The role of fire disturbance on habitat structure and bird communities in South Brazilian Highland Grasslands

Grassland ecosystems cover roughly 40% of the Earth's surface and harbor high and unique biodiversity, shaped by disturbance regimes such as grazing and fire. Fire and grazing alter vegetation structure and habitat complexity, key determinants of bird community composition and diversity. While the effects of fire on plant communities are relatively well studied, how fire disturbance influences bird community patterns remains less clear. In grasslands, fire removes senescent biomass, recycles nutrients, and can increase plant species richness, with frequency influenced by climate and management. Fire promotes habitat heterogeneity by reducing dominance of tall tussock grasses and shrubs that otherwise homogenize structure and reduce biodiversity in the absence of disturbance. Diversity–disturbance relationships (DDR) may be monotonic or peaked; although the intermediate disturbance hypothesis has been debated, peaked responses are common when disturbances are intrinsic to the system. For birds, recently burned, short, litter-free grasslands can increase richness and diversity, partly via increased arthropod prey availability, while prolonged fire exclusion can reduce foraging efficiency for ground feeders. In South Brazilian Highland Grasslands, some bird species (including threatened endemics) prefer frequently burned short vegetation, whereas others select taller vegetation under lower fire frequency; very frequent fires can negatively affect tallgrass-dependent species and fires during the breeding season can reduce reproductive success. Given that fire is traditionally used in regional cattle systems and management remains controversial, especially in protected areas, there is a need for community-level studies. This study evaluates the effect of time since fire on habitat structure (plant community descriptors) and bird community descriptors in South Brazilian Highland Grasslands. The authors hypothesized that with increasing time since fire, sites would show higher vegetation height, lower habitat heterogeneity, lower plant species richness, lower bird taxonomic and functional diversity, and shifts in bird food guild composition, and that habitat structure would directly influence bird community patterns.

The paper synthesizes literature on the role of disturbance (fire and grazing) in shaping grassland biodiversity and structure. It highlights that fire can break dominance of tall tussock species, promote nutrient cycling, and increase plant species richness. DDR can be linear or peaked, with peaked patterns often observed in systems where disturbance is intrinsic. Prior studies show recently burned grasslands increase bird richness and diversity, likely via resource availability, while fire suppression leads to denser vegetation and litter that hinder ground-foraging birds. Raptors are attracted to newly burned areas due to prey availability, and prescribed burns can create feeding opportunities. However, very frequent or rare fires may reduce seed availability for granivores. In South Brazilian Highland Grasslands, species such as Cinclodes pabsti, Anthus nattereri, and Xanthopsar flavus associate with frequently burned shorter vegetation, while Scytalopus iraiensis, Limnoctites rectirostris, Sporophila melanogaster, and S. beltoni are associated with taller vegetation under lower fire frequency. Fire during breeding can negatively affect nesting success. The review underscores gaps in understanding bird community-wide responses to fire and the need to consider interactions with grazing and seasonality.

Study area and design: The study was conducted in South Brazilian Highland Grasslands (Rio Grande do Sul, Brazil; ~29°05'S, 50°05'W) within the Atlantic Forest biome in grassland–forest mosaics. Seven grassland sites (A–G) formed a chronosequence of time since last fire (TSF) ranging from 3 to 180 months: A (5), B (17), C (5), D (180), E (16), F (180), G (3). Sites were in and around Tainhas State Park and within Serra Geral and Aparados da Serra National Parks. All sites had long-term low-intensity cattle grazing (<1.0 AU ha−1); grazing pressure was not directly estimated, assuming similar low levels across sites and stronger detectability of fire effects. Bird sampling: Birds were surveyed during peak breeding (Dec 2017–Feb 2018) using 10-minute, 80 m-radius point counts under suitable morning weather by a single observer. Points were >300 m apart, placed in open grasslands >150 m from edges, fences, or isolated trees. Effort was proportional to fragment size, covering grassland area representatively. A total of 83 points were sampled: A (14), B (14), C (12), D (11), E (12), F (10), G (10). Each point was sampled twice about two months apart; to avoid double-counting, the maximum number of individuals per species across visits at each point was used. Analyses used site-level mean values from point counts. Vegetation and habitat structure: Vegetation was sampled immediately after bird surveys (Feb 2018). Plant species were categorized into three life forms: shrubs, forbs (non-graminoid herbs), and graminoids. Recorded variables included plant species richness (total and by life form); abundance (forbs, shrubs, and Eryngium spp.); percent cover of vegetation, water, rock, and bare soil; and mean vegetation height. Graminoid abundance was not estimated due to dominance and identification challenges. Eryngium individuals, important to birds for nesting and foraging, were counted. At each bird point, four 1 m² quadrats were placed 10 m north, 25 m west, 50 m south, and 75 m east to capture within-point heterogeneity: total quadrats per site: A (56), B (56), C (48), D (44), E (48), F (40), G (40). Eryngium and shrubs were also counted within 10 m buffers around each quadrat. Data processing and analyses: Data were aggregated at the site level. Matrix E contained site means of habitat variables (plant richness and abundances by life form; Eryngium abundance; percent cover of vegetation, water, rock, bare soil; vegetation height). Variance Inflation Factors (VIF < 5) indicated low multicollinearity. Matrix W contained site-level mean bird species abundances. Bird species were assigned to six food guilds (nectarivores, granivores, frugivores, insectivores, carnivores, omnivores), creating trait matrix B (binary traits). Community-weighted mean (CWM) trait values were obtained via B'W to form matrix T (relative food guild contributions per site). Principal Component Analyses (PCA) were performed on E, W, and T; first axes summarized habitat structure and taxonomic and functional bird community structure. Bird taxonomic diversity was calculated as Simpson's 1−D from W; bird functional diversity as Rao's entropy (Gower dissimilarities) from T. Habitat heterogeneity was estimated as Simpson's diversity based on life form composition. Analyses used R. Linear Mixed-Effect Models (nlme) tested effects of log-transformed TSF on plant/habitat descriptors (E), bird taxonomic descriptors (W), and functional descriptors (T), including area (Tainhas vs Serra Geral/Aparados) as a random effect. Both linear and quadratic relationships were fitted and compared via permutation ANOVA; model choice used AICc when significantly different; otherwise, linear models were retained. Pseudo-R² (ML) and p-values were obtained via comparisons to null models. For CWM trait–environment relationships, the multilevel method (MLM3, ter Braak) with max test provided separate p-values for sites and species to control type I error. Canonical Correspondence Analyses (CCA) were run for W and T constrained by a reduced E (mean vegetation height and abundance of forbs, shrubs, and Eryngium within 10 m) to avoid collinearity, with significance assessed by permutation ANOVA.

Sampling yielded 862 individuals from 70 bird species across six food guilds. Insectivores and omnivores comprised 43% and 33% of species, respectively; granivores 11%; carnivores, nectarivores, and frugivores together 13%. Four species were globally threatened (Anthus nattereri, Scytalopus iraiensis, Xanthopsar flavus, Xolmis dominicanus) and three near-threatened (Cinclodes pabsti, Limnoctites rectirostris, Sporophila melanogaster). A. nattereri, C. pabsti, and X. flavus were restricted to recently burned sites; S. melanogaster occurred broadly but was more abundant at longer TSF. Plant and habitat responses to TSF: Most plant community descriptors showed peaked (quadratic) relationships with TSF, with highest values at intermediate TSF, whereas mean vegetation height increased linearly with TSF. - Mean vegetation height ~ TSF: linear model pseudo-R² = 0.820, p = 0.001 (quadratic not better). - Habitat heterogeneity ~ TSF: quadratic pseudo-R² = 0.624, p = 0.033; peaked at intermediate TSF (model comparison p = 0.011). - Total plant species richness ~ TSF: quadratic pseudo-R² = 0.733, p = 0.010 (comparison p = 0.005). - Graminoid richness ~ TSF: quadratic pseudo-R² = 0.689, p = 0.017 (comparison p = 0.013). - Forb richness ~ TSF: quadratic pseudo-R² = 0.605, p = 0.039 (comparison p = 0.032). - Shrub abundance (10 m radius) ~ TSF: quadratic pseudo-R² = 0.921, p < 0.001 (comparison p = 0.005); higher at longer TSF. - Overall habitat structure (PCA1 of E) ~ TSF: linear pseudo-R² = 0.882, p < 0.001. Total plant abundance and Eryngium abundance were not significantly predicted by TSF (p = 0.088 and p = 0.297). Bird community responses to TSF and habitat: - Bird taxonomic diversity ~ TSF: linear pseudo-R² = 0.776, p = 0.001; higher in recently burned sites. - Bird abundance ~ TSF: linear pseudo-R² = 0.593, p = 0.012; higher in recently burned sites. - Bird species composition/structure (PCA1 of W) ~ TSF: linear pseudo-R² = 0.534, p = 0.021. - Bird guild composition (PCA1 of T) ~ TSF: linear pseudo-R² = 0.490, p = 0.030. - Bird taxonomic diversity ~ habitat structure (PCA1 of E): linear pseudo-R² = 0.947, p < 0.001 (quadratic comparison p = 0.002 favored linear); abundance also predicted (pseudo-R² = 0.510, p = 0.026). Bird species composition was predicted by habitat structure (linear pseudo-R² = 0.741, p = 0.002). - Bird functional diversity was not significantly related to TSF (p = 0.698). - CWM traits: granivores’ relative contribution increased with TSF (MLM3 p_sites = 0.012, p_species = 0.036). Multivariate associations (CCA): Habitat variables (vegetation height; abundances of forbs, shrubs, Eryngium) significantly explained variation in bird species composition (χ² = 1.048, F = 2.2189, p = 0.038) and food guild composition (χ² = 0.5922, F = 13.787, p = 0.003). Species associated with taller vegetation and longer TSF included Emberizoides ypiranganus and Sporophila melanogaster; species associated with shorter vegetation, forbs, and shorter TSF included Cinclodes pabsti and Anthus nattereri. Functionally, frugivores were linked to forbs and Eryngium; granivores to taller vegetation and longer TSF; nectarivores and carnivores to recently burned, shorter-vegetation sites; insectivores and omnivores showed broad habitat use.

The study demonstrates that time since fire strongly structures grassland habitat and, in turn, bird communities in South Brazilian Highland Grasslands. Plant richness and habitat heterogeneity peaked at intermediate TSF, while vegetation height and shrub abundance increased with fire exclusion, aligning with theory that fire breaks dominance and maintains structural heterogeneity. These patterns are consistent with other South American grasslands and with generalized DDR findings in systems where disturbance is intrinsic. For birds, responses to TSF were largely linear: recently burned sites supported higher taxonomic diversity and abundance, and TSF and habitat structure both predicted taxonomic composition. Functional composition shifted with TSF, notably with granivores more prevalent at longer TSF, likely reflecting greater seed availability in taller, older stands. CCA confirmed that structural variables (vegetation height, and abundances of forbs, shrubs, and Eryngium) help explain species- and guild-level assemblages, supporting a bottom-up mechanism where fire-driven vegetation changes filter bird communities. The discrepancy between peaked plant responses and linear bird responses suggests that while intermediate TSF maximizes plant heterogeneity, bird community responses are guild- and species-specific, with some threatened species depending on frequently burned, short-vegetation patches, and others on taller vegetation. These findings underscore the conservation value of maintaining a landscape mosaic of patches under different fire frequencies, with careful consideration of seasonal timing to avoid detrimental effects during breeding. The results support inclusion of prescribed fire as a management tool to sustain habitat heterogeneity and conserve both plant and bird diversity.

Habitat variables in grassland vegetation respond to time since fire, and these habitat changes shape bird community patterns. Plant communities and habitat heterogeneity show peaked responses to TSF, whereas bird descriptors respond primarily linearly: bird taxonomic diversity and abundance decrease with increasing TSF, and guild composition shifts, with granivores more prominent at longer TSF. Some bird responses to TSF were not fully explained by measured habitat variables, suggesting potential direct fire effects. Optimal conservation strategies should maintain a mosaic of grassland patches spanning different times since fire to generate diverse habitat structures, with prescribed patch burning integrated into management. Given limited spatial replication, management recommendations should be applied cautiously, and further long-term, regionally replicated studies—considering grazing intensity, fire seasonality, and intervals—are needed to refine fire management policies that balance biodiversity conservation and sustainable use.

The study had a relatively small number of spatial replicates (seven sites) and analyses were conducted at the site level. Grazing pressure was not directly measured; similar low stocking rates were assumed across sites. Biomass and graminoid abundance and seed production were not quantified, limiting mechanistic inference for granivore responses. Potential direct effects of fire on birds (e.g., reproductive impacts, predation) were not explicitly measured. Some plant variables (total plant abundance, Eryngium abundance) showed no significant relationships with time since fire. Findings may be influenced by unmeasured environmental variation and should be generalized cautiously beyond the study area.