1. Facilitation by nurse plants increases understorey diversity and supports ecological communities. In turn, biodiversity shapes ecological networks and enhances ecosystem functioning. However, whether and how facilitation and increased biodiversity jointly influence community structure and ecosystem functioning remains unclear.

2. We performed a field experiment disentangling the relative contribution of nurse plants and increasing understorey plant diversity in driving pollination interactions. Both the presence of nurse shrubs as well as increased understorey plant diversity increased pollinator diversity and visitation rates. While nurse and understorey diversity effects on pollinator visitation rates did not interact, the effects of increasing understorey plant diversity on pollinator diversity were stronger in the absence than in the presence of shrubs, meaning that nurse shrubs attenuated the effects of high understorey diversity and buffered the effects of low understorey diversity.

3. We also found positive complementarity effects among understorey species as well as complementarity between nurse plants and understorey species at high diversity. Results also indicate negative selection effects, suggesting that species with generally few pollinators benefit the most in the polyculture, while a species (possibly the nurse plant) with generally lots of pollinators does not. The corresponding changes in pollination networks with the experimental treatments were due to both changes in the frequency of visits and turnover in pollinator community composition.

4. Synthesis Plant–plant facilitative systems, where a nurse plant increases understorey plant diversity, are common in stressful environments. Here, we show that these facilitative systems positively influence mutualistic interactions with pollinators via both direct nurse effects and indirect positive effects of increasing plant diversity. Conserving and supporting nurse plant systems is crucial not only for maintaining plant diversity but also for supporting ecosystem functions and services.

In order to disentangle the role of direct nurse effects from the indirect effects of the nurse plant on pollinators through increased understorey plant diversity, and to further examine their joint effects on the mutualistic interaction between plants and pollinators, a fully-factorial experimental design was established with two factors (Fig. 1): (i) absence and presence of the legume shrub (open vs nurse), and (ii) understorey plant diversity with zero (0), one (1) and three (3) plant species. This design results in five treatment combinations of open–1, open–3, nurse–0, nurse–1 and nurse–3sp. The treatment combination open–0 was excluded due to the inherent lack of flowers for visitation.

A randomised block design was adopted by grouping together the five treatments and replicating them three times in each block over four blocks, for a total of n = 60 plots. Distance between plots within the same block was approx. 1 m. Blocks were distributed randomly over an area of about 4,800 m². Understorey plant species richness was manipulated with plants of the three species transplanted into pots and then the plots placed beneath Retama or in the open. The density of understorey plants and flowers was kept constant over all understorey diversity levels. Plants were taken from the neighbouring field and replaced every second day to assure fresh flowers were available. Pots were kept aggregated or separated (c. 30 cm apart) below the nurse or in the open. This additional factor (hereafter, aggregation) was replicated per block. In the nurse alone treatment with no understorey species, three empty pots with only soil were placed under the shrub to control for any potential effect of the pots. For the nurse plant treatments, Retama shrubs were chosen of approx. the same size (height 127–178 cm and width 125–220 cm). An area of 1 m² at 1 m height was used as the pollinator observation area in each shrub. Flowers of the surrounding vegetation within at least 1 m around each shrub and open area were cleared with a sickle.

Flower visitation was considered as a proxy for the ecosystem function of pollination (Schleuning et al., 2015; Winfree et al., 2018; IPBES, 2019) as it correlates well with seed production (Losapio & Schöb, 2020; but see Lázaro et al., 2013). Flower visits were documented by sampling, identifying, and recording all insects visiting the flowers of each plant in each plot. Observations were conducted between 9 AM and 7 PM over eight days between 1 May and 14 May 2017, covering the blooming phase of the four plant species. Each plot was observed during three surveys of 20 min randomly allocated over the day. Nurse plants and corresponding open plots were observed simultaneously, thereby reducing the potential confounding effects of changing weather conditions within blocks. Each block was sampled within two days. Pollinators were identified at the species level whenever possible, otherwise to the genus.

See the R script Losapio_JEcol.R