Among land-use changes, pasture and intensive agriculture are the primary drivers of deforestation and the transformation of tropical ecosystems (Brown et al. 2016). The Brazilian Amazon is currently one of the regions experiencing the highest deforestation rates worldwide (INPE 2023). Between 2015 and 2020, approximately 60% of deforested areas were converted to pasturelands or monoculture (e.g., soybean, palm oil, and cocoa) (SEDAP 2020; Venturieri et al. 2022). In freshwater ecosystems, the conversion of forest cover to other land uses around streams increases sediment input, enhances bank erosion and stream widening, and decreases the input of allochthonous organic matter, such as leaves and branches, resulting in lower habitat quality (Nessimian et al. 2008; Dudgeon 2019; Da Silva et al. 2025; Cruz et al. 2025). In addition, the loss of riparian vegetation increases solar radiation exposure, elevates water temperature, and reduces dissolved oxygen concentrations (Monteiro-Júnior et al. 2014; Giehl et al. 2019; Veras et al. 2019). Together, these changes in stream physical structure and water quality affect the composition and dynamics of aquatic communities, promoting either biotic homogenization or increased differentiation, depending on how dispersal processes and environmental filtering generate spatially structured patterns of diversity (Calvão et al. 2023; Bomfim et al. 2024; Brito et al. 2024; Leal-Nascimento et al. 2025; Ribeiro et al. 2025).

Dragonflies and damselflies (Odonata) are widely used as bioindicators for assessing anthropogenic impacts on freshwater biodiversity (Samways et al. 2025). These insects have a life cycle with aquatic larvae and terrestrial flying adults. Adults depend on riparian vegetation and stream habitat integrity due to their ecophysiological, reproductive, and behavioral requirements (De Marco et al. 2015; Miguel et al. 2017; Oliveira-Junior et al. 2019; Calvão et al. 2025). The two suborders, Anisoptera and Zygoptera, exhibit ecological differences related to dispersal ability, thermoregulation, and environmental tolerance, which may result in distinct responses to environmental and spatial gradients (Oliveira-Junior and Juen 2019). Anisoptera are generally larger and have excellent dispersal capabilities, they are typically found in open areas and in environments altered by human activity. In contrast, Zygoptera species are smaller and have limited dispersal ability, they are generally associated with forested areas and are therefore more sensitive to changes in environmental quality.

This is reflected in changes in beta diversity, which represents variation in species composition among sites (Whittaker 1960, 1972). The beta diversity arises from interactions among environmental filtering, dispersal limitation, and stochastic processes, which together shape community assembly across spatial scales (Leibold et al. 2004; Legendre 2014). In general, the beta diversity can be partitioned into two main components: species replacement and richness difference (Podani and Schmera 2011). The replacement consists of a change in species identity, which is generally determined by environmental filtering, interspecific competition, and geographical barriers (Cardoso et al. 2015). In contrast, richness differences correspond to the change in the number of species (loss/gain) between the sampled sites, which most often results from dispersal and extinction (Cardoso et al. 2015).

In this context, investigating how natural environments converted to different land uses differ in terms of species replacement and differences in species richness is essential for understanding the processes that shape biological communities and for informing strategies for biodiversity conservation and ecosystem restoration. In this study, our objective was to assess the effects of land-use change, forest cover, and local and spatial environmental variables on the taxonomic beta diversity of adult Odonata in Amazonian streams, considering total beta diversity, species replacement, and differences in species richness. We hypothesized that (H1) total beta diversity of adult Odonata would be dominated by species replacement due to the combined influence of environmental filtering and dispersal limitation. Regarding land use, we expected that (H2) species replacement would contribute more to total beta diversity than the richness component, particularly in forested and grassland streams, as forest specialists may be replaced by open-area or pasture species without major changes in species numbers. Concerning environmental and spatial drivers, we hypothesized that (H3) local environmental variables (e.g., stream width, depth, and HII – Habitat Integrity Index) would explain a significant portion of variation in total beta diversity, especially in the replacement component, due to their role in filtering species. Finally, we expected that (H4) forest cover and spatial variables would primarily explain richness differences, as changes in landscape structure and spatial configuration may limit dispersal and lead to differences in species richness among sites.

Study area

The study area comprised 50 first- to third-order streams (Strahler 1957) located in the southwestern region of Pará State, northern Brazil, encompassing the municipalities of Altamira, Brasil Novo, Medicilândia, Vitória do Xingu, Senador José Porfírio, and Anapu, in the eastern Amazon (Fig. 1; Supplementary Information: Table S1). The predominant vegetation types in the region are dense and open rainforest (Salomão et al. 2007). The sampled streams were distributed along a gradient of land use and land cover, encompassing different levels of anthropogenic impact. These areas ranged from forest remnants (n = 10) to pasturelands (n = 25) and cocoa plantations (n = 15). Sampling was conducted once in each stream during the dry season, between August 2019 and November 2024. According to the Köppen-Geiger classification, the region’s climate is Tropical Monsoon (Am) (Alvares et al. 2013). The period of highest rainfall generally occurs between December and May, whereas the dry season predominates between July and September. The region receives an average annual rainfall of 1.914 mm, and the yearly average temperature is 26.1 °C.

We collected samples on private rural properties with multiple land uses, including unplanted native forests, cocoa, and pasture. Most of the sampled streams were located in pasture areas with little or no forest cover, though most were surrounded by vegetation. In relation to the streams sampled in cocoa-growing areas, most had multiple land uses in addition to cocoa plantations, such as pasture and remnants of primary and secondary forest. In relation the type of production system, there is no established pattern, plantations vary between agroforestry and monoculture systems, depending on the farm’s production methods. In agroforestry systems, crops are grown under the shade of primary/secondary forest or alongside other fruit-bearing species (cupuaçu tree, acai palm, banana tree). Streams sampled in forested areas had at least 40% forest cover in their surroundings, including unplanted native forest and cocoa. The region is one of the largest cocoa producers in the state of Pará, in addition to having a large land area dedicated to livestock farming (Venturieri et al. 2022). However, there is no information the proportion of land dedicated to cocoa plantations, this is mainly due to cultivation taking place in the understory, heavy cloud cover, and the vast size of the region in the Amazon (Prudente et al. 2020; Venturieri et al. 2022).

Adult Odonata sampling

In each stream, a 100 m stretch was demarcated and subdivided into 20 of 5 m sections (Juen and De Marco 2011; Cezário et al. 2020; Carvalho-Soares et al. 2022). We used the fixed-area sweep method to sample adult Odonata (Cezário et al. 2020). Sampling was conducted during daylight hours, from 10:00 am to 2:00 pm, on sunny days, using an entomological net (250 μm mesh net, 40 cm in diameter, aluminum handle 90 cm long and 65 cm deep) by a single collector for a period of 1 h (3 min in each segment).

After collection, the individuals were stored in separate parchment envelopes labeled with the stream and segment identification (Cezário et al. 2020). In the laboratory, the insects were euthanized with 100% acetone for 12–72 h (Zygoptera and Anisoptera, respectively). After drying, the specimens were stored in plastic bags (10 × 20 cm) (Juen et al. 2025). For species identification, we use specialized taxonomic keys (Borror 1945; Belle 1988, 1996; Garrison et al. 2006, 2010; Lencioni 2005, 2006; Pessacq 2014, 2025; von Ellenrieder 2012). The specimens are deposited in the aquatic insect collection of the Ecology Laboratory – LABECO, Federal University of Pará – UFPA, Altamira, Pará, Brazil.

Stream’s physical structure

In each stream, we measured environmental variables related to the stream’s physical structure (width, depth, and Habitat Integrity Index – HII). We used a tape measure to determine the stream’s width (m) and depth (cm). To assess habitat quality, we used the Habitat Integrity Index (HII) described by Nessimian et al. (2008). This protocol comprises 12 questions on the physical structure and land-use characteristics of the area surrounding the stream. These characteristics include vegetation composition, substrate type, and channel morphology, which collectively reflect the stream’s level of environmental conservation.

The index generates values ranging from 0 to 1, with HII values < 0.69 considered altered and HII > 0.70 considered preserved (Oliveira-Junior et al. 2019). Additionally, HII is related to water quality. Preserved streams are generally associated with high dissolved oxygen values, whereas altered streams are associated with higher electrical conductivity, pH, and water temperature values (Monteiro-Júnior et al. 2014; Giehl et al. 2019; Veras et al. 2019). The HII is a widely used index in studies assessing the impact of environmental changes on stream quality and aquatic insect communities (Brasil et al. 2020b).

Forest cover

In each stream, we analyzed the proportion of land use (%), using buffers with a radius of 200 m, which are commonly used in ecological studies that assess the effects of forest cover on Odonata communities in tropical streams (Oliveira-Júnior et al. 2019; Rocha et al. 2023; Ferreira et al. 2024), to extract the percentage of forest. We obtained land use and land cover data from MapBiomas Collection 9 (https://brasil.mapbiomas.org/colecoes-mapbiomas/), using 30-meter-resolution Landsat satellite images corresponding to the specific sampling year of each stream (2019, 2020, and 2024). We calculated the percentages of forest and pasture cover, the region’s main land uses. We performed the analyses in QGIS Desktop 3.34. The classification of cocoa areas via remote sensing presents a technical challenge due to the dynamics of its spectral signature during their development. In the early stages, generally up to the first year, the crop exhibits reflectance similar to agricultural areas or bare soil due to low biomass density and increased soil exposure. As the crop matures and the canopy closes, its spectral signature becomes similar to that of dense native vegetation, especially in agroforestry systems. For this reason, in the Amazon region, cocoa plantations are classified as forest formation according to the land use classification of MapBiomas (the primary data platform for mapping land use and land cover of biomes in Brazil). This is also due to the extensive cloud cover and vast extent of the Amazon region (Prudente et al. 2020; Venturieri et al. 2022). To try to minimize this limitation, we also included the land use category (Cocoa, Pasture, and Forest) recorded visually in the field at each sampling site. In this context, used forest cover as a predictor variable representing landscape change, lower forest cover is a direct indicator of pasture expansion, while higher forest cover indicates greater environmental preservation.

Spatial predictors

Spatial variables were generated using Principal Coordinates of Neighboring Matrices (PCNM; Borcard and Legendre 2002), which were obtained from the latitude and longitude of each sampled stream. The PCNM axes were generated using the pcnm function from the vegan package. Significant axes were selected according to Moran (p < 0.05) based on distance. Only positive and significant eigenvectors were selected. Subsequently, we used the selected axes in the models to verify the effect of spatial structure on the beta diversity of adult Odonata.

Data analysis

In our analyses, we treated each stream as a sampling unit, yielding 50 sampling units in total. To evaluate the sampling effort, we generated species accumulation curves based on 1.000 permutations using the specaccum function within the vegan package. Furthermore, we calculated the total estimated species richness using the Chao1 estimator to determine the proportion of the community captured during the study. To visually explore and summarize the variation in Odonata species composition across the sampling units, we performed a Principal Coordinate Analysis (PCoA) based on the same Jaccard dissimilarity matrix.

The analyses were conducted separately for the three distinct groups: Odonata order and each suborder separately (Zygoptera and Anisoptera). To understand whether differences in species composition between streams are mainly determined by replacement or loss in species number, we partitioned total beta diversity (β-Total) into replacement (β-replacement) and richness difference (β-richness), following Podani and Schmera (2011). We used a presence-absence matrix of Odonata species and applied Jaccard dissimilarity.

To evaluate the effects of local environment, forest cover, and spatial factors on the beta diversity of Odonata and suborders the Zygoptera and Anisoptera assemblages, we employed Distance-based Redundancy Analysis (db-RDA) coupled with variation partitioning (Legendre and Legendre 2012). We applied a variation partitioning framework based on db-RDA to obtain the unique and shared contributions of three distinct predictor sets in explaining the beta diversity components: (1) local environment (standardized width, standardized depth, and HII); (2) forest cover; and (3) spatial component (significant PCNM axes). This procedure allowed us to decompose the explained variation into pure local environmental effects, pure landscape effects, pure spatial effects, their shared contributions, and the unexplained variation (residuals). The land-use categories (Cocoa, Forest, and Pasture) were not included together in the explanatory models due to their high correlation with local environmental and landscape variables, particularly habitat integrity index (HII) and forest cover. The significance of the global model (including all predictor sets) and the unique (pure) contribution of each fraction were tested using permutation-based anova (9999 permutations). For the pure effects, we used partial db-RDA models, assessing the significance of a given predictor set while conditioning out (controlling for) the effects of the remaining sets.

Finally, to analyze whether changes in land use influence beta diversity and its components, we used a Permutational Multivariate Analysis of Variance (PERMANOVA, Anderson 2001) with 999 permutations. The analyses were conducted using matrices of total beta diversity, species replacement, and difference in species richness, obtained from the decomposition of beta diversity. Land use was used as a categorical predictor variable (Cocoa, Forest, and Pasture). All analyses were conducted in R using (version 4.3.2) the “vegan” and “adespatial” packages.

Description of the odonata community

In the 50 streams, we sampled 912 adult Odonata individuals, identified in nine families (Calopterygidae, Coenagrionidae, Dicteriadidae, Gomphidae, Heteragrionidae, Libellulidae, Polythoridae, Protoneuridae, Pseudostigmatidae), 31 genera, and 82 species (Supplementary Information: Table S2). The suborder Zygoptera represented the majority of individuals (n = 782) and species (n = 49). The most abundant species were Epipleoneura metallica Rácenis, 1955 (n = 81), Mnesarete cupraea (Selys, 1853) (n = 78), and Neoneura sylvatica Hagen, 1886 (n = 72). The Anisoptera suborder represented the lowest abundance (n = 130) and species richness (n = 33). The curve showed a tendency toward stabilization, indicating that most of the common species present at the sites were sampled (Supplementary Information: Fig. S1). The observed richness (82 species) corresponded to approximately 78% of the richness estimated by Chao1 (104.6 species). Of the 82 species collected, 16 were exclusive in pasture streams, 13 species in cocoa streams, and 8 species in forest streams, and 22 species were common among all three-land use. The first two PCoA axes explained 21.1% of the species variation, indicating that there is no regional pattern in the distribution of species across land-use categories (Fig. 2).

Predictors of Odonata beta diversity

We analyzed the total beta diversity across a total of 50 paired streams. We found mean values of total beta diversity of 0.85, 0.88, and 0.92 for Zygoptera, Odonata, and Anisoptera, respectively (Fig. 3). The partitioning of total β-diversity into β-replacement and β-richness components showed that community variation was explained primarily by species replacement (β-replacement), showed a similar pattern across order and suborders (Fig. 3).

The results of distance-based redundancy analysis (db-RDA) and variance partitioning showed that the global model, which included local environmental (width, depth, and HII), spatial predictors, and forest cover variables, explained 11.7% of the variation in total beta diversity (β-Total) of the Odonata assemblages (adj. R² = 0.117; p < 0.001). The largest contribution was from pure spatial structure (PCNM1, PCNM3, PCNM4) responsible for 6.3% of the variation (adj. R² = 0.063; p < 0.001), followed by local (width stream) 2.9% (adj. R² = 0.029; p = 0.003; Fig. 4a). In contrast, the forest cover was not significant (adj. R² = 0.004; p = 0.175).

When analyzing the suborders separately, the total β-Total of Anisoptera and Zygoptera showed distinct responses to the analyzed variables. The global model explained 9.2% of the β-Total of Anisoptera (adj. R² = 0.092; p = 0.007). Pure spatial structure (PCNM4) was the main determinant of β-Total, explaining 5.78% (adj. R² = 0.057; p = 0.017; Fig. 4d). In contrast, local (adj. R² = 0.005; p = 0.361) and forest cover (adj. R² = 0.000; p = 0.444) did not show a significant effect. In relation to the β-Total Zygoptera, the global model explained 13% of the total variation (adj. R² = 0.13; p < 0.001). There was a greater contribution from pure spatial structure (PCNM1, PCNM2, PCNM3, PCNM4), which alone explained 6.49% of the total variation (adj. R² = 0.064; p < 0.001), followed by pure local (width, depth stream) 3.68% (adj. R² = 0.036; p = 0.003; Fig. 4g). On the other hand, forest cover did not show a significant contribution (adj. R² = 0.005; p = 0.189).

In relation to the components of beta diversity, β-replacement and β-richness. The global model explained 12% of the species replacement (β-replacement) in the Odonata (adj. R² = 0.120; p = 0.019). Pure spatial structure (PCNM1, PCNM3) explained 7.6% of the total variation (adj. R² = 0.076; p = 0.038), followed by pure local (width stream) 5.4% (adj. R² = 0.054; p = 0.039; Fig. 4b). On the other hand, forest cover did not show a significant contribution to the β-replacement of Odonata (p = 0.898). Regarding β-replacement of Anisoptera, the global model explained 31.4% of the total variation (adj. R² = 0.314; p < 0.001). Pure spatial structure (PCNM4, PCNM5) was the main determinant of replacement, contributing 18.91% of the explained variance (adj. R² = 0.189; p = 0.006; Fig. 4e). In contrast, the local (adj. R² = 0.081; p = 0.095) and forest cover (adj. R² = 0.032; p = 0.095) did not show a significant effect. The global model did not significantly explain β-replacement in the Zygoptera assemblages (adj. R² = 0.060; p = 0.198). Finally, the β-richness component was not explained by any of the variables analyzed.

The results of the PERMANOVA showed that land use significantly influenced the replacement component of beta diversity in Odonata communities (F = 2.087; p = 0.033). When analyzing Anisoptera and Zygoptera separately, no significant differences were observed between land uses.

Our results showed that change in beta diversity (β-Total) between streams was driven primarily by species replacement (β-replacement), corroborating our first hypothesis. In addition, we identified that only spatial structure and local environment were responsible for determining β-Total and β-replacement, while the difference richness (β-richness) was not explained by any of the analyzed variables. These findings indicate that species replacement plays a key role in the formation of adult Odonata communities along environmental and spatial gradients in Amazonian streams. The high contribution of species replacement to total beta diversity between streams may be the result from environmental filtering, interspecific competition, and geographical barriers (Legendre 2014; Brasil et al. 2018; Oliveira-Junior and Juen 2019). The local environment likely influenced species selection, reflecting the ecophysiological needs of species in relation to environmental conditions, while spatial structure indicate processes such as dispersal limitations or spatially structured environmental gradients. This is consistent with distinct responses observed between Anisoptera and Zygoptera in relation to the analyzed environmental and spatial gradient. These results partially confirm our hypotheses, suggesting that environmental filtering and spatial processes play a greater role in species replacement than in differences in species richness among streams.

The total β-diversity of Odonata and Zygoptera was influenced by both spatial structure and environmental local (particularly stream width and depth, large and medium scale spatial vectors). While the total beta diversity of Anisoptera was only influenced by spatial structure. This pattern suggests that the species of Zygoptera species are more sensitive to environmental variations and have a lower dispersal capacity, making them more susceptible to local environmental filtering and spatial constraints at the regional scale (Alves-Martins et al. 2019). On the other hand, Anisoptera tend to exhibit greater dispersal capacity and higher tolerance to environmental variations, resulting in a more homogeneous distribution across the landscape in response to the environmental gradient. The geographic distance among sites at broad spatial scales can act as a barrier to the dispersal of species with limited dispersal capacity (Leibold et al. 2004). In our study, most of the sampled species belonged to Zygoptera (85%), which generally exhibit limited dispersal ability (Brasil et al. 2018), potentially contributing to the high beta diversity of Odonata observed among the sampled streams, reinforcing the importance of spatial scale in community structuring (Lima et al. 2024; Grönroos et al. 2013). Together, these results suggest that, although both groups are influenced by spatial dynamics, Zygoptera was more strongly associated with local and spatial environmental filtering, while Anisoptera communities proved less predictable, reinforcing the importance of considering ecological differences between suborders when investigating beta diversity patterns in Odonata.

In relation to forest cover and land use variables, although forest cover did not have a significant effect on its own, changes in the landscape may indirectly influence Odonata communities by altering the physical structure of streams. Previous studies have shown that the conversion native forests to other land uses primarily leads to the replacement of species (Carvalho et al. 2018; Mendes et al. 2021; Santos and Rodrigues 2022). High levels of species replacement can occur without changes in species richness (Dolný et al. 2021), because forest-associated species are replaced by species typical of open or pasture areas (Mendes et al. 2019). The absence of an effect on β-richness suggests that environmental changes between streams primarily promoted species replacement, without causing significant differences in the number of species between sites. The results from PERMANOVA support this interpretation, indicating that different land uses led to significant changes in the composition of Odonata communities, primarily related to species replacement. Other studies have shown that streams surrounded by native forest maintain more stable Odonata communities, that is, with less variation in species composition than streams in cocoa plantations and pastures (Santos and Rodrigues 2022). In ecological terms, cocoa plantations can serve as transitional habitats and contribute to the maintenance of species diversity, compared to areas with pastures (Ribeiro et al. 2025), however, they do not replace the original native vegetation.

Finally, the absence and low explanatory power of the analyzed variables and residuals in relation to the predictors of beta diversity and its components may be associated with neutral and stochastic processes influencing the structuring of adult Odonata communities in the studied streams (Oliveira-Junior and Juen 2019; Mendes et al. 2021). Other processes, such as interspecific competition, colonization, and extinction, may also determine beta diversity and its components (Juen and De Marco 2011). Furthermore, additional variables not evaluated in our study (e.g., water physicochemical variables) may have greater explanatory power for beta diversity, β-replacement, and β-richness of adult Odonata (Da Silva et al. 2025). Since variables related to water quality are important for both larvae and adults, we suggest including them in future studies to better understand patterns of beta diversity in Odonata.

Authors

1. Ana Caroline Leal-Nascimento
2. André Ribeiro-Martins
3. Adrielly Souza de Oliveira
4. Gabriel Santos da Silva
5. Francieli F. Bomfim
6. Karina Dias-Silva

Cite this article

Leal-Nascimento, A.C., Ribeiro-Martins, A., de Oliveira, A.S. et al. Local, spatial and land use variables, rather than forest cover, drive Odonata (Insecta) beta diversity in tropical streams. Biodivers Conserv 35, 201 (2026). https://doi.org/10.1007/s10531-026-03408-7

Download citation

• Received: 30 December 2025

• Revised: 19 June 2026

• Accepted: 20 June 2026

• Published: 29 June 2026

• Version of record: 29 June 2026

• DOI: https://doi.org/10.1007/s10531-026-03408-7

Keywords