This is the accepted version of the article: Fu,Yongshuo H.; Zhang, Xuan; Piao, Shilong; [et al.]. «Daylength helps tem- perate deciduous trees to leaf[U+2010]out at the optimal time». Global change biology, Vol. 25, issue 7 (July 2019), p. 2410-2418. DOI 10.1111/gcb.14633 This version is avaible at https://ddd.uab.cat/record/218188 under the terms of the license 1 Daylength helps temperate deciduous trees to leaf-out at the optimal time 2 3 Short title: Daylength guarantees leaf-out at right time 4 5 Authors: Yongshuo H. Fu1,2,*, Xuan Zhang1,Shilong Piao3, 4, 5, Fanghua Hao1, Xiaojun Geng1, 6 Yann Vitasse6, Constantin Zohner7, Josep Peñuelas8,9, Ivan A. Janssens2 7 Affiliations: 8 1 Beijing Key Laboratory of Urban Hydrological Cycle and Sponge City Technology, College 9 of water sciences, Beijing Normal University, Beijing, China 10 2 Department of biology, University of Antwerp, Antwerp, Belgium 11 3Sino-French Institute for Earth System Science, College of Urban and Environmental Sciences, 12 Peking University, Beijing 100871, China 13 4 Key Laboratory of Alpine Ecology and Biodiversity, Institute of Tibetan Plateau Research, 14 Chinese Academy of Sciences, Beijing 100085, China 15 5 Center for Excellence in Tibetan Earth Science, Chinese Academy of Sciences, Beijing 100085, 16 China 17 6 Forest Dynamics Unit, Swiss Federal Institute for Forest, Snow and Landscape Research, 18 WSL, Birmensdorf, Switzerland 19 7 Institute of Integrative Biology, ETH Zurich (Swiss Federal Institute of Technology), 8092 20 Zurich, Switzerland 21 8 CREAF, Cerdanyola del Vallès, Barcelona 08193, Catalonia, Spain 22 9 CSIC, Global Ecology Unit CREAF-CSIC-UAB, Bellaterra, Barcelona 08193, Catalonia, Spain 23 24 Keywords: spring phenology, climate change, daylength, temperature response, deciduous trees 25 Revised manuscript for Global change biology 26 March 15, 2019 1 27 Abstract: 28 Global warming has led to substantially earlier spring leaf-out in temperate-zone 29 deciduous trees. The interactive effects of temperature and daylength underlying this 30 warming response remain unclear, yet need to be accurately represented by Earth 31 System models to improve projections of the carbon and energy balances of temperate 32 forests and the associated feedbacks to the Earth’s climate system. We studied the 33 control of leaf-out by daylength and temperature using data from six tree species 34 across 2377 European phenology observation sites (www.pep725.eu), each with at 35 least 30 years of observations. We found that, in addition to- and independent of the 36 known effect of chilling, daylength correlates negatively with the heat requirement for 37 leaf-out in all studied species. In warm springs when leaf out is early, days are short 38 and the heat requirement is higher than in an average spring, which mitigates the 39 warming-induced advancement of leaf-out and protects the tree against precocious 40 leaf-out and the associated risks of late frosts. In contrast, longer-than-average 41 daylength (in cold springs when leaf-out is late) reduce the heat requirement for leaf- 42 out, ensuring that trees do not leaf-out too late and miss out on large amounts of solar 43 energy. These results provide the first large-scale empirical evidence of a widespread 44 daylength effect on the temperature sensitivity of leaf-out phenology in temperate 45 deciduous trees. 2 46 Introduction 47 The timing of leaf-out co-determines the growth, reproductive success and 48 competitiveness of temperate deciduous trees and thus strongly affects their fitness 49 and distribution (Chuine, 2010). Global warming has led to substantially earlier spring 50 leaf-out (Menzel et al., 2006, Parmesan & Yohe, 2003, Peñuelas & Filella, 2001), 51 although this advance is declining (Fu et al., 2015). These changes in spring 52 phenology may influence terrestrial ecosystem fluxes of carbon, water, nutrient and 53 energy in a short term (Keenan et al., 2014, Myneni et al., 1997, Piao et al., 2017). 54 Mechanistic understanding of the leaf-out process is, however, far from complete 55 (Chuine et al., 2010, Flynn & Wolkovich, 2018, Körner & Basler, 2010, Laube et al., 56 2014, Zohner et al., 2016), which challenges the projected impacts of climate change 57 on ecosystems by dynamic global vegetation models (Richardson et al., 2012). A 58 better understanding of the ecophysiological processes controlling leaf-out phenology 59 is thus essential for improving our understanding of the responses of ecosystems to 60 the ongoing climate change and the subsequent feedbacks to the climate system, as 61 well as explaining the slow-down of the warming-induced advance in leaf out. 62 63 A species’ optimal leaf-out date results from natural selection that optimizes the 64 species’ fitness under given environmental conditions, such as avoiding freezing 65 damage (Lenz et al, 2016), ensuring flowering synchrony among species (Elzinga, 66 Atlan and Biere, 2007; Zohner, Mo & Renner, 2018) and maximizing the length of 67 the remaining season for light and nutrient resources (competition with other trees) as 3 68 well as for tissues maturation (Körner et al, 2016). In temperate and boreal regions, 69 temperature, including both cold winter temperatures (chilling requirement) and warm 70 spring temperatures (heat requirement), and daylength interact to realize leaf out 71 around the optimal date (Flynn & Wolkovich, 2018, Körner & Basler, 2010). 72 Chilling accumulates over autumn and winter, and when the accumulated chilling 73 exceeds the chilling requirement, endodormancy (the first stage of dormancy (Lang, 74 1987)) is broken and buds enter the second dormancy stage: ecodormancy (Chuine & 75 Régnière, 2017, Hänninen, 2016). During ecodormancy, meristem cells begin to 76 grow, a process that is accelerated by warm temperatures and a gradually increasing 77 daylength (Hänninen, 2016). Inter-annual variation of these three environmental 78 drivers is strongly correlated (e.g. a warm winter reduces chilling and increases heat 79 supply, and the earlier leaf-out associates with short daylength). As a result, the direct 80 effect of daylength on spring phenology and its eventual interactions with chilling and 81 the heat requirement remain unclear and highly debated (Chuine et al., 2010, Flynn & 82 Wolkovich, 2018, Körner & Basler, 2010, Laube et al., 2014, Zohner et al., 2016). In 83 this study, based on a large set of in situ phenology observations across Europe 84 (www.pep725.eu), we propose a framework to unravel the effect of daylength on leaf- 85 out phenology of temperate-zone deciduous trees and test the hypothesis that 86 daylength affects the leaf-out process by altering the heat requirement at any given 87 chilling accumulation. 88 4 89 We start by assuming that trees are characterized by an optimal, climate-dependent, 90 daylength (Figure. 1). Occasional late frost events give a competitive disadvantage to 91 individuals that leaf-out earlier than the species’ optimal daylength, while reduced 92 light harvesting gives a competitive disadvantage to individuals that leaf-out later than 93 the optimum period. As such, an optimal date of leaf-out exists for a given species, 94 determined by a trade-off between maximizing annual carbon and nutrient uptake to 95 ensure competitive advantage by lengthening the duration of the ground cover period 96 (earlier leaf out is preferred) and reducing the risk of late frost damage after leaf out 97 (later leaf out gives more security and is thus preferred (Allstadt et al., 2015, Liu et 98 al., 2018) (Figure. 1a). In the long term, carbon assimilation and competitiveness are 99 determined by the lowest of these two cost functions, and the optimal leaf out date 100 occurs where the minimum cost yields the highest carbon assimilation and 101 competitiveness. Experimental studies have revealed a nonlinear relation between 102 accumulated chilling and the heat required for leaves to flush (typically quantified as 103 growing degree day units, GDD) (Figure. 1b). Daylength is hypothesized to act as a 104 cue controlling the sensitivity of meristem cell growth to warm temperatures, thereby 105 altering the apparent relationship between chilling and GDD requirement. Their non- 106 linear relation becomes steeper when days are shorter than optimal (short daylength, 107 Figure. 1c) and less steep when days are longer than optimal (long daylength, Figure. 108 1d). 109 5 110 Two substantially different impacts of sub-optimal daylength on leaf-out date are thus 111 theoretically possible: (i) Shorter than optimal daylength reduces the temperature 112 sensitivity (increases the GDD requirement), thereby avoiding precocious leaf-out that 113 would increase the risk of frost damage, (i.e. the short daylength effect, Figure. 1c). 114 (ii) Longer than optimal daylength increases the temperature sensitivity (reduces the 115 GDD requirement), effectively avoiding belated leaf-out at a time when solar 116 radiation is high and thus ideal for photosynthesis (i.e. the long daylength effect, 117 Figure. 1d). 6 118 Figure 1. (a) Conceptual scheme depicting why an optimal daylength for tree leaf-out 119 exists within which the competitiveness of a species is maximized and how this is 120 realized. Tree competitiveness is increased by earlier start of the growing season, 121 which maximizes annual carbon and nutrient uptake and reduces that of the 122 neighbors, but is subject to a trade-off with avoiding the risk of late frost-induced 123 damage for which a later start of growing season is preferred. The green filled area 124 represents the leaf-out period that ensures the most secure, high competitiveness and 125 carbon uptake. This study provides evidence that the spring leaf-out process requires 126 less warm temperatures (lower heat requirement; GDD), and thus becomes more 127 temperature responsive, as daylength increases. As such, daylength aids in 128 constraining leaf-out within the optimal period in both cold and warm springs. (b) In 129 very warm springs, when leaf-out is early, trees minimize the advance of leaf-out 130 because their temperature sensitivity is low under short daylength. This is reflected in 131 an increased GDD requirement and results in trees being protected against late frost 132 events, i.e. the “short daylength effect”: (c) Under optimal daylength the GDD 133 required for leaf-out is mainly determined by the chilling accumulated during 134 endodormancy; (d) In very cold springs, when leaf-out is late, trees minimize the 135 delay of leaf-out because their temperature sensitivity becomes greater under 136 increasing daylength. This high temperature sensitivity is reflected in the reduced 137 GDD requirement, and protects trees against leafing-out too late, i.e. the “long 138 daylength effect”. 139 7 140 These two postulated consequences of daylength impacts have not been strongly 141 supported by empirical evidence. The short daylength effect in early spring, to our 142 knowledge, has not yet been empirically documented, whereas the long daylength 143 effect in late spring has been experimentally evidenced in earlier studies, albeit only 144 on cut twigs or saplings (Flynn & Wolkovich, 2018, Körner & Basler, 2010, Laube et 145 al., 2014, Malyshev et al., 2018) and not yet on mature trees. We therefore set out to 146 show the consequences of both a short and a long daylength on spring leaf-out of 147 mature trees of temperate deciduous tree species, to assess how widespread these two 148 effects are across these species, to quantify the sensitivities of the GDD requirement 149 to sub- and supra-optimal daylength and, last, to determine the relative importance of 150 chilling and daylength as controls of the leaf-out process. 151 152 Materials and methods 153 We tested the daylength effect on mature trees using data from 2377 sites of the 154 European phenological network (http://www.pep725.eu/) (Templ et al., 2018). The date 155 of leaf-out had been recorded at each site for at least 30 years between 1950 to 2016, 156 but in most cases observations were available for many more years. Six deciduous tree 157 species were selected (for which sufficient observational data were available): Fagus 158 sylvatica (beech), Aesculus hippocastanum (horse chestnut), Betula pendula (birch), 159 Fraxinus excelsior (ash), Quercus robur (oak) and Tilia cordata (lime). In total 509,284 160 individual observations from 12348 site-species combinations at 2377 sites were used. 161 The sites mainly occurred in moderate climates in Central Europe (Supplementary 8 162 Figure 1 and 2). The leaf-out dates were defined based on the BBCH code (Biologische 163 Bundesanstalt, Bundessortenamt und Chemische Industrie, BBCH = 11, first visible 164 leaf stalk) (Templ et al., 2018). We first determined the preseason length for each 165 species at each site as the period before leaf-out for which the partial correlation 166 coefficient between leaf-out and air temperature was highest (Fu et al, 2015). Using 167 this optimal preseason, we then calculated the GDD requirement for each species at 168 each site and in each year. We defined the GDD requirement as an integration of daily 169 mean temperature (Tt) above a temperature threshold (Tth) throughout the preseason 170 with the mean leaf-out dates as the end: 171 𝐺𝐷𝐷 = 𝑇𝑡 − 𝑇𝑡ℎ, 𝑖𝑓 𝑇𝑡 > 𝑇𝑡ℎ (1) 172 where Tth is the threshold temperature for GDD accumulation and Tt is the mean daily 173 temperature. We used a threshold Tth of 5 °C. We also tested a temperature threshold 174 of 0 °C, which produced very similar results. To best the robustness of the results, we 175 further calculated the GDD from the 1st December to the date of leaf-out for each 176 species at each site, and found very similar results (Supplementary Figure 3). We 177 therefore only report results using the threshold of 5 °C and the preseason 178 Chilling occurs at low, yet non-freezing temperatures and the number of days with 179 mean temperature between 0 and 5°C was suggested as a good proxy for chilling 180 accumulation, although inter-species variation in the chilling efficiency of different 181 temperatures is probably high. Chilling requirement is a physiological parameter that 182 corresponds to the amount of chilling needed to break endodormancy and enter the 183 ecodormancy. In the present study, chilling was calculated as the number of days (CD) 9 184 when daily temperature was between 0 and 5 °C from 1 September in the previous year 185 until the day of leaf-out. We also tested another approach, using 0°C and 10 °C as 186 temperature thresholds counting all days with mean temperatures between these 187 thresholds, which produced very similar results. Similar results were also obtained 188 when below-freezing temperatures were included, calculating as the number of days 189 when daily temperature below 5 °C (Supplementary Figure 4) or 7 °C (Supplementary 190 Figure 5), and similar results were obtained. We therefore only report the results based 191 on the chilling accumulation using the 0 °C - 5 °C temperature range. 192 Daylength at the day of leaf-out (DL) was calculated as a function of latitude and DOY: 0.8333π Lπ sin 193 24 DL=24- cos -1 [ 180 + sin 180 sin φ] (2) π Lπ cos 180 * cos φ 194 φ= sin-1(0.29795*cosθ) (3) 195 θ=0.2163108+2*tan-1 (0.9671396* tan(0.0086*(DOY-186))) (4) 196 where L is the latitude of the phenological site. 197 The daily mean air temperature at each site was derived from a gridded climatic data 198 set of daily mean temperature at 0.25ºspatial resolution (approximately 25 km, ERA- 199 WATCH) (Fu et al., 2014). 200 201 The sensitivity of GDD requirement to chilling and daylength 10 202 We calculated cumulative chilling, the GDD requirement and daylength at the day of 203 leaf-out for each year at each site. For each individual tree, we divided the data into 204 four subsets according to chilling accumulation, i.e. case 1: lowest chilling 205 accumulation: CD < CDmean – 1 standard deviation of CD (CDsd); case 2: low 206 chilling accumulation: CDmean-CDsd < CD < CDmean; case 3: high chilling 207 accumulation: CDmean < CD < CDmean + CDsd, and case 4: highest chilling 208 accumulation: CD> CDmean + CDsd. Within each CD subset, we subsequently 209 calculated the GDD requirement for three daylength conditions, i.e. leaf-out under 210 short-daylength conditions (DL < DLmean - 0.75 DLsd), under long-daylength 211 conditions (DL > DLmean + 0.75 DLsd) and under average-daylength conditions 212 (DLmean - 0.5 DLsd < DL < DLmean + 0.5 DLsd). The differences in mean GDD 213 requirement for leaf-out among the DL groups were tested using independent t-tests 214 for each chilling case and each species. Furthermore, we calculated the daylength 215 sensitivity of GDD as the slope of the linear regression between GDD and DL, and 216 then the average of the four daylength sensitivities was determined for each species at 217 each site. Using similar methodology, we divided the data into four subsets according 218 to DL for each individual tree, i.e. case 1: shortest daylength: DL < DLmean - DLsd; 219 case 2: short daylength: DLmean - DLsd < DL < DLmean; case 3: long daylength: 220 DLmean < DL < DLmean + DLsd, and case 4: longest daylength: DL> DLmean + 221 DLsd, and then calculated the chilling sensitivity of GDD for each species at each 222 site. To compare the relative importance of chilling versus that of daylength as 11 223 determinants of the GDD requirement for leaf-out, we first normalized the daylength 224 and chilling sensitivity, respectively, using a min-max normalization for each species: (Si − Smin) 225 Si = (5) (Smax − Smin) 226 Where Si is the daylength or chilling sensitivity at site i, Smin and Smax are the 227 minimum and maximum observed values of the daylength - or chilling sensitivity 228 across all trees of that species. Then, the mean and standard deviation of all 229 normalized sensitivities was calculated for each species. Histograms were used to 230 show the distribution of sensitivities across all trees for each species. 231 232 Results and discussion 233 In agreement with a multitude of previous studies (Cannell & Smith, 1983, Fu et al., 234 2016, Laube et al., 2014), we found that all studied tree species showed lower GDD 235 requirement under higher chilling conditions (Figure 2 and Supplementary Table 1). 236 Interestingly, we also observed that in all six species and within each CD group, the 237 GDD requirement for leaf-out was statistically significantly higher under short- than 238 average-daylength conditions, and significantly lower under long- than average- 239 daylength conditions (using paired t test, P<0.001, Figure 2, and Supplementary 240 Figure 6 and Supplementary Table 2 as an example at high chilling accumulation). 241 Contrasting results of the photoperiod effect were reported in experimental studies 242 (Heide 1993; Laube et al, 2014). However, almost all of these studies are based on 243 cuttings or saplings in manipulative experiments, and using constant day length rather 244 natural continuous changes in day length (e.g. Zohner et al, 2016; Laube et al., 12 245 2014)., and young trees often behave opportunistically and exhibit earlier leaf-out 246 than mature trees of the same species. Using trees of different ontogenetic stages 247 might thus explain part of the differences among previous studies. In the present 248 study, we selected six species that belong to five families (Fagaceae, Betulaceae, 249 Malaceae, Sapindaceae, Oleaceae), some of which phylogenetically quite distinct 250 (Supplementary Figure 4). Because every tested species (all six species for which 251 sufficient observations were available) exhibited very similar daylength responses, we 252 postulate the widespread existence of a daylength effect among temperate zone 253 deciduous tree species, at least among European temperate-zone tree species. In line 254 with our findings, a recent study found consistent daylength effects on leaf-out 255 phenology exists across 28 woody species in a North American temperate forest 256 (Flynn & Wolkovich, 2018). 257 For each individual tree, we calculated the sensitivity of the GDD requirement for 258 leaf-out to changes in daylength. On average across all species and averaged over four 259 different chilling intensities, compared to the GDD requirement under average 260 daylength conditions, each one-hour decrease in daylength (comparable to the 261 observed inter-annual range) increased GDD by 37℃-days, i.e. by 26% 262 (Supplementary Figure 8a and b), while a 1-h increase in daylength decreased the 263 GDD requirement by 31℃-days, i.e. by 22% (Supplementary Figures 8a and b). We 264 further compared the relative sensitivity of the GDD requirement for leaf-out to 265 chilling and daylength (comparing the GDD response to one standard deviation of the 13 266 observed variation in either chilling or daylength), and observed species-specific 267 sensitivity differences among the six study species (Figure. 3). 268 Figure 2. Dependence of GDD requirement of leaf-out on daylength under constant 269 chilling conditions for six tree species. Dependencies are shown for four different 270 chilling intensities (see Methods). 271 272 In detail, sensitivity to daylength was larger than the sensitivity to chilling in four out 273 of six species: Betula pendula, Aesculus hippocastanum, Tilia cordata and Fraxinus 274 excelsior, while no difference was detected in Quercus robur. Fagus sylvatica also 275 exhibited a pronounced sensitivity to daylength, but its sensitivity to chilling was even 276 greater (Figure. 3). Fagus sylvatica is indeed known as a highly chilling-sensitive 277 species (Kramer, 1994, Malyshev et al., 2018). Our findings thus confirm that 278 daylength is an important co-regulator of leaf-out in mature temperate deciduous trees 14 279 (Flynn & Wolkovich, 2018, Körner & Basler, 2010), and further suggest that 280 daylength likely affects the leaf-out process indirectly by altering the non-linear 281 relationship between chilling and GDD requirement. These results also support the 282 hypothesis that the shorter daylength due to earlier leaf-out in spring contributes to the 283 declining apparent temperature sensitivity of leaf-out in European temperate 284 deciduous trees (Fu et al. 2015). These mechanisms are conceptualized in Figure. 4. 285 With climate warming, the GDD requirement increases due to reduced chilling, but 286 GDD supply increases more (Figure. 4, panel a). As a result, GDD supply equals 287 GDD requirement earlier in the year (visualized as an excess GDD supply in Figure 4, 288 panel a), which drives earlier leaf-out. The associated shorter daylength, however, 289 further increases the GDD requirement and thereby restricts the advance of leaf-out 290 (Figure. 4, panel b) and reduces the temperature sensitivity of leaf-out with climate 291 warming. 15 292 Figure 3. Histograms of the Growing Degree Day (GDD) sensitivity to changes in 293 daylength (DL, in red) and in chilling (CD, in blue) across all individual trees of six 294 deciduous tree species. Sensitivity was calculated as the change in GDD per one 295 standard deviation in the observed DL and CD, respectively, and is therefore coined 296 ‘normalized sensitivity’. mean sensitivities and standard deviations (in brackets) are 297 provided. 16 298 Figure. 4. Conceptual scheme summarizing how daylength helps deciduous trees to 299 leaf-out within or close to the optimal period. (a) Leaf-out occurs when the supply of 300 warm temperatures (GDD supply: accumulated daily growing degree days, GDD; 301 black line) equals the GDD requirement (physiological parameter to trigger leaf-out). 302 Note that the X axes indicate the climate from cold to warm spring. The more chilling 303 is accumulated, the lower the GDD requirement (green dashed line; for simplicity 304 reasons we here assume a linear relation). In cold springs, the deficit in GDD supply 305 drives a delay in leaf-out date (blue areas in panels a and b). (b) The associated 306 increasing daylength, however, increases the temperature sensitivity and thereby 307 causes a decline in the GDD requirement, with leaf-out occurring when GDD supply 17 308 equals the declining GDD requirement. In contrast, in warm springs (red areas in 309 panels a and b), GDD supply typically exceeds the chilling-induced GDD requirement 310 earlier in the year, but the short daylength earlier in spring induces a low 311 temperature sensitivity and thereby an increased GDD requirement, which minimizes 312 the warming-induced advance of leaf-out. Note that the X axes indicate the leaf-out 313 timing from early to late. 314 315 Daylength thus acts as an environmental cue, counteracting the advancing impact of 316 global warming, and helping trees to leaf-out close to their optimal date. As daylength 317 increases from early to late spring, we observed that the GDD requirement also 318 decreases under similar chilling conditions. Similar responses, supporting our 319 findings, were previously reported in experimental studies using cuttings, although the 320 daylength difference among treatments was very large in these studies (Zohner et al., 321 2016). There are also other studies that reported a decreased photoperiod effect with 322 increasing chilling accumulation (Laube et al, 2014; Hänninen 2016), which may be 323 because the increasing chilling ensures leaf out at the optimal daylength, when the 324 photoperiod effect is minimal. We observed that the GDD requirement decreases 325 consistently from very short to very long daylength, suggesting that one single 326 physiological mechanism may suffice to explain the protective effect of daylength 327 against both early as well as late leaf-out. We speculate that daylength directly 328 impacts on the temperature sensitivity of cell development to warming. By keeping 329 the calculation of GDD constant with daylength, the increase in temperature 18 330 sensitivity with increasing daylength is mathematically translated into a reduced GDD 331 requirement. However, we acknowledge that the heat signal required by the 332 meristems to initiate leaf-out may not be directly altered by daylength, but that the 333 heat signal reception may become more efficient with increasing daylength. Our data, 334 unfortunately, do not allow unraveling the underlying physiological mechanism. 335 336 We further speculate that the daylength control over the GDD requirement depends on 337 the start date of the ecodormancy phase relative to the date when optimal daylength 338 thresholds are reached or passed. When ecodormancy begins late relative to the 339 optimum daylength thresholds, the temperature sensitivity of cell development is 340 elevated, resulting in reduced GDD requirement to force leaf-out. In contrast, when 341 ecodormancy starts earlier than the target daylength threshold, the temperature 342 sensitivity of cell development may remain low, but not zero, until the date when the 343 optimal daylength threshold is passed. The starting date of ecodormancy, however, 344 cannot easily be determined empirically and is therefore typically ignored in 345 phenology studies, explaining why the relation between daylength and the start of 346 ecodormancy remains poorly understood (Chuine et al., 2016), despite their 347 importance for pushing the field beyond the state of the art (Hänninen, 2016). To our 348 knowledge, neither the start date of ecodormancy, i.e. the start date of GDD 349 accumulation, nor the optimal daylength thresholds have been well studied (but see 350 Chuine et al, 2016). As long as dormancy remains poorly understood, also the 351 estimations of chilling and forcing units remain uncertain. For example, the duration 19 352 of the chilling accumulation period and the start date of the heat accumulation period, 353 as well as their interactions are still unclear. Similarly, the optimal temperature ranges 354 for chilling accumulation and the temperature threshold above which GDD’s start to 355 accumulate, as well as the length of GDD accumulation are poorly understood. 356 Different assumptions can, however, lead to contrasting and sometimes illogical 357 results. For example, an increased GDD requirement is obtained when leaf out is very 358 late and the GDD is calculated over a fixed number of days prior to leaf out (see 359 Supplementary Figure 9). Studies focusing on ecophysiological experiments are thus 360 urgently needed to fully understand spring phenology and enable the development of 361 reliable phenology synthesis studies and –models (Chuine & Régnière, 2017, 362 Hänninen et al., 2019). 363 364 Climate warming-induced spring phenology advances substantially alter regional and 365 global biogeochemical cycles and climate systems (Forzieri et al., 2017, Myneni et 366 al., 1997, Peñuelas & Filella, 2009). However, as the daylength effect reduces the 367 temperature sensitivity of leaf-out in warmer years, slowing down the advancing rate 368 of leaf-out, it thereby also reduces the warming-induced extension of ground cover, 369 and the carbon uptake, evapotranspiration and albedo. This study found that all 370 investigated temperate-zone deciduous tree species (the six species for which 371 sufficient observational data were available) use daylength as a signal to help ensuring 372 that leaf-out occurs close to a species-dependent optimal time of the year, by 20 373 increasing the GDD requirement for leaf-out when daylength is too short, and 374 reducing the GDD requirement as daylength becomes too long. 375 376 References 377 Allstadt AJ, Vavrus SJ, Heglund PJ, Pidgeon AM, Thogmartin WE, Radeloff VC (2015) Spring 378 plant phenology and false springs in the conterminous US during the 21st century. 379 Environmental Research Letters, 10, 104008. 380 Cannell M, Smith R (1983) Thermal time, chill days and prediction of budburst in Picea 381 sitchensis. Journal of Applied Ecology, 951-963. 382 Chuine I (2010) Why does phenology drive species distribution? Philosophical Transactions 383 of the Royal Society B: Biological Sciences, 365, 3149-3160. 384 Chuine I, Bonhomme M, Legave JM, García De Cortázar‐Atauri I, Charrier G, Lacointe A, Amé 385 glio T (2016) Can phenological models predict tree phenology accurately in the 386 future? The unrevealed hurdle of endodormancy break. Global Change Biology, 22, 387 3444-3460. 388 Chuine I, Morin X, Bugmann H (2010) Warming, Photoperiods, and Tree Phenology. Science, 389 329, 277-278. 390 Chuine I, Régnière J (2017) Process-Based Models of Phenology for Plants and Animals. 391 Annual Review of Ecology, Evolution, and Systematics, 48, 159-182. 392 Elzinga JA, Atlan A, Biere A, Gigord L, Weis AE, Bernasconi G. 2007. Time after time: 393 flowering phenology and biotic interactions. Trends in Ecology and Evolution, 22, 394 432-439. 395 Flynn DFB, Wolkovich EM (2018) Temperature and photoperiod drive spring phenology 396 across all species in a temperate forest community. New Phytologist, 219, 1353- 397 1362. 398 Forzieri G, Alkama R, Miralles DG, Cescatti A (2017) Satellites reveal contrasting responses of 399 regional climate to the widespread greening of Earth. Science, 356, 1180-1184. 400 Fu YH, Campioli M, Vitasse Y et al. (2014) Variation in leaf flushing date influences autumnal 401 senescence and next year's flushing date in two temperate tree species. Proceedings 402 of the National Academy of Sciences of the United States of America, 111, 7355- 403 7360. 404 Fu YH, Liu Y, De Boeck HJ et al. (2016) Three times greater weight of daytime than of night‐ 405 time temperature on leaf unfolding phenology in temperate trees. New Phytologist, 406 212, 590-597. 21 407 Fu YH, Zhao HF, Piao SL et al. (2015) Declining global warming effects on the phenology of 408 spring leaf unfolding. Nature, 526, 104-107. 409 Hänninen H (2016) Boreal and temperate trees in a changing climate. Netherlands, Springer 410 Business Media, Dordrecht. 411 Hänninen H, Kramer K, Tanino K, Zhang R, Wu J, Fu YH (2019) Experiments are necessary in 412 process-based tree phenology modelling. Trends in Plant Science, 24, 199-209. 413 Heide O.M. (1993) Day length and thermal time responses of budburst during dormancy 414 release in some northern deciduous trees, Physiologia plantarum, 88, 531–540. 415 Keenan TF, Gray J, Friedl MA et al. (2014) Net carbon uptake has increased through 416 warming-induced changes in temperate forest phenology. Nature Climate Change, 4, 417 598. 418 Körner C, Basler D (2010) Phenology Under Global Warming. Science, 327, 1461-1462. 419 Körner C, Basler D, Hoch G, Kollas C, Lenz A, Randin CF, Vitasse Y, Zimmermann NE (2016) 420 Where, why and how? Explaining the low-temperature range limits of temperate 421 tree species. Journal of Ecology, 104, 1076-1088 422 Kramer K (1994) Selecting a model to predict the onset of growth of Fagus sylvatica. Journal 423 of Applied Ecology, 31, 172-181. 424 Lang GA (1987) Dormancy: a new universal terminology. HortScience, 22, 817-820. 425 Laube J, Sparks TH, Estrella N, Höfler J, Ankerst DP, Menzel A (2014) Chilling outweighs 426 photoperiod in preventing precocious spring development. Global Change Biology, 427 20, 170-182. 428 Lenz A, Hoch G, Körner C, Vitasse Y (2016) Convergence of leaf-out towards minimum risk of 429 freezing damage in temperate trees, Functional Ecology, 9, 1480-1490. 430 Liu Q, Piao S, Janssens IA et al. (2018) Extension of the growing season increases vegetation 431 exposure to frost. Nature Communications, 9, 426. 432 Malyshev AV, Henry HA, Bolte A, Khan MaA, Kreyling J (2018) Temporal photoperiod 433 sensitivity and forcing requirements for budburst in temperate tree seedlings. 434 Agricultural and Forest Meteorology, 248, 82-90. 435 Menzel A, Sparks TH, Estrella N et al. (2006) European phenological response to climate 436 change matches the warming pattern. Global Change Biology, 12, 1969-1976. 437 Myneni RB, Keeling C, Tucker CJ, Asrar G, Nemani RR (1997) Increased plant growth in the 438 northern high latitudes from 1981 to 1991. Nature, 386, 698. 439 Parmesan C, Yohe G (2003) A globally coherent fingerprint of climate change impacts across 440 natural systems. Nature, 421, 37-42. 441 Peñuelas J, Filella I (2001) Responses to a warming world. Science, 294, 793-795. 22 442 Peñuelas J, Filella I (2009) Phenology feedbacks on climate change. Science, 324, 887-888. 443 Piao S, Liu Z, Wang T et al. (2017) Weakening temperature control on the interannual 444 variations of spring carbon uptake across northern lands. Nature Climate Change, 7, 445 359. 446 Richardson AD, Anderson RS, Arain MA et al. (2012) Terrestrial biosphere models need 447 better representation of vegetation phenology: results from the North American 448 Carbon Program Site Synthesis. Global Change Biology, 18, 566-584. 449 Templ B, Koch E, Bolmgren K, Ungersbock M, Paul A, Scheifinger H (2018) Pan European 450 Phenological database (PEP725): a single point of access for European data. 451 International journal of biometeorology, 62, 1-5. 452 Vitasse Y, Basler D (2013) What role for photoperiod in the bud burst phenology of European 453 beech. European Journal of Forest Research, 132, 1-8. 454 Vitasse Y (2013) Ontogenic changes rather than difference in temperature cause understory 455 trees to leaf out earlier. New phytologist, 198, 149-155. 456 Zohner CM, Benito BM, Svenning J-C, Renner SS (2016) Day length unlikely to constrain 457 climate-driven shifts in leaf-out times of northern woody plants. Nature Climate 458 Change, 6, 1120-1123. 459 Zohner, C. M., Mo, L., & Renner, S. S. (2018). Global warming reduces leaf-out and 460 flowering synchrony among individuals. eLife, 7, e40214. 23 461 Acknowledgments 462 This study was supported support by the National Key Research and Development 463 Program of China (2017YFA06036001), the General program of National Nature 464 science foundation of China (No. 31770516) and the 111 Project (B18006) and 465 Fundamental Research Funds for the Central Universities (2018EYT05). Ivan A 466 Janssens and Josep Peñuelas acknowledge support from the European Research 467 Council through Synergy grant ERC-2013-SyG-610028 “IMBALANCE-P”. Ivan A 468 Janssens acknowledges support from the Research Council of the University of 469 Antwerp (Methusalem) and the “ECOPROPHET” project funded by BELSPO (Belgian 470 Science Policy Office; Contract SR/00/334). Constantin M. Zohner acknowledges 471 support from the ETH Zurich Postdoctoral Fellowship Program and the Crowther lab. 472 Our deepest gratitude goes to Dr. Christian Körner and the anonymous reviewers for 473 their careful work and thoughtful suggestions that have helped improve this paper 474 substantially. The authors gratefully acknowledge all members of the PEP725 project 475 for providing the phenological data. Y.H.F. and I.A.J. designed the research and drafted 476 the paper; Y.H.F. and X. Z performed the analysis and all authors contributed to the 477 interpretation of the results and to the text. 24 478 Supplementary Materials 479 Supplementary Legends: 480 Supplementary Table1. The mean GDD requirement of leaf-out in different 481 daylength conditions, i.e. DLearly, short daylength effect, DLmid, average length in 482 daylength and DLlate, long daylength effect, under same chilling accumulation 483 conditions (case) for six tree species. Four chilling condition were studies, e.g. 484 case01, chilling lowest; case02, chilling low, case03, chilling high and case04, 485 chilling highest (see Methods for details). n indicates the number of trees. 486 Supplementary Table 2. The difference in mean GDD requirement of leaf-out 487 between the daylength treatments, e.g. under short daylength: GDDshort DL; under 488 optimal daylength: GDDavgDL; under long daylength: GDDlongDL) under the same 489 chilling conditions in the low chilling accumulation group. 490 Supplementary figure 1. Distribution of the selected phenological sites. 491 Supplementary figure 2. (a) The climate of selected phenological sites. One circle 492 indicates one site. (b) and (c) the annual temperature and precipitation across all 493 selected sites. 494 Supplementary figure 3. Dependence of GDD requirement of leaf-out on daylength 495 under constant chilling conditions for six tree species. Dependencies are shown for four 496 different chilling intensities (see Methods). The GDD was calculated as daily sum of 497 the daily temperature above 50C over the period from 1st September to the date of leaf- 498 out 499 Supplementary figure 4. Dependence of GDD requirement of leaf-out on daylength 500 under constant chilling conditions for six tree species. Dependencies are shown for 501 four different chilling intensities (see Methods). The chilling was calculated as day 25 502 when the daily temperature below 50C over the period from 1st September to the date 503 of leaf-out. 504 Supplementary figure 5. Dependence of GDD requirement of leaf-out on daylength 505 under constant chilling conditions for six tree species. Dependencies are shown for 506 four different chilling intensities (see Methods). The chilling was calculated as day 507 when the daily temperature below 70C over the period from 1st September to the date 508 of leaf-out. 509 Supplementary figure 6. Mean GDD requirement of leaf-out in the daylength 510 treatments (left panels) and under the same chilling (days) conditions (right panels) in 511 the low chilling accumulation group. The numbers in the right panels are the mean 512 chilling days in each of the three daylength regimes in the left panels (using the same 513 colors). The GDD requirement was calculated as explained in Methods. Letters, i.e. a, 514 b and c, indicate statistically significant differences (at P < 0.001). 515 Supplementary figure 7. Phylogenetic tree of European forest tree species and the 516 family names were provided with different color. The figure was modified from study 517 of Sardans et al., 2015. The species that selected in the present study were marked with 518 pink boxes. 519 Supplementary figure 8. Changes in absolute and relative values of GDD requirement 520 for spring leaf-out (across all chilling conditions) under one hour shorter (black) and 521 longer (white) daylengths for each and all of the six studied species. 522 Supplementary figure 9. Dependence of GDD requirement of leaf-out on daylength 523 under constant chilling conditions for six tree species. The GDD was calculated using 524 a fixed length prior to leaf-out for each species at each site. Dependencies are shown 525 for four different chilling intensities (see Methods). 26 526 Supplementary Table 1. The mean GDD requirement of leaf-out in different daylength 527 conditions, i.e. DLearly, short daylength effect, DLmid, average length in daylength and 528 DLlate, long daylength effect, under same chilling accumulation conditions (case) for six tree 529 species. Four chilling condition were studies, e.g. case01, chilling lowest; case02, chilling 530 low, case03, chilling high and case04, chilling highest (see Methods for details). n indicates 531 the number of trees. 532 Chilling DLearly DLmid DLlate DLearly DLmid DLlate Species (n) conditions (hours) (hours) (hours) mean std mean std mean std case01 12.9 13.5 14.0 160.1 60.6 131.3 52.5 85.9 48.1 Aesculus case02 13.1 13.6 14.1 138.9 48.9 107.5 39.3 74.9 36.3 hippocastanum case03 13.0 13.6 14.1 133.9 44.9 99.4 35.6 72.3 34.0 (2069) case04 13.2 13.7 14.1 133.0 49.8 81.6 35.7 66.4 34.1 case01 12.8 13.4 14.0 153.3 60.4 131 50.9 79.8 44.4 Betula pendula case02 13.1 13.7 14.1 134.0 48.9 77.0 32.4 60.8 29.4 (2052) case03 13.1 13.6 14.1 128.9 42.2 94.2 32.6 66.2 30.3 case04 13 13.6 14.1 132.6 45.3 103.4 35.7 68.9 32.2 case01 13.3 13.8 14.2 182.4 77.9 146.8 63.2 113.9 55.2 Fagus sylvatica case02 13.4 13.9 14.3 162.6 59.5 127.4 49.4 96.2 44.7 (1588) case03 13.5 13.9 14.3 159.8 54.5 122.0 43.8 92.5 42.3 case04 13.5 14 14.3 143.7 60.2 102.8 42.8 85.3 41.8 case01 13.7 14.2 14.6 229.0 87.1 199.4 70.4 164.5 62.9 Fraxinus excelsior case02 13.7 14.2 14.6 217.7 67.1 187.1 58.7 146.7 55.0 (1044) case03 13.8 14.3 14.6 214.6 65.9 179.6 55.5 144.7 53.7 case04 13.9 14.3 14.6 181.4 65.6 156.2 54.4 135.3 54.7 case01 13.5 14.1 14.5 215.9 72.9 179.5 58 144.5 55.3 Quercus robur case02 13.6 14.1 14.5 203.3 58.0 159.6 46.9 123.5 43.6 (1686) case03 13.7 14.2 14.5 201.5 62.1 163.7 48.7 124.3 46.8 case04 13.9 14.2 14.6 176.3 60.2 135.8 46.8 113.2 46 case01 13.2 13.9 14.3 231.0 69.4 179.4 61.4 140.0 57.4 Tilia cordata case02 13.4 14.0 14.4 214.3 62.3 159.7 51.4 123.9 46.9 (438) case03 13.5 14.0 14.4 195.4 55.3 153.2 46.2 122.8 46.5 case04 13.5 14.0 14.3 191.7 61.2 138.6 49.2 116.9 51.2 27 533 Supplementary Table 2. The difference in mean GDD requirement of leaf-out between the 534 daylength treatments, e.g. under short daylength: GDDshort DL; under optimal daylength: 535 GDDavgDL; under long daylength: GDDlongDL) under the same chilling conditions in the 536 low chilling accumulation group. 537 GDD shortDL vs. GDDavgDL GDD longDL vs. GDDavgDL Species t P df t P df Aesculus hippocastanum 32.986 <0.001 1928 -45.943 <0.001 1928 Betula pendula 33.118 <0.001 970 -32.853 <0.001 970 Fagus sylvatica 31.793 <0.001 1491 -36.568 <0.001 1491 Fraxinus excelsior 19.682 <0.001 970 -32.853 <0.001 970 Quercus robur 33.999 <0.001 1569 -40.378 <0.001 1569 Tilia cordata 22.646 <0.001 396 -18.238 <0.001 396 28 538 Supplementary figure 1. Distribution of the selected phenological sites. 29 539 Supplementary figure 2. (a) The climate of selected phenological sites. One circle 540 indicates one site. (b) and (c) the annual temperature and precipitation across all 541 selected sites. 30 542 Supplementary figure 3. Dependence of GDD requirement of leaf-out on daylength 543 under constant chilling conditions for six tree species. Dependencies are shown for 544 four different chilling intensities (see Methods). The chilling was calculated as day 545 when the daily temperature below 50C over the period from 1st September to the date 546 of leaf-out. 547 31 548 Supplementary figure 4. Dependence of GDD requirement of leaf-out on daylength 549 under constant chilling conditions for six tree species. Dependencies are shown for 550 four different chilling intensities (see Methods). The chilling was calculated as day 551 when the daily temperature below 70C over the period from 1st September to the date 552 of leaf-out. 553 32 554 Supplementary figure 4. Dependence of GDD requirement of leaf-out on daylength 555 under constant chilling conditions for six tree species. Dependencies are shown for 556 four different chilling intensities (see Methods). The GDD was calculated as daily 557 sum of the daily temperature above 50C over the period from 1st December to the date 558 of leaf-out 559 33 560 Supplementary figure 6. Mean GDD requirement of leaf-out in the daylength treatments 561 (left panels) and under the same chilling (days) conditions (right panels) in the low chilling 562 accumulation group. The numbers in the right panels are the mean chilling days in each of the 563 three daylength regimes in the left panels (using the same colors). The GDD requirement was 564 calculated as explained in Methods. Letters, i.e. a, b and c, indicate statistically significant 565 differences (at P < 0.001). 34 566 Supplementary figure 7. Phylogenetic tree of European forest tree species and the family 567 names were provided with different color. The figure was modified from study of Sardans et 568 al., 2015. The species that selected in the present study were marked with boxes. 35 569 Supplementary figure 8. Changes in absolute (a) and relative (b) values of GDD 570 requirement for spring leaf-out (across all chilling conditions) under one hour shorter (black) 571 and longer (white) daylength for each and all of the six studied species. 36 572 Supplementary figure 9. The distribution of normalized daylength (DL, in red) and chilling 573 (CD, in blue) sensitivity of GDD requirement for leaf-out for all study species. The histograms 574 show the distribution across all sites and the mean sensitivities and standard deviations (in 575 brackets) are provided. The GDD was calculated from the 1st Jan to the date of leaf-out for each 576 year of each species at each site. 37