Leaf shelters function as microclimatic refuges, reducing arthropod exposure to climatic fluctuations of surrounding habitats. Although facilitation is expected to increase under stressful conditions, empirical studies investigating the patterns of variation and magnitude of effects of ecosystem engineering at different spatial and temporal scales are still scarce. In this study we evaluated the facilitation consequences of leaf shelter created by gall-inducers on arthropod communities of Miconia ligustroides (DC.) Naudin (Melastomataceae). We evaluated how such effects change at the leaf and plant levels in an environment subject to strong climatic seasonality. The presence of leaf shelters on M. ligustroides increased arthropod diversity and biomass, modified the species composition at both the leaf and plant levels, and in wet and dry seasons. However, the addition of artificial leaf shelters during the dry season showed greater abundance, richness, and biomass of arthropods when compared to shelters added during the wet season. Regarding the global effects of artificial leaf shelters on the diversity of arthropods associated with M. ligustroides, the dry season showed strong and positive effects, increasing the abundance, richness, and biomass of arthropods by an average of 65% for both years. Our study contributes to a better understanding of the patterns of variation and magnitude of ecosystem engineering at different spatial and temporal scales and provides new insights into the importance of shelters for aridity-sensitive species.

The raw data available here were used to run the data proposed in the following methodology. For more details, see the published article, as well as Table 1 and Supporting Information S1.

Data sampling

Leaf scale

To evaluate the effects of leaf shelters on arthropod secondary colonization, 30 plants spaced approximately 10.0 m from each other were marked in the field in February 2020. All plants were between 1.0 and 1.5 m tall and had no flowers and fruits at the beginning of the study. In each plant, two pairs of leaves were selected as treatments (n = 4, artificial leaf rolls), and two pairs of leaves as control (n = 4, expanded leaves). We used different branches for each pair and marked the branches with colored tags (Figure 3).

The treatment pairs consisted of a rolled leaf simulating a shelter created by D. gallaeformans (Figure 2d). We use young, intact, and fully developed leaves to create these artificial shelters. We opted to use intact leaves to create the shelters because the galled leaves are extremely fragile and break easily when manipulated. Furthermore, a previous study showed that the use of these artificial shelters resembles natural shelters, as they exhibited 79% of species similarity of secondary occupants with natural shelters (Pereira et al., 2021).

Leaves were manually rolled from the abaxial to the adaxial face, from the edges to the midrib, and secured with hairpins, which were painted the same color as the leaves with odorless green spray paint (Figure 2d). These artificial shelters simulate cylindrical rolls, 20 mm in diameter, similar to leaf roll galls naturally created on these plants by the nematode. In previous experiments (Pereira et al., 2021), we showed that 10 days were sufficient for the colonization of these shelters. Thus, after 15 days of setting up the experiment, artificial rolls and expanded leaves were collected, stored in plastic bags with zipper closure, and frozen for later sorting. After this collection, we set up the experiment again and carried out a second collection after 15 days. The treatments were reapplied to the same trees and branches using different pairs of leaves.

To evaluate the effects of shelters over seasons, this experiment was repeated on the same individuals in August 2020 (dry season), as well as in the wet (February) and dry (August) seasons of 2021.

Plant scale

To assess whether leaf roll galls increase arthropod diversity in M. ligustroides, 60 plants at least 10.0 m apart were tagged in January 2020 (wet season) and two groups were selected: i) plants with leaf roll galls (n = 30, control) and ii) plants with galls removed (n = 30, treatment). All plants were between 1.0 and 1.5 m tall and had no flowers and fruits at the time of the study. Individuals of M. ligustroides had an average of ten leaves curled from galls of similar age, and for the plants in the treatment group, we removed all leaves that had galls (as there were no plants naturally without galls). This represents approximately only 1% of biomass removal, as plants on the treatment group had an average of 1,021 (± 106.99 SE) leaves (see Pereira et al. 2021). One month after marking the groups of plants (February 2020) and removing the galls from the treatment group, the entire plants were visually inspected for 25 minutes and the arthropods were collected using entomological forceps and aspirators. Collections were carried out between 10 am and 2 pm, the period when we observed the highest arthropod activity in the studied area. We used zippered plastic bags to collect the shelters and froze them for arthropod screening. To assess the effects of shelters over seasons, this experiment was repeated on the same individuals in July 2020 (dry season), as well as in the wet and dry seasons of the following year (January and July 2021, respectively).

Arthropod screening and identification

We inspected the collected leaves and shelters using a stereomicroscope. All arthropods found were stored in 70% alcohol and identified to the lowest possible taxonomic level or classified into morphospecies (Oliver & Beattie, 1996) to allow the assessment of abundance, richness and dry biomass (in mg). Furthermore, we classified arthropods into feeding guilds as detritivores, herbivores, omnivores, parasites, parasitoids, and predators. All arthropods were dried in an oven at 60°C for 24 hours and weighed on a precision digital scale to estimate biomass (in mg).

Data analysis

We used generalized linear mixed models (GLMMs) to evaluate the effects of shelters on arthropod diversity and biomass at the leaf level (Experiment 1). For this, we used the average values of abundance, richness, and biomass as response variables, each treatment as a fixed factor and individual plants as a random factor.

At the plant level (Experiment 2), the effects of shelters on arthropod diversity and biomass were also assessed by GLMMs. We used the average values of abundance, richness, and biomass as dependent variables, each treatment as a fixed factor and individual plants as random factors. The distribution of errors in the aforementioned analyzes was verified using restricted maximum likelihood (REML).

We used analysis of similarities (ANOSIMs) based on Euclidean distances to evaluate arthropod composition at the leaf and plant levels between different treatments. One-sided significance was calculated by permuting the groups with 9,999 permutations. ANOSIMs paired between all pairs of groups were used as a post-hoc test. To view similarities or differences between treatments, we used non-metric multidimensional scaling (NMDS) analyses, using species abundance for each individual plant sampled. All analyzes mentioned above were performed using the Vegan package (Oksanen et al., 2013) in the R software (R Core Team, 2022).

We used Hedges'd metric to estimate the magnitude of the effects of adding leaf shelters on arthropod abundance, richness and biomass in the dry and rainy seasons of 2020 and 2021 (Hedges & Vevea, 1998). For replicates among treatments, the mean and standard deviation from the 4 leaves per plant, per treatment, were used. Overall effect sizes were calculated on the response variables of arthropod abundance, richness and biomass. Groups were assigned to expanded leaves (control) and leaf rolls (treatment). To estimate the cumulative effect (E++) of treatments, the individual d_i effects were combined using weighted averages and a random model analysis. A positive effect size indicates that abundance, richness, and biomass of arthropods were lower on expanded leaves compared to leaf rolls, whereas a negative effect size implies a lower abundance, richness and biomass of arthropods for the leaf rolls compared to expanded leaves. As a convention, E++ values around 0.2 are considered weak effects, values around 0.5 are considered of moderate magnitude, values around 0.8 are considered strong, and E++ larger than 1.0 are considered very strong (Rosemberg et al., 2000). The cumulative effects were considered significant if the confidence intervals (95%) did not overlap with zero. All analyses were conducted in Metawin 3.0 (Rosemberg, 2023).