Cattle-induced homogenization of microbial communities weakens the influence of geochemical gradients in small water bodies Authors: Zohra Zahir, Michael Mensah, Stephanie P. Flaman, Andrew D.S. Cameron, Kerri Finlay, Britt D. Hall Published: 2026 ABSTRACT Natural ponds and constructed reservoirs (dugouts) provide essential water for agriculture, yet the effects of livestock disturbances on the freshwater microbial ecosystems remain underexplored. In this study, 16S rRNA gene sequencing was combined with water geochemistry analyses to examine the relative influence of livestock and geochemical variables on microbial communities in the Prairie Pothole Region (PPR) of the North American Great Plains. In undisturbed ponds, geochemical factors such as dissolved organic carbon (DOC) and sulfate (SO42−) create strong selective pressures, creating distinct microbial communities specialized in carbon and sulfur cycling, leading to niche differentiation and the dominance of specific bacterial taxa. In contrast, livestock-induced disturbances like manure deposition and physical disruption increase within-pond spatial heterogeneity, weakening local geochemical control, while simultaneously reducing biodiversity and homogenizing communities across ponds despite underlying geochemical differences. One example of cattle-induced disturbances was their impact on relative abundances of multiple microorganisms such as Thiobacillus. Thiobacillus is a major contributor to sulfur cycling, converting hydrogen sulfide (H2S) into SO42− under aerobic conditions and contributing to sulfur sequestration by incorporating SO42− into biomass through assimilatory SO42− reduction. In ponds with cattle activities Thiobacillus relative abundance was decreased and the impact of SO42− and DOC was less prominent. Overall, this study showed that livestock disturbances disrupt local microbial community assembly and weaken the influence of geochemical selection in structuring pond-scale microbial diversity.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-026-55862-6. Introduction Small ponds and agricultural reservoirs host diverse microbial communities responsible for driving essential biogeochemical processes1. Microbial communities in these systems are shaped by nutrient availability2 and redox state3, as well as regional processes like dispersal4 and hydrological connectivity5. The biogeochemical processes in ponds rely on microbial communities to cycle nutrients, decompose organic matter, and transform and detoxify pollutants6,7. Microbes play a critical role in the carbon and sulfur cycles, influencing greenhouse gas dynamics and sulfur compound transformations. Microbial activities modulate the ecological carbon balance of wetlands and impact carbon sequestration8. Similarly, microbial contributions to sulfur cycling ensure the continuous transformation and bioavailability of sulfur for different organisms9. The Prairie Pothole Region (PPR) offers a unique natural laboratory for studying microbial community composition in ponds. The PPR contains millions of shallow wetland ponds that formed after the end of the Wisconsin glaciation approximately 10,000 years ago10. The predominant land use in the PPR is agriculture11. Across the PPR of Canada and USA, livestock production is a large part (8–50%) of the agricultural economy12(Table S1). Agricultural producers in the PPR rely on natural wetland ponds as well as small, constructed freshwater reservoirs, locally known as “dugouts” for irrigation and livestock watering13. These natural wetland ponds and dugouts (hereafter referred to collectively as “ponds”) trap snowmelt runoff, receive precipitation via surface and sub-surface flow over the course of the open-water season, and may receive water from groundwater recharge14. Although these ponds are critical water sources, their microbial communities and geochemical variables remain largely unexplored. In general, natural and constructed PPR ponds have high water and porewater concentrations of dissolved organic carbon (DOC), methane (CH4), and sulfate (SO42−)15,16. The high DOC concentrations can be attributed to the accumulation of organic matter in saturated soil and sediments17,18. The accumulation of organic matter in these ponds is primarily a result of decreased decomposition due to anoxic conditions in both water and soil19. The primary source of SO42− in the PPR is the weathering of underlying geological deposits, which releases SO42− into groundwater that subsequently discharges into ponds20,21. In addition, because these ponds are typically closed-basin systems, dissolved solutes accumulate over time, and evaporation further concentrates SO42−. High SO42− concentrations have long been recognized as a challenge for agricultural water use in Saskatchewan22,23. SO42− concentrations above 1000 mg/L cause illness in cattle and above 2000 mg/L can be deadly. Previous surveys in the Canadian prairies showed that the SO42− concentrations in wetland and agricultural ponds can exceed the accepted threshold (1000 mg/L) of SO42− concentrations in water suitable for livestock24,25. In a recent survey of 100 dugouts in Saskatchewan, SO42− concentrations exceeded 1000 mg/L in approximately 25% of sites, and almost 10% of sites exceeded 2000 mg/L, equating to 35% of these water sources being unsuitable for livestock26. In 2017, 200 cattle died in Southern Saskatchewan due to the consumption of water with a high concentration (24000 mg/L) of SO42–[27, demonstrating the importance of the sulfur cycle and microbial activities for effective water quality management and safeguarding livestock health. In our previous study on prairie wetland ponds (both natural and constructed) located in a research area with no livestock access, we found that there were optima concentrations of DOC and SO42− that correlated with highest richness and phylogenetic diversity, while lowest and highest concentrations of these two compounds correlated with lowest richness and phylogenetic diversity28. These findings were aligned with the intermediate disturbance hypothesis, which suggests that moderate levels of environmental stress or resource availability can maximize diversity by balancing competitive exclusion and niche availability29. In this study, we explored the impact of cattle activities as the environmental stress on geochemical parameters and microbial communities, with a focus on the abundance of sulfur cycling microorganisms. By integrating geochemical data, microbial sequencing, and land-use information, we evaluated (1) the impact of livestock induced disturbances (such as manure and urine deposition and physical disruption) on the microbial community of the pond sediments, (2) the importance of geochemical gradients in shaping the microbial community structures due to these disturbances, and (3) the potential of core microbial taxa as indicators of environmental changes. Together, these findings provide broader insight into how land-use disturbances can alter biogeochemical–microbial interactions in freshwater ecosystems at regional to global scales. Results Microbial diversity was highest in No Cattle ponds We evaluated cattle impacts on geochemistry and microbial community diversity at 32 ponds: 11 with no cattle access (herein “No Cattle”; total of 3 dugouts and 8 wetlands), and 21 ponds with full water access to cattle (herein “Cattle”; total of 19 dugouts and 2 wetlands). There were three ponds located in cattle farms with no direct cattle access (herein “Near Cattle”) but were not included in the main statistical analysis due to the low number of samples. However, they are included in the SI for comparison. We profiled microbial community composition in the saturated soils and submerged sediments by sequencing V4 region of the 16S rRNA gene. Good’s coverage confirmed sufficient DNA sequencing depth for all samples and estimated that approximately 99.9% of amplicon sequencing variants (ASVs) were observed more than once. A total of 77, 339 ASVs were identified in 174 samples across all ponds, distributed as in Fig. S1A. To standardize sequencing depth across samples we used rarefaction which allowed for more robust comparisons of diversity metrics. By rarefying to a uniform sequencing depth, differences in alpha diversity are less likely to be driven by variation in sequencing effort30. We chose a depth of 15,000 reads, taking into consideration the trade-off between minimizing the number of samples and sequences removed and keeping sufficient data for analysis (Fig. S1). Most samples plateaued at lower sequencing depths, indicating that additional sequencing did not increase diversity in this category (Fig. S1). After rarefying, filtering out less prevalent ASVs (those with an abundance of less than 0.001 of the total sample, and chloroplast and mitochondria), and removing potential contamination the number of ASVs decreased to 53,009 and 151 samples remained (Fig. 1). Fig. 1Sampling locations for microbial community analysis. (A) Map of North America. The green shaded area indicates the Prairie Pothole Region (PPR). (B) Map of Saskatchewan. The green shaded area indicates the Saskatchewan portion of the PPR. (C) Locations of natural and constructed ponds that were sampled for geochemistry and microbial community analysis located in southern Saskatchewan. W notation stands for wetland, and D notation stands for dugout. Each pond category (Cattle, Near Cattle, No Cattle) is color-coded accordingly. The contouring lines and colors show the elevation above the sea level calculated using QGIS software. (D) Sample collection strategy. At each pond, three biological replicate samples (co1, co2, co3) were taken from the shallow part of the pond (using a core tube), and another sample was taken from the deep part of the pond using an Ekman sampler (Ek1). Coincidentally, water samples for sulfate (SO4) concentration and dissolved organic carbon (DOC) were taken. A multiparameter meter (YSI) was used to record the pH, conductivity, oxygen percentage, and temperature of the water. Sampling locations for microbial community analysis. (A) Map of North America. The green shaded area indicates the Prairie Pothole Region (PPR). (B) Map of Saskatchewan. The green shaded area indicates the Saskatchewan portion of the PPR. (C) Locations of natural and constructed ponds that were sampled for geochemistry and microbial community analysis located in southern Saskatchewan. W notation stands for wetland, and D notation stands for dugout. Each pond category (Cattle, Near Cattle, No Cattle) is color-coded accordingly. The contouring lines and colors show the elevation above the sea level calculated using QGIS software. (D) Sample collection strategy. At each pond, three biological replicate samples (co1, co2, co3) were taken from the shallow part of the pond (using a core tube), and another sample was taken from the deep part of the pond using an Ekman sampler (Ek1). Coincidentally, water samples for sulfate (SO4) concentration and dissolved organic carbon (DOC) were taken. A multiparameter meter (YSI) was used to record the pH, conductivity, oxygen percentage, and temperature of the water. Species richness (Chao1) was significantly higher in No Cattle ponds compared to Cattle ponds (p-value < 0.0001; Fig. 2A). Similarly, richness and evenness (Shannon) were higher in No Cattle ponds compared to Cattle ponds (p-value < 0.05; Fig. 2B). However, the low number of samples from Near Cattle ponds precluded differentiation of richness and evenness from the other two categories of cattle access (Fig. S2A, B). For comparison, other alpha diversity indices were assessed as well (Fig. S3). Species richness calculated using ACE (similar to Chao1), and Fisher (a parameter from a statistical model assuming taxa follow a log-series distribution), which estimate the number of taxa present in a community, was significantly higher in No Cattle ponds compared to Cattle ponds (p < 0.0001; Fig. S3B, C). Simpson diversity, which emphasizes community dominance and the relative abundance of common taxa, did not differ significantly between Cattle and No Cattle ponds (Fig. S3E). Pielou’s evenness, which measures how evenly individuals are distributed among taxa, was significantly higher in Cattle ponds compared to No Cattle ponds (p < 0.001; Fig. S3F). Fig. 2Alpha diversity analysis based on 16S rRNA gene sequencing of microbial communities across different pond categories during the open water season in 2021 (n = 151). (A) Species richness (Chao1) at a rarefaction depth of 15,000 reads, which led to the removal of 16 samples (151 samples remained). (B) Species richness & evenness (Shannon) at a rarefaction depth of 15,000 reads. (C & D) Comparisons between dugouts and wetland ponds for Chao1 and Shannon indices in each category. (E) Comparisons between dugouts in Cattle and No Cattle ponds and (F) Comparisons between wetland ponds in Cattle and No Cattle. The color scheme corresponds to the pond categories (Cattle: green, No Cattle: purple). Box plots represent the median, quartiles, and outliers. Significance levels: ****p < 0.0001, ***p < 0.001, **p < 0.01, *p < 0.05, ns: non-significant. Statistical tests were performed using the Wilcoxon test and adjusted using the false discovery rate (FDR) correction. Alpha diversity analysis based on 16S rRNA gene sequencing of microbial communities across different pond categories during the open water season in 2021 (n = 151). (A) Species richness (Chao1) at a rarefaction depth of 15,000 reads, which led to the removal of 16 samples (151 samples remained). (B) Species richness & evenness (Shannon) at a rarefaction depth of 15,000 reads. (C & D) Comparisons between dugouts and wetland ponds for Chao1 and Shannon indices in each category. (E) Comparisons between dugouts in Cattle and No Cattle ponds and (F) Comparisons between wetland ponds in Cattle and No Cattle. The color scheme corresponds to the pond categories (Cattle: green, No Cattle: purple). Box plots represent the median, quartiles, and outliers. Significance levels: ****p < 0.0001, ***p < 0.001, **p < 0.01, *p < 0.05, ns: non-significant. Statistical tests were performed using the Wilcoxon test and adjusted using the false discovery rate (FDR) correction. Species diversity in wetland ponds and dugouts was similar in Cattle ponds The water samples in No Cattle ponds had higher SO42− and DOC concentrations than Cattle ponds (Fig. S4) while other parameters such as conductivity, dissolved oxygen (DO) saturation, DO concentrations, pH, Chl.a, and organic matter (OM) of the sediment were not statistically different (Fig. S4). The water samples in wetland ponds of No Cattle ponds had statistically higher SO42− and DOC concentrations than dugouts of No Cattle ponds (Table S2). No other parameters were different among wetlands and dugouts of each category (Table S2). To rule out the possibility that differences in alpha diversity were driven by the distinction between wetland ponds and dugouts rather than the presence or absence of cattle, we analyzed wetland ponds and dugouts separately within each group (No Cattle: 8 wetland ponds and 3 dugouts; Cattle: 2 wetland ponds and 19 dugouts). This separation allowed us to assess whether the observed diversity patterns were primarily influenced by cattle presence rather than pond category. Species richness and evenness (both Chao1 and Shannon) were not significantly different when comparing communities between dugouts and wetland ponds in Cattle ponds (Fig. 2C and D). However, both indices’ values were significantly higher in dugouts than in wetland ponds in the No Cattle ponds (p-value < 0.01; Fig. 2C and D). The comparison between pond type (wetland and dugout) in each category was also performed. The dugouts in No Cattle ponds had significantly higher species richness (Chao1; p-value < 0.001) and higher species richness and evenness (Shannon; p-value < 0.05) than dugouts in Cattle ponds. However, while the species richness (Chao1) was higher in wetland ponds in No Cattle category (p-value < 0.05) than Cattle category, the richness and evenness (Shannon) of wetland ponds in both categories were not statistically significant (Fig. 2E and F). Cattle presence weaken geochemical gradients’ impact in shaping microbial communities in wetland ponds and dugouts at within-pond scale Microbial community composition, known as beta diversity, reflects dissimilarities and differences in taxa and their relative abundances between two environments31. To assess the relative influence of spatial proximity and geochemical variables on microbial community composition, we used principal coordinates of neighbor matrices (PCNM)32. In No Cattle ponds more variability in microbial community composition was explained by geochemical (24%) and spatial (13%) variables compared to the Cattle ponds (10% and 9% respectively; Fig. 3). The residuals, which represent the portion of the variation in microbial community composition which could not be explained by either geochemical or spatial variables, were 56% in No Cattle ponds and 79% in Cattle ponds. Fig. 3Variance partitioning Venn diagram illustrating the relative contribution of geochemical factors (sulfate, dissolved organic carbon, conductivity, water pH, temperature, and chlorophyll-a) and spatial variables using principal coordinates of neighbor matrices to the microbial community compositions (i.e., beta diversity, reflecting differences in taxa and their relative abundances between ponds). The size of each circle represents the magnitude of the respective variable. Residuals represent the portion of the variation in microbial community composition which could not be explained by either geochemical or spatial variables. Variance partitioning Venn diagram illustrating the relative contribution of geochemical factors (sulfate, dissolved organic carbon, conductivity, water pH, temperature, and chlorophyll-a) and spatial variables using principal coordinates of neighbor matrices to the microbial community compositions (i.e., beta diversity, reflecting differences in taxa and their relative abundances between ponds). The size of each circle represents the magnitude of the respective variable. Residuals represent the portion of the variation in microbial community composition which could not be explained by either geochemical or spatial variables. To visualize differences in microbial community composition across pond categories, we performed principal coordinate analysis based on individual samples among ponds (method = PCoA, distance = bray)33. The results showed that samples from Cattle ponds and No Cattle ponds had distinct clusters with fewer samples overlap, whereas samples in Near Cattle ponds had overlap with both categories (Fig. S2C). In this ordination, 17.3% and 11.3% of the variance in microbial community composition was explained by Axis 1 and Axis 2 respectively (Fig. S2C); and rarefaction (depth: 15,000 reads) did not change the ordination (Axis 1: 17.9%, Axis 2: 12.2%, Fig. 4A, B). The microbial communities in these plots were color-coded based on the concentrations of SO42− (Fig. 4A) and DOC (Fig. 4B). A clear visual gradient of the SO42− and DOC concentrations was observed in No Cattle ponds, forming three distinct clusters corresponding to high, medium, and low concentrations. A weaker pattern was also observed in cattle ponds (Fig. 4A and B). Incorporating phylogenetic distances among ASVs using weighted UniFrac analysis (quantitative measure)31, which accounts for phylogenetic relationships and weights branch distances by the relative abundances of ASVs (method = PCoA, distance = UniFrac, weighted = TRUE), yielded similar clustering patterns both at the sample scale (Fig. S5A) and at the pond scale (Fig. S5C). In contrast, unweighted UniFrac analysis (qualitative measure), which considers phylogenetic relationships but does not weight branches by relative abundances, did not produce strong clustering between Cattle and No Cattle ponds or within No Cattle ponds at either the sample (Fig. S5B) or pond scale (Fig. S5D). Fig. 4Bray-Curtis principal coordinate analysis (PCoA) of microbial community composition based on 16S rRNA gene sequencing across two cattle presence categories (A) sulfate (SO42−) and (B) dissolved organic carbon (DOC) concentrations. The color scheme corresponds to concentration gradients (see legend). Ellipses represent 95% confidence intervals around the centroids (solid line = t distribution; dashed line = normal distribution). Bray-Curtis principal coordinate analysis (PCoA) of microbial community composition based on 16S rRNA gene sequencing across two cattle presence categories (A) sulfate (SO42−) and (B) dissolved organic carbon (DOC) concentrations. The color scheme corresponds to concentration gradients (see legend). Ellipses represent 95% confidence intervals around the centroids (solid line = t distribution; dashed line = normal distribution). Geochemical structuring of microbial communities varies with spatial scale in Cattle ponds Because SO42− and DOC gradients strongly structured microbial communities in No Cattle ponds, but these patterns were not as strong in Cattle ponds at the individual sample level, we further examined whether this pattern was influenced by spatial heterogeneity within ponds. Specifically, we asked whether the apparent weakening of geochemical structuring in Cattle ponds resulted from increased variability among samples collected within the same pond. To address this, we repeated the beta diversity analysis (PCoA, method = Bray) at the pond level rather than the individual sample level. This approach also helped account for potential pseudo-replication, as ponds in the No Cattle category were sampled multiple times between May and September, whereas Cattle ponds were sampled only once. For this analysis, replicate samples from each pond were averaged using the merge_samples function from the phyloseq package prior to ordination. At the pond level, No Cattle ponds retained the same three distinct clusters corresponding to high, medium, and low SO42− and DOC concentrations, indicating that geochemical gradients consistently structured microbial communities across ponds (Fig. 4C, D). Cattle ponds showed clearer separation along the SO42− concentration gradient after averaging samples at the pond level, although the clustering remained less distinct than in the No Cattle ponds (Fig. 4C). A weaker separation along the DOC gradient was also observed (Fig. 4D). Microbial community dispersion differs between Cattle and No Cattle ponds The test for homogeneity of multivariate dispersions defined as the average distance of samples to their group centroid in multivariate space (i.e., a measure of within-group variability in community composition)34, revealed differences between Cattle and No Cattle ponds (Fig. S6). Microbial communities’ dispersion in Cattle ponds at both sample scale (mean distance to centroid = 0.58, SD = 0.04; Fig. S6B) and pond scale (mean distance to centroid = 0.554, SD = 0.06; Fig. S6D) showed smaller variability (SD) around the centroid. No Cattle ponds showed a smaller difference in the means, however differences in the amount of dispersion of samples around the mean was greater at both sample scale (mean distance to centroid = 0.508, SD = 0.10; Fig. S6B) and pond scale (mean distance to centroid = 0.47, SD = 0.13; Fig. S6D). Sulfate, DOC, conductivity, and OM were key environmental drivers of community structure To assess how environmental variables might be correlated with microbial community composition, we performed distance-based redundancy analysis (dbRDA)35 at both the sample and pond scales (Fig. 5A and C). In addition to environmental variables, spatial structure was also incorporated by using the first two principal coordinates of neighbor matrices (PCNM1 and PCNM2). At the sample scale, SO42−, DOC, conductivity, and OM were strongly associated with community composition, as indicated by their long vectors oriented toward samples with high values of these variables (Fig. 5A) and supported by permutation (ANOVA) tests (Table S3). Dissolved oxygen (DO) (both concentration and saturation) while being statistically correlated with microbial community (p-value = 0.02 and 0.04 respectively) their vectors were shorter indicating weaker explanatory power (R2 = 0.04 and 0.03 respectively). In contrast, Chlorophyll a (Chl.a), temperature, and pH were weakly associated with community variation (p-values > 0.05). PCNM1 showed a strong directional influence (R2 = 0.80, p-value = 0.001), indicating that spatial processes also contributed to community variation as shown by Fig. 3. At the pond scale, similar patterns were observed, with SO42−, DOC, conductivity, and OM remaining the dominant drivers of community structure (Fig. 5C, Table S3). However, Chl.a shifted to the opposite direction relative to these variables. At the sample scale, DO saturation showed weaker associations with community composition (R2 = 0.05, p-value = 0.44). Fig. 5Distance-based redundancy analysis (dbRDA) of microbial community composition based on 16S rRNA gene sequencing across two cattle presence categories: (A) based on all samples (n = 151); (B) based on average of samples per pond (n = 32). Redundancy analysis (RDA) of microbial community composition based on 16S rRNA gene sequencing across two cattle presence categories: (C) based on all samples (n = 151); (D) based on average of samples per pond (n = 32). Each point represents a sample or a pond (based on the scale) microbial community composition, color-coded by cattle presence. Environmental variables are depicted as arrows, with the length and direction indicating the strength and association with microbial community composition. Top microbial genera (in red) are also mapped (in B and D) to the ordination space, showing their relationships to the environmental gradients. SO42− = Sulfate concentration (mg/L), DOC = Dissolved organic carbon concentration (mg/L), Conductivity (µS/cm), Temp = Temperature (°C), Chl.a = Chlorophyll a (µg/L), DO_sat: Dissolved oxygen saturation %, OM = Organic matter of sediment, PCNM1 and PCNM2 = Principal coordinates of neighbor matrices were derived from PCNM analysis, using pcnm() function from the vegan package. Distance-based redundancy analysis (dbRDA) of microbial community composition based on 16S rRNA gene sequencing across two cattle presence categories: (A) based on all samples (n = 151); (B) based on average of samples per pond (n = 32). Redundancy analysis (RDA) of microbial community composition based on 16S rRNA gene sequencing across two cattle presence categories: (C) based on all samples (n = 151); (D) based on average of samples per pond (n = 32). Each point represents a sample or a pond (based on the scale) microbial community composition, color-coded by cattle presence. Environmental variables are depicted as arrows, with the length and direction indicating the strength and association with microbial community composition. Top microbial genera (in red) are also mapped (in B and D) to the ordination space, showing their relationships to the environmental gradients. SO42− = Sulfate concentration (mg/L), DOC = Dissolved organic carbon concentration (mg/L), Conductivity (µS/cm), Temp = Temperature (°C), Chl.a = Chlorophyll a (µg/L), DO_sat: Dissolved oxygen saturation %, OM = Organic matter of sediment, PCNM1 and PCNM2 = Principal coordinates of neighbor matrices were derived from PCNM analysis, using pcnm() function from the vegan package. The dbRDA analysis included all environmental variables and showed the presence of multicollinearity among predictors. Previous analysis also showed the correlation between DOC and SO42–36. To address this, pairwise correlations were evaluated, and variables exhibiting strong correlations (cor ≥ 0.7), such as DOC, OM, SO42−, and conductivity, were considered to represent similar underlying environmental gradients (Fig. S7A). To reduce redundancy and improve interpretability, only one representative variable from each correlated group was retained, and the dbRDA was subsequently repeated using the reduced set of environmental variables. The overall ordination structure remained consistent between the full and reduced models (Fig. S7B). Microbial taxonomy was different in two pond categories The microbial taxonomy profiles in samples across two pond categories showed differences even at the phyla level according to UniFrac analysis (Fig. S5A). The microbial community taxa differed among pond categories, with the most dominant phyla being Proteobacteria, Bacteroidota, Chloroflexi, Desulfobacterota, Actinobacteriota, and Firmicutes (Fig. 6A, B). In total, 21 of the 69 detected phyla showed statistically significant differences in relative abundance between the two pond categories (Kruskal–Wallis test, FDR-corrected p < 0.05; Table S4). Proteobacteria had statistically higher relative abundance (p.adj-value = 0.001) in No Cattle ponds (27.2%), as compared to Cattle ponds (20.8%; Fig. 5A). Bacteroidota had a similar relative abundance (13.1%) across both categories (p.adj-value = 0.6). Chloroflexi and Desulfobacterota exhibited higher relative abundance in No Cattle ponds (11.2%, 13%; respectively) compared to Cattle ponds (10%, 7.4%). However, this difference was not statistically significant for Chloroflexi (p.adj-value = 0.3), but it was statistically significant for Desulfobacterota (p.adj-value = 0.0001). Cattle ponds were more enriched with Firmicutes (9.7%, p.adj-value = 0.007) and Actinobacteriota (14.7%, p.adj-value = 0.00002) compared to No Cattle ponds. Fig. 6Relative abundance of microbial communities across different taxonomic levels based on 16S rRNA gene analysis and identified using the SILVA V138 database (total samples = 151). (A) Top 10 phyla and their top 5 genera. Each bar plot represents the average microbial community composition of all samples per each pond (32 ponds) with taxa color-coded according to their classification. (B) Heatmap of all phyla (n = 69) based on the log of relative abundance of average microbial community composition of all samples per each pond (32 ponds). (C) The percentage of the top 10 phyla based on each category. The legends in panels A and C provide taxonomic references specific to the respective taxonomic level presented. Relative abundance of microbial communities across different taxonomic levels based on 16S rRNA gene analysis and identified using the SILVA V138 database (total samples = 151). (A) Top 10 phyla and their top 5 genera. Each bar plot represents the average microbial community composition of all samples per each pond (32 ponds) with taxa color-coded according to their classification. (B) Heatmap of all phyla (n = 69) based on the log of relative abundance of average microbial community composition of all samples per each pond (32 ponds). (C) The percentage of the top 10 phyla based on each category. The legends in panels A and C provide taxonomic references specific to the respective taxonomic level presented. To differentiate microbial communities across pond categories and identify microbial biomarkers, we performed linear discriminant analysis (LDA, method = lefse, p-value < 0.05). 1389 taxa (at all taxa levels) were found significant and after p-value adjustment, 1088 taxa were found significant. Minimum LDA score was 1.12 and maximum score was 4.63. At phyla level, 43 out of 69 phyla were found significant (p-value < 0.05; Fig. S8A). Actinobacteriota (class Actinobacteria, order Micrococcales), Firmicutes (class Bacilli, order Bacillales, family Planococcacea), Chloroflexi (class Chloroflexia) and Verrucomicrobiota (class Verrucomicrobiae, order Verrucomicrobiales, family Rubritaleaceae, genus Luteolibacter) appeared as top biomarkers in Cattle ponds. In No Cattle ponds, dominated biomarkers were among phyla Proteobacteria (class Gammaproteobacteria, order Burkholderiales, family Hydrogenophilaceae, genus Thiobacillus), Chloroflexi (class Anaerolineae, order Anaerolineales) and Desulfobacterota (class Desulfobacteria, order Desulfobacterales, family Desulfosarcinaceae) (Fig. S8B). Microbial functional potential showed higher nitrogen cycling pathways than sulfur cycling pathways in the Cattle ponds The prediction of functional profiles based on the prokaryotic communities from sequencing results was also calculated (Fig. S9). Distinct differences were observed across carbon, nitrogen, sulfur, and other metabolic pathways. Within the nitrogen cycle, pathways related to denitrification, nitrate reduction, and nitrite respiration were more abundant in Cattle ponds, whereas nitrification showed higher relative abundance in No Cattle ponds. In the sulfur cycle, pathways associated with sulfur respiration, including sulfate and sulfite respiration, were more abundant in No Cattle ponds. In Cattle pond sulfur oxidation pathways such as anoxygenic photoautotrophy were higher than in No Cattle ponds. For carbon cycling pathways Cattle ponds showed higher relative abundance of fermentation and methanogenesis. No Cattle ponds with high SO42− concentrations had Thiobacillus as their main core microbiome To assess the core microbiome in each pond, we identified microbial taxa that were present in at least 30% of the samples at a minimum relative abundance of 0.0001. Core microorganisms play an important role in the functioning and stability of the ecosystem37. Multiple genera were identified as core microbiome such as Anaerolinea, Desulfobacca, Illumatobacter, Luteolibacter, and Thiobacillus (Fig. 6A). Visually we observed that the genus Thiobacillus was more abundant in No Cattle ponds compared to Cattle ponds, and the presence of Thiobacillus was positively correlated with increased SO42− concentrations (Fig. 7A). To test this statistically, we performed a linear regression analysis by plotting the relative abundance of Thiobacillus against the average SO42− and DOC concentrations across ponds (Fig. 7B, C). We found that while the relative abundance of Thiobacillus positively correlated with SO42− (method = Pearson, No Cattle p-value = 0.007; Cattle p-value = 0.009), and DOC (method = Pearson, No Cattle p-value = 0.002; Cattle p-value = 0.002) concentrations across both pond categories, No Cattle ponds consistently had higher Thiobacillus abundance compared to Cattle ponds at similar SO42− and DOC concentrations (Fig. 7B, C). The linear regression analysis was also repeated for DO concentration, DO saturation, OM %, and Chl.a (Fig. S10A, D). The relative abundance of Thiobacillus was only positively correlated with OM % (method = Pearson, No Cattle p-value = 0.02; Cattle p-value = 0.007). To further analyze the correlations between the core microbiome and geochemical variables, we performed a redundancy analysis (RDA) at both sample and pond scale. Thiocapsa, Illumatobacter, SVA0081 sediment group, and Thiobacillus were all positively correlated with SO42− concentrations, DOC concentrations, OM, and conductivity. However, the longer arrow for Thiobacillus indicates a stronger association with these variables (Fig. 5B and D). Fig. 7Core microbial communities based on genera (total samples = 151, total ponds = 32). (A) Mean relative abundance (%) of core microbial genera across ponds based on 16S rRNA gene analysis and identified using the SILVA V138 database, color-coded on the plot based on cattle presence (Cattle = green, No Cattle = purple) and pond types (dugout = dark grey, wetland = light grey). The genera are color-coded, with core taxa (at least present in 30% of the samples) calculated using core_member() function from microbiome package (v1.26.094. Each bar represents a pond and ponds are ordered based on their average sulfate concentration (mg/L). (B) Mean relative abundance of Thiobacillus plotted against average sulfate (SO42−) concentration (mg/L) and C) dissolved organic carbon (DOC) concentrations (mg/l) across ponds. Data are grouped by the presence of cattle, with each group represented by separate linear regression lines using lm() function from stats package (R base package). Shaded regions represent the 95% confidence interval for the regression model. R and p-value were calculated using stat_cor function (method="pearson”) from ggpubr package (v0.6.0103). Core microbial communities based on genera (total samples = 151, total ponds = 32). (A) Mean relative abundance (%) of core microbial genera across ponds based on 16S rRNA gene analysis and identified using the SILVA V138 database, color-coded on the plot based on cattle presence (Cattle = green, No Cattle = purple) and pond types (dugout = dark grey, wetland = light grey). The genera are color-coded, with core taxa (at least present in 30% of the samples) calculated using core_member() function from microbiome package (v1.26.094. Each bar represents a pond and ponds are ordered based on their average sulfate concentration (mg/L). (B) Mean relative abundance of Thiobacillus plotted against average sulfate (SO42−) concentration (mg/L) and C) dissolved organic carbon (DOC) concentrations (mg/l) across ponds. Data are grouped by the presence of cattle, with each group represented by separate linear regression lines using lm() function from stats package (R base package). Shaded regions represent the 95% confidence interval for the regression model. R and p-value were calculated using stat_cor function (method="pearson”) from ggpubr package (v0.6.0103). Discussion Our study showed that microbial community compositions differ significantly among ponds with or without cattle access. The ponds have diverse depths and chemistries; thus, we were surprised to find that microbial community compositions were strikingly similar among Cattle ponds. The higher Pielou’s evenness observed in Cattle ponds, despite lower richness and Shannon diversity, suggests that livestock disturbance acts as an environmental filter that reduces the dominance of specialized taxa while eliminating rare species, resulting in a more homogenized community with more evenly distributed abundances rather than increased diversity as predicted by the classical intermediate disturbance hypothesis. In a separate study on the impact of geochemical variables on the microbial composition in natural wetland ponds and one dugout (including 8 of our No Cattle ponds28, it was found that SO42− and DOC concentrations aligned with the intermediate disturbance hypothesis, which suggests that moderate levels of environmental stress or resource availability can maximize diversity by balancing competitive exclusion and niche availability29. However, it can be speculated that while the concentrations of SO42− and DOC could have caused the maximum diversity, cattle-induced disturbance probably didn’t let the maximum diversity to occur. We visually observed that cattle disturb sediments and soils and deposit manure and urine into ponds. In a study on impact of herbivores on grassland, it was found that under a single-herbivore disturbance the alpha diversity increased, however under multiple herbivore disturbances alpha diversity’s dependence on environmental variables such as nitrogen were weakened38. They concluded that cumulative disturbance generates spatially heterogeneous nutrient distributions, producing localized hotspots and depleted zones that intensify competition among resource-sensitive taxa and constrain local diversity38. This might also explain the microbial communities in Cattle ponds, where such imbalances can shift communities toward nutrient co-limitation and weaken environmental controls on diversity, ultimately suggesting that microbial communities under intense disturbances converge toward similar community composition. There are a few other studies that have explored the impact of livestock on microbial community diversity in aquatic systems (Table 1). In our study, microbial community richness and evenness in dugouts and wetland ponds in Cattle ponds were similar. This suggests that the differences typically observed between constructed (dugouts) and natural wetlands—often driven by geochemical gradients—may be overridden under cattle disturbance. In other words, livestock activity may act as a homogenizing force, reducing the extent to which environmental variation translates into differences in microbial diversity. However, it is also possible that the physicochemical differences between pond types were insufficient to produce detectable differences in microbial diversity under disturbed conditions (as the difference in geochemical variables were not statistically significant). When comparing wetlands with and without cattle, Chao1 richness differed significantly, whereas Shannon index did not, indicating that cattle presence primarily affected rare taxa rather than overall community structure. In contrast, dugouts showed significant differences in both richness (Chao1) and richness & evenness (Shannon) between Cattle and No Cattle ponds, suggesting a stronger or more consistent response to disturbance in these systems. No previous studies to our knowledge have compared microbial communities in natural wetland ponds and dugouts (or agricultural reservoirs). The contrasting responses observed between wetlands and dugouts further point to the importance of sediment characteristics in mediating disturbance effects. Wetlands are typically richer in organic carbon and more porous (with macropores)39, conditions that can buffer physical disturbance and promote greater resilience of microbial communities. In contrast, dugouts are often dominated by compacted clay sediments with smaller pores (micropores)39, where trampling by cattle may further reduce porosity, limit diffusion, and alter redox conditions. Such physical changes could have stronger impacts on microbial community structure in dugouts, potentially explaining their greater sensitivity to disturbance. Environmental variables such as dissolved oxygen (DO) and chlorophyll a (Chl.a) showed limited influence on microbial community structure. The weak association of DO may reflect similar oxygen conditions in the overlying water column across both pond categories. However, in Cattle ponds, reduced sediment porosity and increased compaction could restrict oxygen diffusion into porewaters, decoupling surface measurements from the conditions experienced by sediment microbial communities. Similarly, the limited effect of Chl.a suggests that primary productivity in the water column may not be a dominant driver of sediment microbial composition, or that its influence is overshadowed by stronger geochemical and disturbance-related processes. The strong contribution of PCNM1 highlights the importance of spatial structure, suggesting that unmeasured spatially structured factors (such as dispersal limitation and hydrological connectivity) also play a significant role in shaping microbial communities beyond local environmental conditions. Future work should therefore explicitly assess sediment properties such as porosity and compaction, as well as porewater chemistry and hydrogeological connectivity. In particular, measurements of dissolved oxygen in sediment porewater would help determine whether disturbance-induced changes in oxygen availability contribute to the observed patterns in microbial community composition. In Cattle ponds, microbial community beta diversity appeared similar despite differences in geochemical conditions, whereas in No Cattle ponds, community composition was more clearly structured by SO42− and DOC gradients. We propose that cattle activity weakens the selective pressure imposed by these environmental gradients through a combination of physical disturbance and uneven nutrient inputs. Grazing, trampling, and the deposition of manure and urine likely generate strong small-scale spatial heterogeneity within ponds, creating localized nutrient hotspots alongside relatively depleted areas38. Such patchiness can disrupt the development of consistent geochemical niches, reducing the strength of environment–community relationships at the scale of individual samples. This disturbance-driven heterogeneity may increase variability among microhabitats while simultaneously promoting convergence at broader spatial scales. Similar to patterns described in disturbed ecosystems40,41, transient increases in response diversity and asynchrony among local patches can weaken coarse scale environmental control over community structure38. In this context, microbial communities in Cattle ponds may be shaped more by stochastic and patch-level processes than by consistent geochemical gradients. This is supported by the observation that signals of SO42− and DOC became more apparent only after averaging samples at the pond scale, suggesting that underlying geochemical controls persist but are masked by within-pond variability. In contrast, No Cattle ponds showed that environmental gradients consistently structured microbial communities regardless of sampling location or time. The relative uniformity of these systems implies stronger environmental filtering and reduced micro-scale heterogeneity. Finally, we recognize that differences in sampling frequency between pond categories introduce the potential for pseudo-replication. However, by comparing patterns at both the sample and pond scales, and by averaging replicate samples, we were able to account for this effect and showed that the observed weakening of geochemical structuring in Cattle ponds is unlikely to be solely an artifact of sampling design. Microbes can convert SO42− to hydrogen sulfide (H2S), a poisonous gas, under reducing conditions (i.e., environments where high-energy electron acceptors such as oxygen and nitrate are depleted, and microbes rely on lower-energy acceptors such as SO42− or carbon dioxide) in sediment42, manure (through the addition of organic sulfur compounds43,44, water columns45, or rumens46. Whether sulfur can affect livestock health as SO42− or H2S, or becomes unavailable via assimilatory SO42− reduction in the sediment, depends on the microbial community and redox conditions of the sediment-water interface47. The wetland sediments, likely experience predominantly reducing (anoxic) conditions under saturation, with rapid oxygen depletion followed by the use of alternative electron acceptors (,NO3- Mn(IV), Fe(III), SO42−, CO2), although high DOC concentrations may promote overlapping redox processes rather than a strictly sequential pathway48–50. The wetland ponds with No Cattle had Thiobacillus as their core microbiome and the presence of Thiobacillus was positively correlated with increased SO42− concentrations, while this was observed in Cattle ponds the abundance of Thiobacillus was overall lower than No Cattle ponds. Species in this genus are chemolithoauthotrophic (microorganisms that obtain energy from inorganic compounds (chemo-), use electrons from inorganic donors (litho-), and fix carbon from CO2 (autotrophic)51, chemolithoheterotrophic (microorganisms that require organic carbon for growth52, and mixotrophic sulfur-oxidizing bacteria that derive energy by oxidizing reduced sulfur compounds, such as sulfide or thiosulfate, and converting them into SO42− and sulfuric acid53,54. However, Thiobacillus only contributes to acidification through sulfuric acid production and contributes to reducing H2S toxicity by oxidizing it to SO42– 55. The conversion of H2S to SO42−, mitigate H2S toxicity in ecosystems, promotes SO42− availability for plant uptake, and enhances soil fertility56. Thiobacillus abundance was lower in Cattle ponds compared to No Cattle ponds, likely due to several disturbances associated with cattle activity. Microbial communities are often found within pores that vary in size, typically between 0.8 and 10 mm in diameter57. One major factor that might impact the Thiobacillus abundance is the disruption of sediment layers (i.e. disruption of pores58 caused by trampling, which leads to the homogenization of the sediment layers. This disturbance disrupts the delicate microhabitats, the oxic-anoxic interfaces, where Thiobacillus typically thrives53. These bacteria require specific environmental conditions (the narrow zones and gradients where sulfide and oxygen coexist59 to carry out sulfur-oxidizing processes, and without these zones of reduced sulfur and oxygen coexisting, they cannot flourish. Another hint about change in the redox condition can be concluded from the potential function analysis. Functional predictions indicated a restructuring of redox processes in Cattle ponds. Nitrogen cycling pathways associated with low-oxygen, particularly denitrification and nitrate reduction, were higher than No Cattle ponds, while nitrification was lower, which might suggest oxygen limitation. In contrast, sulfur respiration pathways were higher in No Cattle ponds, indicating more reducing conditions in those environments. Although sulfur oxidation pathways (e.g., dark sulfide oxidation and anoxygenic photoautotrophy) were present in Cattle ponds, these are typically associated with low-oxygen conditions60. Together, these patterns suggest a shift from strongly reducing, sulfur-dominated conditions toward more redox-variable environments in Cattle ponds. Further investigation is needed to better understand the precise environmental conditions affecting Thiobacillus in cattle-impacted wetland ponds. Experimental approaches, such as sediment core analysis, and oxygen profiling would help clarify the factors limiting Thiobacillus abundance. We acknowledge that our sampling design was unbalanced, with eight wetlands and only three dugouts in the No Cattle category, and 19 dugouts and only two wetlands in the Cattle category. However, this distribution reflects the practical realities of agricultural landscapes in regions such as Saskatchewan, where dugouts are primarily constructed to support livestock watering, while natural wetlands are often avoided due to their typically high SO42− concentrations. As a result, finding wetlands with cattle access, or dugouts without cattle access, is relatively uncommon. Despite this imbalance, the few contrasting cases included in our study were informative: dugouts without cattle showed patterns consistent with other undisturbed systems, while wetlands with cattle still showed signal of disturbance (Fig. S11). This consistency across exceptions supports our overall interpretation that livestock activity, rather than pond type (wetlands vs. dugout) alone, plays a key role in shaping microbial community structure. Table 1Impact of various disturbances on microbial communities.EnvironmentDNA extraction mediumDisturbance typeDisturbance mechanismMicrobial community measurements(disturbed vs. undisturbed)Specific taxa(disturbed vs. undisturbed)ReferencesAgricultural ponds and prairie wetland pondsSaturated soil and sedimentCattleTrampling, urine and manure inputDecreased: Richness (Chao1), Richness and Evenness (Shannon)Decreased: Proteobacteria (Thiobacillus),Desulfobacterota,Increased: Actinobacteriota, Firmicutes, VerrucomicrobiotaThis studyPrairie ponds and springsSoilCattleTrampling, urine and manure inputSimilar: Richness (observed)Decreased: Richness and Evenness (Shannon)Decreased:Proteobacteria, Planctomycetes Increased: FirmicutesNeupane et al.61Arid shrublandsBiocrust or soilHumanFoot tramplingDecreased: BiomassIncreased:Bacterial community diversity (T-RFLP profile)Decreased: Cyanobacteria Increased: Actinobacteriota, Cloroflexi, BacteroidotaKuske et al.62Agricultural landsoilCattle and poultryTrampling, litter, urine, manure inputIncreased: Richness (Chao1), Richness and Evenness (Simpson)Decreased:Actinobacteriota, Verrucomicrobiota Increased: Proteobacteria, FirmicutesYang et al.63RiverWaterYaks and sheepTrampling, urine and manure inputDecreased:Richness (ACE), Richness and Evenness (Shannon)Decreased: Cyanobacteria Verrucomicrobiota Increased: Burkholderiaceae, Acidobacteria, Flavobacteriales, Deltaproteobacteria, RhodobacteralesXu et al.64Grassland (meta-analysis of 95 studies)SoilSheep, cattle, sheep+cattleTrampling, urine and manure inputHigher grazing intensity = lower microbial diversity, lower soil organic carbonNot mentionedXu et al.65Arid shrublandSoilSheepGrazingHigher grazing intensity + less vegetation = lower phylogenetic diversityAbundance not calculatedOlivera et al.[66 Table 1Impact of various disturbances on microbial communities.EnvironmentDNA extraction mediumDisturbance typeDisturbance mechanismMicrobial community measurements(disturbed vs. undisturbed)Specific taxa(disturbed vs. undisturbed)ReferencesAgricultural ponds and prairie wetland pondsSaturated soil and sedimentCattleTrampling, urine and manure inputDecreased: Richness (Chao1), Richness and Evenness (Shannon)Decreased: Proteobacteria (Thiobacillus),Desulfobacterota,Increased: Actinobacteriota, Firmicutes, VerrucomicrobiotaThis studyPrairie ponds and springsSoilCattleTrampling, urine and manure inputSimilar: Richness (observed)Decreased: Richness and Evenness (Shannon)Decreased:Proteobacteria, Planctomycetes Increased: FirmicutesNeupane et al.61Arid shrublandsBiocrust or soilHumanFoot tramplingDecreased: BiomassIncreased:Bacterial community diversity (T-RFLP profile)Decreased: Cyanobacteria Increased: Actinobacteriota, Cloroflexi, BacteroidotaKuske et al.62Agricultural landsoilCattle and poultryTrampling, litter, urine, manure inputIncreased: Richness (Chao1), Richness and Evenness (Simpson)Decreased:Actinobacteriota, Verrucomicrobiota Increased: Proteobacteria, FirmicutesYang et al.63RiverWaterYaks and sheepTrampling, urine and manure inputDecreased:Richness (ACE), Richness and Evenness (Shannon)Decreased: Cyanobacteria Verrucomicrobiota Increased: Burkholderiaceae, Acidobacteria, Flavobacteriales, Deltaproteobacteria, RhodobacteralesXu et al.64Grassland (meta-analysis of 95 studies)SoilSheep, cattle, sheep+cattleTrampling, urine and manure inputHigher grazing intensity = lower microbial diversity, lower soil organic carbonNot mentionedXu et al.65Arid shrublandSoilSheepGrazingHigher grazing intensity + less vegetation = lower phylogenetic diversityAbundance not calculatedOlivera et al.[66 Impact of various disturbances on microbial communities. Decreased: Proteobacteria (Thiobacillus), Desulfobacterota, Increased: Actinobacteriota, Firmicutes, Verrucomicrobiota Similar: Richness (observed) Decreased: Richness and Evenness (Shannon) Decreased: Proteobacteria, Planctomycetes Increased: Firmicutes Decreased: Biomass Increased: Bacterial community diversity (T-RFLP profile) Decreased: Cyanobacteria Increased: Actinobacteriota, Cloroflexi, Bacteroidota Decreased: Actinobacteriota, Verrucomicrobiota Increased: Proteobacteria, Firmicutes Decreased: Richness (ACE), Richness and Evenness (Shannon) Decreased: Cyanobacteria Verrucomicrobiota Increased: Burkholderiaceae, Acidobacteria, Flavobacteriales, Deltaproteobacteria, Rhodobacterales Conclusions The findings of this study showed that livestock disturbance can alter the structure and functioning of pond microbial communities. By disrupting sediment microhabitats and creating spatially heterogeneous nutrient inputs, cattle weaken the influence of natural geochemical gradients, potentially reducing ecosystem services such as sulfide detoxification and nutrient recycling. In this context, microbial communities in the presence of cattle may be shaped more by stochastic and patch-level processes than by deterministic processes, suggesting anthropogenic disturbances can weaken local environmental filters. These patterns highlight the sensitivity of microbial assemblages to land-use practices and reinforce the importance of considering microbial ecology when assessing ecosystem responses to disturbance in agricultural landscapes. This could have far-reaching implications for ecosystem function, particularly in agricultural landscapes where the balance of biogeochemical cycles is crucial for maintaining water quality. Studying small bodies of water is essential to detect and respond to changes in microbial dynamics, water chemistry, and ecosystem health, ensuring the long-term sustainability of these critical environments in the face of ongoing anthropogenic pressures. Materials and methods Field sites Southern Saskatchewan is part of the larger Prairie Pothole Region, which covers approximately 850,000 km2 of the Northern Great Plains67. Annual precipitation in the northern region of the PPR for the period 1994–2017 is approximately 360 mm/year, with around 80 mm/year falling as snow68. The monthly mean air temperature for the same period varies between − 14.7 °C in January and 18.7 °C in July68. Within this expansive region, millions of small productive shallow (depth of less than two meters) wetland ponds of varying sizes (typically between 0.1 and 0.25 ha69) are found70–72. Most of the wetland ponds are temporary, typically retaining water for just one to two months, while some are seasonal, lasting around two to three months71,73. The PPR is also home to many agricultural reservoirs colloquially known as “dugouts” that typically hold water for longer durations, have greater relative depths, and lower concentrations of DOC, total dissolved phosphorus (TDP), and total dissolved nitrogen (TDN) as compared to wetland ponds74. These dugouts were constructed by excavating soil in a depressional area and allowing it to fill via surface, subsurface, or groundwater flow74. The sediment and saturated soil samples’ organic matter (OM) content ranged from 5 to 40%. The wetlands’ sediments were mostly dark in color, and they usually had a rotten egg smell. The dugout’s sediments usually had more clay content. Samples from 8 wetland ponds and 1 dugout were collected between May and September 2021 (once a month) and samples from 4 wetland ponds and 22 dugouts were collected once in June 2021 across southern Saskatchewan in Canada (Fig. 7A and B, Table S4). We retained all samples in our analyses to maintain a relatively balanced sample size between cattle-access and no-cattle ponds. Overall, there were 32 ponds that were used in this study: 11 with no cattle access (“No Cattle”; total of 3 dugouts and 8 wetlands), and 21 ponds with full water access to cattle (“Cattle”; total of 19 dugouts and 2 wetlands). There were three additional ponds located in cattle farms with no direct cattle access ( “Near Cattle”) but were not included in the main statistical analysis due to the low number of samples. However, they are included in the SI for comparison. Additionally, to account for potential effects of sample size on our analyses, we followed the recommendations of Kers & Saccenti75 by conducting power calculations and using multiple alpha and beta diversity metrics to ensure robust statistical interpretation. Sites were as follows: Cattle access 21 sites (19 dugouts and 2 wetland ponds), Near Cattle access 3 sites (1 dugouts and 2 wetland ponds), No Cattle access 11 sites (3 dugouts and 8 wetland ponds) and were located from 532 m to 827 m above sea level on the landscape (Table S5). Sample collection and analysis In Cattle and Near Cattle ponds, all samples were taken from ~ 5 cm water depth (i.e. flooded sites near the edge of the pond) and at the deepest section of the pond (0.06–3.7 m, Fig. 7C). In six No Cattle ponds (located in a conservation area called St. Denis National Wildlife Area), we had only samples from the ~ 5 cm water depth, while for the other five No Cattle ponds (two located in St. Denis National Wildlife Area, three in different farmland locations in Southern Saskatchewan) we had both surface (~ 5 cm water depth) and deep samples (0.5–1.8 m). Sediments from ~ 5 cm water depth were obtained using acrylic core tubes 5 cm in diameter. Samples from the deepest part of the ponds were obtained using an Ekman Grab Sampler76. To minimize cross-contamination, multiple acrylic cores were used during each field trip. Prior to sampling, each core tube was rinsed three times with pond water. The Ekman grab sampler was similarly rinsed three times with pond water before each use. The top 2 cm collected sediment was placed in a Whirl-Pak® bag and immediately stored on ice until arriving at the lab, where they were frozen at − 20 °C for subsequent DNA extraction and geochemical analysis within 6 months. At each sampling site, we collected three replicates. Water for the analysis of [SO42−], [DOC], and chlorophyll a (Chl.a) was collected in a pre-washed bottle rinsed three times with pond water. To minimize sediment resuspension during sampling, water was collected on the side of the pond where wind disturbance was minimal, with the bottle facing towards the wind to avoid sediment particles from mixing. Any water showing visible sediment particles during collection was discarded prior to final sample collection. To avoid sediment resuspension, we collected water upstream of the direction of the wind, ensuring minimal disturbance to the sediment and discarded the initial pond water behind us. After the third rinse, we carefully collected the final sample and avoided any sediment contamination or disruption. Samples were kept on ice, and within 24 h, filtered through Whatman™ GF/C 1.2 μm pore size and, subsequently, through HA Membrane 0.45 μm pore size filters (Millipore, USA) and stored at 4 °C until analysis. DOC and SO42− concentrations were analyzed by standard methods77 on an Aurora 1030D TOC Analyzer (University of Regina) and a SmartChem® 170 Discrete Analyzer (University of Saskatchewan), respectively. Whatman™ filters were wrapped in aluminum foil and stored in the dark at − 20 °C for Chl.a extraction and subsequent analysis using an Agilent 8453 UV-visible spectrophotometer (University of Regina;78). Other water parameters (DO saturation %, DO mg/L, pH, conductivity, temperature) were coincidentally collected from the surface and deep water during the water sampling using a YSI ProQuatro multiparameter meter (Yellow Springs, USA). Loss on Ignition (LOI) analysis was used to determine the organic matter content (%OM) of soil and sediment samples, following the method described by Agriculture and Agri-Food, Canada (2013). In brief, pre-dried samples (at 105 °C) were weighed and combusted in a muffle furnace at 550 °C for 6 h to remove OM. The %OM was calculated as: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\% {\mathrm{OM}}=\left( {{\text{Initial dry weight}} - {\text{Post ignition weight}}} \right)/\left( {{\text{Initial dry weight}}} \right) \times {\text{ 1}}00$$\end{document} DNA extraction and bioinformatics DNA was extracted from 0.25 g of sediment using a DNeasy® PowerSoil® Pro Kit (Qiagen, Germany). All DNA samples were quantified using a Thermo Scientific Qubit 2.0 Fluorometer dsDNA broad-range assay kit. To assess DNA purity, the ratio of absorbance at 260 nm and 280 nm was measured using Thermo Scientific NanoDrop spectrophotometers79. The V4 region of the 16S rRNA gene was amplified using a modified version of a dual-index primer pair 515 F/806R as described in Kozich et al. (2013). Details of PCR conditions, indexing strategy, and library preparation steps are provided in SI (S1 - Sect. 1). Sequencing adaptors and barcodes were bound to ~ 10 ng of amplicon DNA using Nextera™ DNA Prep (Illumina Inc.). 174 libraries were sequenced using Illumina MiSeq V2, 2 × 250 bp paired-end reads to ensure coverage of the whole region of V4 (254 bp) with ~ 46 bp overlap. Library preparation and sequencing were performed by the UBC Sequencing and Bioinformatics Consortium for 90 samples (Vancouver, CAN) and by Institute of Microbial Systems and Society for the other 89 samples (Regina, CAN). The raw sequence reads were deposited in the United States National Center for Biotechnology Information (NCBI) GenBank Short Read Archive (SRA) under accession numbers PRJNA1105329 and PRJNA1203285. Statistical methods After obtaining raw sequence reads from each sample, bioinformatic and statistical analyses were conducted to investigate the microbial communities. First, the forward and reverse reads were merged, denoised, and chimera-filtered using the dada2 pipeline80 within the R framework81(more details in SI). This process generated amplicon sequence variants (ASVs) representing the exact sequences that could be assigned to taxonomies80. The SILVA V138 database was employed to compare the ASVs and assign taxonomic identities80,82,83. A total of 78,177 ASVs distributed across seven taxonomic ranks were identified. In addition to these ASVs, 3648 ASVs could not be confidently assigned to any taxonomic group. Less prevalent ASVs (those with a relative abundance of less than 0.0001 of the total samples (and also chloroplast and mitochondria) were filtered out. The sequenced results were assessed using the decontam package84 to identify and remove potential contamination (more detail in SI; Sect. 4). 838 ASVs were identified as contaminants out of 78,177 total ASVs that corresponds to 1.07%. The investigation of ZymoBIOMICS Microbial Community Standard (D6300) which was included as a positive control showed that the sequencing error was about 1%, however, the sequencing was more biased toward Psedomonas (the observed relative abundance was higher than theoretical), Listeria and Lactobacillus (the observed relative abundance was lower than theoretical abundance, details in SI – Sect. 4). To identify top taxa, a modified version of ggnested (a wrapper function around ggplot2) that allows visualization of the data with a two-level nested, was performed using fantaxtic package85. Sampling depth and Good’s coverage (defined as the proportion of amplicon sequence variants (ASVs) observed more than once, indicating whether sequencing depth was sufficient86) were calculated using formula 100 × (1 - n_sings / n_seqs), where n_sings is the number of singletons and n_seqs is the total number of sequences (Table S687). Samples were then normalized using the rarefy-even-depth() from the phyloseq package (Fig. 1B; v1.44.088) with the sample size (rarefaction depth) of 15,000 with 1000 iteration and 50 steps. The number of ASVs decreased after rarefaction from 77,339 to 56,923 and 16 samples were removed due to low sequence depth (Fig. 1B). The total number of samples therefore decreased from 174 to 158. The Near Cattle samples (n = 6) were also excluded from the main analysis; therefore, the final number of samples was 151. To assess within sample microbial alpha diversity, Chao1 and Shannon’s (and ACE, Fisher, Simpson) indices were calculated using the phyloseq package with plot_richness() function (v1.44.088). Pielou was calculate using (shannon/log observed)89. The significant differences in alpha diversity of the two cattle presence categories were calculated using Wilcoxon test and adjusted using the false discovery rate (FDR) correction. To ensure our sample size was sufficient to detect differences in alpha diversity, we performed a power analysis using the pwr.t.test() function in the pwr package90. The results indicated that a minimum of 64 samples per group was needed to detect a moderate effect size (d = 0.5) with 80% power at a significance level of 0.05. Our sample sizes for Cattle and No Cattle sites exceeded this threshold, which ensured robust statistical comparisons75. However, the power analysis for the Near Cattle ponds showed a power of only 20%, which may limit the ability to detect significant differences in this group and therefore excluded from the main analysis. Principal coordinate analysis (PCoA, method = Bray91 was used to identify variables explaining the most variation in each sample, enabling the visualization of sample dissimilarities using the phyloseq package. Weighted UniFrac analysis was performed using the ordinate() function from the phyloseq package. Weighted UniFrac which account for both phylogenetic relationships and the relative abundance of ASVs ordinate(method = PCoA, distance = UniFrac, weighted = TRUE), whereas unweighted UniFrac takes into account phylogenetic relationships but not relative abundance of ASVs (weighted = FALSE). The normalized option was set to TRUE so that the output values range from 0 to 1. Distance-based redundancy analysis (dbRDA) was performed to investigate relationships between microbial community structure and environmental variables35 using cal_ordination (method = “dbRDA”, measure = “bray”) from microeco package (v1.7.192). Cal_ordination_envfit() were used to fit the environmental vectors to the ordination to obtain the contribution of each variable and cal_ordination_anova() were used which uses ANOVA to test the significance of the variables and axis in ordination. The plot was created using plot_ordination() function. To further investigate the differences in microbial communities, we conducted linear discriminant analysis (LDA) to differentiate microbial communities across pond categories and identifying biomarkers using the microbiomeMarker package93. The function to calculate LDA was run-lefse() with a Wilcoxon cutoff of 0.05 and normalization method of “CPM” (count per million; pre-sample normalization of the sum of the values to 1e + 06.)93 These results were plotted using the plot_abundance(), the plot_heatmap() functions from the same package. Mean relative abundance (%) of core microbial genera (core taxa = at least be present in 30% of the samples) across ponds categorized by cattle presence and pond category was calculated using core_member() function from the microbiome package (v1.26.094). In addition, mean relative abundance of Thiobacillus plotted against average SO42−, DOC, Chl.a, %OM, DO%, and DO concentrations across ponds with each group represented by separate linear regression lines using lm() function from the stats package. Redundancy analysis (RDA) to investigate relationships between microbial community structure and environmental variables was also performed using cal_ordination (method = “RDA”, taxa_level = “Genus” ) and trans_ordination (show_taxa = 10, adjust_arrow_length = TRUE, max_perc_env = 1.5, max_perc_tax = 1, min_perc_env = 0.1, min_perc_tax = 0.1) from the microeco package (v1.7.192). This analysis allowed for the identification of significant correlations between the environmental variables and the top 10 microbial taxa. The prediction of functional profiles based on the prokaryotic communities from sequencing results95,96 was calculated using cal_spe_func (prok_database = “FAPROTAX”) from the microeco package. Pairwise relationships between environmental variables and their correlation across different cattle presence categories were calculated. Scatter plots in the matrix used pairwise relationships between each pair of variables. Density plots were used to show the distribution of each environmental variable by category. Correlation coefficients (Corr) were used to summarize the overall and category-specific correlations between variables. The analysis and plot created using the microeco (v1.7.192) and the GGally (v2.2.197). A variance partitioning Venn diagram showed the relative contribution of geochemical factors (SO42−, DOC, conductivity, DO %, DO mg/L, %OM, water pH, temperature, and chl.a) and spatial variables to the microbial community composition. Geochemical and spatial importance were analyzed using the var_par_pq() function from the MiscMetabar package98, which partitions variance across defined components. Spatial variables were derived from PCNM analysis (principal coordinates of neighbor matrices, using pcnm() function from the vegan package), performed on Euclidean distance matrices calculated from latitude and longitude coordinates using the distm() function based on Haversine formula from the geosphere package (v1.5-2099). Variance partitioning was visualized using plot_var_part_pq() from the MiscMetabar package to highlight the importance of both geochemical and spatial variables in structuring microbial communities. The same statistical analysis for the Near Cattle ponds could not be performed due to the limited number of samples available (ponds = 3). All other graphs were plotted using the ggplot2, grid, corrplot, and lattice packages100–102. All statistical tests were considered significant at p-value ≤ 0.05. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1