Current climate change is disrupting biotic interactions and eroding biodiversity worldwide. However, species sensitive to aridity, high temperatures and climate variability might find shelter in microclimatic refuges, such as leaf rolls built by arthropods. To explore how the importance of leaf shelters for terrestrial arthropods changes with latitude, elevation, and climate, we conducted a distributed experiment comparing arthropods in leaf rolls vs. control leaves across 52 sites along an 11,790 km latitudinal gradient. We then probed the impact of short- versus long-term climatic impacts on roll use, by comparing the relative impact of conditions during the experiment versus average, baseline conditions at the site. Leaf shelters supported larger organisms and higher arthropod biomass and species diversity than non-rolled control leaves. However, the magnitude of the leaf rolls’ effect differed between long- and short-term climate conditions, metrics (species richness, biomass, and body size), and trophic groups (predators vs. herbivores). The effect of leaf rolls on predator richness was influenced only by baseline climate, increasing in magnitude in regions experiencing increased long-term aridity, regardless of latitude, elevation, and weather during the experiment. This suggests that shelter use by predators may be innate, and thus, driven by natural selection. In contrast, the effect of leaf rolls on predator biomass and predator body size decreased with increasing temperature, and increased with increasing precipitation, respectively, during the experiment. The magnitude of shelter usage by herbivores increased with the abundance of predators and decreased with increasing temperature during the experiment. Taken together, these results highlight that leaf roll use may have both proximal and ultimate causes. Projected increases in climate variability and aridity are therefore likely to increase the importance of biotic refugia in mitigating the effects of climate change on species persistence.

Experiments at all 52 sites were conducted following a standardized protocol. Each experiment followed a randomized block design. For each site, we randomly selected 10 to 20 paired trees (hereafter, a pair of plants is referred to as a block). Paired plants within a block were at least 2 m apart. Blocks were at least 6 m apart from each other. In most sites, the experiment was conducted using a single native plant species, typically the locally most common species. However, in some tropical forests, due to high species diversity and low relative density, we used different plant species among blocks, with plants within a pair always belonging tothe same species. We used only broadleaf native plant species that did not exhibit any obvious type of indirect defence (e.g., domatia, extra-floral nectaries, glandular trichomes).

One plant per block was randomly selected as a control plant, while the second plant was used for the leaf-rolling (shelter addition) treatment. On each plant, we selected 5-10 fully expanded leaves without obvious damage (hereafter referred to as ‘sample unit’). Prior to the experiment, any arthropods present on experimental leaves were removed by hand. Then, the leaves of the treatment plant were rolled by hand from the adaxial to the abaxial surface across the leaf axis to form a cylinder approximately 0.6 cm in diameter to mimic shelters built by caterpillars - a phenomenon occurring in at least 17 moth and butterfly families, including Hesperiidae, Nymphalidae, Gelechiidae, Oecophoridae, Lasiocampidae, Pyralidae, Gracillariidae, Tortricidae, Geometridae, Erebidae. The leaf rolls were secured with a metal hairpin. In the control plants, 5-10 unrolled leaves were marked with a metal hairpin. Rolled and control leaves were exposed for 10 days in the field – this was deemed sufficient as previous bioassays showed that leaf shelters can be colonized very quickly (within 24 hours) and that species richness within leaf rolls reaches saturation within a few days. Moreover, leaf abscission was observed in some plant species during the 10-day experiment (e.g., Psychotria, G.Q. Romero, pers. obs.), thus precluding experiments of longer duration. Maximum width of both control and rolled leaves was measured as an estimate of leaf size.

After 10 days of the experiment, we collected rolled and control leaves and stored them grouped by replicates and treatments; both control and rolled leaves were quickly placed into a zip-lock plastic bag and sealed. The leaves were either frozen for later sorting or immediately sorted to collect the invertebrates. We collected all the invertebrates visible to the naked eye (except mites) and stored them in ethanol. Mites could not be considered in the analysis, because identifying both mite taxa and feeding guilds is extremely difficult, with very few specialists worldwide. We identified the invertebrates to the lowest taxonomic level possible and classified them into morphospecies and feeding guilds (i.e., predator, parasitoid, herbivore, detritivore, omnivore). However, the sample size was only sufficiently large for separate analyses of predators and herbivores. Individual body size (dry body mass) was estimated from the dry mass (dried at 70<spanstyle="font-family:cambria; font-size:8pt;="" vertical-align:super"="">oC for 24h) or by measuring total body length and then calculating the dry mass using published taxon-specific allometric equations (Hódar, 1996). Three dependent variables were used for the analyses: species richness, community biomass, and mean individual body size. As expected, total abundance explained much of the species richness (R2 = 0.78, F1,50 = 185.5, P<0.001). Because of the small subsamples (5 to 10 leaves per plant), which would lead to highly stochastic results, we were unable to determine rarefied species richness. Therefore, we focus our analyses on the number of species (richness) and assume that this is a product of increased overall abundance. Arthropod biomass represents the sum of all individual body masses. Arthropod richness and biomass were weighted by the number of leaves sampled per plant (e.g., number of species per leaf), and average individual body size was calculated for each sampling unit by dividing the total biomass for each feeding guild (predator or herbivore) by the number of arthropods within that feeding guild - thus, the sample unit for body mass is the average mass per arthropod upon leaves.