Pre-dispersal strategies by Quercus schottkyana to mitigate the effects of weevil infestation of acorns

We investigated how pre-dispersal strategies may mitigate the effects of weevil infestation of acorns in a population of Quercus schottkyana, a dominant oak in Asian evergreen broad-leaved forests, and assess if weevil infestation contributes to low seedling recruitment. We counted the number of acorns produced, daily from the end of August to mid-late November for 9 years from 2006–2014. We also recorded the rate of acorn infestation by weevils and acorn germination rates of weekly collections. Annual acorn production was variable, but particularly low in 2011 and 2013. There was no trade-off between acorn production and acorn dry mass. However, acorns produced later in the season were significantly heavier. For most years: (i) the rate of weevil infestation was negatively density dependent (a greater proportion of acorns died with increased acorn density), (ii) the percentage germination of acorns was positively density dependent (proportionately more acorns germinated with increased density), and (iii) as the season progressed, the percentage of infested acorns declined while germination rates increased. Finally, (iv) maximum acorn production, percentage infestation and percentage germination were asynchronous. Although pre-dispersal mortality is important it is unlikely to be the primary factor leading to low recruitment of oak seedlings.


Supplemental
. Acorn production 1. Between-year variation in acorn production ( Fig 1A): Results of Likelihood ratio test to determine if the number of acorns produced was different between years. We used generalized linear mixed models with week number as a random effect and Poisson errors. We used week number as a random effect to account for the fact that acorn production is correlated between weeks. Acorn number was square root transformed and we tested for overdispersion in this model; there was none (SSQ residuals/residual df = 1.042, p = 0.223). Thus, we retained the Poisson distribution. Results of cross validation: the test was repeated 100 times using randomly select 90%of our data set, all the iterations achieved a p value of <0.001.

The difference in acorn size between years:
Results of Likelihood ratio test to determine if acorns differ in size between years. We used generalized linear mixed models with week number as a random effect and Gaussian errors. We used week number as a random effect to account for the fact that acorn production is correlated between weeks. Results of cross validation: the test was repeated 100 times using randomly select 90% of our data set, all the iterations achieved a p value of <0.001. 4 4. Acorn size over time (Fig 2A):

Model
Results of Likelihood ratio test to determine if acorn size differs with week number. We used generalized linear mixed models with week number as a fix effect and Gaussian errors. We used week number and year as a random effect to account for the fact that acorn production and size is correlated between weeks. We used year as a random effect to account for differences in production between years, and repeated measures over time. Results of cross validation: the test was repeated 100 times using randomly select 90% of our data set, all the iterations achieved a p value of <0.001.

Model
5. Acorn size and production ( Fig 2B): Results of Likelihood ratio test to determine if acorn size is dependent on the production of acorns. We used generalized linear mixed models with week number and year as a random effect and Gaussian errors. We used week number and year as random effects to account for the fact that acorn production and size is correlated between weeks. We used year as a random effect to account for differences in production between years, and repeated measures over time.  (Fig 1B):

Difference in infestation between years
Results of Likelihood ratio test to determine if the proportion of acorns infested is different between years. We used generalized linear mixed models with year as a fix effect and binomial errors. We used week number and year as a random effect to account for the fact that acorn production is correlated between weeks. Results of cross validation: the test was repeated 100 times using randomly select 90%of our data set, all the iterations achieved a p-value of <0.001. Fig 3A):

Results of Likelihood ratio test to determine if the proportion of acorns infested changes with
week number. We used generalized linear mixed models with year as a random effect and binomial errors. Because the relationship between acorns infection and week number is not monotonic, we used quadratic model. We used year as a random effect to account for differences in production between years, and repeated measures over time. 3. Infestation with production ( Fig 3B):

Model
Results of Likelihood ratio test to determine if the yearly proportion of acorns infested changes with yearly acorn density. We use generalized linear model with binomial errors. We assume the observations are independent between years, because the locations for acorn collection were randomly placed each year.  Table S3. Acorn germination (Fig 1C):

Variation in germination between years
Results of Likelihood ratio test to determine if the proportion of acorns that germinated differed between years. We used generalized linear mixed models with week number as a random effect and binomial errors. We used week number as a random effect to account for the fact that acorn production is correlated between weeks. Results of cross validation: the test was repeated 100 times using randomly select 90%of our data set, all the iterations achieved a p value of <0.001.

Model
2. Germination over time (Fig 3C): Results of Likelihood ratio test to determine if the proportion of acorns that germinated changes with week number. We used generalized linear mixed models with year as a random effect and binomial errors. We used year as a random effect to account for differences in production between years, and repeated measures over time. Results of cross validation: the test was repeated 100 times using randomly select 90%of our data set, all the iterations achieved a p value of <0.001.

Model
3. Germination and acorn density (Fig 3D): Results of Likelihood ratio test to determine if the yearly proportion of acorns that germinated changes with yearly acorn density. We use generalized linear model with binomial errors. We assume the observations are independent between years, because the locations for acorn collection were randomly placed each year.  lines represent least squares fitted values and shaded areas are 95% confidence intervals. The points represent weekly acorn production for each year. Unlike averages across year where linear models were a better fit, quadratic models were fit to data on infestation over time.