Over 70% of the total channel length in all river basins is formed by low order streams, many of which originate on mountaintops. Headwater streams play fundamental roles in processing and transporting terrestrial and aquatic organic matter, often harboring high biodiversity in bottom leaf patches deposited from riparian vegetation. The objective of this study was to assess the variation in taxonomic composition (measured by beta diversity of aquatic macroinvertebrates) among stream sites located in the Espinhaço Meridional Mountain Range, part of a UNESCO Biosphere Reserve in eastern Brazil. We tested two hypotheses. i) Taxa turnover is the main reason for differences in aquatic insect assemblages within stream sites; we predicted that turnover would be higher than nestedness in all stream sites. ii) Stream site altitude and catchment elevation range are the main explanatory variables for the differences in beta diversity; we predicted that local stream site variables would account for only minor amounts of variation. In both dry and wet seasons, we sampled twice in 2 habitat types (5 leaf patches in pools and 5 in riffles) in each of 9 stream sites distributed in 3 different river basins. We computed average pairwise beta diversity among sampling stations and seasons in each stream site by using Jaccard and Bray-Curtis indices, and calculated the percentages of diversity resulting from turnover and nestedness. Finally, we tested the degree that local- or catchment-level predictor variables explained beta diversity. We found that turnover was the main component of beta diversity and that both dissolved oxygen and elevation range best explained Bray-Curtis beta diversity. These results reinforce the importance of leaf patches in montane (sky islands) neotropical savanna streams as biodiversity hotbeds for macroinvertebrates, and that both local and landscape variables explained beta diversity.

In a 100-m long reach of each of the 9 stream sites, 2 sampling stations were set where we collected as targeted samples, with a Surber sampler (30 x 30 cm; 250 μm mesh), 5 leaf patches in riffles and 5 in pools in both dry (July 2006) and rainy (March 2007) seasons, totaling 180 samples. In pool habitats, we used a Surber sampler simply to ensure the same collection area in pools as riffles, placed leaf deposits into the net by hand and transferred the net contents into plastic bags for subsequent processing. Leaf patches were chosen because, as Ligeiro et al. (2020) reported, they concentrate high abundance and diversity of macroinvertebrates, acting as biodiversity hotbeds. Consequently, they limit assemblage responses to the disturbance gradient when this substrate is targeted, as opposed to using multihabitat sampling. At the sites, leaves were washed with stream water over a sieve (mesh size: 250 μm), separated manually and then the macroinvertebrates and fine particulate organic matter were placed in plastic jars and fixed with 70% ethanol. In the laboratory, invertebrates were identified to family using taxonomic keys (Pérez, 1988; Merritt & Cummins, 1996; Fernández & Domínguez, 2001; Costa et al., 2006; Mugnai et al., 2010).