1 This is the submitted version of the following article: Bogdziewicz, Michał; Szymkowiak, 2 Jakub; Fernández Martínez, Marcos; [et al.]. The effects of local climate on the correlation 3 between weather and seed production differ in two species with contrasting masting habit, 4 which has been published in final form at DOI 10.1016/j.agrformet.2019.01.016. 5 © 2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license 6 http://creativecommons.org/licenses/by-nc-nd/4.0/ 7 8 9 10 11 12 13 The effects of local climate on the correlation between weather and seed production differ in 14 two species with contrasting masting habit 15 16 17 Michał Bogdziewicz1, Jakub Szymkowiak2, Marcos Fernandez-Martinez3, Josep Peñuelas4,5, 18 Josep M. Espelta5 19 1 20 Department of Systematic Zoology, Faculty of Biology, Adam Mickiewicz University, 21 Poznań, Poland 2 22 Population Ecology Lab, Faculty of Biology, Adam Mickiewicz University, Poznań, Poland 3 23 Centre of Excellence PLECO (Plant and Ecosystems), Department of Biology, University of 24 Antwerp, Wilrijk, 2610, Belgium. 4 25 CSIC, Global Ecology Unit, CREAF-CSIC-UAB, Bellaterra, 08193 Catalonia, Spain 5 26 CREAF, Cerdanyola del Valles, 08193 Catalonia, Spain 27 28 29 30 1 31 Abstract 32 Many plant species present inter-annual cycles of seed production (mast seeding), with 33 synchronized high seed production across populations in some years. Weather is believed to 34 be centrally involved in triggering masting. The links between meteorological conditions and 35 seeding are well-recognized for some species, but in others consistent correlates have not 36 been found. We used a spatially extensive data set of fruit production to test the hypothesis 37 that the influence of weather on seed production is conditioned by local climate and that this 38 influence varies between species with different life history traits. We used two model species. 39 European beech (Fagus sylvatica) that is a flowering masting species, i.e. seed production is 40 determined by variable flower production, and sessile oak (Quercus petrea) that is a fruit- 41 maturation masting species, i.e. seed production is determined by variable ripening of more 42 constant flower production. We predicted that climate should strongly modulate the 43 relationship between meteorological cue and fruit production in Q. petrea, while the 44 relationship should be uniform in F. sylvatica. The influence of meteorological cue on 45 reproduction in fruiting masting species should be strongly conditioned by local climate 46 because the strength of environmental constraint that modulates the success of flower-to-fruit 47 transition is likely to vary with local climatic conditions. In accordance, the meteorological 48 cuing was consistent in F. sylvatica. In contrast, in Q. petraea the relationship between spring 49 temperature and seed production varied among sites and was stronger in populations at colder 50 sites. The clear difference in meteorological conditioning of seed production between the two 51 studied species suggests the responses of masting plants to weather can be potentially 52 systematized according to their masting habit: i.e. fruiting or flowering. 53 54 Key-words: environmental variability, mast seeding, masting, Moran effect, plant 55 reproduction, seed production 56 2 57 Introduction 58 Masting is characterized by synchronized and highly variable levels of seed production over 59 years within a population or a community (Kelly and Sork 2002). It is a ubiquitous 60 reproductive strategy of plants worldwide that has major cascading effects on ecosystem 61 functioning (Jones et al. 1998, Ascoli et al. 2015, Bogdziewicz et al. 2016, Pearse et al. 2017, 62 Vacchiano et al. 2018). Masting is spatially synchronous, often over large portions of species 63 distributions (Koenig and Knops 1998, Ascoli et al. 2017, Fernández-Martínez et al. 2017a, 64 Vacchiano et al. 2017). The Moran effect (correlated environmental disturbances driving the 65 spatial autocorrelation of ecological phenomena, cf. Koenig 1999) is a major mechanism that 66 can account for large-scale synchrony of reproduction (Koenig and Knops 1998, Kelly and 67 Sork 2002). Specifically, some weather signals (cues) may have large effects on reproduction 68 because selection has favored plants that all respond to the cue in the same way, resulting in 69 high synchrony and individual variability (Kelly et al., 2013, Pearse et al. 2016). 70 Consequently, the correlations between seed production and weather have often been 71 investigated (Crawley and Long 1995, Piovesan and Adams 2001, Kon and Noda 2007, Allen 72 et al. 2012, Koenig and Knops 2014, Fernández-Martínez et al. 2017a, Koenig et al. 2017, 73 Vacchiano et al. 2017). Despite the large effort, consistently linking and predicting the effects 74 of weather on seed production across species has proven to be surprisingly difficult (Crone 75 and Rapp 2014). In particular, consistent and unequivocal links between weather and seed 76 production have not been found in some common and widely studied genera like oaks 77 (Quercus spp.), and the specific meteorological correlations vary among species, and even 78 within species among different studies (Sork et al. 1993, Crawley and Long 1995, Lusk et al. 79 2007, Crone and Rapp 2014, Koenig and Knops 2014, Pérez-Ramos et al. 2015, Koenig et al. 80 2016). Conversely, the link is much more consistent across space and time in some other 81 species, such as New Zealand tussock grasses (Chionochloa sp.) or European beech (Fagus 82 sylvatica) (Piovesan and Adams 2001, Rees et al. 2002, Schauber et al. 2002, Kelly et al. 83 2008, 2013, Tanentzap et al. 2012, Vacchiano et al. 2017). Why the correlations between seed 84 production and weather are consistent in some masting species, and how the irregularity in the 85 responses of other species can be systematized, remains unclear. 86 The lively debate that focused on the mechanistic drivers of masting has begun to 87 formulate a theoretical background that may allow us to resolve the puzzling relationship 88 between weather and masting (Kelly et al. 2013, Crone and Rapp 2014, Koenig et al. 2016, 89 Pearse et al. 2016, Pesendorfer et al. 2016, Bogdziewicz et al. 2017b, a). Specifically, masting 90 species can be broadly divided into two groups: flowering masting species and fruit- 3 91 maturation masting species (Pearse et al. 2016). Even though the division is continuous rather 92 than dichotomous, it may be a useful concept helping us to better understand the impact of 93 weather on plant reproduction. The annual variation in seed production in flowering masting 94 species is largely driven by variable flowering effort (Rapp et al. 2013, Monks et al. 2016, 95 Pearse et al. 2016, Bogdziewicz et al. 2017b). In contrast, the annual variation in flowering 96 effort is less relevant in fruit-maturation masting species, where variation in fruit production 97 is driven by the variable ripening of a more constant flower crop (Espelta et al. 2008, Pérez- 98 Ramos et al. 2010, Pearse et al. 2016, Bogdziewicz et al. 2017b). The theory predicts that the 99 level of synchrony will then depend on the nature of the weather cue and the biological 100 sensitivity of the plant to that signal (Pearse et al. 2016). Here, we further propose that the 101 timing of the cue (i.e. whether it happens before or after flower production) is also relevant. 102 In flowering masting species, selection favored sensitivity to weather signals that 103 trigger flower initiation (Richardson et al. 2005, Smaill et al. 2011, Tanentzap et al. 2012, 104 Kelly et al. 2013). Therefore, in these species, plants are hypersensitive to weather before 105 flowering, and once flowers are initiated, weather is less likely to affect their reproduction 106 (relatively to fruiting masting species). In contrast, fruit-maturation masting species are 107 hypersensitive to weather after flowering, i.e. weather events have strongest influence on fruit 108 production by modulating the success of fruit maturation (Espelta et al. 2008, Koenig et al. 109 2015, Pearse et al. 2016, Bogdziewicz et al. 2017a). Consequently, a large range of weather 110 events can modulate the flower-to-fruit transition, making the responsiveness of the plant to 111 particular weather event more variable. Thus, we hypothesize that in flowering masting 112 species, the effect of weather signal on seeding should be more consistent than in fruit- 113 maturation masting species. In the latter, the link between fruiting and weather should be 114 strongly modulated by local conditions, because these will define the most severe factor 115 limiting the maturation of flowers to fruits (Bogdziewicz et al. 2017a, 2018). For example, 116 drought is the most limiting factor for Quercus spp. individuals in dry and dense 117 Mediterranean forests (i.e. coppices), and seed production is strongly correlated with rainfall 118 (Espelta et al. 2008, Pérez-Ramos et al. 2010, Fernández-Martínez et al. 2012, Bogdziewicz et 119 al. 2017a). In contrast, water is less limiting for Q. petraea in temperate forests, and seed 120 production is thus correlated with spring temperature that determines flowering synchrony 121 and associated pollination efficiency (the phenology synchrony hypothesis; Koenig et al. 122 2015). Nonetheless, both constraints likely operate in both species, just with different strength 123 (Bogdziewicz et al. 2017a). Thus, similar differences may occur within species and among 124 sites. For example, the strength of the positive correlation between spring temperature and 4 125 seed production in case of temperate oaks, could be stronger at colder sites, because these 126 sites may have a generally lower average synchrony of flowering (due to frequent cold 127 springs) (Pessi and Pulkkinen 1994, Zhang et al. 2014). Therefore, the positive effect of 128 increased flowering synchrony and associated pollination efficiency should be more important 129 for seed production. If local climate strongly determines the weather influence on 130 reproduction in fruit-maturation masting species, it would explain the apparent inconsistency 131 of results of past studies. 132 To test how the local climate modulated the meteorological cuing of seed production 133 in two species that belong to the two contrasting masting groups described above, we 134 analyzed data for seed production by the sessile oak (Quercus petraea) and European beech 135 (Fagus sylvatica) at 17 and 19 sites, respectively. Fagus sylvatica is a flowering masting 136 species, and Q. petraea is a fruit-maturation masting species (Nilsson and Wastljung 1987, 137 Bogdziewicz et al. 2017b, Lebourgeois et al. 2018). In F. sylvatica, hot summers a year 138 before seed production increase flower initiation, and seeding consequently correlates 139 positively with the temperatures in the preceding year (Piovesan and Adams 2001, Hacket- 140 Pain et al. 2015, Bogdziewicz et al. 2017b). This relationship is spatially conserved 141 throughout Europe (Vacchiano et al. 2017). In contrast, acorn production in Q. petraea often 142 correlates with current spring temperatures (Kasprzyk et al. 2014, Bogdziewicz et al. 2017b, 143 Caignard et al. 2017, Lebourgeois et al. 2018). The mechanism driving masting is likely 144 through phenology synchrony of flowering (Koenig et al. 2015). In warm springs, flowering 145 in trees within a population is synchronized, which produces high pollination efficiency and 146 success of flower-to-fruit maturation (Bogdziewicz et al. 2017b). Conversely, cold springs 147 lead to desynchronized flowering and pollination failure (Koenig et al. 2012, Pesendorfer et 148 al. 2016, Bogdziewicz et al. 2017b). However, while the link between spring temperature and 149 seeding has been often reported, other studies did not confirm this relationship (Crawley and 150 Long 1995, Wesołowski et al. 2015, Fernández-Martínez et al. 2017a). We hypothesize that 151 such inconsistency is caused by variation in local climate. By considering the contrasting 152 masting strategies of these species, we predict that the response of F. sylvatica to weather will 153 be similar across climatic gradients. Specifically, warmer summers in the year preceding seed 154 dispersal in F. sylvatica will be consistently positively correlated with seed production 155 everywhere. In contrast, the strength of the correlation between spring temperature and seed 156 production in Q. petraea will vary with local climate among the sampling sites. The strength 157 of the positive correlation should be particularly stronger at colder sites. 158 5 159 Materials and Methods 160 We obtained data for seed production for Q. petraea and F. sylvatica from the ICP 161 Forests database (International Co-operative Programme on Assessment and Monitoring of 162 Air Pollution Effects on Forest, operated under the United Nations Economic Commission for 163 Europe (UNECE) Convention on Long-range Transboundary Air Pollution, http://icp- 164 forests.net/). We used data for the carbon content of fruit provided by the same database to 165 calculate fruit production per plot and year expressed as fruit net primary production (units of 166 g C m-2 y-1, i.e. average per unit of surface) (Fernández-Martínez et al. 2017a). These data 167 included fruit production at 17 sites for F. sylvatica and 19 sites for Q. petraea. Seed 168 production was monitored from 2002 to 2008, but the length of the time series differed 169 slightly between the sites (see Figs. S1 and S2 for details). The following analyses were thus 170 based on 95 site-year observations for F. sylvatica and 121 site-year observations for Q. 171 petraea. 172 We extracted climatic data for our study sites from the WorldClim database. This 173 database provides climatic data with a high spatial resolution and contains robust mean 174 monthly climatic data derived from lengthy time series (1950–2000). We used the long-term 175 annual temperature and precipitation means for the study sites as indicators of local climate. 176 We extracted meteorological time series for our forests from the interpolated meteorological 177 data of the MARS unit AGRI4CAST/JRC (http://agri4cast.jrc.ec.europa.eu/), with a 178 resolution of 0.25 × 0.25° (latitude × longitude). This database provided monthly mean 179 temperatures and total precipitation. We used the mean summer temperature (July-September) 180 and mean spring temperature (April-June) for the analysis. 181 182 Statistical analysis 183 We calculated population-level masting metrics, including the coefficient of variation 184 (CV), synchrony between sites (Pearson’s correlation coefficients) and lag-1 temporal 185 autocorrelation of seed production. The CV, synchrony, and their corresponding 95% 186 confidence intervals (CIs) were calculated by bootstrap resampling with 1000 replications. To 187 further explore the spatial synchrony, we also computed a Mantel correlograms of 188 reproduction in both species. 189 We tested our predictions in two steps, each repeated for both F. sylvatica and Q. 190 petraea. First, we tested whether the populations of the two species responded to the 191 meteorological cues suggested by previous studies, i.e. fruit production for F. sylvatica would 192 be strongly correlated with summer temperature in the year preceding seed fall. In contrast, 6 193 fruit production in Q. petraea should be correlated (at the global, complete data-set level) 194 with spring temperature. We built a linear mixed model, with log-transformed fruit production 195 as a response, study site as a random intercept and meteorological variables (i.e. summer 196 temperature for F. sylvatica and spring temperature for Q. petraea) as predictors. Second, we 197 explored whether the links between weather and seeding were conditioned by local climate. 198 We calculated Pearson correlations coefficients between (log-transformed) seed production 199 and meteorological variables separately for each study site. We then used the correlation 200 coefficients as response variables in two linear regression models, each with climatic 201 variables, i.e. either mean annual temperature at the site or mean annual precipitation, 202 included as an explanatory variable. We built separate models because these climatic 203 variables were colinear (r = 0.56). In case of F. sylvatica, we also run analogous analysis but 204 with summer temperature in year T-1 replaced by summer temperature in year T-2, as this is 205 also frequently reported cue for that species (Piovesan and Adams 2001, Vacchiano et al. 206 2017, Hacket-Pain et al. 2018). Results of that analysis are given in the Online Appendix. 207 We used the R lme4 package for the mixed models (Bates et al. 2014). All models 208 were fitted with Gaussian distributions and identity link functions. Model validation by 209 graphical inspection of the residual patterns indicated normality and homoscedasticity. We 210 calculated the R2 for the linear models and calculated marginal R2 (i.e. the proportion of 211 variance explained by fixed effects) and conditional R2 (i.e. the proportion of variance 212 explained by fixed and random effects) for the GLMMs (Nakagawa and Schielzeth 2013, 213 Bartoń 2014). 214 215 Results 216 Both species and all populations had typical masting behavior. Annual seed 217 production was more variable in F. sylvatica, with site-level CV ranging from 1.36 to 2.38 218 (mean = 1.75). Q. petraea was less variable (0.80 – 2.37, mean = 1.36). The negative 219 autocorrelation of seed production ranged from -0.77 to -0.22 (mean = -0.40) for F. sylvatica 220 and from -0.49 to 0.34 (mean = -0.17) for Q. petraea. The among-site synchrony was 0.81 221 (95% CI: 0.68–0.96) for F. sylvatica and 0.18 (95% CI: 0.02–0.37) for Q. petraea. 222 Furthermore, Mantel correlograms suggested that F. sylvatica seed production was 223 consistently highly synchronized among sites at all studied distances (up to 800 km), which 224 was not the case in Q. petraea (Fig. 1). Note, however, that most of the correlations were not 225 significant (Fig. 1). 226 7 227 Figure 1. 228 229 Figure 1. Mantel correlogram of seed production in F. sylvatica (solid line) and Q. petraea 230 (dashed line). The horizontal grey line indicates 0 correlation, while filled circles indicate 231 significant correlations. 232 233 F. sylvatica seed production was strongly correlated with summer temperatures in the 234 year preceding seed fall (β = 0.52, 95% CI range: 0.29 – 0.72, t = 5.48, p < 0.001; Fig. 2A), in 235 accordance with our predictions. The marginal R2 of the model was 0.24, and the conditional 236 R2 was 0.31. In contrast, seed production by Q. petraea was positively but not significantly 237 correlated with spring temperature (β = 0.20, 95% CI range: -0.04 – 0.45, t = 1.63, p = 0.10; 238 Fig. 2B). The marginal R2 of the model was 0.02, and the conditional R2 was 0.02. 239 240 Figure 2. 8 241 242 243 Figure 2. Relationships between a) seed production and previous summer temperature for F. 244 sylvatica and b) seed production and current-year spring temperature for Q. petraea. The lines 245 represent GLMM predictions, and the shaded regions represent 95% confidence intervals. The 246 dashed line represents a nonsignificant relationship. 247 248 The response of the F. sylvatica but not the Q. petraea populations to weather was 249 clearly consistent after the correlations between seeding and meteorological cues had been 250 decomposed to the site level. All correlations for F. sylvatica were positive (range: 0.23–0.92, 251 mean = 0.64; Fig. 3A, B), while the site-level correlation coefficients for Q. petraea ranged 252 from -0.24 to 0.84 (mean = 0.24, Fig. 3C, D). 253 254 Figure 3. 9 255 256 Figure 3. Spatial variation in the response of F. sylvatica and Q. petraea to meteorological 257 cues. A) and B) site-level Pearson correlation coefficients between log-transformed fruit 258 production and previous summer temperature for F. sylvatica. C) and D) site-level Pearson 259 correlation coefficients between log-transformed fruit production and current-year spring 260 temperature for Q. petraea. The vertical dashed lines in b) and d) indicate 0 correlation. 261 262 The strength of the correlation between summer temperature of the previous year and 263 seed production for F. sylvatica did not differ with either the mean annual temperature of the 264 site (β = -0.01, 95% CI range: -0.07 – 0.05, t = -0.35, p = 0.73; Fig. 4A) or the mean annual 265 precipitation (β = 0.0004, 95% CI range: -0.0002 – 0.001, t = 1.29, p = 0.22; Fig. 4B). In 266 contrast, the correlation between spring temperature and seed production for Q. petraea was 10 267 weaker at sites with higher mean annual temperatures (β = -0.21, 95% CI range: -0.40 – -0.04, 268 t = -2.62, p = 0.018; R2 = 0.29; Fig. 4C) with no effects of mean annual precipitation (β = - 269 0.0002, 95% CI range: -0.002 – 0.002, t = -0.18, p = 0.86; Fig. 4D). 270 271 Figure 4. 272 273 Figure 4. Relationships between the site-level Pearson correlation coefficients and mean 274 annual temperature (A and C) and mean annual precipitation (B and C). The lines represent 275 linear model predictions, and the shaded regions represent 95% confidence intervals. The 276 dashed lines represent nonsignificant relationships. 277 278 Discussion 279 The relationship between weather and fruit production for F. sylvatica was 280 consistently positive at all 19 sites studied (Figure 3A). In contrast, the relationship between 281 spring temperature and seed production for Q. petraea was not significant at the whole 282 database level. Yet, decomposition of this relationship to the site-level, hinted that it was 283 because the relationship largely varied among sites (Figure 3C) and was stronger in 284 populations at colder locations. While certainly further studies are warranted, our results are 11 285 consistent with the prediction that the climatic modulation of the responses of flowering 286 masting and fruit-maturation masting species to weather should be different. The flowering 287 masting F. sylvatica is hypersensitive to weather cue that triggers flowering (Bogdziewicz et 288 al. 2017b), possibly through increased expression of genes involved in floral transition 289 (Miyazaki et al. 2014, Pearse et al. 2016). Thus, the general positive effect of temperature on 290 flower initiation should be less susceptible to varying local climate, and therefore more 291 uniform across sites. In contrast, the key meteorological variable in fruit-maturation masting 292 species favors (or prevents) the transition from flower to fruit. The environmental constraint 293 that modulates this success is likely to vary with local conditions, as we observed in Q. 294 petraea. 295 The response of Q. petraea to spring temperature was stronger at colder sites. Low 296 temperatures during flowering are associated in oaks with desynchronized flowering, because 297 meteorological variability in microhabitats is high in cold springs, which leads to a variable 298 onset of flowering (Koenig et al. 2015, Pesendorfer et al. 2016, Bogdziewicz et al. 2017b). 299 The phenological mismatch among individuals within a population leads to relatively lower 300 pollen availability for each tree and therefore lower pollination success and seed production 301 (Koenig et al. 2012, Bogdziewicz et al. 2017b). Warm springs enhance the synchronization of 302 flowering, allowing high pollination success and lead to high fruit production. We thus 303 believe that this variable background synchronization of flowering (Koenig et al. 2015, 304 Pesendorfer et al. 2016, Bogdziewicz et al. 2017b) is the mechanism responsible for the 305 systematic among-site variation of oaks response to spring weather. The average flowering 306 synchrony is likely lower at cold sites (Pessi and Pulkkinen 1994, Zhang et al. 2014), so 307 pollination success would be a strong constraint on seed production. Consequently, the effect 308 of increased spring temperature on fruits production was clear at these sites. In contrast, 309 pollination success probably affects seed production less at warmer sites, because the 310 synchrony of flowering is generally higher, so the correlations would be weaker. As an 311 indirect support, pollen seasons in Pinus sylvestris are longer at northern sites (Pessi and 312 Pulkkinen 1994), and were also found to be shorter at warmer sites in Betlua sp. (Zhang et al. 313 2014). Nonetheless, the opposite can also be true (e.g. Qiu et al. 2018). Thus, direct tests of 314 that hypothesis are necessary, and can include evaluating the effects of spring temperatures 315 and phenological synchrony on oak seeding across elevations. 316 Spring temperature may also condition acorn production in oaks by different 317 mechanisms, e.g. by modulating the acorn development process (Sork et al. 1993, Koenig and 318 Knops 2014). Spring temperature had a negative influence on seed production in 4 of the 19 12 319 populations. Trees may respond to meteorological variables in a symmetric, Gaussian-like 320 fashion, with an optimal response (Fernández-Martínez et al. 2017b). The direction of the 321 response (positive/negative) could then depend on the range of the meteorological variable at 322 different sites. If the meteorological cue has a mode below the optimum in one region, then 323 the correlation between the meteorological variable and the response at that site will be 324 positive (Lusk et al. 2007, Fernández-Martínez et al. 2017b). Likewise, if the mode of the 325 meteorological cue is above the optimum at a site, then the correlation between the 326 meteorological variable and the response in that region will be negative. Our sites at which we 327 observed the negative correlations may operate above the environmental optimum for Q. 328 petraea, e.g. excessively hot springs will lead to flower abortion. F. sylvatica probably has a 329 similar optimum, but it may not be as strongly modulated by local climate as for Q. petraea 330 due to differences in the life-history traits discussed here. Experimental investigations of the 331 influence of weather on flower and seed development in masting plants are rare (Kon and 332 Noda 2007, Kelly et al. 2008, Pérez-Ramos et al. 2010), but will provide necessary insight 333 into the mechanistic links between weather and seeding variation. 334 The spatial consistency in the response of F. sylvatica to weather explains why F. 335 sylvatica populations were on average 4-times better synchronized than Q. petraea (0.81 vs 336 0.18, respectively; for similar result see Nussbaumer et al. 2016). The response of F. sylvatica 337 to meteorological cues was spatially uniform, so all populations within a region fluctuated 338 similarly in response to correlated meteorological conditions (see also Fig. 1). The synchrony 339 of reproduction may then be easily scaled up to a continental scale, as shown in Vacchiano et 340 al. (2017). Conversely, each population of Q. petraea probably responded more in accordance 341 with its local optimum, which consequently lowered the large-scale spatial synchronization. 342 Spatial synchrony of masting has been intensively studied (e.g. Koenig and Knops 1998, 343 Kelly and Sork 2002, Fernández-Martínez et al. 2017, Vacchiano et al et al. 2017) but why 344 species differ in how well they are synchronized is unclear. We propose that synchrony will 345 differ between flowering and fruit-maturation masting species, with the former having a 346 higher average large-scale synchrony due to spatially conserved meteorological cuing, i.e. a 347 regional pattern more consistent with the Moran effect. 348 349 Conclusions 350 The clear difference in meteorological conditioning of seed production between the 351 two species suggests that the responses of masting plants to weather can be systematized and 352 predicted. Meteorological cues in species where annual flowering intensity is the main 13 353 determinant of seed production should include variables associated with resource acquisition 354 and flower initiation, and be generally similar across species ranges (Richardson et al. 2005, 355 Monks et al. 2016, Vacchiano et al. 2017). Meteorological cues in fruit-maturation masting 356 species, though, could be predicted by considering the likely key environmental constraint (so 357 called ‘veto’ cf. Pearse et al. 2016, Bogdziewicz et al. 2018) to fruit maturation for the 358 particular species and region studies. We note, however, that the differentiation between 359 flowering and fruit-maturation masting species is not distinct but a continuum of species with 360 more or fewer mixed strategies (Montesinos et al. 2012, Abe et al. 2016, Pearse et al. 2016). 361 Accurate predictions will therefore need detailed information of species biology. Our research 362 nonetheless provides new insights into the resolution of apparently inconsistent responses of 363 plant reproduction to meteorological variation. Finally, climate appears to condition the 364 species responses to weather stronger in fruit-maturation masting species. It indicates that 365 these species may be more sensitive to global climate change then flowering masting species. 366 367 Acknowledgements 368 We thank Anonymous Reviewers for their helpful comments on the previous versions of this 369 manuscript. MB was supported by the Polish National Science Centre grant Sonatina No. 370 2017/24/C/NZ8/00151. JSz was supported by the Foundation for Polish Science (FNP) 371 scholarship “Start”. MFM is a postdoctoral fellow of the Research Foundation – Flanders 372 (FWO), and JP was supported by the European Research Council Synergy grant ERC-SyG- 373 2013-610028 IMBALANCE-P. 374 375 Competing interest statement 376 Authors declare no competing interests. 377 378 Literature 379 Abe, T. et al. 2016. Parameterisation and validation of a resource budget model for masting 380 using spatiotemporal flowering data of individual trees. - Ecology Letters 19: 1129– 381 1139. 382 Allen, R. B. et al. 2012. Synchronicity, periodicity and bimodality in inter-annual tree seed 383 production along an elevation gradient. - Oikos 121: 367–376. 384 Ascoli, D. et al. 2015. The synchronicity of masting and intermediate severity fire effects 385 favors beech recruitment. - Forest Ecology and Management 353: 126–135. 386 Ascoli, D. et al. 2017. Inter-annual and decadal changes in teleconnections drive continental- 387 scale synchronization of tree reproduction. - Nature communications 8: 2205. 14 388 Bartoń, K. 2014. MuMIn: multi-model inference. R package version 1.10. 0. - See 389 http://CRAN. R-project. org/package= MuMIn in press. 390 Bates, D. et al. 2014. lme4: Linear mixed-effects models using Eigen and S4. - R package 391 version 1: 1–23. 392 Bogdziewicz, M. et al. 2016. How do vertebrates respond to mast seeding? - Oikos 125: 300– 393 307. 394 Bogdziewicz, M. et al. 2017a. The Moran effect and environmental vetoes: phenological 395 synchrony and drought drive seed production in a Mediterranean oak. - Proc. R. Soc. 396 B 284: 20171784. 397 Bogdziewicz, M. et al. 2017b. Masting in wind-pollinated trees: system-specific roles of 398 weather and pollination dynamics in driving seed production. - Ecology 98: 2615– 399 2625. 400 Bogdziewicz M et al. 2018. Correlated seed failure as an environmental veto to synchronize 401 reproduction of masting plants. - New Phytologist 219: 98-108. 402 Caignard, T. et al. 2017. Increasing spring temperatures favor oak seed production in 403 temperate areas. - Scientific Reports 7: 8555. 404 Crawley, M. and Long, C. R. 1995. Alternate bearing, predator satiation and seedling 405 recruitment in Quercus robur L. - Journal of Ecology: 683–696. 406 Crone, E. E. and Rapp, J. M. 2014. Resource depletion, pollen coupling, and the ecology of 407 mast seeding. - Annals of the New York Academy of Sciences 1322: 21–34. 408 Espelta, J. M. et al. 2008. Masting mediated by summer drought reduces acorn predation in 409 Mediterranean oak forests. - Ecology 89: 805–817. 410 Fernández-Martínez, M. et al. 2012. Masting in oaks: disentangling the effect of flowering 411 phenology, airborne pollen load and drought. - Acta Oecologica 43: 51–59. 412 Fernández-Martínez, M. et al. 2017a. The North Atlantic Oscillation synchronises fruit 413 production in western European forests. - Ecography 40: 864–874. 414 Fernández-Martínez, M. et al. 2017b. Nature beyond linearity: meteorological variability and 415 Jensen’s inequality can explain mast seeding behaviour. - Frontiers in Ecology and 416 Evolution 5: 134. 417 Hacket-Pain, A. J. et al. 2015. The influence of masting phenomenon on growth–climate 418 relationships in trees: explaining the influence of previous summers’ climate on ring 419 width. - Tree Physiology 35: 319–330. 420 Hacket-Pain, A. J. et al. 2018. Climatically controlled reproduction drives interannual growth 421 variability in a temperate tree species. - Ecology Letters 21: 1833-1844. 422 Jones, C. G. et al. 1998. Chain reactions linking acorns to gypsy moth outbreaks and Lyme 423 disease risk. - Science 279: 1023–1026. 15 424 Kasprzyk, I. et al. 2014. Relationships among weather parameters, airborne pollen and seed 425 crops of Fagus and Quercus in Poland. - Agricultural and forest meteorology 197: 426 111–122. 427 Kelly, D. and Sork, V. L. 2002. Mast seeding in perennial plants: why, how, where? - Annual 428 Review of Ecology and Systematics 33: 427–447. 429 Kelly, D. et al. 2008. Mast seeding, predator satiation, and temperature cues in Chionochloa 430 (Poaceae). - Population Ecology 50: 343–355. 431 Kelly, D. et al. 2013. Of mast and mean: differential-temperature cue makes mast seeding 432 insensitive to climate change. - Ecology Letters 16: 90–98. 433 Koenig, W. D. 1999. Spatial autocorrelation of ecological phenomena. - Trends in Ecology & 434 Evolution 14: 22–26. 435 Koenig, W. D. and Knops, J. M. 1998. Scale of mast-seeding and tree-ring growth. - Nature 436 396: 225. 437 Koenig, W. D. and Knops, J. M. 2014. Environmental correlates of acorn production by four 438 species of Minnesota oaks. - Population Ecology 56: 63–71. 439 Koenig, W. D. et al. 2003. Dissecting components of population-level variation in seed 440 production and the evolution of masting behavior. - Oikos 102: 581–591. 441 Koenig, W. D. et al. 2012. Stabilizing selection for within-season flowering phenology 442 confirms pollen limitation in a wind-pollinated tree. - Journal of Ecology 100: 758– 443 763. 444 Koenig, W. D. et al. 2015. What drives masting? The phenological synchrony hypothesis. - 445 Ecology 96: 184–192. 446 Koenig, W. D. et al. 2016. Is the relationship between mast-seeding and weather in oaks 447 related to their life-history or phylogeny? - Ecology 97: 2603–2615. 448 Koenig, W. D. et al. 2017. Drivers of synchrony of acorn production in the valley oak 449 (Quercus lobata) at two spatial scales. - Ecology 98: 3056–3062. 450 Kon, H. and Noda, T. 2007. Experimental investigation on weather cues for mast seeding of 451 Fagus crenata. - Ecological Research 22: 802–806. 452 Lebourgeois, F. et al. 2018. Assessing the roles of temperature, carbon inputs and airborne 453 pollen as drivers of fructification in European temperate deciduous forests. - European 454 Journal of Forest Research 137: 349–365. 455 Lusk, J. J. et al. 2007. Correlates of interspecific synchrony and interannual variation in seed 456 production by deciduous trees. - Forest Ecology and Management 242: 656–670. 457 Monks, A. et al. 2016. Resource limitation underlying multiple masting models makes mast 458 seeding sensitive to future climate change. - New Phytologist 210: 419–430. 16 459 Montesinos, D. et al. 2012. Masting uncoupling: mast seeding does not follow all mast 460 flowering episodes in a dioecious juniper tree. - Oikos 121: 1725–1736. 461 Miyazaki, Y. et al. 2014. Nitrogen as a key regulator of flowering in Fagus crenata: 462 understanding the physiological mechanism of masting by gene expression analysis. - 463 Ecology Letters, 17: 1299-1309. 464 Nakagawa, S. and Schielzeth, H. 2013. A general and simple method for obtaining R2 from 465 generalized linear mixed-effects models. - Methods in Ecology and Evolution 4: 133– 466 142. 467 Nilsson, S. G. and Wastljung, U. 1987. Seed predation and cross-pollination in mast-seeding 468 beech (Fagus sylvatica) patches. - Ecology 68: 260–265. 469 Nussbaumer, A. et al. 2016. Patterns of mast fruiting of common beech, sessile and common 470 oak, Norway spruce and Scots pine in Central and Northern Europe. - Forest Ecology 471 and Management 363: 237–251. 472 Pearse, I. S. et al. 2016. Mechanisms of mast seeding: resources, weather, cues, and selection. 473 - New Phytologist 212: 546–562. 474 Pérez-Ramos, I. M. et al. 2010. Mast seeding under increasing drought: results from a long- 475 term data set and from a rainfall exclusion experiment. - Ecology 91: 3057–3068. 476 Pérez-Ramos, I. M. et al. 2015. Environmental drivers of mast-seeding in Mediterranean oak 477 species: does leaf habit matter? - Journal of Ecology 103: 691–700. 478 Pesendorfer, M. B. et al. 2016. Individual resource limitation combined with population-wide 479 pollen availability drives masting in the valley oak (Quercus lobata). - Journal of 480 Ecology 104: 637–645. 481 Pessi, A.-M. and Pulkkinen, P. 1994. Temporal and spatial variation of airborne Scots pine 482 (Pinus sylvestris) pollen. - Grana 33: 151–157. 483 Piovesan, G. and Adams, J. M. 2001. Masting behaviour in beech: linking reproduction and 484 climatic variation. - Canadian Journal of Botany 79: 1039–1047. 485 Qiu, S. et al. 2018. Latitudinal pattern of flowering synchrony in an invasive wind-pollinated 486 plant. - Proceedings of the Royal Society B: Biological Sciences, 285: 20181072. 487 Rapp, J. M. et al. 2013. Sex allocation, pollen limitation and masting in whitebark pine. - 488 Journal of Ecology 101: 1345–1352. 489 Rees, M. et al. 2002. Snow tussocks, chaos, and the evolution of mast seeding. - The 490 American Naturalist 160: 44–59. 491 Richardson, S. J. et al. 2005. Climate and net carbon availability determine temporal patterns 492 of seed production by Nothofagus. - Ecology 86: 972–981. 493 Schauber, E. M. et al. 2002. Masting by eighteen New Zealand plant species: the role of 494 temperature as a synchronizing cue. - Ecology 83: 1214–1225. 17 495 Smaill, S. J. et al. 2011. Climate cues and resources interact to determine seed production by a 496 masting species. - Journal of Ecology 99: 870–877. 497 Sork, V. L. et al. 1993. Ecology of mast-fruiting in three species of North American 498 deciduous oaks. - Ecology 74: 528–541. 499 Tanentzap, A. J. et al. 2012. Soil nutrient supply modulates temperature-induction cues in 500 mast-seeding grasses. - Ecology 93: 462–469. 501 Vacchiano, G. et al. 2017. Spatial patterns and broad-scale weather cues of beech mast 502 seeding in Europe. - New Phytologist 215: 595–608. 503 Vacchiano, G. et al. 2018. Reproducing reproduction: How to simulate mast seeding in forest 504 models. - Ecological Modelling 376: 40–53. 505 Wesołowski, T. et al. 2015. Interannual variation in tree seed production in a primeval 506 temperate forest: does masting prevail? - European Journal of Forest Research 134: 507 99–112. 508 Zhang, Y. et al. 2014. Climate change effect on Betula (birch) and Quercus (oak) pollen 509 seasons in the United States. - International Journal of Biometeorology 58: 909–919. 510 511 512 18 513 514 Online Appendix. Bogdziewicz et al. The effects of local climate on the correlation between 515 weather and seed production differ in two species with contrasting masting habit. 516 517 518 Figure 1S. Fruit production time series for F. sylvatica. Each plot represents a site. 519 520 19 521 Figure 2S. Fruit production time series for Q. petraea. Each plot represents a site. 522 523 524 20 525 Figure S3. A) The relationship between seed production and mean summer temperature two 526 years before seed dispersal (β = -0.62, 95% CI range: -0.82 – -0.42, t = -6.13, p < 0.001). B) 527 Site-level Pearson correlation coefficients between log-transformed fruit production and 528 summer temperature two years before seed dispersal. Both figures for F. sylvatica. 529 530 21