This dataset contains species richness, Sorensen dissimilarity and its components of species turnover (total and proportional) and nestedness-resultant (total and proportional) data for five groups of insects (ants, bees, butterflies, dung beetles, and termites) sampled in 20 sites across three habitat categories: legally protected areas (PRO), reference sites (REF), and forest fragments adjacent to the mudflow (MUD), conducted in the municipality of Brumadinho, Minas Gerais State, southeastern Brazil. The region gained international attention following the collapse of a tailings dam on 25 January 2019, owned by the Brazilian mining company Vale S.A. The original vegetation of the region was predominantly Atlantic Forest, mainly composed of semideciduous forests, with patches of savanna and rocky-shrubby vegetation at higher altitudes, owing to its location near the transition zone between the Atlantic Forest and Cerrado biomes. Eight sites were in legally protected areas (PRO), with which we aimed at understanding the best possible background scenario in the region: three in Special Protection Area of Rio Manso (which sources water to Belo Horizonte Metropolitan region), three in Inhotim Private Reserve, and two in Rola Moça State Park. Six additional sites were classified as reference areas (REF), consisting of secondary forest patches within the landscape unaffected by the tailings but potentially subject to other disturbances.

We sampled six sites in areas adjacent to the mudflow (MUD), within the Ferro-Carvão stream subbasin, which was the area where the mudflow affected secondary forest patches that were directly impacted by the tailings flow, as it removed part of their areas and at least marginally invaded their interior, which we refer to as mud-adjacent sites. Our first sampling campaign was conducted in January 2021, two years after the event. We sampled the areas every three months, totaling nine expeditions and covering two years (2021 – 2023) of sampling effort. Insect sampling was standardized using 400 m transects at each site for all campaigns. We positioned traps in designated areas specific to target taxa (ants, dung beetles, bees, butterflies, and termites) along each transect. For ants, we deployed 10 non-baited pitfall traps at 40 m intervals along each transect, each trap comprising a 9 cm deep, 15 cm diameter container filled with a 250 ml saline-detergent solution (5 % salt, 5 % detergent) to preserve specimens and equipped with rain covers. Ant traps were left in situ for 48 hours per sampling campaign. Bee sampling included Pan traps and aromatic traps, with Pan traps consisting of color-coded containers (yellow, blue, white) filled with a water-salt-soap solution, checked every 24 hours over 48 hours. Complementary aromatic traps were filled with benzyl acetate, eugenol, β-ionone, methyl trans-cinnamate, methyl salicylate, skatole, vanillin, and 1,8-cineole to attract specific groups, such as Euglossina. Butterfly sampling involved baited Van-Someren Rydon traps spaced 50 m apart along transects, checked daily for four days. For dung beetles, baited pitfall traps containing 20 g of a 4:1 pig:human mix of feces were set at 40 m intervals, left in situ for 48 hours to optimize captures. For termites, we conducted active searches in eight sections measuring 5×1 meters, spaced 50 meters apart along each transect, where a trained sampler inspected potential habitats for 30 minutes per section.