AbstractBased on vegetation composition, previous studies of the Pampa biome in southern Brazil have defined seven ecoregions within the area. Here, we test this ecoregion approach studying the semi‐aquatic insect group Odonata in five of these regions, aiming at comparing the ecoregions to the more traditional environmental predictors of water quality and land cover. Based on a data set of occupancy comprising 99 species distributed between 131 localities, a one‐way Permutational Multivariate Analysis of Variance was used to compare differences in the species composition between the ecoregions, followed by a Principal Component Analysis to visualize the variation. The composition varied significantly between all groups tested, and the ordination explained 61.8% of the variance. A partial redundancy analysis of ecoregions, land cover and water quality variables explained 71% of the variance in Odonata community structure. Ecoregion was the most important predictor, followed by water quality and land cover. Within these species assemblies, we could select certain species that were representative of a given ecoregion, to which their distribution within the Pampa biome was entirely or mainly confined. Of 24 representative species 41.7% were rare, while the rest were more abundant and, hence, easier to detect. We suspect that the differences found between the Pampa ecoregions might be due to geology, as such factors may be strong determinants of biodiversity. Specific ecological requirements at the family and genus levels also seemed to act selectively on the species compositions within the ecoregions. Today, the Pampa is highly fragmented due to agricultural activities such as rice cultivation, extensive cattle farming and forest plantations. We suggest that an ecoregion‐based approach to the implementation of conservation measures may be the best way to help these distinct species assemblies survive.