INTRODUCTION

The prairie, once the largest ecosystem in North America, covering about 15% of the continent, created a triangle that spanned from Saskatchewan and Manitoba in Canada, west to the foothills of the Rocky Mountains, south into Texas, and east into Indiana (Robertson et al. 1997; Anderson 2006). Of the original ∼69 million ha of prairie once found across the United States, ∼9 million ha occurred in what is now the state of Illinois (Corbett 2004). As of 2018, only 1010 ha of high-quality prairie remains in Illinois, with most less than 5 ha in size (Corbett 2004). Conversion of land to agricultural uses followed by fire suppression are common reasons for prairie loss (Anderson 1970). With this extreme loss of the prairie ecosystem, establishment and proper management of restorations and prairie remnants is particularly valuable.

Prairie, especially eastern tallgrass prairie which typically receives >800 mm of precipitation each year, is a disturbance-dependent ecosystem. Historically, this grassland ecosystem relied on both fire and grazing disturbances for its existence (Knapp et al. 1999; Collins 2000; Fuhlendorf and Engle 2004). Prior to European settlement, Native Americans were key in maintaining the grass-dominated vegetation of the North American prairie (Curtis 1959; Anderson 1990, 2006). This relationship between Native Americans and the burning of grasslands extends back 30,000 y (Bragg 1995) and was used to prevent woody encroachment of grasslands, increase hunting access, encourage prairie growth, clear areas for agriculture, control insects, and ease traveling (Stewart 1956; Curtis 1959; Pyne 1982; Anderson 2006).

The prevalence of bison on the grasslands of North America has dramatically changed over the past 200 y. Rampant hunting from 1830 to 1880 reduced the estimated number of bison from 30–60 million (Shaw 1995) to a few thousand (Knapp et al. 1999). Therefore, our understanding of the extent of a free-range bison's influence on the tallgrass prairie is limited (McHugh 1972; Flores 1991). Bison restricted in range to large properties demonstrate a diet preference largely composed of grasses, preferably the warm-season prairie grasses, and sedges with less than 10% consumption of woody vegetation and forbs (Knapp et al. 1999). Based on documentation of large numbers of bison east of the Mississippi by French explorers and Jesuit missionaries, researchers estimate that maximal bison populations were found in the prairie peninsula of Illinois during the 17th and 18th centuries (Bamforth 1987; McMillan 2006).

Without fire and grazing, the accumulation of plant litter and dominance of grass reduces forb abundance and diversity, the latter a key element contributing to the diversity of prairies (Collins 1992; Briggs and Knapp 1995; Fuhlendorf et al. 2009). While burning in grasslands favors a homogeneity of dominant grasses and a reduction in woody vegetation, bison grazing reduces grass density and competition thereby increasing forb abundance and diversity (Hartnett et al. 1996; Collins et al. 1998; Knapp et al. 1999; Powell 2006). Fuhlendorf and Engle (2004) stress that, although the effects of bison and fire are often studied individually, the interaction between these disturbances is more important than the sum of either individual effect. Importantly, bison prefer to graze in recently burned areas, regardless of fire season (Knapp et al. 1999; Fuhlendorf and Engle 2004; Vermeire et al. 2004; Allred et al. 2011; Winter et al. 2015). This preference has far-reaching effects for ecosystem processes in the landscape (Archibald et al. 2005; Fuhlendorf et al. 2006; Leonard et al. 2010; Allred et al. 2011). The dominant grasses are suppressed, forbs and bare ground increase, and without fuel, fire is unlikely. Where bison did not graze litter accumulates and these are the patches likely to burn during fire activity. Once burned again, bison return to graze and a shifting patch pattern of burned, grazed, and untouched units occurs (Biondini et al. 1999; Fuhlendorf and Engle 2004). These feedbacks create a “shifting mosaic” and increase the degree of temporal and spatial heterogeneity across the landscape (Hartnett et al. 1996; Knapp et al. 1999; Fuhlendorf and Engle 2004; Powell 2006; Blackburn et al. 2020).

This spatial heterogeneity is challenging to replicate with restored tallgrass prairie, which are often small (tens to hundreds of acres) and in a landscape context of row-crop agriculture. Decades of restoration efforts have gained success in re-establishing native vegetation; however, over time these plantings often became dominated with warm-season native grasses (Grman et al. 2013). While prairies are grass-dominant, many of these plantings were lacking in the flowering plant diversity observed in unplowed native prairies, even with frequent burning. With accruing evidence that bison grazing was a fundamental ecological factor for tallgrass prairie, managers began hypothesizing how reintroducing grazing may sustain plant (and animal) diversity on restored prairies. A handful of restored prairie sites are large enough to accommodate bison reintroduction. Bison have strong matriarchal herd social dynamics across generations (King et al. 2019) and require at least modest-sized herds and large contiguous landscapes for grazing. In addition to numerous intact native prairies, a few larger tallgrass prairie restorations have reintroduced bison, including the Neal Smith National Wildlife Refuge (US Fish and Wildlife Service, Iowa) and Dunn Ranch Preserve (The Nature Conservancy, Missouri). Where introduced, grazing coupled with seed additions have been observed to maintain or increase plant diversity in restored tallgrass prairie (Wilsey and Martin 2015).

In October of 2014, one of the first conservation herds of bison east of the Mississippi River was introduced to The Nature Conservancy's Nachusa Grasslands (hereinafter, Nachusa) to increase plant diversity and heterogeneity. Although bison were historically present in this area of Illinois, The Nature Conservancy's decision to reintroduce bison at Nachusa was first and foremost based in land management rather than simple replication of historical free-ranging bison herds once found in the region. Expectations from land managers include reduced warm-season tallgrass abundance, increased forb abundance and diversity, and niche space for plants and animals that need more open structure (Kleiman 2016). In this study, we ask what effect 5 y of bison grazing has on the plant community at Nachusa. We investigated plant species diversity, composition, abundance, and visual obstruction readings (VOR) in paired grazed and ungrazed plots, including baseline measurements prior to bison reintroduction. Our hypotheses were as follows: (1) As compared to burned plots without grazing, plots with grazing and burning have increased species diversity, greater proportion of forb to grass abundance, and greater proportion of native to nonnative species. (2) VOR are lower in grazed plots compared to ungrazed plots.

METHODS

Nachusa Grasslands is located near Franklin Grove, Illinois (41.883706°N, –89.342410°W) and is owned and managed by The Nature Conservancy (TNC). Nachusa consists of more than 1600 ha of native prairie remnants, restored prairies, savannas and woodlands, and wetlands. More than 700 native plant species and 180 breeding or migrating bird species have been identified on the site. Precipitation averages 675 mm during the growing season (1 April–30 September) and 1030 mm annually. Average annual temperature from 1991 to 2020 was 9.6°C, with an average minimum and maximum temperature of 4.5°C and 14.8°C, respectively (U.S. Climate Normals; NCEI 2022).

In 2014, American bison (Bison bison) were reintroduced to Nachusa in a phased approach. An initial 30 animals were released onto ∼200 ha (the north unit) in fall 2014 and grazed exclusively in the north unit for the 2015 growing season. In addition to 20 calves born in spring 2015, an additional 20 adult animals joined the herd in fall 2015 and the grazing area was expanded to include the 400 ha south unit. Since winter 2015/16, the herd has had full access to the entire 600 ha grazing area. After 2018, the herd has been managed at approximately 100 adult animals.

To examine the impact of bison grazing on plant communities, twenty-two 10 m × 18 m fenced plots (exclosures) and paired unfenced plots were established prior to bison reintroduction (Fig. 1). Fences are electrified metal wire. The plots are stratified to represent the habitat types at Nachusa (Table 1). There are six sampling sites on native prairie (n = 3 in the north unit, n = 3 in the south unit), six sampling sites on older plantings (n = 3 in the north unit, n = 3 in the south unit), four sampling sites on newer plantings (n = 3 in the north unit, n = 1 in the south unit), and three sampling sites on savannas (n = 3 in the south unit). There are three additional sampling sites on wetlands in the south unit, which were not included in the analyses presented here due to the substantially different plant community. This unbalanced design results from uneven representation of these habitat types across the preserve. For example, there are no savanna or wetland habitats in the north bison unit. The stratified experimental design ensures three independent replications of each habitat type, and where possible more than three. All treatment areas but one were included in areas that had been burned within 2 y of sampling.

We used a stratified random approach to placing exclosures on the landscape within each habitat type. Habitat types are not randomly distributed across the site, which constrained a true random placement of sampling sites. Some habitat types only encompass small patches (1–2 ha) within the landscape and exclosures were placed where adequate space was available. Native prairie and wetland habitats at Nachusa primarily exist in patches <5 ha. For each exclosure, three 12.5 m permanent transects inside the fence (ungrazed) are paired with three transects of the same length outside the fence (grazed) ( Supplemental Fig. S1 (24-4 Larson_ Online supplement.docx)). Transect lines are located at least 2 m from fence lines and each transect is 3 m from the adjacent transect. Quarter-square-meter quadrats are placed on the east side of each transect line at five regular intervals (0–0.5 m, 3–3.5 m, 6–6.5 m, 9–9.5 m, and 12–12.5 m) for a total of 15 quadrats in each grazed and ungrazed plot.

Figure 1.

Location and habitat type of exclosures at Nachusa Grasslands.

Species Diversity, Abundance, and Composition

Within each 0.25 m² quadrat, we identified all species rooted in the sub-plot and visually estimated their percent cover for a total of 570 surveyed quadrats. Plots were sampled regardless of whether they had been grazed recently. Plant nomenclature follows Wilhelm and Rericha (2017). Plant community composition and cover were determined in the north unit in 2014 (pre-bison), 2017, and 2019. In the south unit, data were collected in 2015 (pre-bison), 2018, and 2020. Because the bison were introduced to the north unit in fall 2014 and the south unit in fall 2015, this corresponds to year zero, three, and five for each grazing unit. In 2014, Daubenmire cover classes (Daubenmire 1959) were used instead of percent cover. Class 1 is 0–5% coverage, class 2 is 5–25%, class 3 is 25–50%, class 4 is 50–75%, class 5 is 75–95%, and class 6 is 95–100%. To include the 2014 data in our analyses, we converted the 2015–2020 data to Daubenmire classes and then transposed all data to percent cover midpoints. For this analysis, all plants in the Poa genus were converted to Poa spp. and all plants in the Carex genus were converted to Carex spp.

Table 1.

Surveyed exclosure plots and their habitat type, when the units were restored, years since last burned, and burn season. “N” denotes that the plots were in the north bison unit and “S” denotes that plots were in the south bison unit. *Wetland plots were not included in our analyses.

RESULTS

Plant community diversity, nonnative:native species ratio, grass:forb ratio, and VOR did not differ between grazed and ungrazed plots overall. There was an interaction with community type in which grazed plots in savannas had higher grass:forb ratios. This difference was consistent across all sampling years, indicating these differences were present before bison were reintroduced (F_3,100 = 4.77, P = 0.004;  Supplemental Fig. S2 (24-4 Larson_ Online supplement.docx)). Native prairies and older restorations had greater plant Shannon's diversity (F_3,103 = 45.54, P < 0.0001) and grass:forb ratio (F_3,100 = 17.81, P < 0.0001) and lower nonnative:native plant ratio (outlier removed, F_3,85 = 4.12, P = 0.009) compared to savannas and new restorations (Fig. 2). VOR was only collected in year five after bison. Native prairies had the lowest VORs (F_1,30 = 3.97, P = 0.02; Fig. 2), likely reflecting that native prairies on this landscape are all located on the top of dry sandy hills, which limit plant growth. New restorations had the highest VORs, which may reflect more mesic conditions and/or the low grass:forb ratio.

Figure 2.

Plant community Shannon's diversity (A), P < 0.01), nonnative:native plant ratio (B), P = 0.01), grass:forb ratio (C), P < 0.01), and visual obstruction ratio (D), P = 0.02) all varied among the habitat types. Points represent mean values. Error bars are 61 standard error.

Year since bison reintroduction impacted Shannon's diversity and grass:forb ratio. Shannon's diversity was greatest 5 y after bison reintroduction (F_2,103 = 7.84, P = 0.0007; Fig. 3). Grass: forb ratio was lowest 3 y after bison reintroduction (F_2,100 = 6.51, P = 0.004; Fig. 3). Year three sampling was done in 2017 and 2018; both years had above average precipitation during the growing season. Nonnative:native plant ratio was the only response variable to exhibit an interaction. There were no differences in nonnative:native plant ratio in years zero and three; however, in year five, grazed plots had a greater nonnative:native plant ratio than ungrazed plots (F_2,100 = 3.84, P = 0.03; Fig. 3). Analysis within habitat types showed no differences between grazed and ungrazed plots.

Plant community composition was distinct among all prairie habitat types (PERMANOVA, P = 0.001, R² = 0.276; Fig. 4). Time since burn was correlated with overall plant community composition (P = 0.05, R² = 0.07), but only accounting for 7% of differences. Numerous species were distinct indicator species for the three prairie types ( Supplemental Table S1 (23-27 Bach_Online supplement.docx)) including dominant warm-season tallgrasses (Sorghastrum nutans, Andropogon gerardii) in older plantings and native cool-season grasses in remnants (Dichathelium spp., Hesperostipa spartea). Savanna habitats had the greatest variation in composition and did not differ from old restorations or native prairies. Unsurprisingly, savanna habitats were distinguished by several species of trees (Quercus spp., Celtis occidentalis, Ulmus americana), vines (Toxicodendron radicans, Vitis vulpine), and shrubs (Rubus spp.) ( Supplemental Table S1 (23-27 Bach_Online supplement.docx)). In addition, community composition varied in year five after bison reintroduction compared to year zero baseline (P = 0.031, R² = 0.020;  Supplemental Fig. S3 (24-4 Larson_ Online supplement.docx)). These differences were present in both grazed and ungrazed plots. This difference was in part driven by greater variability in plant community.

Figure 3.

Shannon's plant diversity (A) was greatest in year five (P < 0.01). Grass:forb ratio (B) was lowest in year three (P < 0.01). (C) The ratio of non-native:native plants exhibited an interaction between year since bison were introduced and grazing status (P = 0.03). In year five, grazed plots (black) had greater non-native:native plant ratios than ungrazed plots (gray). Points represent mean values. Error bars represent 61 standard error.

DISCUSSION

The results of this study imply that in the first 5 y, bison have had minimal, if any, effects on the vegetation diversity, composition, abundance, and structure at Nachusa Grasslands. Plant communities were distinct across community types, as expected. Across the 5 y, there were some modest changes in plant communities, which were consistent between grazed and ungrazed plots. The one important exception being a greater nonnative:native plant species ratio in grazed plots in year five. These results match those reached by Blackburn et al. (2020) who compared sites within the bison grazing area with sites outside the bison grazing area after the first 3 y of bison reintroduction at Nachusa. These consistent findings using independent experimental designs indicate that bison have had a limited effect on vegetation change at Nachusa since their introduction.

Comparison with Other Bison Reintroduction

Previous studies in the western edge of the tallgrass prairie have shown that bison grazing reduces grass density and competition leading to an increase in forb abundance and diversity (Hartnett et al. 1996; Collins et al. 1998; Knapp et al. 1999; Powell 2006). Impacts on forb and grass abundance were observable at Konza Prairie Biological Station after 6 y with bison stocking rates like Nachusa. In contrast, this study found no change in plant diversity or grazing-correlated changes in grass cover or grass:forb ratio. Explanations may be that at the current stocking rate bison are not impacting grasses and that grasses are benefitting from fire, that the bison have not been on site for sufficient time to have any impact, or that if impact is occurring the sampling design is inadequate to capture it. Baseline plant diversity at Nachusa was greater and contained greater forb cover than baseline measures observed at Konza Prairie. There is likely an upper limit on plant diversity, and it is possible Nachusa diversity baseline is/was closer to this limit. In contrast, Konza Prairie exists in a landscape of intact and historically grazed native grasslands, providing a potential regional native seed source. Nachusa exists in a landscape of row-crop agriculture and passive recruitment of native seed is less likely.

Figure 4.

A three-dimensional non-metric multidimensional scaling (NMDS) ordination visualizes differences in plant community composition among new and old prairie plantings, native prairie, and savanna habitats at Nachusa Grasslands (PERMANOVA P < 0.01). Solid black line shows the correlation of time since last prescribed fire (P = 0.05, R² = 0.07).

The marginal trend of higher average proportion of nonnative to native percent cover in grazed plots in 2019 and 2020 is worth continued monitoring. Bison are known dispersers of graminoid and forb seeds (Rosas et al. 2008) and can introduce nonnative species while moving throughout the bison enclosure on their hair or in dung (Constible et al. 2005; Rosas et al. 2008). In addition, increased bare ground because of trampling and wallowing behaviors (Grudzinski et al. 2016) may be creating germination opportunities. If bison are dispersing seeds of nonnative species, their increased management within the bison unit may be necessary.

Additionally, average VOR were lower in grazed plots for all community types. Bison are important for grassland management because their preferential consumption of dominant grasses results in an overall reduction of aboveground biomass for grassland communities (Knapp et al. 1999). A lower VOR indicates aboveground biomass has been reduced in the grazed plots compared to ungrazed plots. Removal of biomass often results in greater heterogeneity across the landscape (Vinton et al. 1993; Coppedge and Shaw 1998), and increased heterogeneity is beneficial to meeting the needs of a greater variety of fauna (Powell 2006). For example, while declining grassland birds like the Henslow's sparrow (Ammodramus henslowii) require taller vegetation, upland sandpipers (Bartramia longicauda) require short grasses. One of Nachusa's goals in introducing bison was to create a more heterogeneous or patchy landscape. Continued measurement of VOR and aboveground biomass can help land managers know whether they are meeting this goal.