Fig 1.
(A) Location of the five study sites in central Spain, namely the Abadía Retuerta vineyard, the Concejiles and Chaparrito fruit orchards and the Vistalegre and El Peral olive groves. (B) Distribution of the initial 164 nest boxes (black dots) in the Abadía Retuerta vineyard. (C) Distribution of the initial 101 nest boxes (black dots) in the Concejiles fruit orchard. (D) Distribution of the initial 111 nest boxes (black dots) in the Chaparrito fruit orchard. Light gray polygons indicate distant areas without nest boxes used for sentinel caterpillar experiments. Images were taken from [28] under a CC BY license, with permission from Google Maps/Google Earth (URL: https://www.google.es/intl/es/permissions/geoguidelines.html), original copyright 2015. The four panels are not identical to the original images, and are therefore for illustrative purposes only.
Fig 2.
Relevant categories of use for nest boxes installed in the (A) Abadía Retuerta vineyard, (B) Concejiles fruit orchard and (C) Chaparrito fruit orchard, across the four years of study.
Bird breeding refers to verified hatched eggs or recruited fledglings (categories 4–5 of nest box use, see Methods). Bars for insectivorous birds (Paridae) and sparrows refer to categories 2 (finished nest) to 5 (recruited fledglings). Nearly all observations contributing to the “Rodents” bar refer to Garden dormouse (category 6).
Table 1.
AICc-based comparison of alternative models of bird breeding at the Abadía Retuerta vineyard and Concejiles and Chaparrito fruit orchards, as a function of year (Yr), distance to natural or semi-natural vegetation (Dist), and occupancy of adjacent nest boxes (OcAN).
Pairwise interactions of these variables were also included (*). The null model included only a constant term: the mean of observed values. The best models for Abadía Retuerta had no random effects, whereas models for Concejiles and Chaparrito included a random effect of nest box on the intercept. The best models (lowest AICc) are indicated in boldface type. The indicated number of parameters describes the best model, or the most complex model when there were multiple best models.
Fig 3.
Model-averaged estimated effects of (A-C) occupancy by birds of adjacent nest boxes and (D-E) distance to natural or semi-natural vegetation on bird breeding at the (A) vineyard, (B, D) Concejiles fruit orchard, and (C, E) Chaparrito fruit orchard, across the four years of study.
Breeding was measured as the number of laid eggs or chicks/fledglings that were counted at a nest box. A mean distance to natural or semi-natural vegetation of 400 m was used to generate model predictions in panels B and C. Zero occupancy of adjacent nest boxes was used to generate model predictions in panels D and E. Distance to natural or semi-natural vegetation did not apply to the vineyard.
Fig 4.
Predation of sentinel caterpillars samples closer to, or farther away from, active nest boxes and their corresponding paired samples in distant areas without nest boxes at (A) the vineyard, (B) Concejiles fruit orchard and (C) Chaparrito fruit orchard, across the four years of study.
In the vineyard, the sample closer to the active nest box lay <6 m from the box and the sample farther from the active nest box lay >15 m away from the box. In the fruit orchards, the sample closer to the active nest box was 3 m away and the sample farther from the active nest box was 10 m from the box.
Table 2.
AICc-based comparison of alternative models of predation of sentinel caterpillar samples (%) at the Abadía Retuerta vineyard and the Concejiles and Chaparrito fruit orchards, as a function of Treatment (active nest boxes vs. control sites) and Proximity to active nest boxes (close vs. farther away samples).
Interaction of these two variables (*) was also included. The null model included only a constant term: the mean of observed values. All models, including the null model, had the same random structure, representing the effect of nest box on the intercept. The best model (lowest AICc) is indicated in boldface type. The number of parameters, conditional R2 (fixed and random effects) and marginal R2 (only fixed effects) refer to the best model.
Table 3.
Estimated food consumption, expressed in kg per breeding season (b.s.) or kg ha-1 per breeding season, for insectivorous species and field sites.
Values shown are the mean ± standard deviation at each field site across the four breeding seasons.