Mesquites Limit Targeted Grazing Effects on Texas Wintergrass Growth and Reproduction Responses

ABSTRACT In mixed C3/C4 grasslands of the southern Great Plains of North America, honey mesquite (Prosopis glandulosa Torr.) invasion facilitates the dominance of the C3 midgrass Texas wintergrass (Nassella leucotricha [Trin. & Rupr.] Pohl) to the detriment of warm-season grasses. Little is known about the viability of targeted grazing as a treatment method for reducing Texas wintergrass biomass and reproduction, particularly under mesquite canopies. We quantified the effects of targeted grazing treatments (grazed once [G1], grazed twice [G2], and control) and mesquite canopy position (under canopy [UC] and outside of canopy [OC]) on Texas wintergrass basal area, total dry matter (DMY) per basal area, live DMY per basal area, dead DMY per basal area, leaf length, total number of reproductive culms per plant, number of fungus-infected reproductive culms per plant, and rate of reproductive culm development over 2 yr at two study sites. Targeted grazing reduced leaf length and biomass and slowed reproductive development of the wintergrass plants that cattle could more easily access. Environmental factors and differences in mesquite physiognomy at the two study sites likely interacted with grazing treatments, making it challenging to separate the impacts of targeted grazing treatments from other factors. Results suggest mesquite removal is likely necessary to increase cattle access to Texas wintergrass, particularly in stands with multistemmed trees.


Introduction
Many studies have focused on the displacement of C 4 grasslands by C 3 woody plants that gain a competitive advantage as a result of overgrazing by domestic livestock, lower fire frequency, and increasing atmospheric CO 2 levels ( Midgley and Bond 2015 ;Wilcox et al. 2018 ;Wieczorkowski and Lehmann 2022 ). In the southern Great Plains and the southwestern United States, the best example of this is honey mesquite ( Prosopis glandulosa Torr.), a native tree species that has encroached to the point of ecologi-cal dominance in many grassland and savanna areas of the region ( Ellsworth et al. 2018 ;Ansley et al. 2001 ). Mesquite and other encroaching woody species have multiple effects on habitat structure and ecosystem dynamics, including hydrology, nutrient cycling, and microclimate ( Simmons et al. 2008 ;Eldridge et al. 2011 ).
Secondary to woody plant encroachment, and presented less frequently in the literature, is the response of C 3 grasses to biologically fixed atmospheric nitrogen (N 2 ) via woody plant invasion and woody plants' role in altering species composition in mixed C 3 /C 4 grasslands. C 3 grasses typically become dominant first under woody canopies due to microclimatic and soil nutrient modifications and then increase in spaces between trees as woody plant density increases ( Fernandez et al. 2014 ;Ansley et al. 2019 ). In mixed C 3 /C 4 grasslands of the southern Great Plains, mesquite invasion facilitates the dominance of the C 3 midgrass Texas wintergrass ( Nassella leucotricha [Trin. & Rupr.] Pohl) ( Ansley et al. 2004 ;Ansley and Castellano 2006 ;Ansley et al. 2019 ).
Natural resource land managers in the southern Great Plains have mixed attitudes toward Texas wintergrass. Texas wintergrass florets easily penetrate the fleece and skin of sheep and hair-goats and can decrease the value of wool and mohair ( Clendenin 2016 ). On the other hand, Texas wintergrass is one of the most important forages in livestock and wildlife diets in the region, especially during the cool season when C 4 grasses are dormant ( Pinchak et al. 1990 ; Richardson et al. 2008 ;Whitney et al. 2009 ). This is in large part due to the species' prevalence, but because it also has greater crude protein than co-occurring C 4 warm-season species, animals tend to select for it over other available species when it is actively growing during cooler times of the year ( Heitschmidt et al. 1987 ;Pinchak et al. 1990 ;Whitney et al. 2009 ).
Once established, Texas wintergrass yield is generally unaffected by increases in mesquite cover, a contrast to the rapid decline in warm-season midgrass production when mesquite cover becomes 25-40% ( Ansley et al. 2004 ). Mesquite canopy cover limits sunlight penetration to C 4 grasses, and subsequently production, during their growing season since the peak productivity period of C 4 grasses corresponds with full mesquite canopy leaf out ( Ansley et al. 1992 ;Ansley et al. 2004 ). In open grasslands with low mesquite canopy cover ( < 25%), Texas wintergrass coexists with C 4 grasses because it maximizes photosynthesis and growth in early spring before sufficient leaf area of warm-season grasses develops ( Hicks et al. 1990 ).
In addition to prolonging carbon assimilation over a longer time period and wider temperature range than co-occurring C 4 species ( Hicks et al. 1990 ), Texas wintergrass possesses other anatomical and physiological advantages. It uses leaf involution and rolls leaves tightly to reduce water loss from stomata on the upper surface of the leaf ( Anderson and Briske 1990 ). It produces both chasmogamous and cleistogamous seeds ( Call and Spoonts 1989 ). Self-fertilized basal cleistogamous florets assist in seed herbivory avoidance and favor early seed production under stress since they are formed before chasmogamous florets in Texas wintergrass ( Dyksterhuis 1945 ;Campbell et al. 1983 ). The microhabitat created by the parent plant may also provide beneficial conditions for germination and seedling establishment. These adaptations allow Texas wintergrass to thrive in dry, disturbed sites and heavily grazed pastures ( Stubbendieck et al. 2011 ).
A couple of studies have investigated the effectiveness of targeted grazing to reduce Texas wintergrass cover, biomass, and reproductive ability ( Hood 2019 ;Murray et al. 2021 ). Targeted grazing uses livestock to address weed invasions, woody plant encroachment, and flammable fuel accumulation. In contrast with production-oriented grazing, stocking regimes for targeted grazing are chosen to encourage heavy defoliation of select, undesirable species at sensitive growth stages ( Rinella and Bellows 2016 ). For example, targeted cattle grazing has typically been used to control introduced, invasive annual broadleaf weeds and grasses ( Diamond et al. 2012 ;James et al. 2015 ), though it has also been used to reduce dominance of herbaceous perennials ( Miller 1996 ;Launchbaugh and Walker 2006 ;Mantz et al. 2013 ) and some woody species ( Gutman et al. 20 0 0 ;Calleja et al. 2019 ). Two previous studies that used targeted cattle grazing to suppress Texas wintergrass in mesquite/wintergrass-dominated grasslands reported mixed results. Hood (2019) found that applying multi-ple cool-season grazing events reduced Texas wintergrass biomass along with reproductive fitness, indicated by a low number of reproductive culms relative to total biomass (i.e., ∼ 1 reproductive culm: 2 g DM). Hood (2019) applied the same stocking density as we used in this study (i.e., 33 626 kg · ha −1 ). In contrast, Murray et al. (2021) reported that grazing at a density of 1.7 mature cows · ha −1 · d −1 did not reduce Texas wintergrass cover after 2 yr of treatment where targeted grazing was applied in one event per year in mid-April to early May. In combination with the application of glyphosate and prescribed burning, targeted grazing reduced Texas wintergrass cover; however, the application of herbicide alone and herbicide in combination with prescribed fire reduced Texas wintergrass cover by a similar proportion, making it difficult to confirm that grazing provided any additional enhancement to the other suppression tactics.
Most studies of targeted cattle grazing to manage grasses and forbs have been in fairly open grasslands with low woody species density. Murray et al. (2021) applied herbicide, prescribed fire, and grazing treatments following mechanical removal of mesquite trees. Hood (2019) did not apply any treatment to mesquites before targeted grazing. The author also did not take location of herbaceous plants along transects relative to mesquite canopy into consideration when assessing grazing effects. Therefore, the impacts of tree density, structure, and proximity have not been evaluated when determining effectiveness of targeted grazing on suppressing Texas wintergrass.
Increases in brush density reduce cattle utilization of grasses ( Owens et al. 1991 ). Large quantities of forage (i.e., 274 ± 47 to 616 ± 127 kg grass · ha −1 on average in experimental pastures when production was greatest) were required to attract cattle into thick stands of mixed brush consisting of mesquite, Acacia spp., and spiny hackberry ( Celtis pallida Torr.) in a study by Owens et al. (1991) . In contrast, forage biomass had little impact on cattle utilization of areas with low brush abundance. Average brush abundance ranged from 4 092 ± 321 to 6 496 ± 298 stems · ha −1 on average in the experimental pastures used in the study. In addition to density, woody plant morphology affects undercanopy grazing occurrence as well. France et al. (2008) reported angle of accessibility affected the likelihood of undercanopy plants being grazed. Grasses under sagebrush ( Artemisia tridentata Nutt. subsp. wyomingensis Beetle and Young) with narrow, erect canopies were more likely to be grazed than those under spreading, "umbrellashaped" canopies. Thorns, spines, and prickles are additional morphological barriers possessed by many woody species, including mesquite, that can reduce grazing of herbaceous species in close proximity to them ( Milchunas and Noy-Meir 2002 and references therein). For example, Marchi et al. (2019) reported that reductions in palisade grass ( Urochloa brizantha [Hochst. ex A.Rich.] R. Webster) height and dry mass by cattle were proportional to increases in distances from the main stems of shrubs with thorns. Similarly, Rebollo et al. (2002) reported that in grazed pastures blue grama ( Bouteloua gracilis [Willd. ex Kunth] Lag. ex Griffiths) growing within clumps of plains pricklypear ( Opuntia polyacantha Haw.) was taller and produced more seedheads than blue grama outside of pricklypear clumps. Due to reported reductions in grazing under woody canopies due to various physical impedances, we suspect it may become difficult for cattle to access Texas wintergrass under the dense multistemmed mesquites that are often found in the Rolling Plains and Cross Timbers.
The objective of this study was to determine effectiveness of targeted cattle (Bos taurus) grazing on Texas wintergrass biomass, leaf length, and reproductive capacity in a Texas wintergrass/mesquite-dominated savanna. Our specific hypotheses were that, in comparison with grazed Texas wintergrass, ungrazed Texas wintergrass, would 1) have more biomass per basal area (BA) (g · cm −2 ), 2) have longer leaves, 3) produce more repro-ductive culms per plant, and 4) progress faster in reproduction (i.e., more reproductive culms in latter stages of development). We hypothesized that these differences would be greatest between control and G2 plants located outside of mesquite canopies. We also hypothesized that, in comparison with grazed individuals located farther away from mesquite stems ( > 1-m distance from stem), Texas wintergrass individuals submitted to targeted grazing close to mesquite stems ( < 0.5-m distance from stem) would 5) be larger (i.e., larger BA, more biomass per BA), 6) have less reductions in leaf length and biomass per BA as a result of targeted grazing, 7) produce more reproductive culms per plant, and 8) progress faster in reproductive development. , sideoats grama, and vine mesquite ( Panicum obtusum Kunth) also persisted at the site. Mesquite density in the research plots at the Research Unit was 888 trees · ha −1 on average, and most mesquites at this site were multistemmed with an average of three stems per tree. Temperatures in the area typically range from 3.8 °C in January to 29.1 °C in July, and the region receives 720 mm of annual precipitation on average. During the study period, precipitation data were recorded at a micrometeorological station on the Research Unit. Total precipitation received in the 6 mo before wintergrass sampling was greater in 2019 than 2018 ( Table 1 ). In 2018 the site received 77% of the 30-yr average precipitation in the 6 mo before wintergrass sampling. The site received 28% more precipitation than the 30-yr average in the 6 mo before wintergrass sampling in 2019. However, the amount of precipitation received between the initiation of grazing events and the sampling of wintergrass was greater in 2018 than 2019, particularly between the second grazing event and plant harvest. Soils in the research plots were clay loams of the Wichita (fine, mixed, superactive, thermic Typic Paleustalfs) se-ries ( Koos et al. 1962 ). Cattle grazed the Research Unit in a moderate density (i.e., 13 ha · AUY −1 ), rotational system and were not allowed access to the study site for 6 mo before the study.

Site description
Mesquite was the dominant overstory species at the Muse WMA, and the dominant perennial grasses included Texas wintergrass, vine mesquite, sideoats grama, and Eragrostis species. Bromus species were common cool-season annuals at the site. Mesquite density in the research plots at the WMA was 800 trees · ha −1 on average, and most were single stemmed. Average monthly temperatures in the area typically range from 7.1 °C in January to 28.2 °C in July, and the region receives 772 mm of annual precipitation on average. Local precipitation data presented herein to represent the region around the Muse WMA were gathered through the National Oceanic and Atmospheric Administration's online weather data records for Brownwood, Texas ( NWS 2022 ). Similar to the Smith-Walker site, the total precipitation received in this region during the 6 mo before wintergrass sampling at the Muse WMA was greater in 2019 than 2018 (see Table 1 ). In 2018 the site received 64% of the 30-yr average precipitation in the 6 mo before wintergrass sampling. The site received 19% more precipitation than the 30-yr average in the 6 mo before wintergrass sampling in 2019. Contrary to the precipitation pattern at Smith-Walker, the amount of precipitation received between the initiation of grazing events and the sampling of wintergrass at Muse was greater in 2019 than 2018. Soils in the research plots were Sunev clay loam (fine-loamy, carbonatic, thermic Udic Calciustolls) ( Clower 1980 ). The Muse WMA had no livestock grazing between 2008 and when this study was initiated in 2018.

Targeted grazing treatment
We used nine 225-m 2 plots at each location to evaluate three treatments with three replications each. Treatments included grazed once (G1), grazed twice (G2), and control. Temporary electric fencing kept cattle in plots. During grazing events, our target defoliation of Texas wintergrass was 5 cm above soil level. This height dictated the duration of grazing in the treated plots. The duration of individual grazing events ranged between 7 h and 12 h. The first grazing treatments were applied in mid-

Texas wintergrass measurements
Before the first grazing event at each site in 2018, we randomly selected three mesquite trees in each 225-m 2 plot. We set two sampling areas around each mesquite: 1) a radius within 0.5 m of the trunk (under canopy; UC) and 2) the interspace area beyond the dripline of the sampling tree and outside of driplines of neighboring trees (out of canopy; OC). We collected samples in April-May 2018 and 2019 during Texas wintergrass' reproductive stage. Specifically, we harvested one individual Texas wintergrass plant from UC and OC regions around each experimental tree in a plot. Therefore, in each plot we had three subsamples in UC and three in OC areas. Ansley et al. (2021) used a similar approach with three subsamples per treatment · canopy combination in a plot when assessing the interacting effects of mesquite herbicide treatment and canopy microsite on herbaceous production and soil characteristics.
During our sampling, Texas wintergrass plants were distinguished as "individuals" by having ≥ 1-cm spacing from other Texas wintergrass plants. We harvested individual plants by clipping all rooted material (alive and dead) at the ground surface. We then measured the diameter of the plant's base at two locations perpendicular to each other using a caliper. The diameters were used to calculate an individual's BA (cm 2 ).
We separated tissue into live and dead material as Texas wintergrass plants were harvested. Plant material was dried at 60 °C for 48 h before being weighed to calculate dry matter yields (DMY). Live and dead weights were added together to get a total weight for an individual. Weights were divided by BA to calculate weight per BA (g · cm −2 ) for each plant material class (live, dead, total). Reproductive culms were sorted separately from vegetative material to determine density of culms per plant. Reproductive culms that were infected by fungus (Atkinsonella texensis) were counted within the total number of reproductive culms per plant but were not assigned to a phenological reproductive class (explained in detail later). Infected culms were counted as their own class and analyzed separately. We categorized healthy, noninfected reproductive culms into seven classes: R1) early boot; R2) late boot; R3) < 50% of seed head emerged from sheath; R4) > 50% of seed head emerged from sheath; R5) 100% of seed head emerged, seed green; R6) 100% of seed head emerged, seed brown and drying; and R7) seed starting to shatter ( < 50% of seed shattered). None of the reproductive culms that we collected had > 50% of seed shattered.

Statistical analyses
The effects of year; treatment (G1, G2, and control); canopy position (UC and OC); and their interactions were tested on BA, total DMY per BA, live DMY per BA, dead DMY per BA, leaf length, total number of reproductive culms per plant, and number of fungusinfected reproductive culms per plant using linear mixed models (proc mixed procedure, SAS 9.4, SAS Institute Inc., Cary, North Carolina, USA). The effects of year, treatment, canopy position, reproductive stage, and their interactions were tested on the number of healthy reproductive culms within the seven reproductive stages (R1-R7). Sites were analyzed separately due to differences in grazing history. Year, treatment, canopy position, and reproductive stage were considered fixed effects, and individual plots (repli-cates) were considered random within treatments in the models. Differences were considered significant at P ≤ 0.05.
Data were analyzed for normality using the Kruskal-Wallace test. Basal area, total DMY per BA, live DMY per BA, dead DMY per BA, and leaf length were normally distributed, so no data transformations were needed. Total number of reproductive culms per plant, number of reproductive culms per reproductive stage, and number of fungus-infected reproductive culms per plant were square root transformed before analyses to determine effects of fixed effects and their interactions in analysis of variance models. Back-transformed means and 95% confidence limits are reported in the text and figures for those variables that required data transformations. Back-transformed 95% confidence limits are indicated in brackets (i.e., []) when they are reported in the text. Superscripts included in figures and statistical differences discussed in the text reflect analyses of square root −transformed data.
At the Smith-Walker site, year × treatment × canopy interactions were significant ( P ≤ 0.05) for total DMY/BA, live DMY/BA, dead DMY/BA, and leaf length, so we split data by year for further analyses of these variables. Year and/or interactions of year with other effects were significant for total number of reproductive culms, number of fungus-infected culms, and number of reproductive culms per reproductive stage, so we split data by year for further analyses of these variables as well.
Neither year nor any of its interactions were significant ( P > 0.05) for total DMY/BA, live DMY/BA, dead DMY/BA, leaf length, or any of the reproductive variables at the Muse site, but we chose to analyze 2018 and 2019 data separately since the G2 treatment was not applied at the site in 2019. Additionally, analyzing years separately at the Muse site maintained consistency with data analysis for the Smith-Walker site. Treatment × canopy interactions were not significant for total DMY/BA, live DMY/BA, or dead DMY/BA at the Muse site in 2018 or 2019, but we chose to present interaction data to maintain consistency with data presentation from the Smith-Walker site.

Vegetative morphology and biomass at the Smith-Walker Research Unit
Basal area of individual Texas wintergrass plants was similar between canopy positions and years and among the treatments at both sites. In 2018, total DMY/BA was greater for UC plants than OC plants ( Table 2 ) ( P = 0.049); total biomass of UC and OC wintergrass plants was 0.41 ± 0.04 g cm −2 and 0.32 ± 0.03 g cm −2 , respectively. Treatment did not affect total DMY/BA. Live DMY/BA was similar across treatments and in the two canopy positions. Dead DMY/BA was greater in UC plants than OC plants ( P < 0.001); dead biomass of UC and OC wintergrass plants was 0.20 ± 0.02 g/cm 2 and 0.09 ± 0.01 g/cm 2 , respectively. Treatment did not affect dead DMY/BA. There were no treatment × canopy interactions for leaf length, but treatment and canopy individually affected leaf length ( Fig. 1 ). Texas wintergrass leaves were longer in the control and G1 treatments than in the G2 treatment (see Fig. 1 a). Leaves of control plants were > 6 cm longer than those of G2 plants on average. Texas wintergrass plants growing under mesquite canopies had leaves that were 4.5 cm longer than OC Texas wintergrass plants (see Fig. 1 c).
In 2019, there was a treatment × canopy interaction for total, live, and dead DMY/BA (see Table 2 ). Control plants had greater total DMY/BA than G1 and G2 plants in the OC position, but all treatments had similar total DMY/BA in the UC position. In the G1 treatment, UC plants had greater total DMY/BA than OC plants. Live DMY/BA was greater in control plants than grazed plants in the OC position, but there were no differences between treatments in Table 2 Live, dead, and total dry matter yield per basal area (DMY/BA) (g cm − 2 ) of Texas wintergrass plants outside (OC)  the UC position. Dead DMY/BA was greater in control plants than G1 and G2 plants in both canopy positions. Wintergrass plants under mesquite canopies had greater dead DMY/BA than OC wintergrass in the G1 and G2 treatments. G1 wintergrass also had greater live DMY/BA in the UC location than the OC location. There were no treatment × canopy interactions for leaf length, but treatment and canopy individually affected leaf length (see Fig. 1 ). Texas wintergrass leaves were longest in the control treatment (see Fig. 1 b). Leaves of control plants were ∼ 7.5 cm longer than those of grazed plants on average. Texas wintergrass plants growing under mesquite canopies had leaves that were 3.5 cm longer than OC Texas wintergrass plants (see Fig. 1 d).

Texas wintergrass reproduction at the Smith-Walker Research Unit
At the 2018 harvest, G1 plants produced more reproductive culms per plant than control and G2 plants ( Fig. 2 a ). UC and OC plants produced similar numbers of reproductive culms (see Fig. 2  Canopy position did not affect the rate of reproductive culm development, but treatment tended to have an effect ( P = 0.08, see Fig. 3 a). Most of the healthy reproductive culms were in the R1 and R2 stages (early and late boot) (see Fig. 3 a). G1 and G2 plants had more culms in the R1 stage than control plants.
At the 2019 harvest, all treatments produced similar numbers of reproductive culms per plant (see Fig. 2 b). UC and OC plants produced similar numbers of reproductive culms as well (see Fig. 2 Fig. 3 b). The majority of culms in all treatments were in the R3 and R4 reproductive stages. Control and G1 plants had more culms in the R4 stage than G2 plants, but G2 plants had the most culms in the R1 stage. Canopy position had no effect on the rate of healthy reproductive culm development in 2019.

Vegetative morphology and biomass at the Muse WMA
In 2018, wintergrass total DMY/BA and live DMY/BA were similar among treatments and in the two canopy positions at the Muse WMA ( Table 3 ). Dead DMY/BA was greater in control than G2 plants ( P = 0.003); dead DMY/BA averaged 0.392 ± 0.071, 0.249 ± 0.075, and 0.121 ± 0.029 g cm −2 in the control, G1, and G2 treat-ments, respectively. Wintergrass leaf length was not affected by treatment or canopy position ( Fig. 4 a and 4 c ), though there was a tendency for longer leaves in UC plants than OC plants. Wintergrass leaves were 26.59 ± 0.95 cm at sampling in 2018.
In 2019, neither treatment, canopy position, nor their interaction had any effect on wintergrass total DMY/BA, live DMY/BA, or dead DMY/BA ( Table 3 ). In 2019, wintergrass leaf length was not Table 3 Live, dead, and total dry matter per basal area (DM/BA) (g cm −2 ) of Texas wintergrass plants outside (OC)   affected by treatment or canopy position (see Fig. 4 b and 4 d); wintergrass leaves were 28.89 ± 1.18 cm at sampling.

Texas wintergrass reproduction at the Muse WMA
At the 2018 Muse WMA harvest, treatment did not affect the number of reproductive culms ( Fig. 5 a). However, UC plants produced more reproductive culms than OC plants (see Fig. 5  Neither treatment nor canopy position had any effect on the rate of reproductive culm development in 2018. Wintergrass had more reproductive culms in the R6 phase than in the R1 and R2 phases ( Fig. 6 a).
At the 2019 Muse WMA harvest, neither treatment nor canopy position affected the number of reproductive culms produced per plant (see Fig. 5 b and 5 -0.4 to 4.6] culms in the respective treatments were infected by fungus. Both treatment and canopy position affected the rate of reproductive culm development. The majority of healthy reproductive culms were in the R4 stage (see Fig. 6 b, Fig. 7 ). Control plants were slightly further along in reproductive development than G1 plants; control plants had more culms in the R5 stage than G1 plants (see Fig. 6 b). The majority of control plants' culms were in the R4 and R5 stages, while the majority of G1 plants' culms were in the R3 and R4 stages. Reproductive culms of OC plants developed slightly faster than those of UC plants; OC plants had greater numbers of reproductive stalks in the R4 and R5 stage compared with UC plants (see Fig. 7 ). Culms of OC plants were primarily in the R4 and R5 stages, while UC culms were primarily in the R3 and R4 stages.

Targeted grazing and canopy position effects on Texas wintergrass vegetative attributes
Contrary to our hypothesis, UC Texas wintergrass plants did not have larger BA than OC wintergrass plants. Cool-season grasses often have greater cover percentages and biomass production under tree canopies relative to areas outside of canopies ( Vaitkus and Eddleman 1991 ;Bernardi et al. 2016 ;Pilon et al. 2021 ). We suspected similar patterns would also be seen at the individual-plant level.  Mesquite facilitation effects on Texas wintergrass are most evident directly under the canopy, but ameliorating effects carry over into the interspaces as well ( Ansley et al. 2019 ;Ansley et al. 2021 ). In a 9-year study by Ansley et al. (2019) in plots where mesquite canopy cover was 66% and interspaces comprised 34% of the total area, wintergrass production and cover was typically greater than that of C 4 perennial grasses in interspaces. Mesquites' facilitation of wintergrass growth and production in interspaces between trees may have resulted in the lack of differences in the BA of UC and OC wintergrass plants at the Smith-Walker site in this study.
The lack of differences in wintergrass BA between canopy positions at the Muse location may be due to inverse reasons of those at Smith-Walker. At the Smith-Walker location, the mesquite density and canopy coverage were such that we believe interspaces at this site may function more like undercanopy areas than at sites with less tree density and canopy cover. Mesquite canopy cover ≥ 30% decreases C 4 midgrass production ( Ansley et al. 2013 ). While we did not measure canopy cover in this study, the density of trees in the Smith-Walker plots (888 mesquites · ha −1 ) was greater than in the Ansley et al. (2019) study, where the researchers reported a density of 622 mesquites · ha −1 and 66% canopy cover, which resulted in greater wintergrass production and cover than that of C 4 grasses in interspaces. Mesquite density at Muse was approximately 800 · trees ha −1 . Although this was not much lower than the density at Smith-Walker, the single-stemmed mesquites at Muse may have provided less canopy cover. Although we did not measure microclimate conditions in either canopy position at either study site, we suspect both canopy positions at Muse received more light and had greater ambient temperatures than either canopy position at Smith-Walker. We believe that more light and greater ambient temperatures under mesquite canopies at Muse would have resulted in similar microclimate conditions for UC and OC plants at the site; hence similarities in wintergrass BA between the two canopy positions may not be surprising.
Although wintergrass BA did not differ between canopy positions at either site, the interaction of canopy position and grazing treatment did influence wintergrass biomass production. As hypothesized, total biomass per BA was greater in UC plants at the Smith-Walker location. This was primarily driven by greater amounts of dead DMY in UC plants, particularly in the grazed treatments. Live DMY was also greater in UC plants in the G1 treatment in 2019. Greater ratios of dead-to-live biomass per BA in UC plants compared with OC plants indicate a greater amount of wintergrass matured without being consumed at Smith-Walker. Annual and perennial C 3 grasses under tree canopies have delayed phenology compared with plants outside of canopies ( Hicks et al. 1990 ;Jackson et al. 1990 ), due to ameliorating effects of canopies on reducing high temperature, evaporative demand, and light conditions. Soil moisture, organic matter, and N are often greater under woody canopies ( Mordelet and Le Roux 2006 ). Soil N increases under canopies are especially pronounced under the canopies of leguminous trees like mesquites ( Saixiyala et al. 2017 ;Ansley et al. 2021 ). Nutrient enrichment under trees may also be attributed to defecation and urination by livestock and wildlife that select subcanopy areas to rest or ruminate under the shade and/or to graze nutrient-rich forage located there, creating a positive feedback loop and adding to the influence of trees on subcanopy forage quality and attractiveness ( Treydte et al. 2010 ). Although we did not measure soil or forage quality relative to canopy position in this study, other studies' findings of enhanced soil fertility under mesquites in the region ( Dai et al. 2006 ;Ansley et al. 2021 ) indicate that the greater dead-to-live DMY ratios of UC plants were likely due to reduced livestock consumption and subsequent increases in accumulation of dead material.
We have not seen any published literature that reports how differences in shrub or tree physiognomy influence the effectiveness of targeted grazing. However, results in studies of wildlife and lowto moderate-intensity livestock grazing indicate that shrub and tree structure directly affect consumption of plant species under them. Woody species and tall forbs, especially those with physical and chemical defenses, can act as biotic refuges for herbaceous species and young, palatable woody plants, which are vulnerable to browsing ( Saixiyala et al. 2017 ;Kelemen et al. 2019 ;Bustamante et al. 2021 ). Marchi et al. (2019) reported that reductions in Urochloa brizantha height and dry mass by cattle were proportional to increases in distances from the main stems of shrubs and forbs with thorns and/or stinging trichomes. Treydte et al. (2010) reported that grazing wildlife consumed more grass tufts under canopies than outside canopies of trees ≥ 4.4 m tall. This was not seen in small trees ( ≤ 2.2 m tall); the number of grass tufts consumed was similar under and outside of the canopies of small trees. The branching height of the small trees was approximately 0.7 m above the ground, and Treydte et al. (2010) speculated that the reduction in grass consumption under small tree canopies was due to reductions in herbivore access, particularly to larger grazers such as zebras, buffalo, and wildebeests. Grazers tended to target the grasses growing at the canopy edge of these smaller trees, suggesting that grasses growing at the canopy edge and, logically, those under the canopy, likely had greater nutritive value than those farther away from the tree trunks due to facilitation effects of the trees.
From these previous studies of livestock and wildlife grazing behavior in the presence of trees and shrubs and what has been reported about Texas wintergrass nutritive quality and livestock preferences for it, we can draw parallels to the UC and OC wintergrass in our study and assume that the multistemmed mesquites at the Smith-Walker location limited cattle grazing of UC plants. Texas wintergrass in the vegetative growth stage is palatable and selectively grazed by livestock and wildlife ( Pinchak et al. 1990 ;Dillard et al. 2006 ). Texas wintergrass is valuable as cool-season forage on rangelands because its crude protein and digestibility contribute to improved ruminant diet quality, increased animal performance, and reductions in supplementation needs in late winter and spring ( Heitschmidt et al. 1982 ;Angell et al. 1986 ). Multiple studies have reported that grasses under leguminous trees have greater leaf N, P, and digestibility than individuals of the same species growing outside of canopies ( Treydte et al. 2008 ;Treydte et al. 2009 ;Baldassini et al. 2018 ). Texas wintergrass crude protein and digestible organic matter range from 6% to 30%, respectively, in the fall to 14% and 44%, respectively, in the spring ( Huston et al. 1981 ). We suspect that the standing dead material of UC plants during the study period was likely similar in quality to the dormant wintergrass tissue reported for the fall by Huston et al. (1981) . The live, green UC Texas wintergrass should have been as nutritious, if not more nutritious, than the leaf material of OC Texas wintergrass, leading to greater utilization of UC wintergrass in the absence of limited accessibility.
Even with the greater amounts of dead DMY in the UC Texas wintergrass individuals relative to OC wintergrass, the UC wintergrass should still have been relatively palatable and attractive to cattle. First, dead biomass of individual UC wintergrass plants at the Smith-Walker plots averaged 38% of total biomass over all treatments in both years at the time of harvest. We acknowledge that the proportion of dead material at sampling may not exactly reflect the proportion of dead biomass at the time of grazing, since plants were harvested 1 to 2 mo after grazing treatment. Some dead material would have been consumed by cattle, while some may have accumulated between grazing and harvest. Secondly, cattle are less selective and eat more standing dead material than sheep, goats, and deer due to their wider mouths and inflexible upper lips ( Grant et al. 1985 ). Cattle would have consumed some portion of standing dead material in both canopy positions, but accessibility would determine the degree to which it was removed. Most importantly, this was a targeted grazing study, and cattle selectivity was intentionally constrained. Nonetheless, cattle likely consumed the most desirable forage that they could access first, before expending greater amounts of energy and increasing discomfort trying to reach less accessible forage (i.e., Texas wintergrass under multi-stemmed mesquite canopies) ( Owens et al. 1991 ;Marchi et al. 2019 ).
The differences in leaf lengths between the two canopy positions at the Smith-Walker site provide additional evidence to support reductions in cattle access to UC plants by multistemmed mesquites at that location; we observed greater reductions in the leaf length of OC plants than UC plants at this site in both 2018 and 2019. In other studies with grazed sites, leaves were also longer under shrub canopies ( Saixiyala et al. 2017 ). Leaf length was not affected by canopy position at the Muse site, however. Similar to the differences in DMY accumulation in the UC position at the two sites, differences in mesquite physiognomy between the sites likely explain the lack of a canopy effect on leaf length at the Muse site. Cattle likely had greater accessibility around and underneath the site's single-stemmed mesquites, resulting in similar utilization and leaf lengths between UC and OC plants. We expected leaves to be shorter in grazed treatments in both locations; this was true at the Smith-Walker site, but not at the Muse site. Leaves were not shorter in the G2 than control treatment at Muse in 2018, likely because leaf lengths were more variable at the Muse site than the Smith-Walker site, which contributed to the lack of statistical differences in 2018. Individual wintergrass plants were harvested approximately 2 mo after the second grazing event at each site, so plants had a similar amount of time to regrow at the two sites. Additionally, with the greater amount of precipitation between the G2 event and sampling at Smith-Walker compared with Muse, we would have expected a greater amount of regrowth at Smith-Walker and potentially a lack of differences in leaf length between ungrazed and grazed plants due to expedited growth in response to amelioration of soil moisture conditions. Differences in leaf length responses between the two sites may be due to differences in livestock grazing efficiency, which may have resulted in a lack of substantial effects from grazing treatments. Differences in various abiotic (e.g., soil moisture, light, temperature) and/or bi-otic (e.g., competition with other herbaceous species) factors at the sites may have played a role as well.

Targeted grazing and canopy position effects on Texas wintergrass reproduction
The lower numbers of total reproductive culms per plant in 2018 than 2019 at both study sites suggest 2018 precipitation conditions suppressed Texas wintergrass reproduction. Dead-to-live DMY ratios were greater in 2018 than 2019 at both sites. This early initiation of leaf senescence provides additional evidence that resources for reproduction may have been limited in 2018. Total reproductive culms increased threefold and sixfold from 2018 to 2019 at the Smith-Walker and Muse sites, respectively, likely as a result of the above-average precipitation amounts received in both regions during the 2019 study period. Texas wintergrass under mesquite canopies produced more reproductive culms per plant than OC plants at the Muse site in 2018; this same effect was not seen at the site in 2019. During the low precipitation and soil moisture conditions of 2018, the UC microclimate might have been slightly more favorable. It is possible the precipitation was great enough at the Muse site in 2019 that soil moisture and evaporative demand conditions were similar across the two canopy positions, so no facilitative effects from the canopy were seen in that year. Simmons et al. (2008) found that Texas wintergrass reproductive tiller density was not affected by mesquite canopy presence in a relatively dry growing season, but UC wintergrass produced more reproductive tillers in a year with greater annual rainfall. This conflicts with the findings at the Muse site in 2018 and 2019.
Contrary to our hypothesis, targeted grazing had some stimulatory effect on reproductive culm production, though this was only seen with the application of a single targeted grazing event in 1 yr at one site. In conjunction with the greater numbers of reproductive culms produced at Smith-Walker in 2018, G1 plants also tended to have greater total DMY than control and G2 plants, especially live DMY of OC plants, which, based on leaf length, would have been grazed to a greater degree than UC plants. Although a decrease in reproduction is usually one of the goals when applying targeted grazing to undesirable species, some studies have reported that grazing stimulates grass short-term reproduction. Reproductive stimulation by defoliation is more common in C 3 grasses than C 4 grasses ( Anderson and Frank 2003 ). In 2019, grazed plants at both sites tended to have less reproductive culms per plant, suggesting that targeted grazing may reduce wintergrass reproductive success if it is applied over more years, more intensely (i.e., greater stocking density), or more frequently. For example, two grazing periods per year are recommended to prevent cheatgrass ( Bromus tectorum L.) seed production, and ≥ 2 yr of this grazing treatment is required to significantly reduce cheatgrass populations ( Menalled et al. 2017 ). Our results indicated that we might have had more success if we had applied more than two grazing events to wintergrass per growing season.
Additionally, the timing of our grazing events relative to when we sampled wintergrass likely influenced the perceived success of our treatments. In 2018, G2 events were applied approximately 2 mo before wintergrass sampling. In 2019, we applied the G2 event at Smith-Walker and single-grazing event at Muse approximately 1 mo before wintergrass sampling. The reduction in timing between grazing and sampling in 2019 likely resulted in delays in reproductive culm development in grazed plots. These delays were seen at both sites. Though the total number of reproductive culms per plant was not reduced by our grazing treatments, delays in reproductive development provide positive evidence for targeted grazing as a potential Texas wintergrass treatment option. If we were to repeatedly apply targeted grazing, particularly focusing multiple grazing events during the development of repro-ductive culms, we could potentially reduce the total number of culms and seeds produced. Applying targeted grazing before inflorescence emergence from the culm (boot stage) reduced cheatgrass seed bank density ( Diamond et al. 2012 ). Similarly, defoliating perennial forbs such as spotted knapweed ( Centaurea stoebe L.) and sulfur cinquefoil ( Potentilla recta L.) while they were bolting or in flowering stages reduced their seed production ( Benzel et al. 2009 ;Mosley et al. 2017 ). Further study is needed to determine if grazing Texas wintergrass during reproductive phases would reduce reproductive success to a greater degree than grazing during the vegetative stage. Ideally, we would test the application of grazing treatment until Texas wintergrass enters summer dormancy. However, the timeframe may be variable depending on the site and its spring and summer weather conditions. We sampled Texas wintergrass at the Smith-Walker Research Unit almost 1 month later in the spring of 2018 than we did in 2019. Although our sampling was later in 2018, wintergrass reproductive development was delayed at Smith-Walker that year due to drought conditions. Most of the reproductive culms were in the early boot stage. In contrast, we sampled at the Muse site almost 2 wk before the Smith-Walker sampling in 2018, and most of the wintergrass there had culms with 100% of the seed head emerging and seed starting to shatter. Managers attempting to use targeted grazing to reduce Texas wintergrass dominance will have to frequently monitor wintergrass physiognomy and be flexible with the timing of grazing events. Additional logistical challenges associated with repeatedly applying targeted grazing in a production scenario would need to be considered as well. These challenges could include establishing fencing around targeted areas, providing water to grazing animals, or navigating difficulties in predicting quantities of forage, which may either result in a lack of enough livestock to effectively apply targeted grazing or a lack of forage at a time when managers might want to apply targeted grazing.

Implications
Targeted grazing slowed reproductive development and reduced leaf length and biomass in areas where grazing cattle could more easily access wintergrass plants (i.e., outside mesquite canopies). Managers are cautioned, though, that reductions in vegetative vigor and slowed reproduction observed during a single year of treatment will likely not have long-term effects. Reductions in leaf length and biomass in G1 and G2 plants in a single growing season confirmed that plants were grazed, but not necessarily that grazing would have any adverse effects on wintergrass growth that would carry over into the following growing season. Similarly, delayed reproduction in 1 yr may have no residual effects in the next year. Because we harvested whole plants in 2018, we needed to measure new individuals in 2019. The individual plants monitored in the G1 and G2 treatments in 2019 may have been grazed in 2018, but we do not know the extent to which any vegetative or reproductive parameters of these plants were hindered in 2018. To evaluate carry-over effects, a follow-up experiment tracking vegetative and reproductive parameters over multiple growing seasons would need to be conducted. This follow-up experiment should include 1) plants grazed in a single growing season that were monitored within that season along with subsequent ungrazed growing seasons and 2) plants grazed over multiple growing seasons and monitored in every growing season during the experiment.
Our vegetative and reproductive data suggest that mesquite removal may be necessary to increase cattle access to Texas wintergrass, particularly in stands with multistemmed individuals. However, immediate success of targeted grazing and reductions in Texas wintergrass cover, biomass, and reproduction following mesquite removal is not likely. Additionally, tree removal on its own will not change understory species composition. A multi-faceted approach is most likely necessary. If mesquites are not thinned by mechanical or chemical means, reduced access and subsequently reduced Texas wintergrass consumption by cattle under mesquites is likely to perpetuate Texas wintergrass dominance, particularly at sites with predominantly multistemmed trees that impede animal access. Moreover, Texas wintergrass plants that accumulate larger amounts of standing dead material will be less attractive to grazing herbivores than plants with a greater ratio of live-to-dead tissue.