Restoration of Florida scrub vegetation in an old field through 23 years after planting

Florida scrub, a fire‐maintained, xeromorphic shrubland, hosts many rare and declining species, but most scrub habitat has been lost to development and agriculture. Formerly cultivated sites offer an opportunity to restore Florida scrub. In this 23‐year study, we test whether scrub species could be restored to an old field and whether scrub vegetation composition and structure could be reestablished over time. Scrub oaks were planted in a section of an old field in 1992 after the site was cleared and treated with herbicide. Additional oaks and other scrub species were planted in 1993. We determined survival and growth annually of a marked sample of scrub species. We sampled vegetation cover in two height strata annually on 10 permanent line‐intercept transects in the old grove and on 20 transects in adjacent, intact scrub. Initial survival of Quercus geminata exceeded that of Q. chapmanii or Q. myrtifolia. Cover of scrub oaks greater than 0.5 m increased from 1.3% in 1992 to 65.2% in 2010. However, the vegetation structure of large Q. geminata did not resemble that of native scrub. Adaptive management led to cutting large oaks in the fall of 2010, and burning the site in early 2011. Cutting and burning appeared to stimulate sprouting and clonal spread of scrub oaks. Ordination analysis indicated directional change related to increasing scrub oak cover and time since planting but the old field still differed from intact scrub. Vegetation has developed toward scrub composition and structure but exotic grasses persist.


Introduction
Former agricultural lands including croplands and pastures have been abandoned at increasing rates worldwide . Natural revegetation or active restoration of abandoned agricultural lands (hereafter old fields) could be valuable for conservation of species and communities and for ecosystem services. The dynamics of old fields are more complex and variable than indicated by early models (see . Cultivation modifies soil properties and these legacies may last decades to millennia (McLauchlan 2006;Walker & Wardle 2014). Past land use has left legacies in vegetation structure and composition, ecosystem processes, and soils (Foster et al. 2003). Cramer et al. (2008) provided a synthetic framework in which to view the dynamics of old fields contingent on biotic and abiotic legacies of cultivation where recovery of preexisting vegetation becomes increasingly difficult as the biotic and abiotic legacies increase and thresholds are crossed. Where soil modifications are minor, cultivation less intensive, native vegetation is in close proximity, and dispersal of native species can occur or seed banks persist, native species may reestablish and develop toward the historical vegetation with little intervention (Wright & Fridley 2010). With increased scale and duration of agriculture, remnants of native vegetation are reduced, seed banks are lost, and less dispersal of native seed occurs (e.g. Holl 1999;Standish et al. 2007); this may be compounded by introduction of invasive exotic species adapted to the modified conditions resulting from cultivation (e.g. Stylinski & Allen 1999;Hooper et al. 2004;Davis et al. 2005;Midoko-Iponga et al. 2005;Kulmatiski 2006;Tognetti & Chaenton 2012;Kuebbing et al. 2014). Intensification of cultivation (e.g. nutrient additions) may cross abiotic as well as biotic thresholds such that the recovery of historical vegetation is unlikely without extensive restoration. Such sites may persist in a degraded state Author contributions: PAS designed the original research; both authors conducted the field studies, analyzed the data, and wrote and edited the manuscript. or develop as novel combinations of native and exotic species (Hobbs et al. 2006). Substantial areas of Florida scrub have been converted to agriculture, particularly citrus, and subsequently abandoned. With revegetation, these sites could be important in maintaining populations of some rare scrub species (Stephens et al. 2012). Florida scrub is a rare and declining plant community (Myers 1990;Menges 1999) important to a variety of threatened and endangered plant and animal species; at least two-thirds of Florida scrub has been lost to development and agriculture (Christman & Judd 1990;Duncan et al. 2004;Kautz et al. 2007;Weekley et al. 2008). Conservation of remnant scrub vegetation is a matter of state and national concern (Menges 1999).
Scrub plant species establish poorly in old fields; rather, Sabal palmetto (cabbage palm), woody vines, and exotic grasses often dominate (Schmalzer et al. 1994). Restoring scrub habitat involves not only establishing the appropriate vegetation but also developing a habitat structure that can be managed by prescribed fire (Breininger & Schmalzer 1990). For the Florida Scrub-Jay (Aphelocoma coerulescens), a listed species, a shrub vegetation structure with scrub oaks, vegetation height less than 1.7 m, and open, sandy patches is required (Breininger et al. 2014).
Here we report on the long-term changes (23 years after planting) of an old field site where Florida scrub species were planted. We address two overarching questions. Can dominant scrub species be restored to a former agricultural site, and can scrub vegetation composition and structure be reestablished over time? Specifically, we (1) examine the survival and growth of planted scrub species in the old field; (2) determine whether cover of scrub species increases over time in the old field; (3) determine whether cover of exotic grasses decreases over time in the old field; (4) determine whether the height structure of the developing vegetation is similar to that of adjacent intact scrub; and (5) determine whether the composition of the developing vegetation becomes similar to adjacent, intact scrub. The target for this restoration was restoring scrub vegetation that would support the Florida Scrub-Jay. Specifically, the desired vegetation condition would be a shrubland with scrub oaks predominant (≥50% cover), vegetation height of 1.2-1.7 m, minimal exotic species cover, and open, sandy patches present (Breininger & Carter 2003;Breininger et al. 2014). Understanding the complexities of restoring scrub on this old field site will provide valuable insight for future restoration efforts.

Study Area
Kennedy Space Center/Merritt Island National Wildlife Refuge (KSC/MINWR) is located on the Cape Canaveral-Merritt Island barrier island complex on the east coast of central Florida (28 ∘ 38 ′ N, 80 ∘ 42 ′ W). The climate is warm and humid; precipitation averages 131 cm/year, but year-to-year variability is high (Mailander 1990). The wet season extends from May to October.
Multiple dune ridges occur on the barrier island topography (White 1970) with intervening swales of lower elevation. Scrub vegetation occupies the well-drained ridges, pine flatwoods the more poorly drained flats, and graminoid marshes or woody swamps the lower swales (Schmalzer et al. 1999 (Schmalzer & Hinkle 1992b;Schmalzer et al. 1999). Nomenclature follows Wunderlin and Hansen (2011) unless otherwise noted.
The northern part of KSC/MINWR was an early center of citrus agriculture (Davison & Bratton 1986). There were about 1,012 ha in citrus cultivation when lands were acquired for KSC in the early 1960s; cultivation continued for decades thereafter but was abandoned subsequently (USDI 2008 Scrub species were planted in a former citrus grove as part of a scrub habitat restoration plan for KSC/MINWR (Schmalzer et al. 1994). Previously, we reported on the early (1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999) survival, growth, and vegetation development in this site (Schmalzer et al. 2002). Here we extend the record on the survival and growth of scrub plant species and on changes in community composition and structure through 2015.
This study was conducted in a plot of about 5.6 ha that was part of an abandoned grove circa 20.2 ha in size (Fig. 1). This plot was adjacent to extant scrub vegetation cut and burned in 1993 after greater than 40 years of fire suppression (Schmalzer et al. 1994). This grove was established between 1958 and 1965; it was abandoned after a freeze in 1987, and the citrus trees were cleared and burned in 1988 (Table 1). Both the extant scrub and the former grove are on well-drained, sandy soils (Astatula and Paola series, typic and spodic quartzipsamments, respectively; Huckle et al. 1974) associated with and representative of scrub and scrubby flatwoods in the region (Schmalzer et al. 1999(Schmalzer et al. , 2001. Soils of the former grove had higher pH, NO 3 -N, available Cu and Zn but lower organic matter, NH 4 -N, and available Al and Fe than intact scrub (Schmalzer et al. 2002). A sand road divides the site into a larger northern section and a smaller southern section (Fig. 1).  Partial winter (January-February) burn of old field. 2014 Partial January burn southern section of old field.

Site Preparation and Planting
Site preparation began with mechanical removal of small Sabal palmetto that had established in the site followed by two applications of glyphosate herbicide (0.5 L/ha, June 1992; 10.0 L/ha, July 1992). Herbicides were applied to reduce cover of exotic species, particularly grasses, to reduce competition with planted, native species. Species chosen to be planted are the dominant plants of scrub on KSC/MINWR including three scrub oaks (Quercus chapmanii, Q. geminata, and Q. myrtifolia), Serenoa repens, Pinus elliottii, and two ericaceous shrubs (Lyonia fruticosa, Vaccinium myrsinites) (Schmalzer & Hinkle 1992b). Scrub oaks grown for 1 year were planted in August 1992 in rows at a rate of about 988/ha (with Q. geminata > Q. myrtifolia > Q. chapmanii) (Fig. 1). Planting was done in summer during the rainy season with no supplemental watering. An additional 2,000 1-year oaks were planted in late August-September 1993 due to higher than expected initial mortality; only Q. geminata and Q. myrtifolia were included as Q. chapmanii was not available (with Q. myrtifolia > Q. geminata). Other scrub species were grown to 1-gal (3.8 L) pot size and planted between July and August 1993. These included 500 Serenoa, 300 L. fruticosa (coastalplain staggerbush), 200 V. myrsinites (shiny blueberry), and 100 Pinus; these plants were watered when planted but not subsequently. All plants were placed within the same 5.6 ha plot as in 1992 (Fig. 1).
We estimated initial planting densities as sufficient to establish the dominant scrub species on this site with the assumption that clonal spread of oaks, Serenoa, and ericaceous shrubs would increase their cover over time.

Survival and Growth of Planted Species
To determine survival of planted oaks, we tagged 15 oaks in each of 10 selected planted rows ( Fig. 1) in late August 1992 (96 Q. geminata, 41 Q. myrtifolia, and 13 Q. chapmanii); rows were selected in a stratified random manner such that samples occurred north to south and west to east (Fig. 1). From the second planting, we tagged 50 scrub oaks, (20 Q. geminata and 30 Q. myrtifolia) and 50 of the additional scrub species (34 Serenoa, 6 Pinus, 4 Vaccinium, and 6 Lyonia) in the early September 1993 to follow survival and growth. We sampled tagged oaks of the first cohort in April/May 1993 and sampled all tagged individuals from 1994 to 2010. The site did not burn from 1992 to 1999, but burned partially or wholly in prescribed fires six times from 2000 to 2010 (Table 1). These prescribed fires were conducted by USFWS/MINWR staff as part of upland habitat management (USDI 2008); however, many oaks were not top-killed. At each sampling, we determined survival and height of tagged species. In the fall of 2010, large Quercus, Sabal, and some Pinus were cut by chain saw; the site burned in early 2011. This management action was taken because the height structure that had developed was outside the normal range for scrub vegetation and because the oaks (particularly Q. geminata) were growing as single stemmed trees instead of clonal shrubs. We resumed sampling for survival in 2012, but the height of individual scrub oaks was no longer measured because clonal growth made identifying individual stems difficult.

Community Composition and Structure
We established 10 line-intercept transects (15 m length each) (Mueller-Dombois & Ellenberg 1974) in late August 1992 after planting was completed to determine long-term vegetation changes (Fig. 1); transects were located in a stratified random manner. We recorded vegetation cover data by species in two height strata (0-0.5 m and >0.5 m); cover was measured to the nearest 5 cm. We resampled these transects in early fall (late August to early October, typically September) from 1993 to 2015 because the herbaceous vegetation was best developed at that time of year.
Line-intercept cover transects (N = 20, 15 m length each) were established in adjacent scrub in 1992, and vegetation cover sampled as described above. This scrub had never been cleared for agriculture but had been unburned for greater than 40 years. It was cut and burned in 1993 and the transects were sampled from 1994 to 2015.

Data Analysis
We conducted statistical analyses including summary statistics, Chi-square comparisons of survival among oak species, and Spearman rank order correlations (r s ) between selected groups of plants (SPSS ver. 18; IBM, Armonk, NY, U.S.A. www.ibm .com). We used nonmetric multidimensional scaling (NMS) ordination (Kruskal 1964a(Kruskal , 1964b to examine changes in old field community composition over time (PCORD; version 6, MjM Software Design, Gleneden Beach, OR, U.S.A.). NMS is considered the most generally effective method for the ordination of community data (McCune & Grace 2002). We ordinated the greater than 0.5 m strata and 0-0.5 m strata separately, using the Sorenson distance measure. The initial dataset consisted of 237 samples (10 transects sampled 24 times less one transect missing for 3 years). Screening of the greater than 0.5 m data indicated that the 1992 samples (N = 10) did not have sufficient cover to be included in the analysis and screening of the 0-0.5 m data indicated that the 1992 samples were an outlier group as most of the cover was dead grass from the recent herbicide application, and these samples were dropped from further analyses. We also dropped species with less than two occurrences in the greater than 0.5 m or 0-0.5 m data set to avoid outliers. The reduced greater than 0.5 m dataset had 227 samples and 51 species and the 0-0.5 m dataset had 227 samples and 64 species. We compared the mean annual stand composition (>0.5 m) of the old field site to that of the adjacent scrub using NMS; there were 23 samples of the old field site and 22 of the adjacent scrub. Seventy-five species occurred in the combined scrub and old field data. For some results, we grouped native and introduced species of similar life forms (Table 2) to follow dynamics of the vegetation change over time more readily as there were many species involved.

Survival and Growth of Planted Species
By 8 months after planting, scrub oak survival from the first planting was 54.7% overall; survival differed significantly among species ( Fig. 2A; X 2 = 44.6, p < 0.001) with Quercus geminata > Q. chapmanii > Q. myrtifolia. Most oaks that survived to 8 months after planting survived to April 2010, 17.6 years after planting (Fig. S1A, Supporting Information). There was some additional loss of Q. geminata after cutting of large oaks in the fall of 2010 and burning in 2011 and 2013.
Eight months after planting overall scrub oak survival of the second planting was 34.0%; the species differed significantly (X 2 = 10.5, p = 0.001) with survival of Q. geminata > Q. myrtifolia survival (Fig. S1A). The survival of the first oak cohort to 8 months after planting was significantly greater than survival of the second cohort (X 2 = 5.2, p = 0.022). However, neither the survival of Q. geminata (X 2 = 1.2, p = 0.266) nor Q. myrtifolia (X 2 = 0.35, p = 0.522) differed between the two cohorts.
Early height growth of Q. geminata exceeded that of Q. chapmanii and Q. myrtifolia (Schmalzer et al. 2002). From 2000 to 2010, six prescribed fires burned all or part of the site (Table 1). All tagged Q. chapmanii and Q. myrtifolia were top-killed at least once in these fires; however, 66% of Q. geminata were not, and their mean height was 8.0 m in 2010, well outside the normal range of scrub height (Fig. S2A). In contrast, height growth in adjacent, intact scrub reached about 2.0 m before being reduced by prescribed fire (Fig. S2B). Serenoa survival was 100% until April 1995 when it declined to 88.2%; it declined to 61.8% in May 1996 (Fig. S1B). All tagged Serenoa have survived since then. The mortality of Serenoa appeared to be due to rooting of the developing rhizome by feral pigs (Sus scrofa) (Schmalzer et al. 2002). Mean height of Serenoa increased from 19.5 cm (SD = 5.7, N = 34) in 1993 to 154.1 cm (SD = 57.3, N = 21) in 2015. Fire reduced the height of Serenoa but it recovered rapidly.
Most of the loss of Lyonia fruticosa occurred soon after planting (Fig. S1B) with no loss since 1999. The sample size of Vaccinium myrsinites was small (N = 4) and all had died by 2000 (data not shown). Some Vaccinium that were not tagged have survived and spread in the site.

Community Composition and Structure
Immediately after planting total cover was low, there was little cover of scrub oaks (Tables S3 & S4), and no cover of Pinus or Serenoa (Tables S5 & S6). Scrub oaks increased slowly in cover greater than 0.5 m from 1992 through 1999 and then more rapidly to 2010 ( Fig. 2A; Table S3). Much of this increase was due to Q. geminata reaching large size (Fig. S2A). Cutting in 2010 and burning in 2011 reduced oak cover, but this was followed by increases more related to growth of clonal patches of oaks ( Fig. 2A; Table S3). Cover of Pinus greater than 0.5 m increased from zero initially to about 19% in 2015 ( Fig. 2B; Table S5) although pine mortality from fire and cutting has occurred. Cover of Serenoa greater than 0.5 m remained less than 1% through 2002 but has increased since then ( Fig. 2C; Table S5). Fires have reduced Serenoa cover in some years, but cover reestablished rapidly after fire.
Cover of native grasses greater than 0.5 m increased for several years after planting ( Fig. 2D; Table S3). Native grass cover declined from 2005 to 2010 but increased after cutting of large oaks and burning ( Fig. 2D; Table S3). Cover of native forbs greater than 0.5 m increased after planting, decreased from 2003 to 2010, and then increased after the large oaks were cut ( Fig. 2E; Table S3). Cover of native forbs declined with increased cover of scrub oaks (r s = −0.310, p < 0.001).
Exotic grasses increased in cover greater than 0.5 m the year after planting, retained substantial cover for several years, declined from 2002 to 2010, but increased after cutting and burning of the large oaks ( Fig. 3A; Table S3). Cover of exotic grasses 0-0.5 m declined from 2005 to 2010 ( Fig. 3B; Table S4) but has not changed greatly since then.
Not all exotic grasses followed the same pattern. Melinis repens (Natal grass) was relatively abundant from 1993 through 2002 in the greater than 0.5 m strata, but then declined and showed no increase after 2010 ( Fig. 3C; Table S5). Panicum maximum (Guineagrass) remained relatively abundant greater than 0.5 m from 1993 to 2009 and increased after 2010 ( Fig. 3D; Table S5). Cynadon dactylon (Bermudagrass) was relatively abundant 0-0.5 m to 2000 but then declined and showed no subsequent increase ( Fig. 3E; Table S6). Paspalum notatum (bahiagrass) 0-0.5 m increased for several years after herbicide treatment and planting and has fluctuated in a narrow range from 2005 to 2015 ( Fig. 3F; Table S6). Increase in cover of scrub oaks was associated with a decline in cover of exotic grasses (r s = −0.364, p < 0.001) but with increases in cover of woody vines (r s = 0. 475, p < 0.001) and native shrubs (r s = 0.201, p = 0.002).
Ordination Analysis. Temporal trends in vegetation cover were observed in the NMS ordination. The first axis of the NMS was correlated with time since planting (r = 0.597, p < 0.001). Transects sampled in the early years after planting were to the left of the first axis of the old field ordination (Fig. 4A) Table 2. Error bars show ± SE.
( Fig. 4B, left side of first axis), while cover of scrub oaks, Pinus, Serenoa, and woody vines was greater as time since planting increased (Fig. 4B, right side of first axis). The north and south sections of the old field differed in vegetation composition in the 0-0.5 m strata. This was shown by the first axis of the NMS of the old field 0-0.5 m strata (Fig. 5A) and was correlated with position (r = 0.835, p < 0.001). Paspalum notatum occurred at the right side of the first axis (Fig. 5B), shown by the right-most exotic grass point, reflecting its abundance in the southern section of the site. It was abundant from the beginning in the section to the south and remained so throughout the study. Although scrub oaks and slash pine increased in cover greater than 0.5 m south of the sand road, P. notatum remained abundant 0-0.5 m.
The vegetation cover of the old field has remained distinct from the composition of adjacent scrub. In the ordination of the combined old field and scrub data (Fig. 6), composition of the old field shifted with time since planting toward that of the intact scrub but remained distinct. This reflected the increasing cover of scrub oaks, particularly Q. geminata. Variation within the intact scrub was related to time since fire with the most recently burned samples higher on the second axis.

Discussion
Reestablishing scrub vegetation requires that the dominant species, particularly scrub oaks and saw palmetto, survive, grow, and develop a vegetation structure that could be maintained by fire. Planted scrub species showed different patterns of survival but survived in sufficient numbers such that composition of the old field changed over time. Changes from cultivation did not pass abiotic thresholds that would have prevented survival and growth of planted scrub species. Cover of scrub species increased over time, and this resulted in a reduction in the cover of exotic grasses as well as that of native grasses and forbs. The structure that developed in the old field differed from that of intact scrub; the dominant oak species grew as large single stem trees instead of clonal shrubs. Additional management was required to transition to a more clonal shrubland.

Survival and Growth of Scrub Species
Little is known of the seedling establishment of scrub oaks. In this study, Quercus geminata exhibited greater survival than Quercus chapmanii and Quercus myrtifolia in both planted cohorts; 1-year-old Q. geminata were larger than those of the other two oaks, which may have had some effect on survival. Scrub oaks can establish in sandhill vegetation during periods of fire suppression (Guerin 1993;Menges et al. 1993), but exotic grasses may present a more difficult environment for establishment of oak species than the native grasses in sandhills.
We expected clonal spread of scrub oaks to occur after they established. All scrub oaks have extensive underground roots and rhizomes (Guerin 1993), sprout and spread clonally after fire (Menges & Kohfeldt 1995), and reestablish cover quickly (Abrahamson 1984b;Schmalzer 2003). Clonal spread occurred to some extent for Quercus chapmanii and Q. myrtifolia but most Q. geminata grew rapidly in height and became fire resistant more quickly than anticipated. Only after being cut and burned in 2010-2011 did more extensive clonal spread   Table 2. of Q. geminata occur. It may be that Q. geminata allocated more biomass to height growth in the old field initially than to belowground biomass. Quercus geminata can obtain greater height than the other scrub oaks (Nixon et al. 1997); the more open, and possibly less competitive, conditions of the old field compared to intact scrub may have contributed to the greater height growth. Serenoa repens, a rhizomatous, clonal palm, is an important species in scrub, flatwoods, and other vegetation types in Florida (Abrahamson & Hartnett 1990;Myers 1990). The high initial survival seen here is consistent with observations that seedling and adult mortality is low even after drought and fire (Abrahamson 1995;Abrahamson & Abrahamson 2002, 2009. Serenoa grows slowly in intact vegetation (Abrahamson 1995;Abrahamson & Abrahamson 2009), and can reach great age (Takahashi et al. 2011). Serenoa exhibited more rapid growth and clonal spread in this old field than in intact scrub except where surrounded by dense Paspalum notatum (Foster & Schmalzer 2012).
Pinus elliottii var. densa is the common canopy tree in flatwoods in central and southern Florida (Abrahamson & Hartnett 1990). It is capable of rapid growth (Lohrey & Kossuth 1990) as demonstrated in this old field site and will establish in old fields (O'Hare & Dalrymple 2006).  Table 2.

Dynamics of Exotic Grasses
As woody cover increased, cover of exotic grasses declined; however, two species, Panicum maximum and Paspalum notatum, remained common. Panicum maximum is a large, perennial clump grass, and this may make it less likely to be shaded out by developing shrub vegetation. Paspalum notatum is a perennial, rhizomatous, sod-forming grass known to be persistent in pastures (Hamman & Hawkes 2013) and other upland and drained wetland (Toth & van der Valk 2012) sites where it was planted and limits the establishment of native species (Jenkins et al. 2004;Frances et al. 2010;Tucker et al. 2017).

Dynamics of Native Grasses and Forbs
Increasing cover of woody vegetation resulted in a decline in cover of native grasses and forbs suggesting that they do not constitute a substantial barrier to reestablishing scrub species. These native grasses and forbs are characteristic of open and disturbed habitats including old fields and roadsides (Wunderlin & Hansen 2011). Many of these species (or related congeners) were noted in earlier studies of old field succession in north-central Florida (Laessle 1942), north Florida (Gano 1917;Kurz 1945;Kay et al. 1978;Busing & Clebsch 1983), and more broadly in the southeastern United States (Wright & Fridley 2010).

Changes in Community Composition
Ordination analysis of the greater than 0.5 m strata of the old field data showed directional change with time associated with increasing cover of scrub oaks, Serenoa, and Pinus, and decreasing cover of native forbs, native grasses, and exotic grasses. When compared to intact scrub, the old field vegetation moved toward scrub in the ordination space but remained distinct. After 23 years the old field vegetation retains exotic and native species absent from intact scrub. Ordination of the 0-0.5 m strata of the old field data further indicated the intermediate character of the current old field vegetation. A persistent but low-growing exotic grass, P. notatum, was established in the southern section of the planting site and has retained substantial cover through the initial herbicide treatment and growth of scrub species. Ordination of this stratum reflects its continued abundance rather than directional change with time. Scrub vegetation does not develop spontaneously on former agricultural sites. Scrub oaks and Serenoa have relatively large seeds dispersed by animals; therefore, limited dispersal would be expected to be a factor in the lack of natural establishment in cleared areas. Establishment of seedlings in sandy soils, low in nutrients and organic matter, and subject to periodic droughts, may be infrequent. Competition with exotic grasses may further limit establishment in these old fields as found in other systems (Davis et al. 2005;Midoko-Iponga et al. 2005;Kulmatiski 2006;Standish et al. 2007;Tognetti & Chaenton 2012). Limited dispersal of native species and established populations of exotic grasses could be considered a biotic threshold.
Planting and initial reduction in exotic grasses helped overcome that biotic threshold. There has been a clear increase in woody cover of native scrub species along with a reduction in exotic grasses, native grasses, and forbs over time. These changes occurred despite relatively high initial mortality of planted oaks and the unexpected growth of Q. geminata to large trees. A second round of management intervention in 2010-2011 was required to stimulate the sprouting and clonal spread of Q. geminata. The current community composition retains substantial cover of two exotic grasses (P. maximum, P. notatum) and remains distinct from scrub that was never cultivated.
Soils of this old field differed in several parameters from adjacent scrub soils (Schmalzer et al. 2002). These soil legacies did not prevent the survival and growth of planted scrub species, indicating that the site had not passed an abiotic threshold as described by Cramer et al. (2008). Modified soils may have favored invasive exotic grasses (Greenberg et al. 1997;David & Menges 2011) that competed with scrub species. Soil treatments without planting of scrub species affected soil properties but not vegetation on similar sites (Weiler et al. 2013). Persistent soil legacies even after removal of exotic grasses may limit germination of scrub herbs (Hamman & Hawkes 2013).
The future composition and structure of this site is not certain, but the current trajectory is toward greater abundance of scrub species. The clonal growth patterns of oaks and Serenoa should allow them to continue to expand. Shade from oak clones and needle drop from pines are expected to reduce grasses over time. Meiners et al. (2002) noted a decline of exotics over 40 years of succession in a New Jersey old field but found that shade-tolerant exotics remained.
This study illustrates the importance of long-term monitoring in restoration practice. Little was known about reestablishing scrub vegetation at the beginning of this project (Schmalzer et al. 1994). Future restoration should benefit from this experience. We recommend greater initial effort to reduce established populations of exotic grasses, in particular the dense sod formed by P. notatum and the large clumps of P. maximum. Approaches combining tilling to break up the dense rhizomes of P. notatum and repeated herbicide applications have shown some success on other sites (Jenkins et al. 2004;Freeman et al. 2017). Planting larger stock of scrub oaks might reduce initial mortality. Triggering the clonal growth of Q. geminata appears to require coppicing either by fire or cutting. We recommend burning earlier in the vegetation development before Q. geminata becomes fire resistant. A limitation of this study was that it involved only one site. Future studies would benefit from replication; multiple sites could allow testing burning at different intervals after planting.

Supporting Information
The following information may be found in the online version of this article:  Table S1. Mean cover (%) greater than 0.5 m of species groups at the scrub planting site from 1992 through 2015. Table S2. Mean cover (%) 0-0.5 m of species groups at the scrub planting site from 1992 through 2015. Table S3. Mean cover (%) greater than 0.5 m of species at the scrub planting site from 1992 through 2015. Table S4. Mean cover (%) 0-0.5 m of species at the scrub planting site from 1992 through 2015.