et al., 1989; Sogaard et al., 2008). Some models also take into account photoperiod, or have a more physiological, process-orientated basis (Chuine, 2000; Schaber & Badeck, 2003; Blümel & Chmielewski, 2012). But the prediction of onset dates remains largely unsatisfactory (Richardson et al., 2012), leading to a call for even more complex models integrating ecology (Caffarra & Donnelly, 2011; Basler & Körner, 2012) and evolution (Pau et al., 2011).
In 2002, two articles (Fitter & Fitter, 2002; Sparks & Menzel, 2002) suggested that rather than temperature a temperature-related factor or indirect effect – Fitter & Fitter (2002) suggested sunshine hours – might be the key determinant of phenological events. This idea was revisited 2 yr later (Walther, 2004), but to the best of our knowledge was never developed.
Previous experimental attempts to study tree phenology faced technical challenges (if adult trees were investigated), or results were not transferable if seedlings were used since their phenology differs from adult individuals (Augspurger & Bartlett, 2003; Vitasse, 2013). Recent research indicates that the phenology of cuttings is comparable (Laube et al., 2014), if not equivalent (Vitasse & Basler, 2013), to tree phenology. Accordingly, we set up several experiments with cuttings to investigate the influence of air humidity on the spring phenology of tree species.
Our results led to the development of a new framework, which relates winter chilling to tissue desiccation, and spring development to tissue rehydration processes. We suggest that temperature-related air humidity might be the main trigger of the spring development of trees.
Air humidity influences budburst
We set up a climate chamber experiment (Expt 1) with two contrasting relative air humidities (40% and 90%), all other factors (temperature, photoperiod) were identical. Freshly cut dormant twigs of nine different woody species were put into water bottles, and their phenological development was recorded during a 30-d forcing period (see Supporting Information Methods S1, Fig. S1, and Tables S1–S3). Results showed that air humidity influences the budburst of tree species (Fig. 1). In the drier chamber, the median start of leaf unfolding occurred 7 d later than in the more humid chamber (P <0.001, df = 136, linear mixed effect model). This effect was consistent and pronounced for all species (Fig. 1), but with considerable variation. Advance in leaf unfolding was strongest for Cornus mas (median advance 13 d, P <0.001, df = 13), weakest and nonsignificant for Picea abies (1 d). Leafing in the humid chamber was significantly earlier in one or more leafing stages for seven out of nine species (Table S4).
Taking into account the relatively short duration of the experiment (30 d), the advance of 7 d was remarkably high. We are not aware of any study on the effect of air humidity on budburst more recently than 1979 (Düring, 1979). Working with different wine cultivars, that study reported an advance of budburst of between 7 and 14 d at 95% compared to 50% relative humidity; values highly comparable to our results. This magnitude of sensitivity is remarkable, especially since air humidity has been completely ignored as a driver in recent phenological research. It has been reported that coastal populations develop earlier than continental populations (Estiarte et al., 2011), which could result from differences in air humidity. In addition the partial absence of phenological response to urban heat islands might be related to decreased water vapour pressure (Gazal et al., 2008). Other authors (Do et al., 2005; Jochner et al., 2013a) found an influence of air humidity on phenology in tropical environments.
A recent study (Wolkovich et al., 2012) reported a high discrepancy between results of experimental warming studies and real world observations, with experiments systematically underestimating the phenological advance with temperature increase. It is well known that several experimental manipulations of temperature also alter air humidity, especially in commonly used open top chambers which have been shown to decrease relative air humidity (RH) (Marion et al., 1997; Norby et al., 1997). As our results show, reduced air humidity leads to delayed onset, which should counterbalance the effect of rising temperatures in experiments, which might explain this disparity. Thus, a careful monitoring and reporting of changes in air humidity is necessary to increase the comparability of experiments, and to improve the interpretability of results.
Humidity and bud development
Results of Expt 1 indicated that trees are able to sense and react to air humidity, which raised the question of causes and reasons. The literature suggests that spring development is not only restricted by low temperatures, but also by restricted water supply to tree crowns (Table S5). We thus hypothesized that water uptake via aboveground tissue might be involved in spring development. To test this idea, we set up a second experiment (Expt 2). Twigs were cut before budburst, and left without water supply in two humid chambers (box B sprayed daily with water, box A without spraying). The cut stem of each twig was taped to minimize water uptake via stems (details given in the Methods S1, see also Fig. S2).
The cuttings showed a rather normal spring development (Fig. S2). Out of 80 twigs, only seven showed no sign of bud development, with similar results for all species. Budburst was achieved by 43 twigs, 18 were even able to unfold leaves. Following bud development, twigs significantly gained weight (mean increase for twigs with no development 2.3% (n =7), 11.1% for twigs where buds swelled (n =30), 15.6% for twigs that reached budburst (n =25), and 29.1% for twigs that unfolded leaves (n =18)). Differences in increase in mass between different development stages were significant (Kruskal–Wallis test, P =0.001), and rank correlation between phenological development and mass increase was significant (rs = 0.47, P <0.01). Twigs in the box with additional water spraying gained more weight than twigs in the box without spraying: 24.6% cf 7.0% (P <0.001, Kruskal–Wallis test), which corresponds to differences in development stages achieved (78% vs 30% twigs reached budburst).
The mass increase with development could mainly be attributed to onset of photosynthesis with carbon fixation. However, water is the main content of fresh plant biomass (90%). Even if we assume that 70% of initial weight (mean 1.1 g) was water content, and that this water was transported completely to the developing buds, this should allow biosynthesis of c. 0.09 g biomass. Hence it does not explain the mass increase we observed (0.23 g for leaf unfolding, 0.18 g for budburst). Therefore we propose that a considerable proportion of the gain in mass must be related to water uptake. Since absorption via vascular tissue was reduced to a minimum, water must have been absorbed via the plant surface. The literature suggests that this might be possible either via the vapour phase and stomata or cuticula, and/or via small water droplets on the twigs′ surface (Table S6).
Foliar water uptake has been reported for several species before, and it can contribute considerably to plant water supply (Zimmermann et al., 2004; Limm et al., 2009), but research so far has focused on dry or temporarily dry ecosystems. In their review, Zimmermann et al. (2004) already suggested that foliar water uptake might contribute considerably to crown water supply in tall trees. But to the best of our knowledge foliar water uptake has not previously been reported to influence bud development or phenology. Previous findings with respect to bud water contents during winter and spring seem to be in line with our findings (Table S6), although an uptake from the air was never suggested.
Air humidity as a signal of spring
Spring is mainly associated with increasing temperatures and daylengths. But rising absolute air humidity, that is the density of water vapour (in g m−3), is also another feature of spring. During spring in temperate climates, temperature increase leads to higher saturation water vapour (Es), while relative air humidity (RH) is unstable or decreases. Since the temperature dependence of Es is exponential, and the influence of RH on absolute humidity is linear, absolute air humidity increases during spring (Fig. 2).
The temporal increase in spring temperature is highly variable, but extreme warm and cold spells are usually of short duration (a few days). RH reflects those extremes, which means that on exceptionally warm days, RH drops sharply (e.g. day of the year (DOY) 33–38 and 97–112, Fig. 2). Since absolute air humidity is determined by the temperature-dependent Es and RH, the increase of absolute air humidity is relatively smooth. To prove this, we analysed climate data from six meteorological stations across Germany (Methods S1). Coefficients of variation of absolute air humidity proved to be less than one half those of temperature regardless of the time period analysed (Table 1). Sign tests for differences in median CV between absolute humidity and temperature were highly significant (P <0.001), indicating a higher variability in temperature values. Hence the signal given by air humidity shows less fluctuation than the signal given by temperature, and also seems unlikely to expose plants to more frost damage than the temperature signal (Fig. S3).
Table 1. Coefficients of variation (%) for absolute air humidity and temperature during spring and summer
Median CV DOY 1–91
SD DOY 1–91
Median CV DOY 1–151
SD DOY 1–151
Median CV DOY 1–255
SD DOY 1–255
Day of the year (DOY) 1–91, 1 January–1 April; DOY 1–151, 1 January–31 May; DOY 1–255, 1 January–mid-September. n, number of years/station combinations; SD, standard deviation.
Our findings based on six Central European weather stations, suggest that, if plants need to detect spring, or sense temperature, they could rely on the signal given by absolute air humidity. Physiological studies that reported temperature-dependent gene transcriptions, plant hormone concentrations or proteins related to phenology (Kumar et al., 2012) worked with temperature treatments that did not control for absolute air humidity. It is thus not possible from those results to determine whether direct temperature treatments or indirect air humidity changes are the triggers of the physiological processes reported. We therefore suggest that it is not temperature itself, but rather absolute air humidity that might lead to phenological responses.
Chilling as a dehydration process
This assumption raises the question whether winter chilling, which is the second important factor determining the spring development of trees, could also be related to tissue moisture. We set up a chilling experiment with three treatment levels (Methods S1). Together with data from the air humidity Expt 1 and concurrent field observations, five different chilling levels on nine woody species were studied. Forcing requirements until budburst in the different chilling conditions were calculated both on the basis of established thermal time and the alternative of humidity time.
The expected (negative) correlations between thermal time perceived and the number of chill days proved to be significant for only two out of nine species, and results for Picea abies were even contrary to the relationship described in the literature. The established temperature-based approach thus failed to bring experimental and field observations into agreement (Fig. 3). In comparison, strong positive correlations between humidity time and chill days were significant for all species (P-values 0.001–0.015, Spearman's correlation, df = 4), and regression analysis generated high proportions of explained variance for all species (R2 = 0.75–0.97). The consistency between experimental and field observations was not only attributed to a smoother increase of humidity time under field conditions. For humidity time the amount of forcing sums received under experimental conditions complemented the sums already received under field conditions (data not shown). Facing the high discrepancy between experimental and field observations for the temperature based chilling/forcing approach, it is unlikely that this is only due to species-specific differences in estimation of forcing sums. Species-specific optimizations of forcing values, such as varying starting dates or threshold values, were not used with respect to humidity time either, which nevertheless explains a high proportion of variation in observed onset dates.
This suggests that, at least in Central Europe and the set of species used, chilling might be related to tissue moisture contents. The effects of cold periods might be linked to low air humidities during cold periods, which could lead to tissue dehydration. If dormant trees are rather passive, and water transport within trees is hampered, as the literature suggests (Table S5), aboveground tissue moisture should, at least to some extent, follow changes in ambient air humidity. The longer the dry (cold) period, the lower the actual tissue moisture, leading to higher humidity needs to reach tissue moisture required for onset.
Transpiration and likely also reverse transpiration are related to the gradient of water vapour between tissue space and outside air. Usually, the water content of tissue space air is assumed to be saturated at a given tissue temperature (see Peak & Mott, 2011, and references cited therein). However, since active water transport in trees is comparably small during winter and early spring, especially in deciduous trees, and stem/tissue moisture is low, this might also lead to decreased tissue space air humidity.
Further research needs
To conclude, we propose an alternative phenological framework which interprets winter chilling as dehydration, and spring forcing as rehydration of aboveground tissue. Our experiments showed that air humidity influences onset dates, and suggested that air water uptake via aboveground tissue might be involved. Chilling during dormancy might be attributed to dehydration of aboveground tissue, and both dehydration and rehydration processes might suffice to explain phenological onsets.
Although our results led to a rather consistent framework, they are restricted to nine woody species of the Northern Hemisphere flora, and to observations made in highly unnatural conditions. It remains to be validated if air humidity influences budburst in natural conditions (e.g. by testing phenological models based on air humidity). Moreover, the physiological background of the observed responses remains vague, since our assumptions on tissue desiccation and foliar uptake are highly fragmentary. We therefore encourage physiologists to assess possible mechanisms related to absolute air humidity, with respect to foliar uptake, but equally with respect to spring hydraulics of trees or temperature sensors in plants.
The authors thank D. Basler for advice concerning the maintenance of twigs, F. Steinbacher for calibration of climate chambers and technical advice, the team of TUM GHL Dürnast for facilities and help offered. The authors also thank A. Fuchs and H. Rudolf of the Bavarian Forest Administration for kind permission to use trees at Weltwald Freising. Special thanks also to A. Thole and C. Kramer who helped in both setting up the experiment and recording bud development. The authors further thank the German weather service for climate data. With the support of the Technische Universität München – Institute for Advanced Study, funded by the German Excellence Initiative.