Ecological Archives E096192A4
María C. RodríguezRodríguez, Pedro Jordano, and Alfredo Valido. 2015. Hotspots of damage by anatgonists shape the spatial structure of plant–pollinator interactions. Ecology 96:2181–2191. http://dx.doi.org/10.1890/142467.1
Appendix D. Description of the methodology used to estimate the proportion of viable seeds consumed per fruit by predispersal seed predators, plus a table containing the estimated interaction strengths per animal guild.
Methods
Estimation of the proportion of viable seeds consumed per fruit by predispersal seed predators
We defined the proportion of viable seeds consumed by seed predators (P_{sc}) in the infested fruits as P_{sc} = (S_{i} – S_{r}) / S_{i}, where S_{i} is the initial number of viable seeds produced by the fruit, and S_{r} the number of remaining viable seeds after larvae predation.
The estimation of P_{sc} was done in three steps. First, we separated the infested fruits in the respective plant (total n = 185 fruits from 71 plants) from the uninfested ones, and counted the number of remaining viable seeds within the infested fruits (S_{r}). Second, we predicted the initial number of viable seeds present in the fruit before predation (S_{i}) by linear regression. For this, we used data only from uninfested fruits produced by those plants that interacted with seed predators, instead of data from all studied plants. This procedure avoided bias in the estimation of S_{i} caused by mother plant effects. Concretely, we fit a zeroinflated model with the number of viable seeds per uninfested fruit as response variable, and fruit width as explanatory variable (negative binomial family, log link function). After model fitting, we predicted S_{i} for the infested fruits substituting their respective values of fruit width into the regression equation. Third, we estimated P_{sc} using the abovementioned formula after obtaining S_{i} and S_{r}values.
Results
Table D1. Estimates of the interaction strengths between Isoplexis canariensis and its animal mutualists and antagonists, accompanied by the proportion of plants that interacted with each functional group, or any of their animal guilds. See Appendix A for the taxonomic composition of animal guilds.
Interacting agent 
Interaction strength 

Plants 

n_{freq} 
Frequency 
n_{int} 
Intensity 
Frequency × Intensity^{a} 
CV 

n_{pl} 
Proportion 

Mutualists 











Legitimate bird pollinators 
119 
0.028 ± 0.030 
115 
0.162 ± 0.178 
(4.6 ± 6.1)·103 
132.6 

99 
0.83 

Facultative bird pollinators 
119 
0.002 ± 0.003 
117 
0.053 ± 0.136 
(0.3 ± 0.8)·103 
266.7 

35 
0.29 

Legitimate lizard pollinators 
119 
0.002 ± 0.006 
115 
0.031 ± 0.079 
(0.3 ± 0.9)·103 
300 

30 
0.25 

All mutualists 
 
 
 
 
(5.2 ± 6.2)·103 
119.2 

103 
0.87 
Antagonists 











Floral herbivores 
119 
0.66 ± 0.33 
119 
0.18 ± 0.17 
0.16 ± 0.17 
106.3 

112 
0.94 

Nectar larcenists 
119 
0.06 ± 0.14 
119 
0.03 ± 0.09 
0.01 ± 0.04 
400 

24 
0.20 

Predispersal seed predators 
118 
0.09 ± 0.13 
113 
0.40 ± 0.42 
0.08 ± 0.12 
150 

72 
0.61 

All antagonists 
 
 
 
 
0.25 ± 0.24 
96 

115 
0.97 
Notes: Data included in the table are means ± SD. Frequency indicates the proportion of plant censuses (mutualists) or surveys (antagonists) in which the plantanimal interaction was observed. Intensity represents the proportion of flowers probed by mutualists per plant visit, or the proportion of reproductive units (flowers/inflorescences/fruits) damaged by antagonists per plant survey; see Methods in the article. Coefficient of variation (%) is calculated for the product of frequency and intensity. Sample sizes indicate the number of plants in which the interaction frequency (n_{freq}) and intensity (n_{int}) was estimated for the respective animal functional group or guild. a Due to logistic constraints, the data set contained missing values (n = 16 out of 714 values, 15 plants) in the interaction strengths of mutualists (legitimate bird pollinators n = 4, facultative bird pollinators n = 2, legitimate lizard pollinators n = 4) and antagonists (predispersal seed predators n = 6). These values were estimated by multiple imputation using Bayesian regression models (Gaussian family; mi package in R, Su et al. 2011). The sample size n_{pl} indicates the number of plants (out of 119) that had contact with the corresponding animal functional group or guild.
Literature cited
Su, Y. S., A. Gelman, J. Hill, and M. Yajima. 2011. Multiple imputation with diagnostics (mi) in R: opening windows into the black box. Journal of Statistical Software 45:1–31.