Green roofs are a nature-based solution with the potential to support urban arthropod biodiversity. However, evidence remains inconclusive, as many roof characteristics and arthropod groups are still understudied. We used a multi-taxa and functional trait approach focused on soil and ground-dwelling arthropod communities to assess how green roof features related to structure and design, plant community, management, and surrounding landscape influence the species richness and abundance of arthropods (overall and of groups defined by vertical distribution, body size, mobility, and feeding habits). We sampled arthropods (N=7155, 244 species) using pitfall traps in 20 green roofs from central Argentina, and recorded green roof characteristics (i.e., size, age, height, substrate depth and composition, plant species richness and composition, irrigation, and landscape green cover). Substrate depth was the most important structural variable, being positively linked to the species richness of euedaphic and hemiedaphic arthropods and other groups. Additionally, substrates containing plant litter and sand benefited detritivores. Plant composition influenced the richness and abundance of herbivores and parasitoids, probably due to the presence of specific resources. Irrigation increased total abundance and richness of hemiedaphics and small arthropods, although total richness was unaffected. Landscape green cover was positively related to parasitoid and small-sized arthropod abundance, highlighting the role of green spaces as sources of individuals. Our findings demonstrate that green roofs can support diverse communities of soil and ground-dwelling arthropods with varied functional traits, offering practical insights for designing biodiverse green roofs and conserving species within urban ecosystems.

Study area

The study was performed in Córdoba, central Argentina, where 20 green roofs were selected. Five roofs were located in the Paravachasca Valley, eight in the region of Sierras Chicas, and seven in the city of Córdoba (Fig. S1). The green roofs of the Paravachasca Valley and Sierras Chicas corresponded mostly to family homes following bioconstruction techniques. In the city of Córdoba, however, green roofs belonged to family homes and commercial buildings to better represent the green roofs in the city (Suarez et al., 2016). 

Green roof and landscape variables

Structural and design variables were measured on each roof or obtained through interviews with the roof owners. For green roof plant community variables, we sampled all the plant species on each roof for 15 minutes, collecting specimens that were later herbarized at the laboratory. Plant species were identified following the criteria of the Darwinian Botanic Institute (Zuloaga & Morrone, 2017), and the species richness for each roof was calculated. In addition to species richness, we also considered plant species composition as an explanatory variable of arthropod communities. Therefore, we performed a Principal Component Analysis (PCA; Hotelling, 1933) on plant communities (see Statistical analyses). Finally, to measure the surrounding landscape in terms of green cover, we calculated the percentage of vegetated areas in circular buffers of 600 m radius, a scale that influenced urban arthropods in previous studies (e.g., MacIvor, 2016). Sentinel images of 10 m resolution (bands 4 and 5; acquired on March 14, 2017) were used to calculate the Normalized Differential Vegetation Index (NDVI), which was then used to estimate the percentage of green cover (Purevdorj et al., 1998). Table 1 summarizes local and landscape-level characteristics registered along the 20 green roofs.

Soil and ground-dwelling arthropod sampling and identification

Sampling took place during February and March of 2017, at the end of the summer, a period in which high arthropod activity is recorded and accumulated rainfall favors a higher plant cover on the roofs. On each green roof, soil and ground-dwelling arthropods were sampled using three pitfall traps (100 ml plastic cups, 5.5 cm high and 7.5 cm diameter) filled with 50 ml of 20% ethylene glycol as a preservative. Pitfall traps are widely used in studies comparing arthropod communities, especially for passively collecting soil and ground-dwelling species (Southwood, 1978). Traps were buried at the ground level and separated by at least four meters to cover the entire area of the green roofs. After seven days, the content of each trap was filtered and stored in labeled plastic recipients with 70% ethanol. Samples were taken to the laboratory and preserved for further analysis.

Arthropod identification was performed using dichotomous keys for orders and families. Individuals were then classified at the morphospecies level (hereafter, species), which is widely supported as a measure of species richness when large numbers of individuals are collected and in regions where taxonomic knowledge is limited (Obrist & Duelli, 2010). On the other hand, four functional traits were evaluated for each species: vertical distribution, body size, mobility, and feeding habit. Vertical distribution was related to the position of arthropods in the soil profile, where species are classified into euedaphic, hemiedaphic, or epiedaphic (see above). Regarding body size, the body length of each species was measured using a stereoscope with a measuring eyepiece, and species were classified as small (less than 2.5 mm), medium (2.5 to 6 mm), or large (more than 6mm), following similar criteria to previous studies (e.g., Schweiger et al., 2005). Mobility was categorized into high (species with well-developed functional wings), intermediate (soil and dwelling species with relatively large dispersal ability), and low (soil and dwelling species with limited dispersal). Finally, for feeding habits, arthropods were classified into herbivores, predators, parasitoids, omnivores, and detritivores by considering the dominant family habits (or a more detailed taxonomic category in cases of families with multiple feeding habits). For groups with varying feeding habits across life stages, the stage with the most relevant habit from a functional point of view was considered (e.g., gnats from the family Sciaridae were considered detritivores due to the habits of their larvae). The assignment of species to functional groups was based on specific literature for insects (Triplehorn et al., 2005), springtails, and mites (Bokhorst et al., 2012; Momo & Falco, 2009; Önen & Koç, 2011; Potapov et al., 2020).