Annual variability in sound acorn production was regulated by a generalist seed predator weevil in Quercus serrata


 Highly variable and synchronous seed production within a population (‘masting’) could be from either synchronised high annual variability in floral initiation (‘flower masting’) or synchronised floral abortion until maturity (‘fruit maturation masting’). We investigated the demographic processes of the female organs from flowering to seed maturity, including each type of insect damage identified, in Quercus serrata in six individuals within a stand from 2014 to 2020, western Japan. Although the annual production of sound acorns was significantly correlated with that of female flowers, the annual variability in sound acorn production within an individual was significantly higher and their synchrony increased, compared to those of female flowers. The annual production of female flowers was positively correlated with the temperature difference in April between the previous and flowering years. However, their fluctuation was low, which was neither affected by seed and flower production in the previous year nor contributed to predator starvation. Key-factor analyses revealed that reproductive loss due to oviposition and sap suction by Mechoris ursulus , a generalist seed predator weevil for oak species, was the largest and most important factor that contributed to the annual variation in the total pre-dispersal loss of Q. serrata . The survival rate from female flowers to sound acorns was strongly predicted by the temperature in June, corresponding to the emergence of adult M. ursulus . This study suggests that highly variable and synchronous sound seed production can be proximately regulated by seed predation when the main predator is a generalist.


28
Highly variable seed production, which synchronises among individuals within a population, We selected six Q. serrata individuals (diameter at breast height (DBH) ± SD, 35.4 ± 6.8 cm; 128 height, 18.4 ± 2.2 m; canopy area, 45.2 ± 42.2 m 2 ). These trees were within a natural stand 129 where a number of Q. serrata trees were mixed. Regarding fagaceous species within the stand, 130 Q. glauca, Q. sessilifolia, and Castanopsis cuspidata were also mixed to a lesser extent. We set of damage other than being broken (exhibiting a gnaw mark, a penetrating hole, larval frass, 141 deformation, and/or discolouration; mainly caused by insects), and (D) the organs that were 142 broken (mainly caused by vertebrate predation), as described by Hirayama et al. (2017). The organs classified into category (D) were low in our study system (0-1.4 % of all female 144 reproductive organs falling during one year in each tree). 145 When we classified the organs into category (C), we dissected them and observed them 146 using a stereomicroscope. Based on larval faeces, feeding traces, penetration and oviposition 147 scars, and/or larvae inside, we classified the damage into the following types: (a) sap suction 148 by adult weevils, mainly by M. ursulus, (b) damaged by moth larvae, (c) oviposition by M. 149 ursulus, (d) oviposition by weevils other than M. ursulus, mainly by K. rectirostris, (e) damaged 150 by cynipid wasps, and (f) others, as described by Hirayama et al. (2017). In addition, we reared 151 the larvae from the female reproductive organs and observed the feeding and/or oviposition considered. We also calculated the difference between a year and its previous year for all 217 variables, according to methods described by Kelly et al. (2013), who suggested that the 218 differential cue might cause masting behaviour. We checked for multicollinearity among all 219 variables using a variance inflation factor (VIF), where a VIF > 10 was considered to indicate 220 multicollinearity (Quinn and Keough 2002). If one variable was highly correlated with another 221 variable based on the VIF value, we adopted one variable of the two (Online Resource 1). We then performed GLMM models, which were a log link and negative binomial error distribution, 223 with the annual number of female flowers per individual (non-log-transformed) as the response 224 variable, each adopted weather variable as the explanatory variable, and the individual tree as 225 a random factor. We selected five weather variables based on a comparison of the models using were less than 10, which confirmed that multicollinearity among these weather variables was 240 not a serious concern. We constructed GLMM models, which were a log link and negative 241 binomial error distribution, with the annual number of sound acorns (non-log-transformed) as 242 the response variable and each weather variable as the explanatory variable. The annual number 243 of female flowers was an offset variable, and the individual trees were treated as random factors.

244
We determined the best model, which describes the relationship between the survival rate of 245 female flowers and weather variables using the same method in the analyses of annual 246 variability in female flower number as mentioned above.

248
Relative contribution of each type of reproductive loss during the pre-dispersal phase 249 Key factor analysis (Podoler and Rogers 1975) was used to estimate the relative contribution 250 of factors that influenced plant reproductive potential to the observed variation in total pre-251 dispersal losses. The reproductive losses due to factor and the total pre-dispersal loss (K)

252
were calculated for each year in each individual tree as follows: where is the number of female reproductive organs that remained after mortality due to 256 factor i (the number per total area of 1.5 m 2 per individual), and n represents the number of 257 factors. We added 1 to those numbers before log-transformation to avoid excluding zero data.

258
The slope of the regression of on K was used to determine the relative impact of each factor reproductive loss due to abortion of pistillate flowers and immature acorns was assumed to 275 occur before insect damage in the analysis.

276
All analyses were performed using R 4.1 (R Core Team 2021).

279
Annual variability in the numbers of female flowers and sound acorns 280 The annual numbers of sound acorns exhibited higher inter-year variation than those of female 281 flowers for all individual trees (Fig. 1). for the number of sound acorns was significantly 282 higher (1.51) than that for the number of female flowers (0.50) based on their 95% confidence 283 intervals using the bootstrap method (Table 1). The annual number of sound acorns was  Table 2).

308
For the survival rate from female flowers to sound acorns per individual, in the models 309 selected based on AIC values, the five weather variables which were highly related to them  Table 2).
Except in 2014 and 2020, when the production of sound acorns was high for all individuals, the 318 number of female organs tended to be greatly reduced by sap suction by adult weevils (Fig. 4,   319 stage 3-4) and by oviposition by M. ursulus for all individuals (stage 6-7). Reproductive losses 320 due to abortion of pistillate flowers and immature acorns (stage 1-2) and oviposition by weevils 321 other than M. ursulus, mainly by K. rectirostris (stage 5-6), were not significantly related to 322 total pre-dispersal loss, while reproductive losses due to almost all types of insect damage were 323 significantly related to the total pre-dispersal loss (Table 3). Among the types of insect damage, 324 the slope of the regression on the total pre-dispersal loss was the highest in oviposition by M. 325 ursulus, followed by sap suction by adult weevils, mainly by M. ursulus (Table 3 and Online   326 Resource 10).

329
Although the annual production of female flowers per Q. serrata individual significantly 330 influenced the annual production of sound acorns (Fig. 2), the annual fluctuation of female 331 flower production was significantly lower than that of sound acorn production ( Fig. 1 and Table   332 1). Previous studies have suggested that the annual variability of flower production is affected pre-dispersal loss was significantly influenced by almost all types of insect damage. Therefore, synchronised floral or seed abortion until maturity in Q. serrata is not caused by plant internal 366 mechanisms, such as pollination failure nor resource limitation, but by insect damage.

367
Among the types of insect damage, reproductive loss, owing to oviposition and sap 368 suction by M. ursulus, was the largest and most important factor that contributed to the annual 369 variation of total pre-dispersal loss in Q. serrata (Fig. 4, Table 3, and Online Resource 10). In      Individual trees were treated as a random factor.

Annual variability in sound acorn production was regulated by a generalist seed predator weevil in Quercus serrata
Kimiko Hirayama 1* , Kenta Mizo 1 , Manaka Tatsuno 2 , Mizuki Yoshikawa 2 , Chieri Tachikawa 2 Online resource 1 Data of weather variables which were analysed in relation to the production of female flowers. Temperature is in ℃; precipitation is in mm. Variables marked *1, *2, and *3 were highly correlated with each other and had been found to have a high VIF value (>10). We adopted one variable from each pair.
Online Resource 2 Data of weather variables which were analysed in relation to survival rate from female flowers to sound acorns. Temperature is in ℃; precipitation is in mm. When the number of flowers was zero, the data were excluded.