Assessing the biodiversity and the impact of pollinators on carob production

sampling weeks), tree sex (Male or Female), tree side (N, S, E, W), A. mellifera presence and wasp presence (present or absent). Interactions between the fixed factors were not modelled in either analysis, as their inclu- sion led to model convergence problems. Tree was included as a random factor for both analy- ses. Including both plot and tree nested within plot as random factors resulted in a singular model fit, indicating that the random effect structure was too complex to be supported by the data. Model fitting was carried out using the Bobyqa algorithm with the maximum number of iterations set to 2*109. The significance of fixed effects was assessed using F-tests, with denomi- nator degrees of freedom obtained by running the model in lme as a linear mixed effects model (package nlme) [38], as discussed by Bolker [39]. Model diagnostics were performed in the DHARMa package using the function simulateResiduals [40]. The function runs tests for correct distribution (KS test), dispersion and residuals, as well as the Levene test for homoge- neity of variance. All models fitted the data well. For the carob pollination study, the data on the number of pods produced in open vs wind polli- nation were analyzed in a generalized linear mixed effects model framework using the package glmmTMB (function glmmTMB) [41]. Preliminary data analyses showed that the pod number data were over-dispersed, with greater variation than that predicted by the poisson model. The package glmmTMB allows for the modeling of data using a log link, and a negative binomial distri-

bution (family nbinom1, where the variance increases linearly with the mean). As fixed effects we included Treatment (open–wind pollination), carob grove (SE Lowlands 1, NW Coast 4, S. Plateau 1) and their interaction, as well as the number of individual inflorescences per inflorescence group. Tree and branch nested within tree were incorporated in the model as random effects. The signifi- cance of fixed effects was assessed using the Anova function of the car package [42], which applies Wald chi-square tests. We point our readers to the ongoing discussion about the use of p-values and the different options available to evaluate models at the GLMM FAQ site [39]. The data on pod weight were analyzed in a linear mixed effects framework (function lmer) in the package lme4 [37]. The model included treatment (open–wind pollination), carob grove (SE Lowlands 1 vs NW Coast 4 vs S. Plateau 1) and their interaction as fixed effects. The data from SE Lowlands were not included in the analysis as there were only five pods produced by a total of three carob trees in the wind pollination treatment. Tree, branch nested within tree, and inflorescence group nested within branch within tree were incorporated in the model as random effects. Degrees of freedom for F-tests were estimated with Satterthwaite’s approxima- tion as implemented in the ANOVA function of the package lmerTest [43]. Model diagnostics were performed in the DHARMa package [40] using the function simu- lateResiduals, as outlined in the bee presence-absence study. All models fitted the data well,

with the exception of pod weight data where there was a slight departure from variance homo- geneity of the simulated residuals, which did not seem to be a major issue after checking the plot of simulated scaled residuals. Results Bee presence-absence study The overwhelming majority of insect visitors to carob flowers belonged to the order Hyme- noptera, i.e. domesticated and wild bees and wasps (Figs 2 and 3), while negligible visits by Lepidoptera or Diptera (Syrphidae) were detected. Apis mellifera was the most common spe- cies, on both male and female trees, followed by wild bees and wasps (Fig 2). As shown on Fig 3, there was a trend of more insect presences at the beginning of carob blooming. Identified wild bees belonged to ten different species, representing three families: Halictidae with Lasioglossum sp. and Ceylalictus variegatus, Andrenidae with Andrena bicolor, Andrena PLOS ONE Biodiversity and the impact of pollinators on carobs PLOS ONE | https://doi.org/10.1371/journal.pone.0291431 October 20, 2023 5 / 16