There is a consensus that diversity enhances ecosystem functioning (

It is well known that ecosystem productivity is positively associated to species diversity (

In addition to plant diversity, greater temporal stability of floral resources in diverse communities (

To evaluate whether there is an effect of flower diversity on pollinator reproduction it is necessary to disentangle the effect of flower abundance, as it could be positively correlated with flower richness, as it happens with biomass in plant communities (

Our aim is to study the effect of flower diversity and temporal stability of floral resources on the reproduction of a cavity nesting bee assemblage from the Monte desert in Argentina. Based on the above arguments, we expected to find that flower diversity and temporal stability of floral resources correlates positively to three estimates of bee reproduction at the population and community levels: average number of brood cells per nest per site, total number of brood cells per site, and total number of nests per site. We also expected to find a positive correlation between the strength of the reproductive output-diversity correlation and the degree of generalization of each bee species.

This study was conducted in the Monte desert in Villavicencio Nature Reserve, located ca. 40 km north of Mendoza city, Argentina, during the 2008 flowering season (15 October–8 December 2008; authorized by Dirección de Recursos Naturales Renovables de la Provincia de Mendoza, approval numbers 1130 and 646). We worked in fourteen 100 × 200 m rectangular study sites (minimum and maximum distance between them were 1.11 and 14.13 km, respectively). These sites lie at 1,100–1,500 m above sea level, at the ecotone between the Monte desert and the Prepuna biomes (

We placed trap nests in six points per plot as shown in

Floral resource availability was studied using flower density, as flowers represent the resource packages encountered by pollinators as they forage (see also

To evaluate the effects of flower diversity and temporal stability of floral resources on bee fitness and to assess the influence of other ecological factors on this relationship, we used structural equation models (hereafter SEM). We built a general initial model to explore the effects of flower richness, flower abundance, time elapsed since the last fire, elevation, and temporal stability in flower production on bee reproductive parameters. We estimated bee productivity at the community level using two proxies: total number of brood cells per site, and total number of nests per site. To evaluate reproduction at the species level we used three proxies: average number of brood cells per nest, total number of brood cells per site, and total number of nests per site. To estimate the average number of brood cells per nest we used only data of sites where species were present; for the total number of brood cells per site and the total number of nests per site we used data of all sites, as the absence of a species in a site represented zero abundance. Flower richness was used as a proxy of flower diversity; it was rarefied to remove the effect of flower abundance. Flower abundance was estimated as flower density per site. Time elapsed since the last fire was provided by park rangers (E. L. Stevani, personal communication). Temporal stability in flower production along the season was calculated as the inverse of its coefficient of variation (see

(A) Complete model. (B) Nested models generated by removing variables with non-significant effects or small path coefficients that were non-significant. Model 2 was selected by ΔAIC for all bee species.

We evaluated alternative SEM models using a d-separation test (_{i}_{i}_{w}

To evaluate whether the effect of flower diversity becomes stronger with increasing pollinator generalization, we performed Spearman’s rank correlations between the path coefficient representing the effect of flower richness on each of the three bee reproductive parameters mentioned above, and two measures of the corresponding species degree of generalization. We estimated the degree of diet generalization of each bee species using the species degree and Simpson’s diversity index; degree is simply the number of food species consumed from all sites polled, whereas Simpson’s index is a function of the number of food items and the proportion in which they were consumed. We used rarefaction to estimate both measures of generalization to make them comparable among bee species, as the number of brood cells was highly variable among nests. A positive correlation between the path coefficient of flower richness on bee reproduction and generalization would support our hypothesis that the reproduction of generalist pollinators is enhanced by flower richness.

All analyses were performed using R statistical software (

We recorded 598 occupied trap nests by 11 solitary bee species (

We used in this study the species that had more than 30 nests.

Bee species | Occupied trap nests |
---|---|

6 | |

54 | |

31 | |

39 | |

222 | |

17 | |

74 | |

3 | |

59 | |

88 | |

5 |

The complete model assessing the influence of multiple ecological factors on the potential relationship between flower diversity and bee reproduction at a community level (Model 1,

Model | Variables | Path coefficients | |
---|---|---|---|

1 | Elevation → Flower richness | 0.39 | 0.15 |

Elevation → Flower abundance | −0.51 | 0.06 | |

Flower richness → Stability | −0.54 | 0.04 | |

Elevation → Total brood cells | −0.59 | 0.03 | |

Elevation → Total nests | −0.57 | 0.04 | |

Flower abundance → Total brood cells | 0.26 | 0.43 | |

Flower abundance → Total nests | 0.24 | 0.48 | |

Flower richness → Total brood cells | 0.27 | 0.80 | |

Flower richness → Total nests | 0.38 | 0.23 | |

Stability → Total brood cells | 0.35 | 0.35 | |

Stability → Total nests | 0.30 | 0.35 | |

Time elapsed since last fire → Total brood cells | 0.35 | 0.28 | |

Time elapsed since last fire → Total nests | 0.45 | 0.14 |

The complete model assessing the influence of multiple ecological factors on the potential relationship between flower diversity and each bee species fitness (Model 1,

In each box plot, the middle line indicates median, box limits are the first and third quartiles, whiskers indicate most extreme points ≤ 1.5 times the interquartile range, and circles indicate outliers of the seven path coefficients of the corresponding effect. Model 2 describes the effect of flower diversity (estimated using flower richness), temporal stability of flower production along the flowering season (estimated as the inverse of the coefficient of variation of the weekly flower abundance mean), and elevation (m above sea level) on three bee reproductive variables: “Average cells,” the average number of brood cells per nest per site; “Total cells,” the total number of brood cells per site; and “Total nests,” the total number of nests per site. The ordinates represent the path coefficients; the abscissa represent the effect of the above ecological variables on bee reproductive variables. Blue error bars are the ninety-five percent confidence limits of path coefficients obtained from bootstrap sampling of the distribution of path coefficients.

The effect of flower diversity on pollinator reproduction was unrelated to pollinator generalization for any of the bee reproductive variables and generalization indexes used (

Generalization index | Reproductive variable | Correlation coeficient ( |
||
---|---|---|---|---|

Degree | Average number of cells per nest | −0.21 | 0.66 | 7 |

Degree | Total number of cells per site | 0.42 | 0.35 | 7 |

Degree | Total number of nests per site | 0.39 | 0.39 | 7 |

Simpson’s diversity index | Average number of cells per nest | 0.14 | 0.78 | 7 |

Simpson’s diversity index | Total number of cells per site | 0.46 | 0.30 | 7 |

Simpson’s diversity index | Total number of nests per site | 0.28 | 0.55 | 7 |

Contrary to our expectations, we found no effects of flower diversity and flower abundance on bee reproduction, either at the community or at the species level. Thus, flower diversity did not matter for the reproduction of the solitary bees studied here. Considering the ecosystem functioning context where relationships are commonly saturating (

An explanation of the negative effect of temporal stability on brood cell production concerns a compensatory behavior of females to avoid parasitism. In sites with high temporal stability in flower production, females might lay fewer eggs per nest while building more nests, so as to maximize larval survival per site. This reasoning makes two implicit assumptions. First, that the bee species are parasitized, which we indeed observed for many of the bee species studied here. Second, that nesting sites are not limited for the population. In fact, the trap nest sampling with replacement highly increased the nest availability in our study sites. If this mechanism were responsible for the observed negative effect of temporal stability in flower production on the average number of brood cells per nest, we would expect the number of cells per site to be either unrelated to temporal stability or to be higher in the more temporally stable sites, and the number of nests to be higher in the more temporally stable sites, as females would be laying eggs at their maximum capacity but distributing them in more nests. Matching these expectations, the total number of brood cells per site was unrelated to temporal stability (

An alternative explanation of the negative effect of temporal stability on brood cell production per nest could be that elevation might be weakening the effects of other variables on pollinator reproduction. This is particularly likely considering the positive direct effect of elevation on flower diversity, the positive indirect effect of elevation on stability, and the negative effect of elevation on bee reproduction (

We found no support for the idea that generalist bees are more favored in their reproduction by flower diversity than specialized ones, despite bee species included in this study having contrasting degrees of feeding specialization. Again, we think the negative effect of elevation on bee reproduction can be responsible for this unexpected result. It seems reasonable to think that species will respond idiosyncratically to flower diversity and stability when there is context dependency, given our finding of no general effects of flower diversity on bee reproduction.

Although there is a consensus that diversity promotes ecosystem-level productivity (

The variables in brackets are the independent ones and variables in {} are the conditional ones.

Trap nests consist of wood pieces with a longitudinal hole of three different diameter where bee species nest. Each occupied trap nest constitutes one bee nest.

Model 1 describes the effect of flower diversity (estimated using flower richness), flower abundance (estimated using flower density) and temporal stability of flower production along the flowering season (estimated as the inverse of coefficient of variation of the weekly flower abundance mean), fire (estimated as time elapsed since last fire), and elevation (estimated as meters above the sea). Y axis represents the path coefficients that rank from 1 to −1, and the x axis represents the effect of above mentioned ecological variables on three bee reproductive variables: A is the average brood cell number per nest per site, B is the total number of brood cells per site, and C is the total number of nests per site. In blue color are the ninety-five percent confidence limits of path coefficients obtained from bootstrap sampling of the distribution of path coefficients.

Column names: “codigo.i” refers to bee species, “sitio” refers to site names, “mean.no.celd” is the main number of brood cells per nest per site, “celd.tot” is the total number of cells, “nidos.tot” is the total number of nests.

Column names: “sitio” refers to site name, “abund.flores” is the flower abundance, “raref” is the flower richness rarefied, “altitud” is the altitude, “anio.q” is the time elapsed since last fire. The other columns were not used in this study.

Column names: “id.trampa” is a identification number for bee individuals, “codigo.i” is the bee species, “site” is name of the site, “codigo.p” is the plant species consumed, “sum.sp” are the number of pollen grains observed, “sum.tot” is the total pollen grains observed for that bee individual (this was used to calculate the proportion of use), “prop” is the proportion of use of a given plant by a bee individual.

We thank the administration of Villavicencio Natural Reserve for permission to conduct this study, the park rangers for help to find appropriated study sites in the field, Arturo Roig for help with bee identifications, Leticia Escudero, Nydia Vitale and Georgina Amico for laboratory assistance, and members of the Ecological Interactions Lab for helpful comments on the manuscript. JD is a postdoctoral fellow and DPV a career researcher with CONICET.

The authors declare that they have no competing interests.

The following information was supplied relating to field study approvals (i.e., approving body and any reference numbers):

Dirección de Recursos Naturales Renovables de la

Provincia de Mendoza 1130 and 646.

The following information was supplied regarding data availability:

The raw data was supplied as