Crown fires remove a fire-sensitive canopy dominant from oak-juniper woodlands: results from long-term monitoring of wildfires

Background In central Texas, re-sprouting oaks ( Quercus spp.) co-occur with non-resprouting Ashe juniper ( Juniperus ashei ) in a mosaic of fire-dependent (oak savanna) and fire-sensitive (oak-juniper woodland) habitats. The region’s mature woodlands are the only nesting habitat for the endangered golden-cheeked warbler ( Setophaga chrysoparia ). We studied long-term recovery of woodland structure and species composition after single and repeated crown fires on three soil types (mesa, slope, and deep savanna soils). Results On once-burned sites, density and basal area of non-juniper trees (all woody species except juniper) reached or exceeded unburned levels after 14–24 years, indicating successful recruitment (24 years vs unburned, mesa: 481 ± 254 vs 155 ± 137 stems ha −1 , 2 ± 1 vs 1 ± 2 m 2 ha −1 ; slope: 910 ± 330 vs 251 ± 103 stems ha −1 , 5 ± 2 vs 3 ± 2 m 2 ha −1 ). Ashe juniper, however, remained mostly absent from burned woodlands (juniper tree density, 24 years vs unburned, mesa: 6 ± 10 vs 691 ± 410 stems ha −1 ; slope: 20 ± 17 vs 731 ± 183 stems ha −1 ) and total basal area was 47–87% lower than in unburned areas. In formerly fire-suppressed savannas, non-juniper tree density exceeded unburned levels and juniper density recovered the most (24 years vs unburned, non-juniper: 679 ± 250 vs 251 ± 103 stems ha −1 ; juniper: deep: 50 ± 71 vs 317 ± 297 stems ha −1 ). Juniper trees were still absent from twice-burned sites in year 11 and understory density was recovering more slowly, at least on slopes (one fire: 224 ± 206 stems ha −1 ; two fires: 26 ± 47 stems ha −1 ). Juniper recovery was correlated with distance to the wildfire perimeter, suggesting that regeneration is limited in part by dispersal. Conclusions We found successful recruitment of resprouting hardwood species after one and two crown fires, likely due to the low deer densities at Fort Cavazos. In fire-suppressed oak savannas, a single crown fire did not restore savanna structure and Ashe juniper is slowly re-establishing. Long-term restoration of a savanna on these soils will require additional treatments, like repeated prescribed fire. In oak-juniper woodlands, crown fires removed the fire-sensitive Ashe juniper from canopy co-dominance for decades, making the woodlands unsuitable as habitat for golden-cheeked warblers. Given the long-term consequences of crown fires for golden-cheeked warbler habitat, existing mature oak-juniper woodlands should be protected from crown fire.


Background
Fire shapes the distribution of ecosystems as well as the structure and species composition of plant communities (He et al. 2019;Hanberry et al. 2020).Species in fireprone systems are usually adapted to the historical fire regime, which can range from frequent, low-severity surface fires to infrequent, high-severity crown fires (Keeley et al. 2011).Common adaptations to fire include the ability to resprout after fire, a persistent seedbank, and morphological adaptations that prevent fire damage, such as thick bark (Keeley 2006;Hmielowski et al. 2014;Varner et al. 2016).Feedbacks between species composition and fire history can create mosaics with sharp boundaries between fire-dependent grasslands or savannas and firesensitive woodlands or forests (Peterson and Reich 2008;Hoffmann et al. 2009;Trauernicht et al. 2012).
In central Texas, re-sprouting oaks (Quercus spp.) cooccur with the non-resprouting Ashe juniper (Juniperus ashei J. Buchholz) in a complex mosaic of habitats that include grasslands, oak-dominated savannas and shrublands, and mixed woodlands co-dominated by oaks and junipers.The pre-settlement fire regime differed among these ecological communities (Fowler and Carden 2024).Grasslands, savannas, and open woodlands were maintained by relatively frequent fire (Murray et al. 2013) but are now threatened by fire exclusion that is causing woody plant encroachment, especially by junipers (Yang and Crews 2020).In contrast, old-growth woodlands require long fire-free intervals to achieve co-dominance by Ashe juniper (Diamond 1997;Reemts and Hansen 2008;O'Donnell 2019) and are now threatened by clearing for grazing and urban development (Duarte et al. 2013;Rhodes et al. 2021).The woodlands, which are co-dominated by juniper, oaks, and other hardwoods, are distinct from juniper-encroached and juniper-dominated grasslands and savannas.The mixed woodlands are often found on slopes and shallow soils, especially in the topographically complex eastern Edwards Plateau, while grasslands and savannas can be found on deeper soils and are more abundant in the flatter western Edwards Plateau (Bray 1904;Fowler and Carden 2024) (Fig. 1).
Management of a complex ecosystem mosaic requires fire use in some habitats and fire protection in other habitats.As in other regions, grasslands and savannas in central Texas are often managed with prescribed fire to prevent woody plant encroachment (e.g., Ansley et al. 2021).If woody plant encroachment is advanced, restoration may require mechanical or chemical treatments, sometimes in combination with prescribed fire under extreme drought or heat conditions (Rollins and Bryant 1986;Ansley et al. 2006;Twidwell et al. 2016).Management goals for grasslands and savannas are influenced by economic needs (primarily increasing forage for cattle), by interest in increasing water infiltration, and by conservation needs of grassland-dependent birds (Banta and Slattery 2011;Rosenberg et al. 2019;Scholtz et al. 2021).
In contrast to grasslands and savannas, many of the oak-juniper woodlands in central Texas have been protected from fire because they serve as the exclusive nesting habitat for the endangered golden-cheeked warbler.Golden-cheeked warblers require the presence of both Ashe juniper and oaks (along with other hardwoods) for nesting material and food (Pulich 1976;Beardmore 1994;Trumbo et al. 2021).Warbler abundance usually peaks when juniper occupies around half of the woodland canopy but birds avoid edges and canopy gaps (DeBoer and Diamond 2006;Peak and Thompson 2013).Managers of these woodlands must balance the habitat requirements of golden-cheeked warblers with a lack of oak recruitment and high fire risk.High deer densities limit the recruitment of oaks and other hardwoods to the canopy (Russell and Fowler 2004;Andruk et al. 2014;Van Auken et al. 2023).Low-to moderate-severity prescribed fires in woodlands can reduce understory density and increase oak regeneration (primarily resprouting after canopy mortality), but do not reduce fuel loads (Yao et al. 2012;Reidy et al. 2016Reidy et al. , 2021)).However, any mortality of canopy trees reduces habitat quality for golden-cheeked warblers (Reidy et al. 2016(Reidy et al. , 2021)).
In contrast to low-severity prescribed fires, high-severity crown fires cause major changes in oak-juniper woodland structure and composition.We have previously shown that Ashe juniper was largely absent for the first 10 years after a single crown fire (Reemts and Hansen 2008) and that repeated crown fires slowed short-term juniper recovery even more (Reemts and Hansen 2013).
Here we provide a long-term follow-up to our original studies (24 years after the first fire and 11 years after the second fire) to examine the long-term consequences of crown fires in oak-juniper woodlands and understand the implications for golden-cheeked warbler habitat and savanna restoration.Specifically, we wanted to address three major questions: (1) Have burned transects recovered (i.e., have they become similar to unburned transects)?(2) Are once-and twice-burned transects recovering at similar rates?(3) Does recovery vary by soil type, especially when comparing shallow soils unburned slope where juniper has dark green foliage while oaks and other hardwoods have light green foliage; unburned deep soil site with small juniper on the left and large oak on the right.Middle row, from left to right-twice-burned mesa, twice-burned slope, and once-burned deep soil site.The standing dead stems in the mesa and slope photos show that height and size of regrowth after the first fire.Ashe juniper is absent from all three burned photos.Bottom-landscape context of the habitats, with frequently burned oak savanna in foreground and mature oak-juniper woodlands on slopes and mesas in the background (mixed woodlands on slopes and mesas) and deeper soils (savannas)?

Study area
Fort Cavazos Military Reservation (formerly known as Fort Hood) is an ~ 86,000 ha Army installation in Bell and Coryell Counties, TX (Fig. S1).Common plant communities include oak-Ashe juniper woodlands and oak shrublands on slopes and mesa tops.Grasslands are found in the valleys, on rolling lowlands, and on slopes with very high (~ annual) fire frequency.Savannas dominated by Texas live oak (Quercus fusiformis Small) and post oak (Q.stellata Wagnenh.)occur in the lowlands and on deeper soils on the mesas.Fort Cavazos includes breeding habitat for a federally endangered songbird (the golden-cheeked warbler) and a recently delisted songbird (the black-capped vireo, Vireo atricapilla).

Fire descriptions
On 21 February 1996, military training activities started three grassland fires.Due to hot, dry, and windy conditions, the fires moved into adjacent woodlands, where they burned mostly as crown fires.By the time the fires were controlled on 7 March, they had burned more than 4000 ha of woodlands.The second fire, which burned from 3 to 12 April 2009, was smaller (608 ha; 360 ha overlapping with first fire, Fig. S2).The Palmer Z-index was extreme in February 1996 and mid-range in April 2009.Burn severity in the 2009 fire was lower than in the same areas during the 1996 fire, perhaps due to a combination of less extreme drought and lower fuel loads (Reemts and Hansen 2013).

Vegetation sampling
Following the first fire, we randomly located 101 permanent transects in moderately to severely burned oak-juniper woodlands.We sampled most of these "onceburned" transects annually from 1996 to 2002 and again in 2005, 2010, and 2020; a few transects were first sampled in 1997 (Table S1).After the second fire, we sampled all "twice-burned" transects annually from 2009 to 2011 and again in 2020.We sampled "unburned" transects in mature woodlands located adjacent to the burned areas on similar slopes and soils in 2001, 2005, 2010, and 2020.While the fire history of these unburned woodlands is not known, all are present in aerial imagery taken in 1938 and appear similar in maturity to the woodlands subsequently burned in 1996 (Fig. S2).Tree skeletons in the burned areas were also of similar size to trees in the unburned areas.
Transects were classified into three groups based on topographic position and soil type, which influence the dominant plant communities (Fig. 1, Fig. S2)."Mesa" transects, found on flat-topped hills, were on Eckrant cobbly silty clay soils with shallow slopes (< 8° slope, Low Stony Hill ecological site).On Fort Cavazos, the vegetation on this soil type includes shin oak (Quercus sinuata Walter var.breviloba [Torr.]C.H. Mull) shrublands and shin oak-Ashe juniper woodlands."Slope" transects were on sloped Real gravelly clay loam soils (5-30° slope, Steep Adobe ecological site) and were slightly more mesic.This soil series supports a more diverse woodland, codominated by Buckley oak (Q.buckleyi Nixon and Dorr) and Ashe juniper, that also includes shin oak, Texas ash (Fraxinus albicans Buckley), and Texas live oak in the canopy.In mesic canyons on the same soils, Ashe juniper dominance decreases and additional species, such as bigtooth maple (Acer grandidentatum Nutt.) and chinkapin oak (Q.muehlenbergii Engelm.), are added.Finally, "deep" soil transects were on flat sites with Evant silty clay (1-3° slope, Redland ecological site).These soils are slightly deeper than mesa soils (15-33 cm to bedrock vs 10-30 cm for mesa soils) and have fewer bedrock outcrops.Savannas on these soils are dominated by post oak and blackjack oak (Q.marilandica Muencch.);woodlands are created by juniper encroachment.
Transects were 110 m long and followed either a random bearing (for level sites) or were perpendicular to the slope (i.e., followed contour lines).Plots were located at 10 m intervals along the transects and 7 of the 11 possible plots were randomly selected for sampling.From 1996 to 2002, we re-randomized plot selection every year.Starting in 2005, we used the plot locations from the first sampling year on each transect.
We sampled trees in 10 m × 10 m plots.From 1996 to 2002, we recorded the stem density of trees in 5-or 10-cm dbh classes.Starting in 2005, we recorded dbh to the nearest 0.1 cm for each tree to track tree size in greater detail.If saplings or trees branched above the root crown, only the largest stem was measured and counted.In our study site, Ashe juniper is the only species that usually has multiple branches rather than multiple stems; Ashe juniper is therefore under-represented when density is converted to basal area.

Statistical analyses
While we sampled multiple transects in the burned areas, these transects represent subsamples of the two wildfire events and we lack true replication (Wester 1992).However, our data are still useful to document vegetation responses for comparison with future studies elsewhere in the region.
To analyze woodland structure, we calculated averages by transect for stem density of the understory (seedlings, shrubs, and saplings; all stems < 5 cm dbh or < 1.8 m tall) and overstory (trees, all stems ≥ 5 cm dbh), as well as tree basal area.Basal area, which is the sum of the crosssectional trunk area per sampled area, incorporates tree sizes and provides additional information about forest structure when combined with tree density.Calculations were done for Ashe juniper and for all other species (referred to as "non-juniper" species).
We conducted separate analyses of juniper and nonjuniper species for understory stem density, tree stem density, and tree basal area.Because data from earlier years are already published (Reemts andHansen 2008, 2013), we focus here on years 9-24 post fire (for onceburned sites) and year 11 post-fire (for twice-burned sites); analyses of the full dataset are provided in the supplement.In the analyses with twice-burned data, we defined year "10" as year 9 for once-burned transects and year 11 for twice-burned transects (Table S1).Analyses with twice-burned data only include mesa and slope transects, because no deep soil transects were burned in the second fire (Fig. S2).We did not analyze juniper tree density or basal area in twice-burned sites because there were no juniper trees in our transects.For comparisons with unburned sites, we always used the data from 2020.
Data were analyzed using linear mixed-effects models (lmerTest package, Kuznetsova et al. 2017) with a Gaussian distribution.To account for repeated sampling, all models included a random intercept for transect.We designed models to address our focal questions of recovery (i.e., similarity to unburned sites) as well as the influence of soil type and number of fires on recovery.For analyses focused on once-burned sites, we used time since fire (categorical variable with values of 9, 14, 24, and unburned), soil type (mesa, slope, and deep), and their interaction as fixed effects.For analyses focused on twice-burned transects, we used the number of fires (once, twice, or unburned), soil type (mesa and slope), and their interaction as fixed effects; only data from year ~10 was included from burned sites.
While the models address the overall influence of our independent variables, we also wanted to understand the data in more detail.For this reason, we used contrasts that we defined a priori (not based on linear model outcomes), focusing only on comparisons of interest.For once-burned transects, we compared soil types within each time since fire (e.g., mesas and slopes in year 9) as well as burned and unburned transects within a soil type (e.g., year 9 vs unburned for mesas, see Table S3 for full list of contrasts).For twice-burned transects, we compared once-and twice-burned transects in year ~10 within the same soil type, as well as twice-burned and unburned transects within the same soil type (see Table S5 for full list of contrasts).Contrasts were calculated using the multcomp package (Hothorn et al. 2008), which accounts for multiple comparisons by adjusting P values in a single step based on the joint t-distribution of the linear function.
Finally, we examined the relationship between understory stem density and distance to the wildfire perimeter for once-burned transects, conducting separate analyses for Ashe juniper and non-junipers.We used data from all years  in a linear mixed model with soil, distance to fire edge, and their interaction as fixed effects.We also included a random intercept effect for transect.Because only 6 transects in twice-burned areas had understory juniper by year 11, the analysis could not be repeated for the second fire.

Understory density of non-juniper species recovered to unburned levels
In once-burned transects, understory stem density of all species except Ashe juniper ("non-juniper" species) was higher than unburned transects in years 9 and/or 14 in all three soil types, but then became similar to unburned sites (Fig. 2, Table S3).In twice-burned transects, understory stem density was similar to once-burned and unburned transects (Fig. 2, Table S5).Density was similar among soils in all years (Tables S3 and S5).

Forest structure changed: more but smaller trees of non-juniper species
In once-burned transects, tree density increased above unburned levels in mesa and slope transects in years 14 and/or 24 (Fig. 2, Table S7).Density in twice-burned transects was similar to unburned transects (Table S9).Tree density differed among soils: mesas had lower density than slopes in all years for once-and twice-burned transects, and lower density than deep soils in year 14 (Tables S7 and S9).
In contrast, tree basal area in once-burned transects was lower than unburned sites in years 9 and/or 14 for slopes and deep soils (Fig. 2, Table S11).Basal area in twice-burned transects was lower than unburned transects on mesas and slopes (Table S13).Like tree density, basal area differed among soils for once-burned transects: mesas had lower basal area than slopes and deep soils in years 14 and/or 24 (Tables S11 and S13).
Juniper trees were absent from all once-burned transects until at least year 14 and tree density remained significantly lower than unburned transects in slope and mesa transects (Fig. 3, Table S19).Juniper basal area was also significantly lower on burned than unburned transects on slopes and mesas in all years (Fig. 3, Table S21).On deep soils, where some unburned transects did not have any juniper trees, where juniper tree density in year 24 was lower but not significantly different from unburned transects (p = 0.10) and basal area was lower Fig. 2 Oak-juniper woodland response to one or two crown fires compared to unburned woodlands at Fort Cavazos, TX, USA, for all woody species except Ashe juniper.Top-understory stem density (stems < 5 cm dbh or < 1.8 m tall).Middle-tree stem density (stems ≥ 5 cm dbh).Bottom-tree basal area (stems ≥ 5 cm dbh).* indicates a significant difference between burned and unburned data.Mesas are dominated by shin oak; slopes are dominated by Buckley oak; deep soils are dominated by post oak but not different for all years (Fig. 3, Tables S19 and S21.In twice-burned transects, no trees were present by year 11.

Junipers, but not other species, are more common close to wildfire edges
In once-burned transects, Ashe juniper understory density increased near the wildfire edge (Fig. 3, Table S22).The increase was steepest in deep soils, although these soils also had the lowest maximum distance to the wildfire perimeter (deep: 344 m; slope: 438 m; mesa: 575 m for mesas).Non-juniper understory density was not significantly correlated with distance to wildfire edge (Table S23).

Discussion
Single or repeated crown fires changed oak-juniper woodland structure and removed Ashe juniper, a nonresprouting canopy co-dominant, for at least 24 years after wildfire at Fort Cavazos in central Texas.The density and basal area of non-juniper species (all other woody species, generally resprouters) recovered to preburn values by year 24.However, in the absence of Ashe juniper, overall burned woodland basal area was 47-87% lower than in unburned areas, depending on the soil type.3 Ashe juniper response to one or two crown fires compared to unburned woodlands at Fort Cavazos, TX, USA.Top-understory stem density (stems < 5 cm dbh or < 1.8 m tall).Middle-tree stem density (stems ≥ 5 cm dbh).Bottom-tree basal area.Juniper trees were absent from once-burned sites until year 14 and were absent from twice-burned sites in all years.* indicates significant difference between burned and unburned data; ° indicates that no juniper trees were found A second crown fire had little effect on recovery rates for non-juniper species, but Ashe juniper recovered even more slowly after the second fire: understory stem density in year 11 was 50-90% lower than after the first fire.
Unlike elsewhere in central Texas and in other oakdominated forests (e.g., Andruk et al. 2014;Alexander et al. 2021;Van Auken et al. 2023), oaks and other hardwood trees recruited successfully after one and two crown fires.We observed that the vast majority of this recruitment came from resprouting and not from seed.In most oak forests, recruitment is limited by deer herbivory and by lack of fire, which allows fire-sensitive mesophytic species to crowd out shade-intolerant oaks (Habeck and Schultz 2015;Hanberry et al. 2020).Fire clearly played a role in stimulating resprouting at Fort Cavazos, but in other oak woodlands in Texas and the USA, high levels of deer browsing can negate the effects of prescribed fire (McEwan et al. 2011;Andruk et al. 2014;Cory and Russell 2022).At Fort Cavazos, sustained hunting pressure had reduced deer density to ~6 deer 100 ha −1 at the time of the first crown fire (Wolverton et al. 2009).This density is lower than the density of 20 deer 100 ha −1 that allowed successful oak recruitment at the nearby Kerr Wildlife Management Area and much lower than the 30+ deer 100 ha −1 densities at sites with no oak recruitment (Russell and Fowler 2002).The presence of saplings in unburned areas, in contrast to the long-term lack of saplings in many other central Texas woodlands (Russell and Fowler 1999;Van Auken et al. 2023), also suggests that low deer numbers contributed to successful oak regeneration.These results support the current regional approach of managing deer density or excluding deer to promote recruitment of oaks and other hardwoods (Redick and Jacobs 2020).
While hardwood recruitment is desirable in oak and oak-juniper woodlands, abundant resprouting is problematic for savanna restoration.In our study, the pre-fire vegetation on deep soils was juniper-encroached and firesuppressed post oak savanna with the lowest basal area of juniper and the highest basal area of other tree species (Figs. 2 and 3).By 24 years after a single crown fire, overall stem density of resprouting species was higher than in unburned areas and juniper density was no longer statistically different, although still lower (Figs. 2 and 3), indicating that these woodlands did not regain or maintain a savanna structure.In fire-suppressed oak savanna elsewhere in the USA, repeated low-intensity prescribed fire is usually combined with mechanical thinning or brush control to restore the desired open structure (Glennemeier et al. 2020;Vander Yacht et al. 2020).Increasingly, managers are also testing high-intensity or extreme prescribed fire to control resprouting woody plants (Haney et al. 2008;Twidwell et al. 2016).However, our data and other studies show that a single fire, even if it is of high intensity, cannot restore savanna structure.Instead, highintensity fire must either be repeated or its effects must be maintained by other treatments like frequent lowintensity fire (Haney et al. 2008;Dey et al. 2017).
The virtual elimination of a reseeding canopy dominant after crown fire has been noted in other studies.In the pine-oak forests of the southwestern USA and Mexico, for example, crown fires can eliminate ponderosa pine (Pinus ponderosa Lawson & C. Lawson) from large areas, shifting the community from pine forest to oak shrubland (e.g., Barton and Poulos 2018;Coop 2022).Regeneration of pines depends on distance to unburned sites and topography and is hampered by competition from resprouting oaks and unfavorable weather (Haffey et al. 2018;Rodman et al. 2020;Stevens et al. 2021).In central Texas, loblolly (Pinus taeda L.) regeneration was lower in high-severity burn patches than in moderate-severity patches (Cooper et al. 2018).As with ponderosa pine, loblolly regeneration is limited by dispersal, competition with resprouting oaks, and xeric conditions in burned sites.
In our study, Ashe juniper regeneration was likely influenced both by dispersal limitation and by competition with resprouting species.After fire, Ashe juniper understory stem density was higher near wildfire edges (Fig. 4) while the density of resprouting species did not vary with distance (Table S13).Ashe juniper is dispersed by birds and mammals (Chavez-Ramirez andSlack 1993, 1994) but seed distribution varies by species, with flocking birds and mammals creating highly clumped patterns (Chavez-Ramirez andSlack 1993, 1994).Studies of other juniper species have found variable dispersal efficiencies, ranging from 3 m per year for Utah juniper (Juniperus osteosperma (Torr.)Little) to more than 12 m for eastern redcedar (Juniperus virginiana L., Holthuijzen and Sharik 1985;Neupane and Powell 2015).Even though Ashe juniper is dispersing more quickly than other juniper species (up to 200 m in the first year and 500 m by year 24), the higher understory density near the wildfire edge suggests that dispersal limitation may slow woodland recovery.
After the second fire, recovery of Ashe juniper may be proceeding even more slowly than after the first fire.On slopes, understory stem density was 90% lower a decade after the second fire compared to a similar period after the first fire.In other systems such as Mediterranean pine forests, Australian eucalyptus forests, and California chaparral, repeated wildfires can reduce or eliminate obligate reseeders if fires recur before plants mature enough to produce more seed (e.g., Zedler et al. 1983;Fairman et al. 2015;Taboada et al. 2017).In our sites, the slower recovery must be due to other factors, because juniper did not reseed successfully after even one fire.
One potential reason is the smaller sample size of twiceburned transects (19 vs 65 once-burned at the equivalent time), which makes it less likely that widely dispersed seedlings fall in a transect.Another reason may be the initial higher density of resprouters after the second fire that provided stronger competition with juniper seedlings (Fig. S4, Reemts and Hansen 2013).Previous studies have found that Ashe juniper seedling establishment and survival are highest in shaded conditions but growth is highest at canopy edges and in open grasslands (Jackson and Van Auken 1997;Van Auken et al. 2004), so dense hardwood resprouts after crown fire likely reduce growth of any junipers that establish in the burned areas.Longer monitoring will be needed to confirm that juniper recovery remains slower after repeated fires.
Ashe juniper is a critical habitat component for the endangered golden-cheeked warbler, which nests only in central Texas.Warblers used shredding bark from mature junipers (20+ years old) for nest building and both adults and fledglings forage in juniper canopies (Pulich 1976;Kroll 1980;Trumbo et al. 2021).After the first crown fire, warbler abundance in the burned areas dropped by 85% for at least 10 years, with the remaining birds using only unburned fragments and wildfire edges where juniper was still present (Baccus et al. 2017).Given that warbler abundance peaks when Ashe juniper occupies around half of the woodland canopy (DeBoer and Diamond 2006;Peak and Thompson 2013), the burned woodlands at Fort Cavazos will remain unsuitable as warbler habitat for many decades.

Conclusions
Vegetation in central Texas is a complex mosaic of firedependent and fire-sensitive habitats.At Fort Cavazos, those vegetation types include oak-dominated savannas (sometimes encroached by Ashe juniper after fire suppression) and mixed woodlands that are co-dominated by oaks and Ashe juniper.In contrast to other oak woodlands in Texas and elsewhere in the USA, we found successful recruitment of oaks and other resprouting hardwood species after one and two crown fires, likely due to the low deer densities at Fort Cavazos.In fire-suppressed oak savannas, a single crown fire did not restore savanna structure and Ashe juniper is slowly re-establishing.Full restoration of a savanna on these soils will require additional treatments, like repeated prescribed fire.In formerly mature oak-juniper woodlands on slopes and mesas, crown fires removed the fire-sensitive Ashe juniper from canopy co-dominance for decades, making the woodlands unsuitable as habitat for the endangered golden-cheeked warbler.Given the long-term consequences of crown fires for golden-cheeked warbler habitat, existing mature oak-juniper woodlands should be protected from crown fire.This project was funded by the U.S. Army through cooperative agreements DPW-ENV-97-A-0001, DPW-ENV-02-A-0001, and DPW-ENV-07-A-0001 with The Nature Conservancy of Texas, as well as the Department of Defense Legacy Resource Management Program.The content of this manuscript does not

Fig. 1
Fig.1Woodland types at Fort Cavazos, TX, USA.Top row, from left to right-unburned mesa with bedrock outcrop and large Ashe juniper trees; unburned slope where juniper has dark green foliage while oaks and other hardwoods have light green foliage; unburned deep soil site with small juniper on the left and large oak on the right.Middle row, from left to right-twice-burned mesa, twice-burned slope, and once-burned deep soil site.The standing dead stems in the mesa and slope photos show that height and size of regrowth after the first fire.Ashe juniper is absent from all three burned photos.Bottom-landscape context of the habitats, with frequently burned oak savanna in foreground and mature oak-juniper woodlands on slopes and mesas in the background

Fig.
Fig.3Ashe juniper response to one or two crown fires compared to unburned woodlands at Fort Cavazos, TX, USA.Top-understory stem density (stems < 5 cm dbh or < 1.8 m tall).Middle-tree stem density (stems ≥ 5 cm dbh).Bottom-tree basal area.Juniper trees were absent from once-burned sites until year 14 and were absent from twice-burned sites in all years.* indicates significant difference between burned and unburned data; ° indicates that no juniper trees were found

Fig. 4
Fig. 4 Ashe juniper understory stem density (stems < 5 cm dbh or < 1.8 m tall) in once-burned transects increased near the edge of the 1996 wildfire at Fort Cavazos, TX, USA.Points are slightly transparent to show overlap