Stability of C 3 and C 4 Grass Patches in Woody Encroached Rangeland after Fire and Simulated Grazing

: As the woody legume, Prosopis glandulosa (honey mesquite) has encroached into grasslands and rangelands in the southern Great Plains, USA, two grass species, C 4 shortgrass, Buchloe dactyloides (buffalograss), and C 3 mid-grass, Nassella leucotricha (Texas wintergrass), have increased in dominance. Occurrence of more productive C 4 mid-grasses and herbaceous diversity have declined. We measured effects of various combinations of spring clipping (to simulate cattle grazing) and summer and/or winter ﬁre treatments on the stability of monoculture patches of these two grass species over an eight-year period, with the goal of reducing Nassella and increasing C 4 mid-grass cover. All ﬁre treatments top-killed most Prosopis trees that subsequently resprouted. Buchloe cover declined in the No Clip + No Fire treatment but remained intact with clipping and/or ﬁre. Frequent clipping reduced Nassella cover across all ﬁre treatments. Buchloe encroachment into Nassella patches was greatest in the Clip + Alternate Season ﬁre treatment. C 4 mid-grass cover increased to 15–25% in Nassella patches in several ﬁre-only or Clip + Fire treatments; greatest gains were observed in treatments that included summer ﬁre. In contrast, C 4 mid-grass gains were lower in Buchloe patches. These results suggest that C 4 mid-grass restoration was linked with treatments that reduced Nassella cover.


Introduction
Grasslands in the western portion of the southern Great Plains (SGP), USA, are defined as "southern mixed" grasslands, with C 4 grasses being dominant and C 3 grasses contributing a smaller proportion of ANPP [1][2][3].This region does not support C 4 tallgrass species; rather, the C 4 component is composed of rhizomatous or stoloniferous C 4 shortgrasses (i.e., sod grasses and lawn grasses) and structurally larger C 3 or C 4 perennial "mid-grasses" that are taller than the C 4 shortgrasses but shorter than the C 4 tallgrasses [4,5].
Several combinations of the competitive effects of these three species are suspected to have caused the decline of C 4 mid-grasses as a percentage of total ground cover and/or biomass production of individual plants in the SGP [5,31].We hypothesize that prior to the European settlement that introduced domestic cattle grazing, much of the SGP frequently burned [32,33] and was open grassland-dominated by C 4 mid-grasses with smaller portions of C 4 shortgrasses, like Buchloe.Nassella was likely present but, as it was a C 3 species, probably only occurred as a small fraction of the grassland community.Soil isotope research has shown that many areas in the SGP now dominated by woody plants were once mostly C 4 grasslands [34,35].As Prosopis WPE advanced, so did Nassella [21].Additionally, in recent decades, the grazing management practice of heavy livestock grazing on native pasture in summer months, followed by moving livestock to dual-use Triticum aestivum (wheat) in winter may have promoted Nassella advancement, as it removed the grazing of Nassella during its growth period [36].
The presence of larger C 4 mid-grasses in SGP grasslands increases landscape structural heterogeneity that enhances species diversity and ecosystem services.The numerous grass species within this type (see Appendix A) enhance the habitat for gallinaceous and other grassland obligate bird species [37,38] and cover for other wildlife species, including Odocoileus virginianus (white-tailed deer) and Antilocapra americana (pronghorn antelope) [11].The greater root mass of C 4 mid-grasses compared to Buchloe or Nassella may also increase the potential for carbon sequestration and more diverse soil microbiomes [12,13,35].In addition, the greater inherent above-ground production potential compared to Buchloe or Nassella increases forage availability for livestock and other herbivores [5,8].Buchloe and Nassella are also important forage sources for herbivores [36,39,40], but Nassella has expanded to undesirable levels [18,36].Thus, in a Prosopis-encroached landscape where Buchloe and Nassella dominate the understory, a management goal would be to increase herbaceous diversity, in part, by reducing (but not eliminating) Nassella cover, maintaining or only slightly reducing Buchloe cover, and increasing C 4 mid-grass cover into these monoculture patches.In addition, if quantities of C 4 mid-grass propagules are limited (possible under advanced WPE), the herbaceous successional pattern that is formed after the woody plants are suppressed or removed might be from Nassella to C 4 shortgrasses to C 4 mid-grasses [11].In this case, managers would view the encroachment of Buchloe into Nassella patches as an acceptable trend.The key to achieving these goals will be to develop a sustainable management strategy that reduces the cover and/or density of the woody overstory.Here, we focused on the potential of prescribed fire to reduce Prosopis dominance.
Most prescribed fires in the SGP occur during the dormant season (February-March) [41,42].They are usually safer and more manageable compared to summer fires in areas dominated by Nassella, which is dormant and highly combustible in the summer months.However, as Nassella is green during the late winter and early spring, the level of Prosopis suppression (i.e., "top-kill") from the prescribed fires during this time can be highly variable, as green Nassella growing beneath Prosopis canopies limits the ability of fire to scorch Prosopis stems.As a result, summer fires usually achieve greater Prosopis suppression than do winter fires [43].However, since even the most extreme summer fires rarely cause whole plant mortality (i.e., "root-kill") in Prosopis [43][44][45][46], it is necessary to apply frequent burns to maintain the suppression of Prosopis regrowth [7,43].However, due to the strong association between Prosopis and Nassella in the SGP, any treatments that only top-kill Prosopis and allow it to resprout to heights > 2 m will likely maintain Nassella beneath or near the Prosopis canopy [47].
Our objective was to quantify the impact of repeated seasonal fires and simulated grazing via clipping on the stability of Buchloe and Nassella monoculture patches that dominate the grass community under advanced WPE.In addition, we wished to determine the degree to which these treatments can reduce, but not eliminate, Nassella cover and can facilitate the advancement of structurally larger C 4 mid-grasses into Buchloe and Nassella patches, thus enhancing ecosystem diversity.

Site Description
Our research occurred on a 200 ha savanna in North Central Texas, USA (+33.852603,−99.436555; elev.384 m), near the north-south mid-point in the southern Great Plains [48][49][50] and northern edge of the subtropical climate zone.The mean annual precipitation (30 year; 1977-2006) is 698 mm and the mean annual air temperature is 17.1 • C. Mean daily air temperatures range from 35.9 • C in July to −2.4 • C in January [51].Soils range from 1 to 2 m deep clay loams of the Tilman series on 1-3% slopes.The ecological site description is Clay Loam R078CY096TX [52].
The pre-treatment overstory at the point of study initiation consisted of 2-3 m tall Prosopis at 30-50% canopy cover.No fires occurred on the site for at least 40 years prior to this study.The herbaceous layer included C 3 and C 4 perennial grasses, dominated by C 3 Nassella in Prosopis subcanopy microsites, and Nassella and C 4 Buchloe in the intercanopy spaces between Prosopis trees.Buchloe was recently described as Bouteloua dactyloides [30]; however, we followed the recommendations made by the authors of [40,53,54] to retain Buchloe as the genus descriptor.Several C 4 perennial mid-grass species and annual grasses comprised a small component of the herbaceous layer [53,54] (Appendix A).Other than Nassella, the most common C 3 perennial grass species include Poa arachnifera and Elymus smithii; however, these species comprise < 1% of the total cover.This site is occasionally dominated by C 3 annual grasses.Common forbs include Ambrosia psilostachya, Amphiachyris dracunculoides, Conyza canadensis, Croton texensis, Liatris mucronata, and Solanum elaeagnifolium [36,55].The typical growing season for C 3 grasses is from February to May, and for C 4 grasses, from April to September.Between 1950 and 1991, the site experienced moderate continuous cow-calf cattle grazing at a stocking rate of 12 ha/cow [56].However, we removed cattle grazing for the duration of the study period.

Fire Treatments
The study had three replicate plots (1 to 5 ha each) in each of six treatments: (1) no fire, (2) repeated winter fires in 1991, 1993, and 1995 (3WF), ( 3) alternate season fires in winter 1991, summer 1992, and winter 1994 (3AF), ( 4) alternate season fires compressed temporally in winter 1993, summer 1993, and winter 1996 (3AFC), (5) two summer fires in 1992 and 1994 (2SF), and (6) two summer fires compressed temporally in 1993 and 1994 (2SFC) (Table 1).Winter fires were conducted in late January to mid-March when Prosopis trees were leafless.Summer fires were conducted in September when Prosopis were in full foliage.We identified each fire application in each treatment as a fire "step".We conducted all fires as prescribed head fires [41], with one exception.In the 2SFC treatment, an area adjacent to our fire plots burned in a wildfire on 24 August 1993.This area, originally a no fire treatment, was subsequently divided into three replicate plots and the second fire step (s94) was burned in summer 1994.
Except for the first fire in the 2SFC treatment, we measured the air temperature, relative humidity (RH), and wind speed on site a few minutes prior to each fire.Weather data for the first fire in the 2SFC treatment were obtained from NOAA weather records.We measured the herbaceous fine fuel amount (grass and forb standing crops and litter) in each plot within one day prior to burning at 10 randomly located 0.25 m 2 quadrats in intercanopy spaces between Prosopis trees.We also measured the fire temperatures and fireline intensity in each treatment step.Details are available in the study published by the authors of [43] and in Table S2.In summary, the air temperature, RH, peak fire temperature, and fireline intensity were greater during the summer than winter fires.The wind speed and fine fuel were found to be similar among the winter and summer fires.Most Prosopis were completely top-killed by all fire treatments, the more effective being those with summer fires (Table S3).Following the completion of all fire treatments by 1996, <3% of Prosopis were root-killed and all top-killed Prosopis had multi-stemmed resprouting that grew to 2.0-2.5 m height by 2000 [43].

Patch Identification and Clip Treatments
Within each fire plot, we located 2 or 3 patches (each 6-15 m 2 in size) dominated by Buchloe and another 2 or 3 patches dominated by Nassella in intercanopy spaces between Prosopis in the summer of 1992.Within each patch, two 0.6 × 0.9 m (0.54 m 2 ) quadrats were permanently marked.One quadrat of each pair was not clipped and was only exposed to the fire treatments.We clipped all vegetation in the other quadrat of each pair to a 2 cm height.Our original protocol was to clip twice each year, once in spring (May) and again in mid-summer (July).We followed this plan in 1993.However, after observing that Nassella had minimal regrowth from May to July, we assumed our clipping regime was too severe to meet long-term herbaceous diversity goals and changed our protocol to only clip once each May from 1994 to 1997.For the remainder of this study, we refer to the two clipping treatments from 1993 to 1997 as "No Clip" or "frequently clipped" treatments.We chose not to clip in spring 1998 due to severe drought.
We changed our protocol in 1999 and implemented a "universal" clip on all patches, previously clipped or not clipped, in the spring of 1999 to determine whether there were any residual effects of the previous fire treatments on patch tolerance to clipping.Thus, we divided this study into two "Phases".Phase 1 included responses to the frequent clipping and fire treatments from 1992 to 1998.Phase 2, from 1998 to 2000, included the universal clip in 1999 with a final assessment in 2000.
The percent basal ground cover of each herbaceous species, bare ground, and litter were determined within each quadrat via visual estimates each year from 1992 through to 2000, except for 1995.In addition, data collection in the "No Fire" patches did not begin until the spring of 1993.We measured cover during the fall before any winter fire treatments, or in the summer a few days prior to any summer fire treatments, with one exception.In 1997, we measured cover in late spring before the clip treatment.The reason for this was that we observed a sharp decline in Nassella cover in some of the treatments in 1996 and we wanted to observe any changes in spring 1997.We organized cover values into seven groups: Buchloe, Nassella, C 4 mid-grasses, C 3 annual grasses, forbs, litter, and bare ground.We identified litter as any dead herbaceous material on the surface that was not physically connected to any living plant species.All cover values in each quadrat added to 100%.

Analysis
For Phase 1 (1992-1998), treatment effects were analyzed in what were initially Buchloe or Nassella patches using a 3-way analysis of variance with the SAS PROC MIXED procedure [57] that tested the main effects and interactions of the fire treatment (6 levels), clip treatment (2 levels), and year (6 levels; no data from 1995).PROC MIXED contains both fixed-and random-effects parameters and is thus more flexible and more widely applicable in repeated measures situations than a general linear model that only tests fixed effects [57].We pooled patch subplots within each replicate fire plot prior to analysis; thus, n = 3. Response variables were the basal covers of Buchloe, Nassella, C 4 mid-grasses, C 3 annual grasses, forbs, litter, and bare ground.We conducted a second 2-way PROC MIXED analysis to test the main effects and interactions of the fire treatment and clip treatment on each functional group only in the year 1998.Where appropriate, we compared the cover of functional groups averaged over the main effects of the clip treatment, fire treatment, or year.Finally, to address the observed sharp decline in Nassella cover in some treatments in 1996, we conducted a 3-way analysis of variance with SAS PROC MIXED to compare the responses to treatments in the fall of 1996 vs. the spring of 1997.
Analysis for Phase 2 included a 3-way analysis with SAS PROC MIXED (main effects and interactions of clip, fire, and year) from 1998 to 2000, and a second 2-way analysis with SAS PROC MIXED (effects of clip and fire treatments) that only included the year 2000.In these analyses, any fire treatment effects were residual, since all fire applications had ended by 1996.In all analyses, mean separations were via least squares (p ≤ 0.05).We transformed all percentage data prior to analysis using the ArcSin transformation.

Precipitation
The annual precipitation was >20% above average in five of the ten years from 1991 to 2000 and was >20% below average in only two of the ten years (Figure 1).The C 3 growing season precipitation was >20% above average in four of the ten years, and >20% below average in three of the ten years (Figure 1B).The C 4 growing season precipitation was >20% above average in three of the ten years, and >20% below average in three of the ten years (Figure 1C).However, the three years where the C 4 growing season rainfall was >20% below average occurred in the last five years of the study after all fire treatment steps had been applied.The year 1995 had three times the average C 4 growing season rainfall (Figure 1C).A drought year occurred in 1996, where the greatest deviation from the average occurred during the C 3 growing season (Figure 1B).

Phase 1-Herbaceous Cover Responses
The main effects of clip treatment, fire treatment, and year during Phase 1 (1992-1998) were significant (p ≤ 0.05) for many variables in Buchloe and Nassella patches (Table 2).The two-way interactions that were significant for the most variables in Buchloe and Nassella patches were clip × year and fire × year.Significant three-way interactions occurred for C 3 annual grasses and forbs in Buchloe patches, and for litter and bare ground in Nassella patches.Only the main effects of clip, fire, or year were found to be significant for C 4 mid-grasses in Buchloe patches.The main effects of fire and year, and a clip × fire interaction were found to be significant for C 4 mid-grasses in Nassella patches.
Buchloe cover in Buchloe patches initially ranged from 58 to 83% (Figures 2 and 3).During Phase 1 (from 1992 to 1998), other than a temporary decline immediately after some fire treatments, Buchloe cover gradually declined to the greatest degree in fire treatments that were not clipped.Nassella encroachment into Buchloe patches from 1992 to 1998 was greatest in the No Clip + 3AF and No Clip + 2SFC treatments (Figures 2E and 3E, respectively).
Nassella cover in Nassella patches initially ranged from 43 to 55% (Figures 4 and 5).Frequent clipping, either in the no fire treatment or in any of the fire treatments, reduced Nassella cover to ~15-25% by 1998.Nassella cover sharply declined between 1994 and 1996 in some no clip fire treatments (Figures 4A, 5A, and 5C, respectively).However, by 1997, cover in all these no clip treatments increased.

Phase 1-Herbaceous Cover Responses
The main effects of clip treatment, fire treatment, and year during Phase 1998) were significant (p ≤ 0.05) for many variables in Buchloe and Nassella patche 2).The two-way interactions that were significant for the most variables in Buc Nassella patches were clip × year and fire × year.Significant three-way interact curred for C3 annual grasses and forbs in Buchloe patches, and for litter and bare in Nassella patches.Only the main effects of clip, fire, or year were found to be sig for C4 mid-grasses in Buchloe patches.The main effects of fire and year, and a cl interaction were found to be significant for C4 mid-grasses in Nassella patches.The fire treatment code is presented in Table 1.  1.
Buchloe encroachment into Nassella patches occurred in several fire treatments if frequent clipping was included but was minimal in all fire treatments without clipping.Greatest gains in Buchloe cover in Nassella patches occurred in the clipped + 3AF (Figure 4F) and clipped + 3AFC (Figure 5B) treatments.In these treatments, Buchloe cover exceeded that of Nassella.
C 4 mid-grass cover, averaged over the fire treatments and Phase 1 years (1992-1998), was significantly (p ≤ 0.05) greater in no clip (3.24 ± 0.53%) than frequently clipped (1.07 ± 0.20%) in Buchloe patches.There was no significant difference in C 4 mid-grass cover between clip treatments in Nassella patches, although their mean values (5.62 ± 0.83% in no clip; 5.73 ± 0.83% in frequently clipped) were greater than that found in Buchloe patches, indicating a greater advancement of C 4 mid-grasses into Nassella than Buchloe patches.4A, 5A, and 5C, respectively).However, cover in all these no clip treatments increased.
Buchloe encroachment into Nassella patches occurred in several fire treatmen quent clipping was included but was minimal in all fire treatments without c Greatest gains in Buchloe cover in Nassella patches occurred in the clipped + 3AF 4F) and clipped + 3AFC (Figure 5B) treatments.In these treatments, Buchloe c ceeded that of Nassella.
C4 mid-grass cover, averaged over the fire treatments and Phase 1 years (199 was significantly (p ≤ 0.05) greater in no clip (3.24 ± 0.53%) than frequently clippe 0.20%) in Buchloe patches.There was no significant difference in C4 mid-grass c tween clip treatments in Nassella patches, although their mean values (5.62 ± 0.83 Regarding the main effect of fire, averaged over the clip treatments and Phase 1 years, C 4 mid-grass cover was greater in the no fire treatment and 3AF than in the 3WF and 2SF fire treatments in Buchloe patches (Table 3).In Nassella patches, C 4 mid-grass cover was greater in the alternate season fire treatments (3AF and 3AFC) than in the no fire treatment, 2SF, and 2SFC.The significant two-way fire × clip interaction found with C 4 mid-grasses in Nassella patches in Phase 1 (Table 2) occurred because the C 4 mid-grass cover was either greater in the no clip treatment compared to the frequently clipped treatments in some fire treatments (Figures 4 and 5) but was greater in the frequently clipped treatment in the 2SFC fire treatment (Figure 5E,F).The fire treatments are as described in Table 1.The fire treatment code is presented in Table 1.
Regarding the main effect of fire, averaged over the clip treatments and C4 mid-grass cover was greater in the no fire treatment and 3AF than in the fire treatments in Buchloe patches (Table 3).In Nassella patches, C4 mid-gr greater in the alternate season fire treatments (3AF and 3AFC) than in th ment, 2SF, and 2SFC.The significant two-way fire × clip interaction found grasses in Nassella patches in Phase 1 (Table 2) occurred because the C4 m was either greater in the no clip treatment compared to the frequently clipp in some fire treatments (Figures 4 and 5) but was greater in the frequently ment in the 2SFC fire treatment (Figure 5E,F).1.
Regarding the main effect of year, C 4 mid-grass cover progressively increased in Buchloe patches during the Phase 1 years but only reached 4.1% by 1998 (Table 3).In contrast, C 4 mid-grass cover increased to its highest level in Nassella patches, reaching 11.6% by 1998.Thus, C 4 mid-grass advancement starting from near zero in each case was three times greater in the Nassella than Buchloe patches during Phase 1.
C 3 annual grass and forb cover each remained mostly below 10% in all treatments from 1992-1998, with a few exceptions (Figures S1-S4).In Buchloe patches, C 3 annual grass and forb cover increased to one-year peaks of 14 and 30%, respectively, in 1994 in the No Clip + 2SFC treatment (Figure S2E).This peak occurred 1 year after the first summer fire in 1993.In Nassella patches, forb cover increased to 15 and 16% in the No Clip + No Fire and No Clip + 2SF treatments, respectively, in 1996 (Figures S3A and S4C), and to 18% in the frequently clipped + No Fire treatment in 1997 (Figure S3B).
The percent bare ground temporarily exceeded litter cover from 1993-1996 in all the frequently clipped fire treatments (not counting the no fire treatment) in both the Buchloe and Nassella patches (Figures S5-S8).The percent bare ground exceeded litter cover more in the Nassella than Buchloe patches.The greatest difference occurred in 1994 in the frequently clipped + 3AFC fire treatment in Nassella patches (64% bare ground vs. 1% little cover; Figure S8B).However, bare ground decreased to 14% in this treatment by 1998.
The three-way analysis that compared the responses to treatments in the fall of 1996 vs. the spring of 1997 is in Table S4.In Buchloe patches, the effect of year was significant (p ≤ 0.05) for all variables except Buchloe and bare ground.In Nassella patches, the effect of year was significant for Nassella, C 3 annual grasses, litter, and bare ground.Nassella also had a significant clip × year interaction.Nassella cover increased from 1996 to 1997 to a much greater degree in plots that were not clipped compared to frequently clipped plots (Figures 4 and 5).

End of Phase 1 Herbaceous Cover (1998)
The two-way analysis of the effects of the clip and fire treatments in 1998 revealed that the main effect of clip treatment was significant (p ≤ 0.05) for Nassella, C 4 mid-grasses, forbs, and bare ground in Buchloe patches, and for Buchloe, Nassella, forbs, litter, and bare ground in Nassella patches (Table 4).The main effect of fire was significant for Nassella, C 3 annual grasses, and litter in Buchloe patches, and C 3 annual grasses, forbs, litter, and bare ground in Nassella patches.Regarding the main effects of clip treatment, Nassella cover in Nassella patches was greater in the no clip (40.8 ± 3.1%) than the frequently clipped (18.7 ± 2.0%).In contrast, Buchloe cover in Nassella patches was greater in frequently clipped (25.0 ± 3.8%) than no clip (5.2 ± 2.5%) (Figure 6).In Buchloe patches, there was no effect of clipping or fire on Buchloe cover (59.4 ± 2.4%; averaged across all clip and fire treatments).Nassella cover in Buchloe patches was greater in no clip (9.0 ± 2.1%) than frequently clipped (1.9 ± 0.8%) (Figure 7).
Regarding the main effects of fire treatment, Nassella encroachment into Buchloe patches was greater in two of the six no clip fire treatments (3AF and 2SFC) (Figure 7).Both of these treatments included summer fire.The two-way clip × fire interaction was nearly significant (p = 0.0574; Table 4) for Nassella cover in Buchloe patches.The higher Nassella cover of the No Clip + 3AF and No Clip + 2SFC treatments did not occur under frequent clipping.
There was little evidence of C 4 mid-grasses advancing into Buchloe patches.However, when averaged over all fire treatments, C 4 mid-grass cover was three times greater in no clip (6.3 ± 1.7%) than frequently clipped (2.0 ± 0.5%).C 4 mid-grass cover in Nassella patches was found to be similar across the clipping and fire treatments (average 4.1 ± 1.0%).
Buchloe patches was greater in no clip (9.0 ± 2.1%) than frequently clipped (1.9 ± 0.8% (Figure 7).Buchloe patches was greater in no clip (9.0 ± 2.1%) than frequently clipped (1.9 ± 0.8% (Figure 7).The C 3 annual grass cover at the end of Phase 1, averaged over clip treatments, was greater in 3WF than in the 3AF and 2SFC fire treatments in Buchloe patches (Table 5).The C 3 annual grass cover in Nassella patches was greater in no fire, 3WF, and 2SF than the 3AF and 3AFC fire treatments.Forb cover was over two times greater in the frequently clipped treatments than in the no clip treatment in both Buchloe and Nassella patches.Averaged over clip treatments, forb cover was greater in the 3WF than in the no fire treatment in Buchloe patches, with all other fire treatments intermediate.Forb cover was greater in the 2SF treatment than in the no fire, 3WF, and 3AF fire treatments in Nassella patches.Litter cover was greater in no clip than the frequently clipped treatment in Nassella patches but not in Buchloe patches (Table 5).Litter cover, averaged over clip treatments, was greater in no fire than the 3AF, 2SF, and 2SFC fire treatments in Buchloe patches, and was greater in no fire than all other fire treatments in Nassella patches.Bare ground cover was six and ten times greater in the frequently clipped than the no clip in the Buchloe and Nassella patches, respectively.Bare ground was greater in no fire than 3AF and 2SF in Buchloe patches and was greater in 3AFC than 3AF and 2SF in Nassella patches.

Phase 2-Herbaceous Cover Responses
The three-way analysis of clip × fire × year during Phase 2 (1998-2000) revealed that the main effect of clip treatment was significant for more response variables in Buchloe patches than in Nassella patches (Table 6).The main effect of fire treatment was significant for all response variables except C 4 mid-grass cover in Nassella patches.There were significant clip × fire interactions for Nassella cover in Buchloe patches, and Nassella, C 4 mid-grass, and bare ground cover in Nassella patches.The two-way analysis of the effects of the clip and fire treatments at the end of Phase 2 in 2000 revealed significant main effects of the clip and/or fire treatment for Buchloe, Nassella, C 4 mid-grass, and litter cover in Buchloe patches, and for Buchloe, Nassella, and litter cover in Nassella patches (Table 7).There was one significant clip × fire interaction with Nassella cover in Buchloe patches.Buchloe cover remained higher after the 1999 clip in patches that had been previously frequently clipped and burned compared to clipping alone, even though the last fire step was conducted in 1996 (Figure 8A).The 1999 clip treatment that occurred during a drought period caused large declines in Buchloe cover by the year 2000 in Buchloe patches that we previously had not clipped.Nassella encroachment into Buchloe patches after the 1999 clip was greatest in the one clip in the 1999 + 2SFC treatment (Figure 8B).period caused large declines in Buchloe cover by the year 2000 in Buchloe p previously had not clipped.Nassella encroachment into Buchloe patches aft was greatest in the one clip in the 1999 + 2SFC treatment (Figure 8B).  1.
The 1999 clip either did not change or increased Nassella cover in Nass the year 2000 (Figures 4 and 5).The one exception was the frequently clipp ment (Figure 5B).Buchloe cover sharply declined in Nassella patches in most that were previously frequently clipped (Figures 4F and 5B,D,F).
Forb cover increased to 15-20% in most treatments in the Buchloe and N in 2000 after the 1999 clip (Figures S1-S4).In most of these treatments, this w level of forb cover during the entire study.Bare ground cover increased by points in most treatments in Buchloe patches from 1998 to 2000 (Figures creases in bare ground from 1998-2000 were lower in Nassella patches (Figu

Discussion
Long-term responses of monoculture patches of Buchloe and Nassella clipping treatments in this study must be interpreted in the context that al occurred in intercanopy spaces between mature Prosopis trees that were to  1.
The 1999 clip either did not change or increased Nassella cover in Nassella patches by the year 2000 (Figures 4 and 5).The one exception was the frequently clipped 3AFC treatment (Figure 5B).Buchloe cover sharply declined in Nassella patches in most fire treatments that were previously frequently clipped (Figures 4F and 5B,D,F).
Forb cover increased to 15-20% in most treatments in the Buchloe and Nassella patches in 2000 after the 1999 clip (Figures S1-S4).In most of these treatments, this was the highest level of forb cover during the entire study.Bare ground cover increased by >10 percentage points in most treatments in Buchloe patches from 1998 to 2000 (Figures S5 and S6).Increases in bare ground from 1998-2000 were lower in Nassella patches (Figures S7 and S8).

Discussion
Long-term responses of monoculture patches of Buchloe and Nassella to the fire and clipping treatments in this study must be interpreted in the context that all grass patches occurred in intercanopy spaces between mature Prosopis trees that were top-killed via the fire treatments and subsequently resprouted.Prosopis root-kill was <3%, so Prosopis plant density remained unchanged.These effects were common to all grass patches, except for the no fire treatment where Prosopis remained as mature trees.

Responses in Buchloe Patches
The occurrence of Buchloe in the SGP is not well understood, nor is the interaction of Buchloe and C 4 mid-grasses under encroachment via Prosopis or other woody species.The relative percentages of cover of Buchloe and C 4 mid-grasses in the SGP prior to the European settlement are unknown; what is known is that the region was once dominated by C 4 vegetation that we assumed was grasses [35].The average rainfall in the SGP of 500-700 mm should support C 4 mid-grasses instead of Buchloe on deeper soils.However, extended droughts can occur, and these events temporarily increase the percentage of land area covered by C 4 shortgrasses [3].Turfgrass studies have found Buchloe to be very drought tolerant [58][59][60].Native grassland and rangeland studies (and unquantified observations) describe Buchloe as tolerant of heavy grazing and drought [30,39,40,54,61,62].An earlier study in the north Texas portion of the SGP found that Buchloe was common in heavily grazed pastures, while C 4 mid-grasses dominated ungrazed exclosures and moderately grazed pastures [4].Whether Buchloe displaced C 4 mid-grasses during the process of Prosopis encroachment is not known, but likely occurred [6,63].
Buchloe is one of the few C 4 grasses that persists within intercanopy spaces in dense Prosopis stands, but rarely occurs directly beneath Prosopis canopies due to insufficient light.Increases in Prosopis cover only slightly reduced Buchloe production in intercanopy spaces [15].As a short-statured sod grass, Buchloe root systems avoid competition with Prosopis lateral roots for soil moisture as they occupy shallower soil spaces above Prosopis lateral roots [64].This may be an example of Walter's two-layer hypothesis [65,66].Prosopis lateral roots into intercanopy spaces tend to track at between 0.5 and 1 m soil depth [64,67].Thus, they likely compete for soil moisture with grass species with deeper roots, such as C 4 mid-grasses, but not with shallow-rooted sod grasses [47].
Some studies have found that Buchloe is very resistant to single winter or spring fires [41,61].Our study demonstrates that Buchloe is tolerant of repeated summer and alternate season fires, although some treatments caused a steep but temporary decline in Buchloe cover.Another study indicated that prescribed fires conducted during late winter and early spring favor C 4 over C 3 grass species, as C 4 species are dormant during the fire [68].Other studies found that burning during the summer or fall may harm certain C 4 grasses, such as Bouteloua curtipendula [41,69].A literature review demonstrated the difficulty in predicting the responses of C 3 and C 4 grasses to seasonal fires [42].The amount of time post-fire is an important criterion as some C 4 grasses, such as Bouteloua curtipendula or B. gracilis, may require more than one growing season to fully recover from a summer fire [61,70,71].
Our results indicate that Buchloe thrives under frequent defoliation from clipping or fire.A summer fire may temporarily set it back, but it quickly recovers and seems tolerant of fire in any season.Buchloe patches during the average precipitation years were not disturbed by frequent defoliation.This suggests that Buchloe may have gained a competitive edge over C 4 mid-grasses when early settlers introduced livestock grazing in the late 1800s.
Our results suggest that Buchloe requires a light saturated environment and frequent defoliation to maintain patch dominance, since Buchloe cover declined to the greatest extent in the No Clip + No Fire treatment.However, we must view this in the temporal and spatial contexts of when and where Buchloe becomes dominant in the successional process during WPE [7].Buchloe and similar C 4 short-grass species (Appendix A) are the last of the C 4 grasses that persist under advanced WPE [14].What our study demonstrated is that from this temporal point in the successional process, with most C 4 tall grasses and midgrasses gone and a woody cover of > 50%, Buchloe is dominant in spaces between Prosopis.However, as our study revealed, Buchloe cover will begin to decline without defoliation disturbance and/or some level of Prosopis canopy reduction.If either of these disturbances occur, Buchloe patches will remain stable.Most importantly, Buchloe patches remained stable in cases where clipping was the only disturbance and the Prosopis cover remained intact.By 1998, there were no treatments with frequent clipping + fire or fire alone (and all these treatments top-killed most Prosopis) that yielded a significantly higher Buchloe cover than the frequently clipped + No Fire treatment.After severe drought, however, clipping alone or clipping plus the previous fire treatment was not able to maintain Buchloe patch integrity and percent bare ground increased.This may have been due to the combination of drought and competition from Prosopis, either as large trees in the no fire treatment or as regrowth in the fire treatments [61].
No clipping + 3WF, 3AF, and 2SFC fire treatments favored Nassella encroachment into Buchloe patches.Common to these three fire treatments was that they initially severely reduced Buchloe cover the first year post-fire.This likely provided the opening for Nassella cover to increase.Nassella encroachment into Buchloe patches was not as great in the other fire treatments (2SF and 3AFC) as these treatments did not reduce Buchloe cover as much as Correctdid the other three fire treatments.This response was independent of the fire effects on Prosopis as all five fire treatments top-killed most Prosopis.Thus, a combination of no clipping with certain extreme fire treatments that temporarily severely reduce Buchloe cover may provide an opening for Nassella encroachment into Buchloe patches.Most resource managers would view this trend as undesirable [18,36].The clipping regime implemented in this study (single yearly clipping event aside from two clips in the first season) limited Nassella encroachment into Buchloe patches across all fire treatments.
Regarding the responses in Phase 2 of this study that included the universal clip in the spring of 1999 following a severe drought, there appeared to be a residual effect of clipping + some previous fire treatments or clipping alone on Buchloe tolerance to the 1999 clipping.In these treatments, Buchloe maintained a higher cover and limited Nassella encroachment into Buchloe patches by the next year.Moreover, the fire treatments in combination with previous clippings that generated the greatest Buchloe cover in 2000 included extreme combinations of winter and summer fires or repeated summer fires (3AFC, 2SF, and 2SFC; Figure 8A).However, it was clear that fire alone without the previous clipping treatments did not have the same effect.Reasons for these responses are not known.Previous defoliations from clipping or fire may have conditioned Buchloe for greater tolerance to the 1999 clipping.In addition, the more extreme fire treatments (3AFC, 2SF, and 2SFC) may have added to Buchloe tolerance to the 1999 clip by reducing competition from other species or generating a more favorable soil nutrient balance.We noted that effects of fires that occurred from 4 years to 6 years earlier coupled with spring clipping that had ended 2 years earlier (in 1997) had a residual effect on Buchloe tolerance to the 1999 clipping.This points to the need for assessing fire effects over longer periods than one-or two-years post-fire.
While previous fire treatments plus frequent clipping or frequent clipping alone prevented Nassella encroachment into Buchloe patches after the Phase 2 1999 clip, Nassella encroachment did increase in Buchloe patches that were exposed to fire treatments but had not been clipped until 1999.For example, by the study's end in 2000, Nassella cover in Buchloe patches had increased to 30% in the 1999 Clip + 2SFC treatment and exceeded Buchloe cover in Buchloe patches that initially had >80% Buchloe cover (Figure 3E).This response may have been due to a legacy effect of the two consecutive summer fires in 1993 and 1994 that severely, but temporarily, reduced Buchloe cover and initiated a gradual, yet consistent, trajectory of Nassella encroachment in this treatment.These results reveal that a robust patch of a drought and grazing-tolerant perennial species, such as Buchloe, that is highly adapted to the region can succumb to another perennial species, perhaps facilitated via a regrowing woody overstory, after only a few severe defoliation events (in this case, two summer fires and one clipping in 8 years).
The advancement of C 4 mid-grasses into Buchloe patches was minimal except, similar to Nassella encroachment, where the fire treatment temporarily severely reduced Buchloe cover.However, this occurred only with the No Clip + 3AF (Figure 2E) and No Clip + 2SFC (Figure 3E) fire treatments and not the No Clip + 3WF treatment (Figure 2C) that facilitated Nassella encroachment.This may have occurred because the Buchloe cover was not as severely reduced by the 3WF treatment (from 73 to 50%) compared to the 3AF (from 78 to 39%) and 2SFC (from 79 to 30%) treatments.In both treatments (No Clip + 3AF and No Clip + 2SFC), the level of Nassella encroachment remained greater than that of the C 4 mid-grasses.This may be due to Prosopis post-fire regrowth that favored Nassella over the C 4 mid-grasses, either via lower root competition for water (Nassella avoids competition by mostly growing before Prosopis leaf emergence in the spring) [12,21], or N fixation by Prosopis that favors the C 3 grasses [24].The only exception to this trend was in the No Clip + No Fire treatment, where C 4 mid-grass encroachment into Buchloe patches slightly exceeded that of Nassella (Figure 2A).We have no explanation for this, as we would have predicted this treatment to be more favorable to Nassella than to the C 4 mid-grasses.

Responses in Nassella Patches
Nassella, a C 3 plant adapted to the hot summer conditions in the SGP, thrives in dense Prosopis stands, and especially beneath Prosopis canopies due to effects of Prosopis on the microenvironment (N fixation in soils; slight cooling of air temperature and reduction in sun intensity) [17].As Prosopis density increases, these beneficial effects to Nassella extend into gaps between Prosopis canopies, probably because of partial shading.Top-killing Prosopis temporarily removes the beneficial effects of the Prosopis canopy for Nassella [5], but resprouting Prosopis plants can reestablish this effect from within 5 to 10 years in this region [21,43].
While many taxonomy books and the USA Federal plants database refer to Nassella as native to the SGP [19,40,54,55], this is only assumed based on what early naturalists in the region observed.No studies to date have identified the paleo-ecological origin of Nassella.There are many Hesperostipa (formerly Stipa) species in the northern Great Plains, one of the most extensive being H. comata.Nassella (formerly Stipa leucotricha) may be a variant of a southward migration of H. comata, as their geographical ranges overlap in northwest Texas [54].However, it is possible that Nassella was not present in the SGP until Prosopis migrated from South and Central America into the SGP, as this species has been included in the South American genus Nassella [55].Phylogenetic cladograms show a very distant relationship between the Hesperostipa and Nassella genera [72] and reveal a North American origin of Hesperostipa and a South American origin of the Nassella genera [73].
In central Argentina, 4-6 m tall Prosopis flexuosa facilitated the growth of the unpalatable C 3 grass Stipa ichu beneath the P. flexuosa canopy, while C 4 grasses were more frequent in interspaces [26].In semiarid Prosopis caldenia forests of central Argentina, native unpalatable C 3 grasses, S. ichu and Nassella tenuissima, expanded in C 3 /C 4 mixed grassland that had mostly been previously C 4 grasses [29].The analysis conducted by the authors of [29] revealed a strong association between P. caldenia and these two grass species.P. glandulosa is most closely related genetically to P. nigra, P. flexulosa, and P. caldenia, all from Argentina [74], and this supports the theory that Prosopis ancestors originated in South America and migrated north.The association of S. ichu and N. tenuissima with P. flexulosa and P. caldenia [26,29], combined with the theory that Prosopis ancestors migrated from South to North America, suggests that the ancestral variant of Nassella came from South America in paired association with Prosopis.
Nassella was tolerant of winter or summer fires in our study.This contrasts with the studies published by the authors of [41] and [68], who indicated that Nassella was severely harmed by intense winter or early spring fires.However, two studies in the Texas portion of the SGP found that the Nassella standing crop recovered to levels in unburned plots within one or two growing seasons after a summer or winter fire [20,75].
The annual spring clipping regime imposed as part of Phase 1 in the current study reduced Nassella cover in all fire treatments.Nassella cover in Nassella-dominant patches was best maintained under no clipping.Another study in the SGP found that repeated defoliation in spring suppressed, but did not eliminate, Nassella [18].However, the Phase 2 portion of our study, where all patches previously clipped or not clipped were clipped in 1999, generated a sharp increase in Nassella cover from 1999 to 2000 in most of the fire and clipping treatments in Nassella patches (Figures 4 and 5), and as previously mentioned, in some of the Buchloe patches (Figures 2C and 3E).We hypothesize that the size of postfire regrowth Prosopis facilitated Nassella growth during Phase 2. These results reveal the difficulty in permanently reducing Nassella cover if Prosopis are not root-killed and resprouting after top-kill increases to large-sized trees.
One unexpected observation in Nassella patches was the large decline in Nassella cover in some no clip treatments, and especially the No Clip + No Fire treatment, from 1994 to 1996 (Figures 4 and 5).High rainfall in 1995 caused rapid Nassella growth in Nassella patches [75].Thus, the decline in cover occurred between the 1995 and 1996 growing seasons.Reasons for this decline are unknown, but they may relate to excessive litter accumulation from the 1995 growth that limited growth in 1996 [76][77][78].In addition, C 3 growing season precipitation in 1996 was only 35% of the average.Thus, spring drought combined with shading from excessive litter from the previous year's growth may have killed new tillers attempting to emerge through the litter.
Buchloe encroachment into Nassella patches was driven by the clipping treatment and was independent of fire treatment, except for the clipped + alternate season (3AF) fire treatment.This treatment yielded the greatest encroachment of Buchloe into Nassella patches, increasing to 44 and 43% cover in 1997 and 1998, respectively, after initially being <2% cover (Figure 4F).Nassella cover declined in this treatment from 45% initially to 18% in 1998.In addition to the frequent clipping treatment that was more advantageous to Buchloe than Nassella across all fire treatments, we hypothesize that a combination of factors related to the 3AF alternate season fire treatment may have facilitated this conversion from Nassella to Buchloe dominance.First, this fire treatment contained one summer fire that was more effective than winter fires at reducing Prosopis cover, and this increased available sunlight for Buchloe growth.Second, with only one summer fire that occurred when Buchloe was physiologically active instead of two summer fires, as occurred in the 2SF and 2SFC fire treatments, the 3AF fire treatment likely had a less negative effect on Buchloe.Third, the second late winter fire in this treatment in 1994 may have inhibited Nassella growth that spring without negatively affecting Buchloe, which was dormant at that time.This combination of effects gave Buchloe the greatest advantage to overtake portions of the Nassella patches.
While frequent clipping in spring had a greater effect on reducing Nassella cover than did any combination of the summer and winter fire treatments, and stimulated Buchloe encroachment into Nassella patches in the 3AF fire treatment, it did not increase the ability of C 4 mid-grasses to encroach into Nassella patches compared to no clipping.This suggests that C 4 mid-grasses were more sensitive to clipping effects than was Buchloe.However, the alternate season fire treatments (3AF and 3AFC) increased C 4 mid-grass advancement into Nassella patches (Table 3).These results suggest that in Prosopis-dominated grassland that has degraded to mostly a Nassella understory, the most optimal grazing and fire management strategy for increasing C 4 mid-grasses may be to incorporate alternate season fires with livestock grazing that does not impose heavy grazing pressure on C 4 mid-grasses in the spring following a summer fire.Full recovery of some C 4 mid-grass species after a summer fire may take two growing seasons [70,71].
Earlier research conducted on the same site as this study (and before any fire treatments occurred) concluded that this region of the SGP would always be dominated by Buchloe, regardless of the presence of Prosopis [5].We have subsequently found that alternate season fire treatments in the absence of grazing significantly increased the cover of C 4 mid-grasses [31,43].This included C 4 mid-grass species, such as Bouteloua curtipendula and Panicum obtusum, that are valued as forage for cattle.However, it also included other C 4 mid-grass species, such as Aristida purpurea, Bothriochloa laguroides, Hilaria mutica, and Sporobolus compositus, that are not optimal for cattle grazing but are valuable for other grassland ecosystem services, such as wildlife cover, grassland obligate bird species, soil stability, wind erosion mitigation, and/or carbon sequestration [36][37][38]40,54].Since gains in C 4 mid-grass cover under certain clipping and fire treatments were greater in Nassella than in Buchloe patches, we hypothesize that the potential for the landscape-scale transition towards C 4 mid-grasses under the right combination of fire and defoliation from grazing may depend, in part, on the percentage of the total area dominated by Nassella.Reversing this logic, our results also suggest that during the decades-long process of Prosopis WPE into the region, the increase in Nassella, as facilitated via Prosopis encroachment [21], came more at the expense of C 4 mid-grasses and not C 4 shortgrasses, like Buchloe.
One fire treatment, the 2SFC, that included two intense summer fires in consecutive years appeared to stimulate Nassella cover in Buchloe patches in the absence of clipping (Figures 3E and 8B).From 1992 to 2000, Nassella cover increased from 0 to 30% in this treatment.This result emphasizes the need for including grazing after planned or unplanned summer fires.These fires will likely top-kill Prosopis, but the consequences without post-fire grazing could be to increase rather than decrease Nassella cover.Frequent spring clipping in this fire treatment limited Nassella cover to <3% in Buchloe patches.
Forb cover increased with frequent clipping in both Buchloe and Nassella patches, with repeated winter fires in Buchloe patches, and with repeated summer fires in Nassella patches.While the range of forb basal cover shown in Table 5 appears to be small (0.42 to 6.38%), many forb species in the region, such as Amphiachyris dracunculoides, have small diameter stems that support large canopies [36,55].A three-to fifteen-fold increase in forb basal cover in some fire treatments likely increased floral diversity in the Buchloe or Nassella patches.

Conclusions
This study demonstrated that Nassella cover can be reduced, but not eliminated, by frequent spring clipping (as simulated grazing) combined with repeated winter, repeated summer, or alternate season fire treatments.We also found that Buchloe can encroach and become the dominant grass in what were initially Nassella-dominated patches under certain combinations of frequent clipping and seasonal fires.Specifically, Buchloe encroachment into Nassella patches was greatest in the frequently clipped + alternate season fire treatment (Figure 4F).We found little evidence of Nassella advancement into Buchloe patches under frequent clipping and fire.However, in the absence of clipping, some fire treatments, coupled with Prosopis post-fire regrowth, stimulated Nassella encroachment into Buchloedominant patches (see Figure 3E).
Gains in the C 4 mid-grass cover under certain clipping and fire treatments were greater in Nassella than in the Buchloe patches.We note, however, that advancement of the C 4 mid-grasses into Nassella patches was not rapid nor extensive, and only occurred under a complex set of clipping and fire treatments.Thus, the potential of clipping (or grazing) and prescribed fire for rapidly increasing C 4 mid-grass abundance (and associated secondary increases in diversity, e.g., enhanced wildlife habitat) may be limited under an advanced stage of WPE via a fire-resistant woody species such as Prosopis.
Forb basal cover increased with frequent clipping and repeated winter fires in Buchloe patches, and repeated summer fires in Nassella patches.This likely increased ecosystem floral diversity.However, the repeated summer fire treatments that best increased forb cover in Nassella patches differed from one alternate season fire treatment (3AF) that yielded the greatest increase in C 4 mid-grass cover (and lowest increase in forb cover).Our study illustrates that different manipulations of clipping (as a proxy for grazing) and fire can profoundly disrupt the stability of monoculture patches of two grass species once the Prosopis woody overstory that reinforces such stability is suppressed.

Supplementary Materials:
The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/d15101069/s1.Table S1: List of grass species alphabetically by scientific name found at the site and their classification into functional groups.All nomenclature by Stubbendieck et al. (2017) [53].Table S2: Fire treatments in the study, and fire steps within each treatment.All fires were conducted in the 1990's (e.g., w91 = winter 1991).Table S3: Pre-burn weather and herbaceous fine fuel, and peak fire temperature and fire intensity of each fire step in each fire treatment.Except for the two bottom rows, all values are means (±s.e.) of 3 plots (36 total prescribed fires and 1 wildfire).The last two rows are means of the replicates of all winter or summer fire steps (±s.e.).The s93 fire step in the 2FX treatment was due to an August 24, 1993 wildfire; fire temperature and intensity were not recorded.Table S4: Mesquite root-kill, top-kill and stand foliage reduction in response to each step of each fire treatment.All values are means ± standard error (n = 3).Means with similar letters within each column are not significantly different (p ≤ 0.05).Table S5: Proc Mixed 3-way analysis for initially buffalograss (Buda) and Texas wintergrass (Nale) patches from fall 1996 to spring 1997.Shaded values are p ≤ 0.05.C4M = C 4 mid-grasses, C3A = C 3 annual grasses.S2.Vertical bars are standard error.Figure S7.C 3 annual grass (C3A) and forb cover in patches initially dominated by buffalograss in response to fire treatments No Fire, 3WF and 3AF in winter ('f') or summer ('F') and clipping (c).Right panels show the frequent clipping treatments of Phase 1 (1992)(1993)(1994)(1995)(1996)(1997)(1998).Right and left panels show the Phase 2 clipping treatment in 1999.Treatment code is in Table S2. Figure S8.C 3 annual grass (C3A) and forb cover in patches initially dominated by buffalograss in response to fire treatments 3AFC, 2SF and 2SFC in winter ('f') or summer ('F') and clipping (c).Right panels show the frequent clipping treatments of Phase 1 (1992)(1993)(1994)(1995)(1996)(1997)(1998).Right and left panels show the Phase 2 clipping treatment in 1999.Treatment code is in Table S2. Figure S9.C 3 annual grass (C3A) and forb cover in patches initially dominated by Texas wintergrass in response to fire treatments No Fire, 3WF and 3AF in winter ('f') or summer ('F') and clipping (c).Right panels show the frequent clipping treatments of Phase 1 (1992)(1993)(1994)(1995)(1996)(1997)(1998).Right and left panels show the Phase 2 clipping treatment in 1999.Treatment code is in Table S2. Figure S10.C 3 annual grass (C3A) and forb cover in patches initially dominated by Texas wintergrass in response to fire treatments 3AFC, 2SF and 2SFC in winter ('f') or summer ('F') and clipping (c).Right panels show the frequent clipping treatments of Phase 1 (1992)(1993)(1994)(1995)(1996)(1997)(1998).Right and left panels show the Phase 2 clipping treatment in 1999.Treatment code is in Table S2. Figure S11.Litter and bare ground cover in patches initially dominated by buffalograss in response to fire treatments No Fire, 3WF and 3AF in winter ('f') or summer ('F') and clipping (c).Right panels show the frequent clipping treatments of Phase 1 (1992Phase 1 ( -1998)).Right and left panels show the Phase 2 clipping treatment in 1999.Treatment code is in Table S2. Figure S12.Litter and bare ground cover in patches initially dominated by buffalograss in response to fire treatments 3AFC, 2SF and 2SFC in winter ('f') or summer ('F') and clipping (c).Right panels show the frequent clipping treatments of Phase 1 (1992)(1993)(1994)(1995)(1996)(1997)(1998).Right and left panels show the Phase 2 clipping treatment in 1999.Treatment code is in Table S2. Figure S13.Litter and bare ground cover in patches initially dominated by Texas wintergrass in response to fire treatments No Fire, 3WF and 3AF in winter ('f') or summer ('F') and clipping (c).Right panels show the frequent clipping treatments of Phase 1 (1992)(1993)(1994)(1995)(1996)(1997)(1998).Right and left panels show the Phase 2 clipping treatment in 1999.Treatment code is in Table S2. Figure S14.Litter and bare ground cover in patches initially dominated by Texas wintergrass in response to fire treatments 3AFC, 2SF and 2SFC in winter ('f') or summer ('F') and clipping (c).Right panels show the frequent clipping treatments of Phase 1 (1992Phase 1 ( -1998)).Right and left panels show the Phase 2 clipping treatment in 1999.Treatment code is in Table S2.
Author Contributions: Conceptualization, methodology, and formal analysis R.J.A. and W.E.P.; investigation, R.J.A.; resources, R.J.A. and W.E.P.; data curation, R.J.A. and W.E.P.; writing-original draft preparation, R.J.A.; writing-review and editing, R.J.A. and W.E.P.; supervision, R.J.A.; project administration, R.J.A.; funding acquisition, R.J.A.All authors have read and agreed to the published version of the manuscript.

Figure 2 .
Figure 2. Grass cover in patches initially dominated by Buchloe in response to the fire treatm fire, 3WF, and 3AF in winter ('f') or summer ('F') and clipping (c).Left panels (A,C,E) sh sponses to not clipping from 1992-1998 and a single late season clipping 1999-2000.Right (B,D,F) show the frequent clipping treatments of Phase 1 (1992-1998).The vertical dotted lin rates Phase 1 and Phase 2. Phase 2 includes the 1999 clip on all treatments.C4M = C4 mid-The fire treatment code is presented in Table1.

Figure 2 .
Figure 2. Grass cover in patches initially dominated by Buchloe in response to the fire treatments no fire, 3WF, and 3AF in winter ('f') or summer ('F') and clipping (c).Left panels (A,C,E) show responses to not clipping from 1992-1998 and a single late season clipping 1999-2000.Right panels (B,D,F) show the frequent clipping treatments of Phase 1 (1992-1998).The vertical dotted line separates Phase 1 and Phase 2. Phase 2 includes the 1999 clip on all treatments.C4M = C 4 mid-grasses.The fire treatment code is presented in Table1.

Diversity 2023 ,Figure 3 .
Figure 3. Grass cover in patches initially dominated by Buchloe in response to the fire tr 3AFC, 2SF, and 2SFC in winter ('f') or summer ('F') and clipping (c).Left panels (A,C,E) sponses to not clipping from 1992-1998 and a single late season clipping 1999-2000.Rig (B,D,F) show the frequent clipping treatments of Phase 1 (1992-1998).The vertical dotted rates Phase 1 and Phase 2. Phase 2 includes the 1999 clip on all treatments.C4M = C4 mid-

Figure 3 .
Figure 3. Grass cover in patches initially dominated by Buchloe in response to the fire treatments 3AFC, 2SF, and 2SFC in winter ('f') or summer ('F') and clipping (c).Left panels (A,C,E) show responses to not clipping from 1992-1998 and a single late season clipping 1999-2000.Right panels (B,D,F) show the frequent clipping treatments of Phase 1 (1992-1998).The vertical dotted line separates Phase 1 and Phase 2. Phase 2 includes the 1999 clip on all treatments.C4M = C 4 mid-grasses.

Figure 4 .
Figure 4. Grass cover in patches initially dominated by Nassella in response to the fire fire, 3WF, and 3AF in winter ('f') or summer ('F') and clipping (c).Left panels (A sponses to not clipping from 1992-1998 and a single late season clipping 1999-2000 (B,D,F) show the frequent clipping treatments of Phase 1 (1992-1998).The vertical do rates Phase 1 and Phase 2. Phase 2 includes the 1999 clip on all treatments.C4M = CThe fire treatments are as described in Table1.

Figure 4 .
Figure 4. Grass cover in patches initially dominated by Nassella in response to the fire treatments no fire, 3WF, and 3AF in winter ('f') or summer ('F') and clipping (c).Left panels (A,C,E) show responses to not clipping from 1992-1998 and a single late season clipping 1999-2000.Right panels (B,D,F) show the frequent clipping treatments of Phase 1 (1992-1998).The vertical dotted line separates Phase 1 and Phase 2. Phase 2 includes the 1999 clip on all treatments.C4M = C 4 mid-grasses.The fire treatments are as described in Table 1.Table 3. C 4 mid-grass cover in Buchloe and Nassella patches in response to the main effect of the fire treatment (3 left columns) and year (3 right columns) during Phase 1 (1992-1998).Means with similar letters in each column are not significantly different at p ≤ 0.05.

Figure 5 .
Figure 5. Grass cover in patches initially dominated by Nassella in response to the 3AFC, 2SF, and 2SFC in winter ('f') or summer ('F') and clipping (c).Left panels ( sponses to not clipping from 1992-1998 and a single late season clipping 1999-20 (B,D,F) show the frequent clipping treatments of Phase 1 (1992-1998).The vertical d rates Phase 1 and Phase 2. Phase 2 includes the 1999 clip on all treatments.C4M = The fire treatment code is presented in Table1.

Figure 5 .
Figure 5. Grass cover in patches initially dominated by Nassella in response to the fire treatments 3AFC, 2SF, and 2SFC in winter ('f') or summer ('F') and clipping (c).Left panels (A,C,E) show responses to not clipping from 1992-1998 and a single late season clipping 1999-2000.Right panels (B,D,F) show the frequent clipping treatments of Phase 1 (1992-1998).The vertical dotted line separates Phase 1 and Phase 2. Phase 2 includes the 1999 clip on all treatments.C4M = C 4 midgrasses.The fire treatment code is presented in Table1.

Figure 6 .
Figure 6.Percent cover of Nassella in Nassella patches (A) and Buchloe in Nassella patches (B) in th fire and clipping treatments at end of Phase 1 in 1998.Means with similar letters within each pan (A or B) are not significantly different at p ≤ 0.05.Horizontal lines are means across all fire trea ments; an asterisk indicates a significant difference (p ≤ 0.05) within each panel.The fire treatmen code is outlined in Table1.

Figure 7 .
Figure 7. Percent cover of Nassella in Buchloe patches in the clipping and fire treatments at the en of Phase 1, 1998.Means with similar letters are not significantly different at p ≤ 0.05.Horizontal line are means across all fire treatments; an asterisk indicates a significant difference (p ≤ 0.05) withi each panel.The fire treatment code is outlined in Table1.

Figure 6 .
Figure 6.Percent cover of Nassella in Nassella patches (A) and Buchloe in Nassella patches (B) in the fire and clipping treatments at end of Phase 1 in 1998.Means with similar letters within each panel (A or B) are not significantly different at p ≤ 0.05.Horizontal lines are means across all fire treatments; an asterisk indicates a significant difference (p ≤ 0.05) within each panel.The fire treatment code is outlined in Table1.

Figure 6 .
Figure 6.Percent cover of Nassella in Nassella patches (A) and Buchloe in Nassella patches (B) in th fire and clipping treatments at end of Phase 1 in 1998.Means with similar letters within each pane (A or B) are not significantly different at p ≤ 0.05.Horizontal lines are means across all fire treat ments; an asterisk indicates a significant difference (p ≤ 0.05) within each panel.The fire treatmen code is outlined in Table1.

Figure 7 .
Figure 7. Percent cover of Nassella in Buchloe patches in the clipping and fire treatments at the en of Phase 1, 1998.Means with similar letters are not significantly different at p ≤ 0.05.Horizontal line are means across all fire treatments; an asterisk indicates a significant difference (p ≤ 0.05) withi each panel.The fire treatment code is outlined in Table1.

Figure 7 .
Figure 7. Percent cover of Nassella in Buchloe patches in the clipping and fire treatments at the end of Phase 1, 1998.Means with similar letters are not significantly different at p ≤ 0.05.Horizontal lines are means across all fire treatments; an asterisk indicates a significant difference (p ≤ 0.05) within each panel.The fire treatment code is outlined in Table1.
Figure 7. Percent cover of Nassella in Buchloe patches in the clipping and fire treatments at the end of Phase 1, 1998.Means with similar letters are not significantly different at p ≤ 0.05.Horizontal lines are means across all fire treatments; an asterisk indicates a significant difference (p ≤ 0.05) within each panel.The fire treatment code is outlined in Table1.

Figure 8 .
Figure 8. Percent cover of Buchloe in Buchloe patches (A) and Nassella in Buchloe p fire and clipping treatments at end of Phase 2 in 2000.Means with similar letters w (A or B) are not significantly different at p ≤ 0.05.Horizontal lines are means acr ments; an asterisk indicates a significant difference (p ≤ 0.05) within each panel.T code is outlined in Table1.

Figure 8 .
Figure 8. Percent cover of Buchloe in Buchloe patches (A) and Nassella in Buchloe patches (B) in the fire and clipping treatments at end of Phase 2 in 2000.Means with similar letters within each panel (A or B) are not significantly different at p ≤ 0.05.Horizontal lines are means across all fire treatments; an asterisk indicates a significant difference (p ≤ 0.05) within each panel.The fire treatment code is outlined in Table1.
Figure S1.Proc Mixed 3-way analysis for initially buffalograss (Buda) and Texas wintergrass (Nale) patches from fall 1996 to spring 1997.Shaded values are p ≤ 0.05.C4M = C 4 mid-grasses, C3A = C 3 annual grasses.Figure S2.Percent cover of Texas wintergrass in Texas wintergrass patches among fire and clipping treatments, 1996-2000.Treatment code is inTable S2.Vertical bars are standard error.Figure S3.Percent cover of Texas wintergrass in buffalograss patches among fire and clipping treatments, 1996-2000.Treatment code is in Table S2.Vertical bars are standard error.Figure S4.Percent cover of buffalograss in Texas wintergrass patches among fire and clipping treatments, 1996-2000.Treatment code is in Table S2.Vertical bars are standard error.Figure S5.Percent cover of C 4 mid-grasses in buffalograss patches among fire and clipping treatments, 1996-2000.Treatment code is in Table S2.Vertical bars are standard error.Figure S6.Percent cover of C 4 mid-grasses in Texas wintergrass patches among fire and clipping treatments, 1996-2000.Treatment code is in Table

Table 1 .
Fire treatments in the study, and fire steps within each treatment.All fires were conducted in the 1990s (e.g., w91 = winter 1991).

Table 2 .
PROCBuchloe cover in Buchloe patches initially ranged from 58 to 83% (Figures 2 a During Phase 1 (from 1992 to 1998), other than a temporary decline immediatel some fire treatments, Buchloe cover gradually declined to the greatest degree in fir ments that were not clipped.Nassella encroachment into Buchloe patches from 1992 was greatest in the No Clip + 3AF and No Clip + 2SFC treatments (Figures 2E a respectively MIXED 3-way analysis for the initial Buchloe and Nassella patches in Phase 1, 1992-1998.Shaded values are p ≤ 0.05.C4M = C 4 mid-grasses, C3A = C 3 annual grasses, and df = degrees of freedom.

Table 3 .
C4 mid-grass cover in Buchloe and Nassella patches in response to the main treatment (3 left columns) and year (3 right columns) during Phase 1 (1992-1998).M ilar letters in each column are not significantly different at p ≤ 0.05.

Table 5 .
Main effects of clip treatment (averaged over fire) and fire treatment (averaged over clip) on the cover of C 3 annual grasses (C3A), forbs, litter, and bare ground in Buchloe and Nassella patches in 1998.Means ± standard error with similar letters within each patch and treatment group (clip or fire) are not significantly different at p ≤ 0.05.

Table 7 .
PROC MIXED 2-way analysis for the initial Buchloe and Nassella patches at the end of Phase 2 in 2000.Shaded values are p ≤ 0.05.C4M = C 4 mid-grasses, C3A = C 3 annual grasses, and df = degrees of freedom.