In tropical forest canopies, host tree characteristics shape epiphyte communities, and both host tree characteristics and epiphytes determine invertebrate communities, e.g., ants. Tree height is among the factors most often mentioned as a strong predictor for both bromeliad and ant communities. However, many factors interact dynamically in shaping the tree-bromeliad-ant association. Here, we investigated the effects of the host tree and canopy structural characteristics on both bromeliads and ants. We sampled bromeliads on 180 trees and ants in 360 bromeliads living on a subset of 60 of those trees in a continuum of native forest fragments and agroforestry systems in the Atlantic Forest, a Brazilian biodiversity hotspot. We found that the host trees’ crown area had a positive effect on the abundance of bromeliads. Also, the introduced tree species had higher abundance of bromeliads than native tree species. Moreover, we found that the ant species composition was different between native and introduced trees. In addition, we observed a positive effect of the size of bromeliads and tree crown height on ant species richness on both bromeliad and tree scales, but there was no effect of tree height. Taken together, our findings highlight the importance of trees with large crowns for the maintenance of bromeliads, which are also associated with richer ant communities in both Atlantic Forest and agroforestry systems. These results emphasize the importance of trees with larger crowns for biodiversity conservation in both native forest fragments and agroforestry systems.

Study area

The study was carried out in a continuous patch of preserved fragment of Atlantic Forest and cacao agroforestry farm located in the micro-basin of the Una River in the Una municipality of the state of Bahia in northeastern Brazil (15° 17’ S, 39° 07’ W). This region is part of the Central Atlantic Forest Corridor, one of the main areas of endemism of the Atlantic Forest (Myers et al., 2000). According to the Köppen classification (Alvares et al., 2013), the climate of the region is hot and humid (Af), with no defined dry season. The annual precipitation is greater than 1,300 mm and can reach 1,600 – 1,800 mm in rainy years (Mori 1989). The average annual temperature is 24.5 ºC, with warmer periods between October and April (maximum of 38 ºC), and periods with markedly lower temperatures between June and August (minimum of 7 ºC) (Mori 1989). The annual average relative humidity is 85%. The soil is typical Dystrophic Red-Yellow Latosol (Ker 1994) and the dominant vegetation in the region is South Bahia Hygrophilous Forest (Gouvêa et al., 1976), or South Bahia Humid Forest, characterized by the presence of tall, broad-leaved evergreen trees with large numbers of lianas and epiphytes.

Sampling design

We sampled ants and bromeliads during the hottest and rainiest season of two consecutive years, from October 2013–April 2014 and from October 2014–February 2015. Each of the three sampling points (Figure 1) contained a fragment of Atlantic Forest and an area of cacao agroforestry. Although the agroforestry system supported high biodiversity, this habitat differs greatly from the preserved Atlantic Forest. Agroforestry areas contained few native tree species, the trees are more isolated, and the understory is almost completely occupied by cocoa trees. However, cacao agroforestry system is a suitable habitat for many species of native fauna, which occupy these areas to feed and nest. In the preserved Atlantic Forest, only native tree species were found. Some native trees had been selectively logged in the past, but the fauna and flora are currently protected by strict conservation policy. We chose different species of host trees (phorophytes). In the fragments of Atlantic Forest, we considered native trees, while in the cacao agroforestry matrix (Johns 1999), we included native trees as well as three introduced species: jackfruit (Artocarpus heterophyllus), erythrina (Erythrina spp.) and rubber tree (Hevea brasiliensis). We selected these introduced species because they are commonly used as shade trees in the mosaic of the cacao agroforestry landscape (Cassano et al., 2014). The fruits of jackfruit are widely used as food by the local population, as well as by native animal species such as the golden-headed lion tamarin, Leontopithecus chrysomelas Kuhl 1820, which forages and lives in these agroforests (Oliveira et al., 2011). Erythrina trees have been used as shade trees in cacao plantations since the 1960s based on the recommendation of the Centro de Pesquisa da Lavoura Cacaueira (CEPLAC), the Center for Research on Cacao Farming that guides the decisions of local farmers (Alvim 1966). Rubber tree was also introduced in the 1960s (Alvim & Nair 1986), and its use has now been strongly stimulated because the extraction of rubber in addition to coca increases the economic viability of farms through the diversification of crops.

First, we selected 60 central point trees in the continuum of preserved fragment of Atlantic Forest and cacao agroforestry areas. These trees were the emergent ones whose tree-crowns were directly exposed to sunlight, with a minimum circumference at breast height (CBH) of 130 cm, i.e., about 41 cm of diameter at breast height (DBH). They are at a minimum of 50 m apart from the next central point tree and having at least six bromeliads in the crown of the tree. These central point trees consisted of 24 individuals of native species (12 on Atlantic Forest and 12 on agroforestry system), and 36 individuals of the introduced species—12 erythrinas, 12 jackfruits and 12 rubber trees in the agroforestry system. We then selected two additional trees, close to the central point trees, that reached the canopy and had ≥130 cm of circumference at breast height (CBH), i.e., about 41 cm of diameter at breast height (DBH). This resulted in 60 central-point trees plus 120 study trees, for a total of 180 trees. We identified all trees to the species (or morphospecies) level.

Sampling of bromeliads and ants

We counted and identified all bromeliads with a total length (from root to leaf end) greater than 10 cm in each tree crown of the 180 selected trees. We conducted these counts from the ground, which although it can have some shortcomings (Flores-Palácios & García-Franco, 2001), is an efficient way to sample a large number of trees. Since we used the same ground-based approach across all samples, we were able to obtain sampling units that were comparable across all of our samples. We used the “Encyclopaedia of Bromeliads” (version 3.1; Gouda et al., 2012) to identify all bromeliads to the species or morphospecies level whenever possible. Within the genera of bromeliads inventoried, some morphospecies were grouped into complexes because they could not be identified to the species level. Of the 180 trees sampled, we excluded three tree individuals (1.7%) from the analyses: this consisted of one individual of Spondias mombin (Anacardiaceae), which was an exotic species not included among the list of three introduced species for sampling, and two rubber trees in which part of the sampled data was lost.

Only the 60-original central-point trees were used to study the ant fauna associated with bromeliads. We climbed each of these 60 central trees using an adaptation of the single rope climbing technique (Perry, 1978), in which one additional rope is added (double rope) in order to improve the climber’s movement within tree-crowns. We collected six bromeliads within a six-meter horizontal distance from the center of the tree crown, prioritizing those that were the largest and most accessible to the climber and as far from each other as possible, for a total of 360 individual bromeliads. We used the “canopy hamper” method to collect the bromeliads and the associated ant fauna together with all of the associated organic matter and suspended soil (for details of this sampling technique, see: DaRocha et al., 2015a, Delabie et al., 2021). The ants associated with the suspended soil were extracted by direct collection using a Winkler trap while still in the field (Bestelmeyer et al., 2000). The collected ants were preserved in 92% alcohol in the field and then transported to the laboratory where they were mounted and when possible, identified to the species level using the entomological reference collection (CPDC) of the Centro de Pesquisa do Cacau CEPEC.

Vegetation variables

We measured tree- and canopy-related variables. Tree height of the 177 trees was estimated using a stick of known length by a single observer. In order to obtain the height of the tree crown, we first estimated the height of the first fork of the trees. Then, we obtained the height of the tree crown by subtracting the height of the first fork from the total height of each of the 177 trees. To obtain the area of the tree crown we measured the extension of tree crown from the ground, i.e., the projection of the tree crown on the ground. We quantified the percentage of canopy cover for each tree using hemispheric photography with a digital camera and an 8-15 mm 180º fisheye lens (Rich 1990). Three hemispherical photographs were taken per tree, with the camera positioned at 1.8 m above the ground and 5 m away from the trunk of the tree, at 0 º, 120 º and 240 º. The percentage canopy cover per tree was estimated from the photographs using the RT4Bio package (Reis-Jr et al., 2015) in the R environment (R Core Team 2017).

To understand the relationship between bromeliad size and ant species richness, we determined the size of the bromeliads from which we sampled ants. The size (volume) of each of the 360 bromeliads collected were determined from the measurements in centimeters (cm) of the length of the innermost leaf of the rosette and the diameter of the rosette. We approximated the volume of the central part of the rosette to that of a cylinder (cm³), calculated as: V = πr² * h, where: r is the rosette radius (cm) and h is the length of the most internal leaf (cm) (DaRocha et al., 2015a).