1. Study area
We conducted our study in the Capão do Leão municipality, located in Rio Grande do Sul state (RS), within the Pampa biome. This area is characterized by its unique and diverse grassland habitats (locally known as campos) that potentially harbor a high species biodiversity. The water bodies we studied were surrounded by native grasslands, dunes, sparse shrubs, and Restinga forest fragments (Renner et al. 2018; Garcia Junior et al. 2019). Anthropogenic activities in the landscape included crops and forestry with soy, corn, Pinus sp., Eucalyptus sp. plantations, and pasture. Our diverse study setting facilitated examining how environment and area affected Odonata diversity because we selected lentic water bodies constructed by humans to water crops and wetlands (locally known as banhados), both eventually used by cattle and horses (Renner et al. 2018).
The water bodies sampled were within the limits of the Mirim São Gonçalo basin and in three different properties: the Horto Botânico Irmão Teodoro Luis (HBI), Empresa Brasileira de Pesquisa Agropecuária (EMB), and Centro Agropecuário da Palma (CAP) (Figure 1; Supplementary Information – Table S1). The climate is classified as Cfa (humid subtropical; Koppen), with an altitude of 21 m above sea level. Summers have an average annual temperature between 18° and 20°C and a mean annual precipitation of 1,500 mm (Kuinchtner & Buriol 2001).
2. Odonata Sampling
We selected 12 water bodies along a rural road, ranging from HBI to CAP. We sampled adult dragonflies in January, February, and March 2021 (summer), coinciding with the peak activity of insects, from 9:00 A.M. to 4:00 P.M. Two or three researchers employed entomological nets collecting dragonflies around the water body margins in the presence of sunlight (Renner et al. 2015). Notably, the sampling time varied with water body area (30 min for small water bodies and 90 min for large water bodies; Supplementary Information - Table S1) (Renner et al. 2015; Pires et al. 2019).
We placed the specimens in entomological envelopes, identified them with the locality and sampling occasion, and took them to the laboratory for species identification. The taxonomist Diogo Vilela identified all the specimens by using identification keys (Lencioni 2005, 2006, 2017; Garrison, 2006, 2010). We deposited them in the entomological collection of Laboratório de Ecologia de Lepidoptera (LELep) belonging to the Museu de Ciências Naturais Carlos Ritter at the Universidade Federal de Pelotas (UFPel). We assessed the extinction risk of each dragonfly species by checking the redlists provided by ICMBio (2023) and IUCN (2022).
3. Environmental variables
We created ortho-mosaics using the images captured with a UAV (Unmanned Aerial Vehicle) Fimi X8 SE 2020 to measure the vegetation cover and water body area. To create the ortho-mosaic, we processed the images captured at 120 m height in Agisoft software (Agisoft LLC 2021). We defined 30 m buffers and grids with 900 ⨉ 900 m² containing squares with 30 ⨉ 30 m² where we measured vegetation cover. We based buffer size on the area where we collected and observed dragonflies interacting with vegetation. For the vegetation coverage, we established five categories: the proportion of tree (TRE), plantation (PLA), shrub (SHR), herbs (HER – including herbaceous and gramineous plants), and aquatic vegetation (AQU – including flooded vegetation and macrophytes) inside the buffer area. We also measured the proportion of bare ground (BAR) and water surface (WAT), which are parts of the environment that do not contain plants.
We first uploaded the ortho-mosaic as a raster layer in QGIS software (QGIS Team version 3.22.10 2021). Then, we delimited the water body margin, calculated the area, and constructed the buffer and grid as vectorial layers. We used marking points comprising only the buffer area in each grid to choose random squares. We chose ten squares to mark polygons corresponding to the vegetation cover, each represented by a code and a different color. Afterward, we estimated the area of each polygon, the total area of the square, and the percentage that the polygon occupied in each square. We then calculated the percentage and the total area of each vegetation cover.
4. Data analysis
We organized and tabulated the insects and EH data for statistical analysis in R (version 4.2.3; R Core Team, 2022). We first calculated the sample coverage and the q0 representing species richness with the q statistic, using the ‘iNEXT’ package (Chao et al. 2014; Hsieh et al. 2020). Using ‘BiodiversityR’ (Kindt & Coe 2005), we estimated vegetation cover heterogeneity (our proxy to EH) for each water body with the Shannon index.
To test the relation between species richness and abundance with vegetation cover heterogeneity and water body area, we used Linear Models (LMs). We checked for potential spatial autocorrelation in LMs residuals with Moran’s I. We evaluated assumptions of normality in the residual distribution and the homogeneity of variances with quantile plots and scatter plots of the residuals and fitted values (Zuur et al., 2009), respectively. We used the performance package (Lüdecke et al. 2021) to assess whether there were influential observations with Cook’s distance and leverage functions, and we verified multicollinearity between the predictor variables using the Pearsons correlation and Variance Inflation Factor (VIF) (Zuur et al. 2010). We considered a VIF equal to three as a threshold of multicollinearity (Zuur et al. 2009). The LMs residuals did not violate the assumptions of normality and homogeneity of variances, and there was no spatial autocorrelation.