This is the accepted version of the article: Bogdziewicz, Michał; Szymkowiak, Jakub; Bonal, Raúl; [et al.]. «What drives phenological synchrony? Warm springs advance and desynchronize flowering in oaks». Agricultural and forest meteorology, Vol. 294 (Nov. 2020), art. 108140. DOI 10.1016/j.agrformet.2020.108140 This version is avaible at https://ddd.uab.cat/record/232216 under the terms of the license 1 What drives phenological synchrony? Warm springs advance and desynchronize flowering in 2 oaks 3 4 Running head: Warm springs desynchronize flowering 5 6 Bogdziewicz M1,2*, Szymkowiak J3, Bonal R4, Hacket-Pain A5, Espelta JM2, Pesendorfer M6, 7 Grewling L7, Kasprzyk I8, Belmonte J9, Kluska K8, De Linares C9,10, Penuelas J2,11, 8 Fernandez-Martinez M12 9 10 1 Department of Systematic Zoology, Faculty of Biology, Adam Mickiewicz University, 11 Poznań, Poland 12 2 CREAF, Cerdanyola del Vallès 08193, Catalonia, Spain 13 3 Population Ecology Lab, Faculty of Biology, Adam Mickiewicz University, Poznań, Poland 14 4 Forest Research Group, INDEHESA, University of Extremadura, Plasencia, Spain 15 5 Department of Geography and Planning, School of Environmental Sciences, University of 16 Liverpool, Liverpool, UK 17 6 Institute of Forest Ecology, Department of Forest and Soil Sciences, University of Natural 18 Resources and Life Sciences, Vienna, Austria 19 7 Laboratory of Aerobiology, Department of Systematic and Environmental Botany, Faculty 20 of Biology, Adam Mickiewicz University, Poznań, Poland 21 8 Department of Environmental Monitoring, Institute of Biology and Biotechnology, 22 University of Rzeszow, Zelwerowicza 4, 35-601 Rzeszów, Poland 23 9 Institute of Environmental Science and Technology, Universitat Autònoma de Barcelona, 24 Cerdanyola del Vallès 08193, Catalonia, Spain 25 10 Department of Animal Biology, Plant Biology and Ecology, Universitat Autònoma de 26 Barcelona, Cerdanyola del Vallès 08193, Catalonia, Spain 27 11 CSIC, Global Ecology Unit, Bellaterra 08193, Catalonia, Spain 28 12 PLECO (Plants and Ecosystems), Department of Biology, University of Antwerp, 2610 29 Wilrijk, Belgium 30 *corresponding author: michalbogdziewicz@gmail.com 31 32 Word count: 3692; Figures: 5; Tables: 1; Online Supplement: Yes 1 33 Summary 34 Annually variable and synchronous seed production, or masting, is often correlated with 35 environmental factors and in oaks involves differential pollination success that depends on 36 phenological synchrony in flowering. The synchronization of phenology of flowering was 37 thought to be driven by temperature during flowering (microclimatic hypothesis). We tested 38 an alternative, whereby phenological synchronization is driven by the timing of the onset of 39 flowering (photoperiod-sensitivity hypothesis). This hypothesis assumes that flowering 40 synchrony is driven by interaction between daylength and temperature, and individual 41 variation in sensitivity to daylength as a phenological cue. We used long-term (23-26 years) 42 records of airborne pollen in Quercus robur, Q. petraea, Q. ilex, and Q. humilis. Late pollen 43 seasons were short, as predicted by photoperiod-sensitivity hypothesis. The onset of pollen 44 seasons was delayed as preseason temperatures cooled over the last three decades at our 45 Mediterranean sites, which was paralleled by shortening in pollen seasons, providing 46 additional support for the photoperiod-sensitivity hypothesis. Global warming under the 47 microclimatic hypothesis is predicted to lead to less frequent reproductive failures and thus 48 decreased variability and synchrony of mast seeding. In contrast, warming under the 49 photoperiod-sensitivity hypothesis should advance the onset of and desynchronize flowering, 50 a pattern supported by our data. This pattern suggests that global warming will lead to more 51 frequent vetoes and more stochastic and variable patterns of oak reproduction. 52 53 Key words: flowering onset, flowering synchrony, global warming, microclimatic variation, 54 phenology, photoperiod sensitivity, pollen 55 56 2 57 Introduction 58 Masting, or mast seeding, is the highly variable and synchronized production of seeds 59 by plant populations (Crone and Rapp, 2014; Kelly, 1994), that is a widespread reproductive 60 strategy in perennial plants (Fernández-Martínez et al., 2019; Kelly and Sork, 2002; 61 Tanentzap and Monks, 2018). The variable allocation of resources associated with masting 62 affects plant growth, the population dynamics of plants and animals, macronutrient cycling, 63 carbon stocks, forest regeneration, future species composition, and risk of disease in humans 64 (Bogdziewicz et al., 2016; Clark et al., 2019; Hacket‐Pain et al., 2018; Ostfeld and Keesing, 65 2000). Masting functionally depends on economies of scale, i.e. individual plants that 66 reproduce when other plants are also flowering have lower costs for each surviving offspring 67 (Bogdziewicz et al., 2020c; Kelly, 1994). The proximate mechanisms driving masting, i.e. 68 how annual variability in seed production and synchronization among individuals happen, 69 remain poorly understood (Bogdziewicz et al., 2020a; Pearse et al., 2016). 70 Masting is frequently correlated with environmental cues such as temperature or 71 rainfall (Bogdziewicz et al., 2020b; Koenig and Knops, 2000; Pérez‐Ramos et al., 2015; 72 Schermer et al., 2019), but little attention has been paid to the mechanisms driving these 73 connections (Bogdziewicz et al., 2017a; Kelly et al., 2013; Koenig et al., 2015). Flowering 74 and pollination dynamics are hypothesized to provide the mechanistic link for the observed 75 relationship between weather and population-level seed production (Hanley et al., 2019; 76 Koenig and Knops, 2013; Nussbaumer et al., 2018). The phenological synchrony hypothesis 77 proposes that weather drives pollen limitation by determining the annual differences in the 78 synchrony of flowering within a population (Koenig et al., 2015). The original formulation of 79 the phenological synchrony hypothesis states that the phenology of flowering is driven by 80 temperature during flowering (microclimatic hypothesis). Specifically, flowering is more 81 synchronous in warm years when microclimatic conditions are more homogeneous, 82 conditions that lead to lower variability in flowering time (Koenig et al., 2015). We tested an 83 alternative hypothesis, whereby phenological synchronization is driven by the timing of the 84 onset of flowering. 85 This alternative, known as the photoperiod-sensitivity hypothesis, assumes that 86 flowering synchrony is driven by interaction between daylength and temperature, and 87 individual variation in sensitivity to daylength as a phenological cue (Bogdziewicz et al., 88 2020a). Days in cold years are already long at the onset of warmer spring temperatures, 89 reducing the effect of plant daylength sensitivity on flowering time and increasing population- 90 level flowering synchrony (Fu et al., 2019c; Zohner et al., 2018). In contrast, days in warm 3 91 years are still short at the onset of warmer spring temperatures, preventing plants sensitive to 92 daylength from flushing and flowering. Leaf-out and flowering in warm years thus advance 93 only in individuals insensitive to daylength, which extends population-level flowering time 94 and reduces synchrony. Experiments have confirmed large intraspecific variation in daylength 95 sensitivity within populations of some species (Zohner et al., 2018). This response may 96 consequently decrease the population-level synchrony of flowering when days are short 97 (warm years, early spring) and increase the synchrony of flowering in late springs (cold years, 98 late spring). 99 These two processes linking variation in weather with pollen limitation lead to 100 contrasting predictions of the effects of global warming on flowering synchrony in plants and 101 thus their reproductive success. Global warming under the microclimatic hypothesis should 102 increase the frequency of years favorable for pollination, because warmer weather should 103 generally lead to more frequent highly synchronized flowering, which in turn should reduce 104 the interannual variability of seed production (Koenig et al., 2015). In contrast, global 105 warming under the photoperiod-sensitivity hypothesis would lead to a lower frequency of 106 years favorable for pollination, because warmer years would generally advance flowering in 107 some individuals and thus desynchronize it at a population level. More frequent failures of 108 pollination will increase the stochasticity and interannual variability of seed production 109 (Koenig et al., 2015; Schermer et al., 2020). A better understanding of the ecophysiological 110 processes controlling flowering phenology is thus essential for improving our understanding 111 of the responses of trees and forests to the ongoing climate change. 112 We used long-term (23-26 years) records of airborne pollen concentrations from 113 several locations in Europe to test the hypothesis that flowering synchrony in four oak species 114 (Quercus robur, Q. petraea, Q. ilex, and Q. humilis) is determined by the timing of flowering 115 onset. We used airborne pollen concentrations data as proxy of flowering times, using the 116 length of the pollen season as an index of phenological synchrony in pollen release within the 117 population. The link between acorn production and flowering synchrony, as measured by the 118 length of the pollen season, has already been established in the oak populations we study here 119 (Bogdziewicz et al., 2017a; Bogdziewicz et al., 2017b). Following relationships should hold 120 under the photoperiod sensitivity hypothesis. First, the advanced onset of pollen production 121 should lead to long pollen seasons. An advanced onset of pollen seasons should be in turn 122 driven by high temperatures, as established by many studies of this topic (Fu et al., 2015; Liu 123 et al., 2016; Peaucelle et al., 2019; Zohner et al., 2018). Second, we explored the trends 124 (1994-2019) in timing of pollen seasons onset and their length. Here, advances in the timing 4 125 of pollen seasons onset driven by temperature should reduce synchrony within populations, 126 leading to longer pollen seasons. 127 128 Methods 129 Study species 130 We investigated the relationships between temperature, calendar day, and airborne 131 pollen concentrations for four oak species that are common throughout Europe. We present 132 data for populations of two temperate oak species in Poland, Q. petraea and Q. robur, and 133 two Mediterranean oak species in Spain, Q. ilex and Q. humilis. All species are large 134 broadleaved trees that often dominate forests in their native ranges. They are pollinated by 135 wind and generally self-incompatible. Flowers are produced and fertilized in spring and 136 develop into fruits in the same year in which they were pollinated. Our previous study found 137 that synchronous flowering was positively correlated with seed production in these 138 populations (Bogdziewicz et al., 2017a; Bogdziewicz et al., 2017b). 139 140 Pollen and meteorological data 141 Data for airborne pollen for Q. petraea and Q. robur were collected at two pollen- 142 monitoring sites in Poland for 1997-2019: Poznań (52°25′N, 16°53′E) and Rzeszów (50°01′N 143 22°02′E). Data for Q. ilex and Q humilis were collected at three sampling sites in Spain: 144 Barcelona (41°23′N, 2°90′E) for 1994-2019, Bellaterra (41°30′N, 2°60′E) for 1994-2019, and 145 Huecas in Toledo province (39°59′N, 4°13′W) for 2008-2019 (Fig. S1). Pollen grains could 146 only be identified to genus, so the data from Poland included both Q. robur and the closely 147 related species Q. petraea, and we analyzed the data for these two species at the community 148 level. The pollen grains of an evergreen oak (Q. ilex) in Spain are distinguished from those of 149 a deciduous species (Q. humilis). Other oaks in the study area included Q. coccifera 150 (evergreen) and Q. suber (deciduous), but their densities were much lower than those of the 151 dominating Q. ilex and Q. humilis. 152 Pollen grains were collected using Hirst traps (Hirst, 1952) specifically designed to 153 record the concentration of atmospheric particles as a function of time. The Hirst type trap is a 154 standard sampling method in pollen monitoring studies. For instance, in the European 155 Aeroallergen Network, all monitoring stations use the Hirst type volumetric trap (Galan., et 156 al. 2014). Moreover, the pollen concentrations used in this study, despite being produced in 157 different palynological laboratories, are fully comparable, as they have been obtained 158 following the standards of the European Aerobiology Society (Galan., et al. 2014). 5 159 The traps were located at a building roofs (10 – 30 m a.s.l.), and contained a built-in 160 vacuum pump, a wind-oriented vane, and a clockwork-driven drum mounted with transparent 161 adhesive tape that served as a medium for collecting the pollen. The sampler drum was 162 changed weekly, and the tape was divided into seven segments (corresponding to 24-h 163 periods). The pollen was then stained with basic fuchsine and transferred to microscope slides 164 (Scheifinger et al., 2013). The pollen grains were identified to genus under a light microscope 165 based on distinct morphological features. More than 10% of the total surface area of all slides 166 was investigated following the recommendation of the European Aerobiology Society (Galán 167 et al., 2014). We derived the length of the pollen season from the raw data for each study year 168 as a measure of flowering synchrony (Bogdziewicz et al., 2017a; Bogdziewicz et al., 2017b; 169 Lebourgeois et al., 2018). The length of the pollen season in Poland was determined using the 170 95% method, whereby the season started when 2.5% of the total yearly pollen was collected 171 and ended when 97.5% was collected (Goldberg et al., 1988). We used a cutoff of 80% for 172 Spain due to the higher number of days with very low concentrations of pollen at the end of 173 the season, which may have been due to the redistribution of pollen rather than additional 174 pollen release (Bogdziewicz et al., 2017a; Fernández-Martínez et al., 2012). Data for daily air 175 temperature and precipitation were obtained from meteorological stations within 10 km of the 176 pollen-monitoring sites. 177 178 Statistical analysis 179 We began our analysis by asking whether the timing of pollen seasons onset was 180 driven by preseason temperatures. For simplicity, we used mean daily temperatures from 181 January to April for all sites and years; the average day of flowering onset was in April for all 182 species (Fig. S2). Calculating these over different time periods (e.g. January – March or 183 February – April) do not change conclusions of the analysis. We did not use more 184 complicated approaches, like the growing degree days requirement adjusted per each site, as 185 this may lead to illogical results when the flowering onset is very late (Fu et al., 2015, 2019c), 186 which was the case in our dataset (unpublished). We constructed a linear mixed model 187 (LMM), with the day of the start of the pollen season (day of the year) as a response variable 188 and species, average preseason temperature, and their interaction as fixed effects. We next 189 tested our prediction that advanced pollen season onset would lead to longer pollen seasons 190 using a LMM that featured the length of the pollen seasons (in number of days) as a response 191 variable and species, day of onset, and their interaction as fixed effects. All models included 192 site as an intercept-only random effect to control for repeated sampling (Zuur et al., 2009). 6 193 We explored whether the length of the pollen season was better predicted by the 194 timing of the season onset, as predicted by the photoperiod-sensitivity hypothesis, or by the 195 temperature during pollen season, as predicted by the microclimatic hypothesis, by building a 196 set of LMMs that included combinations of three predictors: the day of pollen season onset, 197 mean temperature during the pollen season, and species. We included site as a random 198 intercept. We then compared the models using standard Akaike information criteria (AICs) 199 (Anderson and Burnham, 2004). We calculated the mean temperature during each pollen 200 season as the average of daily temperatures from the first until the last day of that pollen 201 season. 202 Finally, we used our long-term records to identify temporal trends in pollen season 203 onset and length of the pollen season for 1994 (Spain) or 1997 (Poland) to 2019. We built two 204 LMMs, one with the timing of the onset of the pollen season and the other with the length of 205 the pollen season as the response. We used the interaction between year and species as a fixed 206 effect and included site as an intercept-only random effect in both LMMs. 207 We explored models with temporal autocorrelation structures (lag-1 autocorrelation) 208 and retained those where dAIC < 2. We fitted all models using the glmmTMB package 209 (Brooks et al., 2017) in R using restricted maximum likelihood (REML). We tested for the 210 statistical significance of fixed factors using the Wald type II chi-square test. Model 211 validation by graphical inspection of residual patterns indicated normality and homogeneity. 212 We calculated conditional (i.e. variance explained by the complete model) and marginal (i.e. 213 variance explained by the fixed factors) R2 for the models using the sjstats package (Lüdecke, 214 2018; Nakagawa and Schielzeth, 2017) and selected models based on the AICs using the 215 MuMIn package (Bartoń, 2018). 216 217 Results 218 The timing of pollen season onset was negatively correlated with the average 219 preseason (January-April) temperatures (χ2 = 41.09, p < 0.001), which varied among species 220 (preseason temperature  species interaction: χ2 = 27.63, df = 2, p < 0.001). Pollen season 221 onset for Q. robur and Q. petraea advanced by 1.06 d (SE = 0.56) per 1 °C increase in the 222 mean temperatures (Fig. 1). This effect was stronger in the Mediterranean species, advancing 223 by 6.63 d (SE = 1.05) for Q. humilis and by 4.97 d (SE = 0.99) for Q. ilex per 1 °C increase in 224 the preseason temperatures (conditional R2 = 0.82, marginal R2 = 0.58). 225 The length of the pollen season was negatively correlated with the timing of the 226 season onset (χ2 = 27.25, p < 0.001), as predicted by the photoperiod-sensitivity hypothesis 7 227 (Fig. 2). A 1-d delay in flowering onset shortened the pollen season by 0.59 d (SE = 0.23) for 228 Q. robur and Q. petraea, 0.29 d (SE = 0.15) for Q. humilis, and 0.62 d (SE = 0.13) for Q. ilex 229 (conditional R2 = 0.60, marginal R2 = 0.20). 230 The timing of pollen season onset was a better predictor of the length of the pollen 231 season than the mean temperature during the season, supporting the photoperiod-sensitivity 232 hypothesis (Table 1). Models that included the timing of pollen season onset and the 233 temperature during the pollen season, or only the timing of pollen season onset, received 234 similar AIC support (ΔAIC = 1.29). In contrast, the model that included temperature during 235 the pollen season as the only predictor received little support given the data (ΔAIC = 16.43). 236 In fact, when both predictors when included, the temperature was no longer significant 237 predictor of the pollen season length (β = 0.23, SE = 0.59, p = 0.70), while the onset was (β = 238 -0.51, SE = 0.11, p < 0.001). These analyses indicated that the timing of flowering onset was 239 a stronger driver of the length of the pollen season than the temperature during the pollen 240 season. 241 We detected significant trends in both the onset of pollen seasons (χ2 = 6.70, p = 242 0.009) and its duration (χ2 = 24.08, p < 0.001), which varied among the species (species  243 onset of flowering: χ2 = 10.86, df = 2, p = 0.004; species  length of the pollen season: χ2 = 244 13.57, df = 2, p < 0.001). The timing of pollen season onset for Q. robur and Q. petraea did 245 not advance in the last decades (p = 0.12, Fig. 3). In contrast, the onset was significantly 246 delayed by 0.45 d y-1 (SE = 0.15) for Q. humilis and by 0.35 d y-1 (SE = 0.15) for Q. ilex. The 247 trend of delays in the pollen seasons onset shortened their length by 1.05 d y-1 (SE = 0.18) for 248 Q. ilex (Fig. 4) but not for Q. humilis (p = 0.12), in accordance with the predictions of the 249 photoperiod-sensitivity hypothesis. The length of the pollen season did not consistently 250 change with time (p = 0.64) for Q. robur or Q. petraea, where no temporal trend of the onset 251 of flowering was detected. The contrasting phenological trends in Poland and Spain were 252 likely due to contrasting changes in temperatures the last ca. 30 years in these regions. The 253 mean temperatures for January-April did not change in Poland but decreased in Spain (Fig. 254 S3). We note, however, that the general trend during the last century in Spain was positive, 255 although it slowed and reversed in the last decades (Fig. S4). 256 257 Discussion 258 Our results indicated that a late pollen seasons onset shortened their length for all 259 species we studied, as predicted by the photoperiod-sensitivity hypothesis. The pollen season 260 in warm years started earlier, which was correlated with long population-level flowering 8 261 times. Our analyses also suggested that high temperatures during the pollen seasons were not 262 the main driver of their durations, as suggested by the microclimatic hypothesis (Koenig et 263 al., 2015) or by studies linking flowering synchrony with pollen limitation and seed 264 production (Bogdziewicz et al., 2017a; Bogdziewicz et al., 2017b; Lebourgeois et al., 2018). 265 Both processes can act together in driving the synchrony of flowering. However, determining 266 whether microclimate or photoperiod sensitivity is the main driver, is important for our 267 understanding of the processes that drive the phenological synchrony of flowering in plants, 268 because they generate contrasting predictions of the effects of changes in global temperature 269 on flowering synchrony and consequently seed production. Warming under the photoperiod- 270 sensitivity hypothesis should lead to advanced flowering onsets and desynchronized 271 flowering, a pattern supported by our data. 272 High preseason temperatures advanced onset of pollen seasons in all species, leading 273 to long flowering seasons, supporting the predictions of the photoperiod-sensitivity 274 hypothesis. Temperature and daylength in temperate and boreal regions interact to cause leaf‐ 275 out around the optimal date (Flynn and Wolkovich, 2018; Fu et al., 2019a, 2015). The optimal 276 timing is a trade-off between occasional late frosts and the harvesting of light (Liu et al., 277 2018). Daylength is hypothesized to act as a cue controlling the sensitivity of the growth of 278 meristem cells to warm temperatures, thereby conditioning the relationship between 279 temperature and phenology (Fu et al., 2019c). Shorter than optimal daylength reduces 280 temperature sensitivity, thereby allowing plants to avoid precocious leaf‐out that would 281 increase the risk of frost damage. A longer than optimal daylength in turn increases 282 temperature sensitivity, allowing leaf‐out when high solar radiation is optimal for 283 photosynthesis (Flynn and Wolkovich, 2018; Fu et al., 2019a; Malyshev et al., 2018). The 284 sensitivity of leaf unfolding to photoperiod was recently reported to vary between species (Fu 285 et al., 2019a). Moreover, different intraspecific sensitivities to daylength desynchronized leaf- 286 out and subsequently flowering in several temperate species, including oaks (Zohner et al., 287 2018). We thus hypothesized that the negative correlation between pollen seasons onset and 288 their length documented by our data was due to the advance of flowering in oaks insensitive 289 to daylength, but not in individuals sensitive to daylength. Experiments and individual-level 290 phenological observations are now required to test causality (Bogdziewicz et al., 2020a). For 291 example, in the absence of individual plant flowering data, a long pollen season can be caused 292 by either each tree releasing its pollen over a longer period or because trees are less 293 synchronized with each other. In addition, experiments that simulate early (short days, high 294 temperatures) and late (long days, high temperatures) springs in greenhouse conditions should 9 295 lead to desynchronization under the photoperiod-sensitivity hypothesis but not the 296 microclimatic hypothesis. 297 The long-term changes in preseason temperature that affected the phenology of pollen 298 production provides additional support for the photoperiod-sensitivity hypothesis. We found 299 no trends in Poland and delayed pollen seasons onset in the Mediterranean oaks. The 300 difference between regions in these trends of flowering paralleled the trends in preseason 301 temperatures that decreased in Spain in the last 26 years but did not change in Poland. 302 Importantly, the trend of delayed pollen phenology was paralleled by the shortening of the 303 pollen seasons for Q. ilex, as predicted by the photoperiod-sensitivity hypothesis. The changes 304 in pollen production phenology for Q. humilis, though, did not lead to trends in the length of 305 the pollen seasons, despite the similar rate of change in the timing of the pollen season onset. 306 We hypothesized that this result follows different sensitivity to daylength in the two oak 307 species. Recent analyses found large interspecific variations in the effects of photoperiod on 308 the sensitivity of plants to changes in preseason temperature (Fu et al., 2019a, 2019b; Zohner 309 et al., 2018, 2016). In support, our models indicated that the effect of phenology of pollen 310 season onset on the length of the pollen season was twice as large for Q. ilex than Q. humilis. 311 Our results suggest that the interspecific variation in phenological sensitivity led to 312 species-specific effects of long-term temperature trends on flowering synchrony. Possible 313 consequences include asymmetrical effects on pollen limitation, as predicted by the 314 phenological synchrony hypothesis (Koenig et al., 2015; Pesendorfer et al., 2016; Wagenius 315 et al., 2020), but also on changes in gene flow and biotic interactions (Elzinga et al., 2007; 316 Jump and Peñuelas, 2005). The changes in flowering synchrony induced by temperature are 317 also likely to be associated with similar changes in the phenology of leaf unfolding (Zohner et 318 al., 2018). Leaf-out synchrony in turn affects insect herbivory (Pearse et al., 2015), which 319 may strongly affect seed production; insect herbivory in Q. ilex can decrease seed set by half 320 (Canelo et al., 2018), which can create differences in the responses of species to global 321 warming that will have asymmetrical effects on fecundity in different plant species. Thus, a 322 comprehensive understanding of the species‐specific differences of the effects of global 323 warming on flowering is essential for improving our understanding of the responses of forests 324 to the ongoing climatic change (McDowell et al., 2020). 325 Our results provide support for a novel hypothesis linking variation in weather with 326 flowering synchrony in masting oaks. Flowering synchrony drives pollen limitation in oaks, 327 which in turn is believed to interact with the dynamics of plant resources in driving mast 328 seeding (Koenig et al., 2015; Pesendorfer et al., 2016; Schermer et al., 2019). Desynchronized 10 329 flowering vetoes reproduction, which forces plants to conserve resources for subsequent 330 years, so more frequent interference leads to more stochastic and variable patterns of 331 reproduction (Pearse et al., 2016; Schermer et al., 2020). To the extent that phenological 332 synchrony is involved in determining variable seed production, global warming under the 333 microclimatic hypothesis is predicted to lead to less frequent vetoes (more frequent high 334 flowering synchrony years), thereby decreasing masting intensity (Koenig et al., 2015). We 335 found support for the hypothesis that predicts the opposite: global warming will lead to more 336 frequent vetoes, i.e. more frequent desynchronized flowering. These effects will depend on 337 the regional trends in temperature and will be asymmetrical among species, as our results 338 demonstrate. Studies of long-term trends in fecundity have reported both increases and 339 decreases in mean reproduction in forest species, but we do not know what drives these 340 differences (Bogdziewicz et al., 2020c; Mutke et al., 2005; Pesendorfer et al., 2020; Redmond 341 et al., 2012; Richardson et al., 2005). A mechanistic understanding of the links between 342 climatic variation and plant reproduction will help to identify these drivers. 343 344 Acknowledgements 345 We thank two Anonymous Reviewers for their comments on the previous version of this 346 manuscript. MB was supported by the (Polish) National Science Centre grant numbers 347 2018/28/U/NZ8/00003 (Uwertura) and 2017/24/C/NZ8/00151 (Sonatina). MFM is a 348 postdoctoral fellow of the Research Foundation-Flanders (FWO). MBP is supported by the 349 Austrian Science Foundation (FWF) project number P30381. This work contributes to the 350 ICTA-UAB ‘Unit of Excellence’ (MinECo, MDM2015-0552). JP acknowledges the financial 351 support from the European Research Council Synergy grant ERC-SyG-2013-610028 352 IMBALANCE-P). 353 354 Contributions 355 MB conceived the study, run the analysis, and drafted the manuscript. All authors collected 356 data, participated in the evaluation of the results, contributed to text editing and approved the 357 final version. 358 359 Literature 360 Anderson, D., Burnham, K., 2004. Model selection and multi-model inference. Second. NY: 361 Springer-Verlag 63. 11 362 Bogdziewicz, M., Ascoli, D., Hacket‐Pain, A., Koenig, W.D., Pearse, I., Pesendorfer, M., 363 Satake, A., Thomas, P., Vacchiano, G., Wohlgemuth, T., Tanentzap, A., 2020a. From 364 theory to experiments for testing the proximate mechanisms of mast seeding: an 365 agenda for an experimental ecology. Ecology Letters 23, 210–220. 366 https://doi.org/10.1111/ele.13442 367 Bogdziewicz, M., Fernández-Martínez, M., Bonal, R., Belmonte, J., Espelta, J.M., 2017. The 368 Moran effect and environmental vetoes: phenological synchrony and drought drive 369 seed production in a Mediterranean oak. Proceedings of the Royal Society B: 370 Biological Sciences 284, 20171784. https://doi.org/10.1098/rspb.2017.1784 371 Bogdziewicz, M., Kelly, D., Tanentzap, A.J., Thomas, P.A., Lageard, J.G.A., Hacket-Pain, 372 A., 2020b. Climate Change Strengthens Selection for Mast Seeding in European 373 Beech. Current Biology . https://doi.org/10.1016/j.cub.2020.06.056 374 Bogdziewicz, M., Kelly, D., Thomas, P.A., Lageard, J.G.A., Hacket-Pain, A., 2020c. Climate 375 warming disrupts mast seeding and its fitness benefits in European beech. Nature 376 Plants 6, 88–94. https://doi.org/10.1038/s41477-020-0592-8 377 Bogdziewicz, M., Szymkowiak, J., Kasprzyk, I., Grewling, \Lukasz, Borowski, Z., Borycka, 378 K., Kantorowicz, W., Myszkowska, D., Piotrowicz, K., Ziemianin, M., 2017. Masting 379 in wind-pollinated trees: system-specific roles of weather and pollination dynamics in 380 driving seed production. Ecology 98, 2615–2625. 381 Bogdziewicz, M., Zwolak, R., Crone, E.E., 2016. How do vertebrates respond to mast 382 seeding? Oikos 125, 300–307. 383 Brooks, M.E., Kristensen, K., van Benthem, K.J., Magnusson, A., Berg, C.W., Nielsen, A., 384 Skaug, H.J., Machler, M., Bolker, B.M., 2017. glmmTMB balances speed and 385 flexibility among packages for zero-inflated generalized linear mixed modeling. The R 386 journal 9, 378–400. https://doi.org/10.3929/ethz-b-000240890 387 Canelo, T., Gaytán, Á., González‐Bornay, G., Bonal, R., 2018. Seed loss before seed 388 predation: experimental evidence of the negative effects of leaf feeding insects on 389 acorn production. Integrative Zoology 13, 238–250. https://doi.org/10.1111/1749- 390 4877.12292 391 Clark, J.S., Nuñez, C.L., Tomasek, B., 2019. Foodwebs based on unreliable foundations: 392 spatiotemporal masting merged with consumer movement, storage, and diet. 393 Ecological Monographs 89, e01381. https://doi.org/10.1002/ecm.1381 394 Crone, E.E., Rapp, J.M., 2014. Resource depletion, pollen coupling, and the ecology of mast 395 seeding. Annals of the New York Academy of Sciences 1322, 21–34. 12 396 Elzinga, J.A., Atlan, A., Biere, A., Gigord, L., Weis, A.E., Bernasconi, G., 2007. Time after 397 time: flowering phenology and biotic interactions. Trends in Ecology & Evolution 22, 398 432–439. https://doi.org/10.1016/j.tree.2007.05.006 399 Fernández-Martínez, M., Belmonte, J., Maria Espelta, J., 2012. Masting in oaks: 400 Disentangling the effect of flowering phenology, airborne pollen load and drought. 401 Acta Oecologica 43, 51–59. https://doi.org/10.1016/j.actao.2012.05.006 402 Fernández-Martínez, M., Pearse, I., Sardans, J., Sayol, F., Koenig, W.D., LaMontagne, J.M., 403 Bogdziewicz, M., Collalti, A., Hacket-Pain, A., Vacchiano, G., Espelta, J.M., 404 Peñuelas, J., Janssens, I.A., 2019. Nutrient scarcity as a selective pressure for mast 405 seeding. Nature Plants 1–7. https://doi.org/10.1038/s41477-019-0549-y 406 Flynn, D.F.B., Wolkovich, E.M., 2018. Temperature and photoperiod drive spring phenology 407 across all species in a temperate forest community. New Phytologist 219, 1353–1362. 408 https://doi.org/10.1111/nph.15232 409 Fu, Y.H., Geng, X., Hao, F., Vitasse, Y., Zohner, C.M., Zhang, X., Zhou, X., Yin, G., 410 Peñuelas, J., Piao, S., Janssens, I.A., 2019a. Shortened temperature-relevant period of 411 spring leaf-out in temperate-zone trees. Global Change Biology 25, 4282–4290. 412 https://doi.org/10.1111/gcb.14782 413 Fu, Y.H., Piao, S., Zhou, X., Geng, X., Hao, F., Vitasse, Y., Janssens, I.A., 2019b. Short 414 photoperiod reduces the temperature sensitivity of leaf-out in saplings of Fagus 415 sylvatica but not in horse chestnut. Global Change Biology 25, 1696–1703. 416 https://doi.org/10.1111/gcb.14599 417 Fu, Y.H., Zhang, X., Piao, S., Hao, F., Geng, X., Vitasse, Y., Zohner, C., Peñuelas, J., 418 Janssens, I.A., 2019c. Daylength helps temperate deciduous trees to leaf-out at the 419 optimal time. Global Change Biology 25, 2410–2418. 420 https://doi.org/10.1111/gcb.14633 421 Fu, Y.H., Zhao, H., Piao, S., Peaucelle, M., Peng, S., Zhou, G., Ciais, P., Huang, M., Menzel, 422 A., Peñuelas, J., Song, Y., Vitasse, Y., Zeng, Z., Janssens, I.A., 2015. Declining global 423 warming effects on the phenology of spring leaf unfolding. Nature 526, 104–107. 424 https://doi.org/10.1038/nature15402 425 Galán, C., Smith, M., Thibaudon, M., Frenguelli, G., Oteros, J., Gehrig, R., Berger, U., Clot, 426 B., Brandao, R., EAS QC Working Group, 2014. Pollen monitoring: minimum 427 requirements and reproducibility of analysis. Aerobiologia 30, 385–395. 428 https://doi.org/10.1007/s10453-014-9335-5 13 429 Goldberg, C., Buch, H., Moseholm, L., Weeke, E.R., 1988. Airborne Pollen Records in 430 Denmark, 1977–1986. Grana 27, 209–217. 431 https://doi.org/10.1080/00173138809428928 432 Hacket‐Pain, A.J., Ascoli, D., Vacchiano, G., Biondi, F., Cavin, L., Conedera, M., 433 Drobyshev, I., Liñán, I.D., Friend, A.D., Grabner, M., Hartl, C., Kreyling, J., 434 Lebourgeois, F., Levanič, T., Menzel, A., Maaten, E. van der, Maaten‐Theunissen, M. 435 van der, Muffler, L., Motta, R., Roibu, C.-C., Popa, I., Scharnweber, T., Weigel, R., 436 Wilmking, M., Zang, C.S., 2018. Climatically controlled reproduction drives 437 interannual growth variability in a temperate tree species. Ecology Letters 21, 1833– 438 1844. https://doi.org/10.1111/ele.13158 439 Hanley, M.E., Cook, B.I., Fenner, M., 2019. Climate variation, reproductive frequency and 440 acorn yield in English Oaks. J Plant Ecol 12, 542–549. 441 https://doi.org/10.1093/jpe/rty046 442 Hirst, J.M., 1952. An Automatic Volumetric Spore Trap. Annals of Applied Biology 39, 257– 443 265. https://doi.org/10.1111/j.1744-7348.1952.tb00904.x 444 Jump, A.S., Peñuelas, J., 2005. Running to stand still: adaptation and the response of plants to 445 rapid climate change. Ecology Letters 8, 1010–1020. https://doi.org/10.1111/j.1461- 446 0248.2005.00796.x 447 Kelly, D., 1994. The evolutionary ecology of mast seeding. Trends in Ecology & Evolution 9, 448 465–470. https://doi.org/10.1016/0169-5347(94)90310-7 449 Kelly, D., Geldenhuis, A., James, A., Penelope Holland, E., Plank, M.J., Brockie, R.E., 450 Cowan, P.E., Harper, G.A., Lee, W.G., Maitland, M.J., 2013. Of mast and mean: 451 differential-temperature cue makes mast seeding insensitive to climate change. 452 Ecology Letters 16, 90–98. 453 Kelly, D., Sork, V.L., 2002. Mast seeding in perennial plants: why, how, where? Annual 454 review of ecology and systematics 33, 427–447. 455 Koenig, W.D., Knops, J.M., Carmen, W.J., Pearse, I.S., 2015. What drives masting? The 456 phenological synchrony hypothesis. Ecology 96, 184–192. 457 Koenig, W.D., Knops, J.M.H., 2013. Large-scale spatial synchrony and cross-synchrony in 458 acorn production by two California oaks. Ecology 94, 83–93. 459 https://doi.org/10.1890/12-0940.1 460 Koenig, W.D., Knops, J.M.H., 2000. Patterns of Annual Seed Production by Northern 461 Hemisphere Trees: A Global Perspective. The American Naturalist 155, 59–69. 462 https://doi.org/10.1086/303302 14 463 Lebourgeois, F., Delpierre, N., Dufrêne, E., Cecchini, S., Macé, S., Croisé, L., Nicolas, M., 464 2018. Assessing the roles of temperature, carbon inputs and airborne pollen as drivers 465 of fructification in European temperate deciduous forests. Eur J Forest Res 137, 349– 466 365. https://doi.org/10.1007/s10342-018-1108-1 467 Liu, Q., Fu, Y.H., Zhu, Z., Liu, Y., Liu, Z., Huang, M., Janssens, I.A., Piao, S., 2016. Delayed 468 autumn phenology in the Northern Hemisphere is related to change in both climate 469 and spring phenology. Global Change Biology 22, 3702–3711. 470 https://doi.org/10.1111/gcb.13311 471 Liu, Q., Piao, S., Janssens, I.A., Fu, Y., Peng, S., Lian, X., Ciais, P., Myneni, R.B., Peñuelas, 472 J., Wang, T., 2018. Extension of the growing season increases vegetation exposure to 473 frost. Nature Communications 9, 1–8. https://doi.org/10.1038/s41467-017-02690-y 474 Lüdecke, D., 2018. sjstats: Statistical functions for regression models. R package version 0.14 475 3. 476 Malyshev, A.V., Henry, H.A.L., Bolte, A., Arfin Khan, M.A.S., Kreyling, J., 2018. Temporal 477 photoperiod sensitivity and forcing requirements for budburst in temperate tree 478 seedlings. Agricultural and Forest Meteorology 248, 82–90. 479 https://doi.org/10.1016/j.agrformet.2017.09.011 480 McDowell, N.G., Allen, C.D., Anderson-Teixeira, K., Aukema, B.H., Bond-Lamberty, B., 481 Chini, L., Clark, J.S., Dietze, M., Grossiord, C., Hanbury-Brown, A., Hurtt, G.C., 482 Jackson, R.B., Johnson, D.J., Kueppers, L., Lichstein, J.W., Ogle, K., Poulter, B., 483 Pugh, T.A.M., Seidl, R., Turner, M.G., Uriarte, M., Walker, A.P., Xu, C., 2020. 484 Pervasive shifts in forest dynamics in a changing world. Science 368. 485 https://doi.org/10.1126/science.aaz9463 486 MuMIn, B.K., 2018. multi-model inference. R package version 1.15. 6. 2016. 487 Mutke, S., Gordo, J., Gil, L., 2005. Variability of Mediterranean Stone pine cone production: 488 Yield loss as response to climate change. Agricultural and Forest Meteorology 132, 489 263–272. https://doi.org/10.1016/j.agrformet.2005.08.002 490 Nakagawa, S., Schielzeth, H., 2017. A general and simple method for obtaining R2 from 491 generalized linear mixed-effects models. Methods in Ecology and Evolution 133–142. 492 https://doi.org/10.1111/j.2041-210x.2012.00261.x@10.1111/(ISSN)2041- 493 210X.STATSTOO 494 Nussbaumer, A., Waldner, P., Apuhtin, V., Aytar, F., Benham, S., Bussotti, F., Eichhorn, J., 495 Eickenscheidt, N., Fabianek, P., Falkenried, L., Leca, S., Lindgren, M., Manzano 496 Serrano, M.J., Neagu, S., Nevalainen, S., Pajtik, J., Potočić, N., Rautio, P., Sioen, G., 15 497 Stakėnas, V., Tasdemir, C., Thomsen, I.M., Timmermann, V., Ukonmaanaho, L., 498 Verstraeten, A., Wulff, S., Gessler, A., 2018. Impact of weather cues and resource 499 dynamics on mast occurrence in the main forest tree species in Europe. Forest 500 Ecology and Management 429, 336–350. https://doi.org/10.1016/j.foreco.2018.07.011 501 Ostfeld, R.S., Keesing, F., 2000. Pulsed resources and community dynamics of consumers in 502 terrestrial ecosystems. Trends in Ecology & Evolution 15, 232–237. 503 Pearse, I.S., Funk, K.A., Kraft, T.S., Koenig, W.D., 2015. Lagged effects of early-season 504 herbivores on valley oak fecundity. Oecologia 178, 361–368. 505 https://doi.org/10.1007/s00442-014-3193-2 506 Pearse, I.S., Koenig, W.D., Kelly, D., 2016. Mechanisms of mast seeding: resources, weather, 507 cues, and selection. New Phytologist 212, 546–562. https://doi.org/10.1111/nph.14114 508 Peaucelle, M., Janssens, I.A., Stocker, B.D., Descals Ferrando, A., Fu, Y.H., Molowny-Horas, 509 R., Ciais, P., Peñuelas, J., 2019. Spatial variance of spring phenology in temperate 510 deciduous forests is constrained by background climatic conditions. Nature 511 Communications 10, 1–10. https://doi.org/10.1038/s41467-019-13365-1 512 Pérez‐Ramos, I.M., Padilla‐Díaz, C.M., Koenig, W.D., Marañón, T., 2015. Environmental 513 drivers of mast-seeding in Mediterranean oak species: does leaf habit matter? Journal 514 of Ecology 103, 691–700. https://doi.org/10.1111/1365-2745.12400 515 Pesendorfer, M.B., Bogdziewicz, M., Szymkowiak, J., Borowski, Z., Kantorowicz, W., 516 Espelta, J.M., Fernández‐Martínez, M., 2020. Investigating the relationship between 517 climate, stand age, and temporal trends in masting behavior of European forest trees. 518 Global Change Biology 26, 1654–1667. https://doi.org/10.1111/gcb.14945 519 Pesendorfer, M.B., Koenig, W.D., Pearse, I.S., Knops, J.M.H., Funk, K.A., 2016. Individual 520 resource limitation combined with population-wide pollen availability drives masting 521 in the valley oak (Quercus lobata). Journal of Ecology 104, 637–645. 522 https://doi.org/10.1111/1365-2745.12554 523 Redmond, M.D., Forcella, F., Barger, N.N., 2012. Declines in pinyon pine cone production 524 associated with regional warming. Ecosphere 3, art120. https://doi.org/10.1890/ES12- 525 00306.1 526 Richardson, S.J., Allen, R.B., Whitehead, D., Carswell, F.E., Ruscoe, W.A., Platt, K.H., 2005. 527 Climate and Net Carbon Availability Determine Temporal Patterns of Seed 528 Production by Nothofagus. Ecology 86, 972–981. https://doi.org/10.1890/04-0863 529 Scheifinger, H., Belmonte, J., Buters, J., Celenk, S., Damialis, A., Dechamp, C., García- 530 Mozo, H., Gehrig, R., Grewling, L., Halley, J.M., Hogda, K.-A., Jäger, S., Karatzas, 16 531 K., Karlsen, S.-R., Koch, E., Pauling, A., Peel, R., Sikoparija, B., Smith, M., Galán- 532 Soldevilla, C., Thibaudon, M., Vokou, D., de Weger, L.A., 2013. Monitoring, 533 Modelling and Forecasting of the Pollen Season, in: Sofiev, M., Bergmann, K.-C. 534 (Eds.), Allergenic Pollen: A Review of the Production, Release, Distribution and 535 Health Impacts. Springer Netherlands, Dordrecht, pp. 71–126. 536 https://doi.org/10.1007/978-94-007-4881-1_4 537 Schermer, É., Bel‐Venner, M.-C., Fouchet, D., Siberchicot, A., Boulanger, V., Caignard, T., 538 Thibaudon, M., Oliver, G., Nicolas, M., Gaillard, J.-M., Delzon, S., Venner, S., 2019. 539 Pollen limitation as a main driver of fruiting dynamics in oak populations. Ecology 540 Letters 22, 98–107. https://doi.org/10.1111/ele.13171 541 Schermer, É., Bel‐Venner, M.-C., Gaillard, J.-M., Dray, S., Boulanger, V., Roncé, I.L., 542 Oliver, G., Chuine, I., Delzon, S., Venner, S., 2020. Flower phenology as a disruptor 543 of the fruiting dynamics in temperate oak species. New Phytologist 225, 1181–1192. 544 https://doi.org/10.1111/nph.16224 545 Tanentzap, A.J., Monks, A., 2018. Making the mast of a rainy day: environmental constraints 546 can synchronize mass seeding across populations. New Phytologist 219, 6–8. 547 https://doi.org/10.1111/nph.15219 548 Wagenius, S., Beck, J., Kiefer, G., 2020. Fire synchronizes flowering and boosts reproduction 549 in a widespread but declining prairie species. PNAS 117, 3000–3005. 550 https://doi.org/10.1073/pnas.1907320117 551 Ziska, L.H., Makra, L., Harry, S.K., Bruffaerts, N., Hendrickx, M., Coates, F., Saarto, A., 552 Thibaudon, M., Oliver, G., Damialis, A., Charalampopoulos, A., Vokou, D., 553 Heiđmarsson, S., Guđjohnsen, E., Bonini, M., Oh, J.-W., Sullivan, K., Ford, L., 554 Brooks, G.D., Myszkowska, D., Severova, E., Gehrig, R., Ramón, G.D., Beggs, P.J., 555 Knowlton, K., Crimmins, A.R., 2019. Temperature-related changes in airborne 556 allergenic pollen abundance and seasonality across the northern hemisphere: a 557 retrospective data analysis. The Lancet Planetary Health 3, e124–e131. 558 https://doi.org/10.1016/S2542-5196(19)30015-4 559 Zohner, C.M., Benito, B.M., Svenning, J.-C., Renner, S.S., 2016. Day length unlikely to 560 constrain climate-driven shifts in leaf-out times of northern woody plants. Nature 561 Climate Change 6, 1120–1123. https://doi.org/10.1038/nclimate3138 562 Zohner, C.M., Mo, L., Renner, S.S., 2018. Global warming reduces leaf-out and flowering 563 synchrony among individuals. eLife 7, e40214. https://doi.org/10.7554/eLife.40214 17 564 Zuur, A., Ieno, E.N., Walker, N., Saveliev, A.A., Smith, G.M., 2009. Mixed effects models 565 and extensions in ecology with R. Springer Science & Business Media. 566 18 567 Table 1. Model selection. Models are ranked by Akaike’s information criterion adjusted for 568 small sample size (AICc), and w indicates model weight. Each model is a Gaussian LMM, 569 with ar1 temporal autocorrelation structure and plot included as a random intercept. Model df LogLik AICc ∆AIC w Flowering onset + species 7 -586.35 1187.4 0 0.66 Flowering onset + temperature during 8 -585.89 1188.7 1.29 0.34 flowering + species Temperature during flowering + species 7 -594.56 1203.9 16.43 0 Null model (random effect only) 4 -610.17 1228.6 41.18 0 570 571 19 572 Figure 1. Warmer preseason temperatures advance the timing of pollen season onset (day of 573 the year, DOY). The lines are based on significant LMM predictions, and the shading 574 indicates the 95% confidence intervals, β shows effect sizes and assocaited statistics. Points 575 are the per-site, per-year observations, based on 22-yr in Q. robur (2 sites, Poland), 25-yr for 576 Q. humilis (2 sites, Spain), and 25-yr data set of pollen records in Q. ilex at two sites and 12 577 years at one site (all Spain). 578 579 580 20 581 Figure 2. Late pollen seasons are short. The lines are based on significant LMM predictions, 582 and the shading indicates the 95% confidence intervals, β shows effect sizes and assocaited 583 statistics. Points are the per-site, per-year observations, based on 22-yr in Q. robur (2 sites, 584 Poland), 25-yr for Q. humilis (2 sites, Spain), and 25-yr data set of pollen records in Q. ilex at 585 two sites and 12 years at one site (all Spain). 586 587 588 21 589 Figure 3. Temporal changes in the timing of pollen season onset. The colored lines represent 590 sites, prediction lines are based on an LMM, and the shading indicates the 95% confidence 591 intervals, β shows effect sizes and assocaited statistics. The model was based on a 22-y data 592 set at two sites in Poland (Poznań, yellow; Rzeszów, blue) for Q. robur and Q. petraea, 25-y 593 data set for Q. humilis (two sites, Spain), and 25-y data set of pollen records at two sites 594 (Barcelona, orange; Bellaterra, light blue) and a 12-y data set at one site (Huecas, green) for 595 Q. ilex. 596 597 598 599 600 22 601 Figure 5. Temporal changes in the length of the pollen seasons. The colored lines represent 602 sites, prediction lines are based on an LMM, and the shading indicates the 95% confidence 603 intervals, β shows effect sizes and assocaited statistics. The model was based on a 22-y data 604 set at two sites in Poland (Poznań, yellow; Rzeszów, blue) for Q. robur and Q. petraea, 25-y 605 data set for Q. humilis (two sites, Spain), and 25-y data set of pollen records at two sites 606 (Barcelona, orange; Bellaterra, light blue) and a 12-y data set at one site (Huecas, green) for 607 Q. ilex. 608 609 610 611 23