Contents lists available at ScienceDirect

Forest Ecology and Management

journal homepage: www.elsevier.com/locate/foreco

Site preparation for longleaf pine restoration on hydric sites: Standdevelopment through 15 years after planting

Connor D. Croucha,1, Benjamin O. Knappa,⁎, Susan A. Cohenb, Michael C. Stambaugha,Joan L. Walkerc, G. Geoff Wangd

a School of Natural Resources, University of Missouri, 203 Anheuser-Busch Natural Resources Building, Columbia, MO 65211, USAb Institute for the Environment, UNC Chapel Hill, 100 Europa Dr. Suite 490, Chapel Hill, NC 27517, USAcUS Department of Agriculture, Forest Service, Southern Research Station, Clemson University, 233 Lehotsky Hall, Clemson, SC 29634, USAdDepartment of Forestry and Environmental Conservation, Clemson University, 261 Lehotsky Hall, Clemson, SC 29634, USA

A R T I C L E I N F O

Keywords:Pinus palustrisGrass stageSilvicultureHerbicideBeddingSoil manipulationStem analysis

A B S T R A C T

Longleaf pine (Pinus palustris Mill.) restoration is an important land management goal throughout the south-eastern U.S. On hydric sites within the Atlantic Coastal Plain, restoration may involve site preparation prior toplanting to overcome challenges to seedling establishment, such as abundant competition and poor soil drainage.Investment in site preparation assumes that treatments will result in long-term benefits to stand development,yet lasting impacts of site preparation on longleaf pine are not well understood. We sampled longleaf pineplantations in Onslow County, North Carolina through three years and at 15 years after site preparation andplanting. The eight study treatments we tested include an untreated control, six combinations of two vegetationcontrol treatments (chopping or herbicide) with three soil manipulation treatments (mounding, bedding, or flat-planting [no treatment]), and a chopping-herbicide-bedding treatment. Our findings indicate that site pre-paration significantly improved survival and growth of longleaf pine through 15 years. Herbicide resulted ingreater growth, higher survival, and earlier grass stage emergence than chopping. Similarly, soil manipulationtreatments resulted in improved stand establishment outcomes relative to flat-planting (no treatment). Effects ofsite preparation treatments on diameter growth were observed early and maintained through the end of thestudy period, while effects on survival were not observed within the first three years. Differences in stand heightamong treatments were more strongly driven by growth rates following grass stage emergence than timing ofemergence. Our results demonstrate that site preparation improves longleaf pine stand establishment on hydricsites, although the intensity of site preparation treatments recommended for restoration depends on manage-ment objectives.

1. Introduction

Restoration of longleaf pine (Pinus palustris Mill.) ecosystems is animportant land management objective throughout the southeastern U.S.Prior to European settlement, longleaf pine occupied an estimated 37million hectares (Frost, 2006). By 2010, however, one of North Amer-ica’s most extensive forest ecosystems had been reduced to approxi-mately 4.5% of its pre-settlement range (Oswalt et al., 2012). This re-duction was due to a number of anthropogenic factors, including timberharvest and naval stores production, land conversion for urban andagricultural uses, grazing by hogs and other livestock, and, perhapsmost importantly, fire exclusion (Frost, 2006; Landers et al., 1995). In

addition to this dramatic reduction in areal extent, the cultural, eco-logical, and economic value of longleaf pine contribute to interest in itsrestoration. Longleaf pine ecosystems include some of the most diverseplant communities in North America (Walker and Peet, 1983; Walkerand Silletti, 2006), and many plant species within these ecosystemsdemonstrate high endemism and rarity (Glitzenstein et al., 2001;Walker, 1998, 1993). In addition, longleaf pine produces high qualitytimber and suitable habitat for both game species and wildlife species ofconservation concern (Brockway et al., 2015; Landers et al., 1995;Mitchell et al., 2006). Longleaf pine forests may be simultaneouslymanaged for timber and biodiversity (Freeman and Jose, 2009; Mitchellet al., 2006), and this pairing of economic and ecological benefits has

https://doi.org/10.1016/j.foreco.2020.117928Received 7 November 2019; Received in revised form 20 January 2020; Accepted 21 January 2020

⁎ Corresponding author.E-mail addresses: connor.crouch@nau.edu (C.D. Crouch), knappb@missouri.edu (B.O. Knapp), susanac@email.unc.edu (S.A. Cohen),

stambaughm@missouri.edu (M.C. Stambaugh), joanwalker@fs.fed.us (J.L. Walker), gwang@g.clemson.edu (G.G. Wang).1 Current affiliation: School of Forestry, Northern Arizona University, 200 E Pine Knoll Dr, Flagstaff, AZ 86011, USA.

Forest Ecology and Management 461 (2020) 117928

0378-1127/ © 2020 Elsevier B.V. All rights reserved.

T

been critical to increasing interest in longleaf pine restoration acrossthe southeastern U.S. (Landers et al., 1995; Mitchell et al., 2006).

Restoration often occurs on sites where mature longleaf pine is nolonger present or is a minor component of the overstory, which ne-cessitates the use of artificial regeneration. This practice has become socommon that over 25% of contemporary longleaf pine forests wereestablished through artificial regeneration (Oswalt et al., 2012). Amajor impediment to successful establishment of longleaf pine is itsintolerance to competition, particularly as a grass stage seedling (Boyer,1990; Brockway et al., 2006). During this stage, seedlings allocategrowth to root development rather than stem elongation and can beoutcompeted and overtopped by fast growing vegetation (Boyer, 1990).This challenge is amplified on hydric sites, such as flatwoods, where soilmoisture limits seedling growth and competition from woody plants isabundant (Brockway et al., 2006; Glitzenstein et al., 2003; Peet, 2006).Longleaf pine would have naturally regenerated on hydric sites under aregime of frequent surface fire (Frost, 2001; Peet, 2006), but sites withheavy competition from woody plants and a history of fire exclusionoften require site preparation to successfully establish planted seedlings(Brockway et al., 2006).

Prior to artificial regeneration of southern pines, herbicide or me-chanical treatments are commonly used to reduce the abundance ofcompeting vegetation. Herbicides are favored because they can targetspecific plant groups and do not promote sprouting of woody compe-tition, as occurs with mechanical treatments (Fox et al., 2007; Loweryand Gjerstad, 1991; Walker and Silletti, 2006). Studies have con-sistently observed increased longleaf pine seedling growth when her-bicide is applied prior to planting (Addington et al., 2012; Boyer, 1988;Freeman and Jose, 2009; Knapp et al., 2006), while herbicide effects onearly survival have been mixed (Addington et al., 2012; Boyer, 1988;Freeman and Jose, 2009; Knapp et al., 2006; Loveless et al., 1989).There has been less research on mechanical site preparation, such asdrum chopping, for vegetation control during longleaf pine establish-ment. Knapp et al. (2006) found that herbicide resulted in greater rootcollar diameter growth and grass stage emergence than chopping. Thepoor performance of chopping in that study may be due to the fact thatonly a single pass of the chopper was applied. Boyer (1988) found that asecond pass of chopping or harrowing resulted in higher survival oflongleaf pine seedlings and increased grass stage emergence comparedto a single pass.

On hydric sites, longleaf pine seedling establishment is furthercomplicated by poor soil drainage that can inhibit seedling growth andsurvival (Brockway et al., 2006; Fox et al., 2007). Soil manipulationtreatments can improve the rooting environment for planted seedlingsby increasing aeration (Johnson and Gjerstad, 2006; Londo and Mroz,2001; Lowery and Gjerstad, 1991) and by providing additional vege-tation control through mechanical disturbance (Johnson and Gjerstad,2006). Bedding and mounding treatments have been shown to con-sistently improve the growth of slash pine (Haywood, 1987; Outcalt,1984; Pritchett, 1979) and longleaf pine seedlings (Knapp et al., 2006;Loveless et al., 1989). In addition, bedding has been shown to increasesurvival of slash pine seedlings (Pritchett, 1979), although neitherbedding nor mounding were reported to improve survival of longleafpine seedlings within the first few years after planting (Knapp et al.,2006; Loveless et al., 1989).

Investment in site preparation assumes that treatments will result inlong-term benefits to stand establishment, yet many studies report onlyshort-term responses from within a few years after treatment. Threepossible long-term growth responses to site preparation have been de-scribed (Morris and Lowery, 1988; Nilsson and Allen, 2003). Type Aresponse occurs when growth gains from initial treatment continue toincrease over time. Type B response occurs when initial growth gainsfrom the treatment are maintained through time but do not continue toincrease. Type C response occurs when initial growth gains from thetreatment are subsequently lost. Nilsson and Allen (2003) described afourth trajectory, Type D, which occurs when growth of an untreated

stand surpasses that of the treated stand. Studies on slash and loblollypine have reported conflicting results on long-term growth responsesfollowing site preparation (Nilsson and Allen, 2003; Outcalt, 1984;Zhao et al., 2009). Although short-term studies on longleaf pine havegenerally found that site preparation improves seedling growth, it is notknown if those early increases result in meaningful differences in standdevelopment. Given the unique seedling development of longleaf pine,reports from within a few years of planting may not provide enoughtime for seedlings to emerge from the grass stage. Thus, it is not knownif site preparation may result in (1) earlier grass stage emergence, (2)improved growth following grass stage emergence, or (3) a combina-tion of the two.

The objective of this study was to quantify responses of plantedlongleaf pine to site preparation treatments through 15 years on poorlydrained, hydric sites. The study served as a follow up to an earlier studyby Knapp et al. (2006, 2008) that monitored seedling growth and sur-vival through three years after planting. We addressed three specificquestions: (1) How does site preparation affect longleaf pine tree sizeand density at the stand level 15 years after planting? (2) Are earlygains obtained from site preparation maintained beyond the seedlingestablishment phase and into stand development? (3) Are differences instand development among treatments the result of earlier emergencefrom the grass stage, increased growth rates after grass stage emer-gence, or a combination of the two?

2. Methods

2.1. Study area

This study was conducted on Marine Corps Base Camp Lejeune(34°36′N, 77°24′W) in Onslow County, North Carolina. Camp Lejeune islocated within the Atlantic Coastal Flatwoods section of the OuterCoastal Plain Mixed Forest province (McNab et al., 2007). The studysites were located on Leon sand (sandy, siliceous, thermic Aeric Ala-quod), a poorly drained Spodosol with a fluctuating water table thatmay be at or near the surface (Barnhill, 1992; NRCS, 2014). Spodosolsare the primary soil order of flatwoods and are characterized by theirsandy, acidic, and infertile nature (Brockway et al., 2015; Peet, 2006).Leon sand is one of the most extensive soil series on Camp Lejeune(Barnhill, 1992; Frost, 2001) and is of large extent throughout theSoutheast (NRCS, 2014). Timber production has been a primary landuse objective throughout Camp Lejeune’s history (MCBCL, 2015), andthe sites selected for this study were previously used for slash and lo-blolly pine plantations, which were harvested six months to two yearsprior to treatment installation. In recent decades, Camp Lejeune hasprioritized restoration of longleaf pine on hydric flatwoods sites(MCBCL, 2015). Although hydric flatwoods are less commonthroughout the range of longleaf pine than mesic and xeric sites(Kirkman et al., 2017), longleaf pine restoration on hydric sites iscommon (e.g., Freeman and Jose, 2009; Glitzenstein et al., 2003; Hierset al., 2012; Jose et al., 2010; McCaskill and Jose, 2012) and is parti-cularly important because of the diverse plant communities that occuron such sites (Frost, 2001; Peet, 2006; Walker and Peet, 1983).

The natural vegetation community on frequently burned Leon sandin this area is species-rich longleaf pine wet savanna (Frost, 2001). Thisbilayered community consists of an overstory dominated by longleafpine and an understory containing a diverse array of grasses, sedges,forbs, and fire-dwarfed shrubs (Frost, 2001). Wiregrass (Aristida strictaMichx.) dominates the herbaceous layer, and bluestems (Andropogonspp. and Schizachyrium spp.) are also common (Frost, 2001; Peet, 2006).The estimated pre-settlement fire return interval was 1–3 years, andwith such frequent fire Leon sand supports rare species such asroughleaf loosestrife (Lysimachia asperulifolia Poir.) and Venus flytrap(Dionaea muscipula Ellis) (Frost, 2001). These sites also typically containother insectivorous plants, such as pitcher plants (Sarracenia spp.) andsundews (Drosera spp.) (Frost, 2001; Peet, 2006). Common shrubs

C.D. Crouch, et al. Forest Ecology and Management 461 (2020) 117928

2

include inkberry (Ilex glabra (L.) A. Gray), dwarf and blue huckleberries(Gaylussacia dumosa and frondosa (Andrews) Torr. & A. Gray), and a fewblueberry species (Vaccinium spp.) (Frost, 2001; Peet, 2006).

2.2. Experimental design

This study utilized a randomized complete block design, consistingof eight site preparation treatments replicated across five blocks for atotal of 40 experimental units. Study treatments were randomly as-signed to experimental units with dimensions of 55 m × 86 m(~0.5 ha), with 20 m buffers between units to prevent treatmentoverlap and reduce edge effects. Study treatments consisted of combi-nations of two vegetation control treatments (herbicide and chopping)and three soil manipulation treatments (mounding, bedding, and flat-planting [no additional treatment]) to create a 2 × 3 factorial design ofsix treatments. The two additional study treatments were an untreatedcontrol (no vegetation control and flat-planting) and a bedding treat-ment with both levels of vegetation control. The eight resulting studytreatments (Table 1) are herein referred to by their initials: F (flat-planting and no vegetation control), HF (herbicide and flat-planting),CF (chopping and flat-planting), HM (herbicide and mounding), CM(chopping and mounding), HB (herbicide and bedding), CB (choppingand bedding), and CHB (chopping, herbicide, and bedding).

Prior to site preparation, all blocks were harvested and sheared toremove standing vegetation. Study treatments were then applied insummer 2003. Vegetation control treatments were applied first, fol-lowed by the soil manipulation treatments. The chopping treatment wasimplemented using a 2.4 m Lucas Drum Chopper pulled by a TD15Dresser crawler tractor. The herbicide treatment included 0.70 kg/ha ofimazapyr and 0.56 kg/ha of triclopyr, which were mixed and broadcastat a rate of 280 L/ha. Mounds approximately 1.2 m wide were createdusing a New Forest Technology custom mounding bucket on aCaterpillar 320BL excavator. For consistency, the mounds were createdin rows, rather than in the discontinuous, random distribution that iscommonly associated with mounding site preparation. Beds approxi-mately 2.1–2.4 m wide were created using a Rome 6-disc BeddingHarrow with three discs on each side. All blocks were burned inOctober or November 2003 to remove remaining vegetation and furtherprepare the sites for planting. Container-grown longleaf pine seedlingsfrom locally collected seed were hand planted in December 2003. Therewas variation in initial planting density across treatments, particularlyon mounded treatments where space between mounds resulted in lessplanting space than bedded and flat-planted treatments. All blocks wereburned again in March 2006, and after that, two more burns occurred at4-7-year intervals, depending upon the block.

2.3. Data collection

After planting, a full census was conducted to record the number ofseedlings planted in each experimental unit. A sub-sample of 45 seed-lings from each experimental unit was then randomly selected for re-peated growth and survival measurement. Survival was measured

through two years after planting, with two measurements periods inAugust 2004 and 2005. Root collar diameter (RCD) was measuredthrough three years after planting, with four measurement periods oc-curring in June and December 2004, December 2005, and December2006 (see Knapp et al., 2006, 2008 for more details).

In summer 2018, a sample population of trees was selected fromeach experimental unit to quantify stand density and tree size. Fivecircular subplots were systematically located within each experimentalunit, with one subplot located at the experimental unit center and eachof the other four subplots spaced between each respective corner andthe center. Each subplot had a radius of 10 m, resulting in approxi-mately a third of each experimental unit being sampled. Within eachsubplot, every tree was identified to species and measured for heightand diameter at breast height (DBH). There was some volunteer seedingin of longleaf, slash, and pond pine (Pinus serotina Michx.); however,this was always a minor component of the developing stands and ouranalyses considered only the planted longleaf pine.

In summer and winter 2017, we conducted stem analysis on arandom sample of longleaf pine trees from each experimental unit toquantify annual height growth of individual trees. We established atransect running longways across the middle of each rectangular ex-perimental unit and located six equally spaced points along thetransect. We then selected the longleaf pine nearest each point for de-structive sampling for stem analysis. Each selected tree was felled in thefield and cut into sections at 10 cm intervals from the ground line to1.5 m and every 25 cm thereafter. These intervals gave us a fine re-solution toward the base of the tree, where transitions between growthyears were closer. Each sample was prepared by sanding with pro-gressively finer grit, following the protocol of the Missouri Tree-RingLaboratory (J. Marschall, personal communication, 2017), and tree agewas determined for each section (at known heights) using a stereomicroscope to count annual rings.

2.4. Data analysis

We fit linear mixed models to each stand-level response variable ofinterest (i.e., total height, DBH, trees per hectare, grass stage seedlingsper hectare, and basal area), with treatment as a fixed effect and thehierarchical, nested sampling design (i.e., subplots within experimentalunits within blocks) as a random effect. We used a height of 15 cm(from ground line to terminal bud) as the threshold to classify grassstage emergence (Boyer, 1988; Wang et al., 2016). We conducted one-way analysis of variance (ANOVA) on each of the fitted mixed modelsto test for differences among the eight study treatments for each re-sponse variable of interest. When one-way ANOVA indicated thattreatments significantly differed (α = 0.05), post-hoc Tukey-adjustedpairwise comparisons were used to determine which treatments sig-nificantly differed. Additionally, we used a 2 × 3 factorial ANOVA,excluding the F and CHB treatments, to determine the effect of in-dividual factors (e.g., chopping vs. herbicide) on each response variableof interest and to test for interactions between the vegetation controltreatments and soil manipulation treatments. With significant treat-ment or interaction effects, we used post-hoc Tukey-adjusted pairwisecomparisons to determine which individual factors significantly dif-fered, as appropriate.

To understand how longleaf pine responses to site preparationchanged through time, we integrated the stand-level measurementstaken 15 years after planting with data collected in the original study,which measured seedlings through three years after planting (from2003 through 2006). We computed two response variables of interest:survival and relative diameter. Survival in the initial two-year mea-surement period was calculated by applying the survival rates from sub-sampled seedlings to the number of planted seedlings recorded in theinitial census. Survival at 15 years was calculated by scaling longleafpine density within the subplots up to the experimental unit level andthen dividing those estimates by the number of seedlings planted in

Table 1Summary of site preparation treatments implemented in the study.

Treatment Vegetation Control Soil Manipulation

Chop Herbicide Flat Mound Bed

Flat [control] (F) XChop/flat (CF) X XHerbicide/flat (HF) X XChop/mound (CM) X XHerbicide/mound (HM) X XChop/bed (CB) X XHerbicide/bed (HB) X XChop/herbicide/bed (CHB) X X X

C.D. Crouch, et al. Forest Ecology and Management 461 (2020) 117928

3

each experimental unit. To integrate root collar diameter measurements(RCD) from 2003 to 2006 with DBH measurements recorded in 2018,we calculated the respective diameter (RCD or DBH) for each treatmentrelative to the untreated control (F). Relative diameter was calculatedfor each measurement period using the following equation:

=

−relative diameter treatment mean control meancontrol mean

Therefore, a treatment mean diameter that is twice as large as thecontrol mean diameter results in a relative diameter value of one. We fitlinear mixed models to both survival and relative diameter and thenused repeated measures ANOVA to test for differences among treat-ments and through time and to test for interactions between treatmentand time. Significant main effect and interaction terms (α= 0.05) fromrepeated measures ANOVA were followed by post-hoc Tukey-adjustedpairwise comparisons to determine which treatments and time periodssignificantly differed.

For the stem analysis data, we determined the age at each heightlevel by counting annual rings and adjusting internode tree heightsusing the Carmean method (Carmean, 1972; Dyer and Bailey, 1987).Tree height growth through time was reconstructed for the six sampledtrees per experimental unit. Patterns of mean height and mean annualheight growth through time were determined for each treatment, withnon-overlapping 95% confidence intervals indicating statistically sig-nificant differences in growth patterns. We computed two additionalresponse variables of interest: age of emergence from the grass stageand mean rate of height growth after emergence. Age of emergencefrom the grass stage was determined by the number of years a tree tookto reach 15 cm in height (Boyer, 1988; Wang et al., 2016). The rate ofgrowth after emergence was determined by computing the mean annualheight growth after the tree exceeded 15 cm in height. We fit linearmixed models for both age of emergence and post-emergence growthrate. We then conducted one-way ANOVA, followed by Tukey-adjustedpairwise comparisons, on both mixed models to test for differencesamong the eight study treatments. We also conducted a 2 × 3 factorialANOVA, excluding the F and CHB treatments and followed by Tukey-adjusted pairwise comparisons, to determine the effects of individualtreatment levels on the response variables.

Data were analyzed in R (R Core Team, 2018). The lme function inthe nlme package (Pinherio et al., 2018) and the glmer function in thelme4 package (Bates et al., 2015) were used for fitting linear mixedmodels, with the former being used in most cases and the latter beingused to specify a Poisson distribution where necessary. The anovafunction in the stats package (R Core Team, 2018) was used for runningANOVA tests. The lsmeans function from the emmeans package (Lenth,2019) and the CLD function from the multcompView package (Graveset al., 2015) were used for post-hoc, Tukey-adjusted pairwise compar-isons. When factorial or repeated measures ANOVA resulted in a sig-nificant (α = 0.05) interaction term, we used the testInteraction func-tion in the phia package (De Rosario-Martinez, 2015) to conduct post-hoc interaction significance tests. The dplyr package (Wickham et al.,2019) was used for data manipulation, the ggsurvplot function in thesurvminer package (Kassambara and Kosinski, 2018) was used forplotting survival, and the ggplot2 package (Wickham, 2016) was used tocreate all other figures.

3. Results

3.1. Stand-level effects of site preparation at year 15

Fifteen years after planting, site preparation resulted in significantdifferences in mean height (p < 0.001) and DBH (p < 0.001) amongthe eight study treatments (Fig. 1A and B). We found that CHB (chop,herbicide, and bed), HM (herbicide and mound), and HB (herbicide andbed) resulted in the greatest tree size at age 15, with mean heightsexceeding 5 m and mean DBH near 8 cm. These three treatments

resulted in significantly greater heights than HF (herbicide and flat-planting), CF (chop and flat-planting), and F (flat-planting and no ve-getation control) and significantly greater DBH than CF and F. In con-trast, CF and F resulted in the smallest trees, with mean heights of lessthan 2 m and mean DBH less than 5 cm. Based on factorial ANOVA,both soil manipulation and vegetation control treatments affectedheight (p < 0.001) and DBH (p ≤ 0.001) of planted longleaf pine(Table 2), and there was no significant interaction between the treat-ment types for height (p = 0.469) or DBH (p = 0.773). Bedding andmounding resulted in greater height and DBH than flat-planting (no soilmanipulation), with differences in height and DBH exceeding 2 m and2 cm, respectively. Herbicide resulted in greater height and DBH thanchopping, with differences in height and DBH exceeding 1 m and 1 cm,respectively.

Site preparation treatments also resulted in significant differences inlongleaf pine tree density (p < 0.001) and grass stage seedling density(p = 0.001) at age 15. HB, CHB, CB (chop and bed), and HF (herbicideand flat–planting) had greater than three times the number of trees/hacompared to F (Fig. 1C), while F resulted in more grass stage seedlings/ha than CHB and HM (Fig. 1D). F averaged 28 grass stage seedlings/ha,while grass stage seedlings were absent from all but one experimentalunit for CHB and HM. Based on factorial ANOVA, soil manipulation(p = 0.010) and vegetation control treatments (p = 0.026) affected thenumber of trees/ha, and there was no significant interaction betweenthe treatment types (p = 0.659). Bedding resulted in greater treedensity than mounding, and herbicide resulted in greater tree densitythan chopping 15 years after planting (Table 2). For grass stage seedlingdensity, differences among soil manipulation treatments were sig-nificantly different (p = 0.006), but there was no significant differencebetween vegetation control treatments (p = 0.264) and no significantinteraction between treatment types (p = 0.162). Flat-planting resultedin greater grass stage seedling density than mounding and bedding(Table 2).

Basal area, which integrates tree size and density at the stand level,was also significantly different among the eight study treatments(p < 0.001). Mean basal area for CBH and HB, which approached3.5 m2/ha, was significantly greater than that of CM, CF, and F, whichwere near or below 1 m2/ha (Fig. 1E). Soil manipulation (p = 0.009)and vegetation control treatments (p = 0.001) affected basal area, andthere was no significant interaction between the treatment types(p = 0.848). Bedding resulted in significantly greater basal area thanflat-planting, and herbicide resulted in significantly greater basal areathan chopping 15 years after planting (Table 2).

3.2. Effects of site preparation through time

Repeated measures ANOVA indicated that time significantly af-fected longleaf pine survival (p < 0.001) through 15 years afterplanting, while treatment did not (p = 0.281). Survival decreasedthrough time, and mean survival across all treatments was 72.5% ateight months after planting, 59.1% at 20 months, and 32.7% at15 years. The interaction between time and treatment approachedsignificance (p = 0.061), likely because survival did not significantlydiffer among treatments after the first and second growing seasons(Knapp et al., 2006) but did differ in year 15 according to one-wayANOVA (p = 0.002). At this measurement period, the herbicide treat-ments – HB, HM, HF, and CHB – resulted in higher survival compared tothe untreated control (Fig. 2).

The interaction between treatment and time was significant(p < 0.001) for relative diameter of longleaf pine through 15 yearsafter planting. There were no significant differences among treatmentsat seven or 12 months after planting (p ≥ 0.855), but there were dif-ferences at two, three, and 15 years after planting (p < 0.001). Aftertwo years, CHB and HB consistently had greater relative diameter thanCF and F (Fig. 3). By year 15, CHB, HM, HB, CB, and CM had sig-nificantly greater relative diameter than CF and F. Notably, there were

C.D. Crouch, et al. Forest Ecology and Management 461 (2020) 117928

4

no significant changes in relative diameter for any individual studytreatment after year two (p = 0.382).

3.3. Effects of site preparation on height growth reconstruction

One-way ANOVA indicated that there were significant differences inage of emergence from the grass stage among the eight study treatments(p < 0.001). CF, which had a mean emergence age of 9.25 years afterplanting, emerged significantly later than the other seven treatments(Fig. 4A). The other treatments had emergence ages ranging from 3.72to 5.29 years and did not significantly differ from each other. Based onfactorial ANOVA, soil manipulation (p < 0.001) and vegetation

control (p < 0.001) affected emergence age. The interaction betweenthe treatment types approached significance (p = 0.054) because CFresulted in much later emergence than any of the other treatments.Bedding and mounding resulted in earlier grass stage emergence thanflat-planting, while herbicide resulted in earlier emergence than chop-ping (Table 3).

One-way ANOVA also found significant differences in mean annualheight growth after grass stage emergence among the eight studytreatments (p < 0.001). HM and HB, with mean height growth ex-ceeding 0.5 m/year, grew significantly faster than F and CF, which hadmean growth rates of 0.33 m/year and 0.23 m/year, respectively(Fig. 4B). CF had significantly lower annual height growth than all

(A) (B)

(C) (D)

(E)

Fig. 1. Longleaf pine stand-level responses to site preparation study treatments 15 years after application. Treatments are defined in Table 1. Bars show treatmentmeans and standard errors; points show means for each experimental unit. Same letters indicate no significant difference among treatments within a responsevariable based upon Tukey-adjusted pairwise comparisons; p-values are from one-way ANOVA tests.

C.D. Crouch, et al. Forest Ecology and Management 461 (2020) 117928

5

treatments except for F and CM. Based on factorial ANOVA, soil ma-nipulation (p < 0.001) and vegetation control (p < 0.001) sig-nificantly impacted post-emergence growth, while the interaction be-tween the two treatment types was insignificant (p = 0.195). Beddingand mounding resulted in a greater annual growth rates than flat-planting, and herbicide produced greater annual growth than chopping(Table 3).

Stem analysis allowed us to reconstruct height growth of eachsampled longleaf pine through time (Fig. 5). The year in which eachtreatment line begins on the x-axis indicates the earliest grass stageemergence age of trees sampled in that treatment. All treatments weresignificantly taller through time than CF, due to that treatment’s lateemergence and slow growth. For the last four years of growth, HB wassignificantly taller than F, and during that same period, HM approachedsignificantly greater height than F. Stem analysis also allowed us toreconstruct mean annual height growth after grass stage emergence foreach treatment (Fig. 6). For years 1–3, HB had significantly greaterheight growth than F, while HM had significantly greater height growththan F during years 2–5. Annual height growth continued to increaseslightly over time for F, while growth generally levelled off after

Table 2Least square means of longleaf pine stand-level responses to individual sitepreparation treatment levels 15 years after application.

Treatment Meanheight (m)

Meandiameter atbreast height(cm)

Meantrees perhectare

Mean grassstageseedlings perhectare

Meanbasalarea(m2/ha)

Flat 2.59 b 5.77 b 383 ab 20.4 a 1.01 bMounding 4.79 a 7.79 a 321 b 8.3 b 1.75 abBedding 5.02 a 7.78 a 535 a 4.5b 2.55 ap-value < 0.001 <0.001 0.010 0.006 0.009Chopping 3.50 b 6.46 b 349 b 13.2 a 1.08 bHerbicide 4.76 a 7.76 a 477 a 8.9 a 2.46 ap-value < 0.001 0.001 0.026 0.264 0.001

Same letters indicate no significant difference within a treatment type and re-sponse variable based upon Tukey-adjusted pairwise comparisons; p-values arefrom 2 × 3 factorial ANOVA tests.

Fig. 2. Mean survival for each study treatment (defined in Table 1) at threepoints in time: 8 months, 20 months, and 15 years after planting.

Fig. 3. Diameter of each study treatment (defined in Table 1) relative to thecontrol (F) at five points in time, from seedling establishment into stand de-velopment.

(A)

(B)

Fig. 4. Tree-level responses to site preparation study treatments 15 years afterapplication. Treatments are defined in Table 1. Same letters indicate no sig-nificant difference among treatments within a response variable based uponTukey-adjusted pairwise comparisons; p-values are from one-way ANOVA tests.

Table 3Least square means of tree-level responses to individual site preparationtreatment levels 15 years after application.

Treatment Age of grass stage emergence Mean annual height growth (m/year)

Flat 7.31 a 0.330 bMounding 4.33 b 0.459 aBedding 4.48 b 0.486 ap-value < 0.001 <0.001Chopping 6.47 a 0.357 bHerbicide 4.27 b 0.493 ap-value < 0.001 <0.001

Same letters indicate no significant difference within a treatment type and re-sponse variable based upon Tukey-adjusted pairwise comparisons; p-values arefrom 2 × 3 factorial ANOVA tests.

C.D. Crouch, et al. Forest Ecology and Management 461 (2020) 117928

6

4–5 years for the other treatments.

4. Discussion

Our results show that site preparation had strong effects on longleafpine stand development through 15 years after treatment (Fig. 7).Among all treatments, the untreated control (F) generally resulted inthe smallest height and DBH, the lowest tree density, and the greatestnumber of seedlings still in the grass stage, although CF did not differfrom the untreated control. Thus, all methods of site preparation con-sidered in this study, aside from a single pass of chopping, improvedlong-term longleaf pine stand establishment compared to no treatment.Other studies have also found significantly improved growth of longleafpine seedlings (Addington et al., 2012; Freeman and Jose, 2009; Knappet al., 2006) with site preparation. Moreover, our results demonstrate apattern of increasing magnitude of response with increasing site pre-paration intensity, which is consistent with other studies that have

examined longleaf pine responses to vegetation control and soil ma-nipulation (Knapp et al., 2006; Loveless et al., 1989). Study treatmentsthat resulted in the greatest tree size and density included both herbi-cide and soil manipulation, while treatments that produced inter-mediate size and density included either herbicide or soil manipulation.The treatments that resulted in the smallest tree size and lowest density– CF and F – included neither herbicide nor soil manipulation.

Chopping was not effective for improving conditions for longleafpine establishment. This is consistent with other studies, which havefound that chopping resulted in reduced longleaf pine seedling growth(Freeman and Jose, 2009; Knapp et al., 2006) and increased woodyvegetation (Miller, 1980) compared to other vegetation control treat-ments. Although chopping causes minimal soil disturbance, has lowimpact on herbaceous vegetation, and immediately reduces standingwoody vegetation, a major drawback is that it promotes sprouting ofwoody species (Lowery and Gjerstad, 1991). Freeman and Jose (2009)suggested that chopping plus fire may be sufficient for longleaf pine

Fig. 5. Reconstruction of mean height growth after grass stage emergence for each study treatment (defined in Table 1). Individual treatment panels include a line forthe treatment mean and an envelope of the 95% confidence interval. The plot with all eight treatments shows only treatment means.

C.D. Crouch, et al. Forest Ecology and Management 461 (2020) 117928

7

establishment in flatwoods, even though seedling growth may be slow.However, our findings did not support this expectation through15 years after planting on similar flatwoods sites. Perhaps this con-trasting result is due to interactions between treatments and fire, but ananalysis of such interactions is beyond the scope of this study becausewe lack long-term, replicated data on fire behavior. The contrastingresult might also be due to a fire return interval that was too long tocontrol abundant woody competition in our study. The first burn oc-curred two years after site preparation, but the next two burns occurredat 4-7-year intervals, depending upon the block. This is much less fre-quent than the study sites’ estimated pre-settlement fire return intervalof 1–3 years (Frost, 2001). The need for frequent fire, especially duringrestoration on flatwoods sites, is emphasized because even a slight re-duction in frequency could allow shrub encroachment (Glitzenstein

et al., 2003).Our findings show that herbicide applied for site preparation can

have lasting improvements on growth and survival of planted longleafpine. Although previous studies have consistently observed increasedlongleaf pine seedling growth (Addington et al., 2012; Boyer, 1988;Freeman and Jose, 2009; Knapp et al., 2006; Loveless et al., 1989) andearlier grass stage emergence (Boyer, 1988; Freeman and Jose, 2009;Loveless et al., 1989) as a result of herbicide application for site pre-paration and release, few studies have reported responses beyond sixyears after planting. Additionally, previous studies have reported thatlongleaf pine seedling survival is either unaffected (Addington et al.,2012; Loveless et al., 1989) or reduced by herbicide (Boyer, 1988;Freeman and Jose, 2009) within the first few years following planting.Although survival did not differ among treatments through two years

Fig. 6. Mean annual height growth after grass stage emergence for each study treatment (defined in Table 1). Individual treatment panels include a line for thetreatment mean and an envelope of the 95% confidence interval. The plot with all eight treatments shows only treatment means. Lines are smoothed on a three-yearmean to reduce noise. Treatment lines are cut off after all five experimental units are no longer represented in the data.

C.D. Crouch, et al. Forest Ecology and Management 461 (2020) 117928

8

after planting in our study (Knapp et al., 2006), herbicide contributedto the highest survival among treatments by year 15. This finding em-phasizes the importance of a longer-term perspective when evaluatingresponses to site preparation and suggests that treatments that en-courage early grass emergence may be important for increasing long-term survival.

Soil manipulation treatments resulted in increased longleaf pinegrowth compared to flat-planting (no treatment) through 15 years aftertreatment, which is consistent with previous studies of early growth forslash (Haywood, 1987; Outcalt, 1984; Pritchett, 1979) and longleafpine seedlings (Knapp et al., 2006; Loveless et al., 1989). Survival anddensity responses to soil manipulation in our study were less clear.Bedding has been shown to increase survival of slash pine seedlings(Pritchett, 1979) but not of longleaf pine seedlings (Knapp et al., 2006;Loveless et al., 1989), while mounding has not been shown to increasesurvival of slash (Haywood, 1987) or longleaf pine seedlings (Knappet al., 2006). In our study, treatments with soil manipulation tended tohave higher survival in year 15 than the untreated control, yet onlythose treatments that included herbicide were significantly greater thanthe control. Stand density in year 15 on bedded sites was significantlygreater than mounded sites, but this was likely due to lower initialplanting density on mounds (mean = 886 trees/ha) compared to beds

(mean = 1,371 trees/ha) because survival in year 15 was similaramong bedded and mounded treatments.

Stem analysis indicated that herbicide and soil manipulation treat-ments provided both immediate (earlier grass stage emergence) andlong-term (increased growth rate after emergence) advantages forlongleaf pine establishment. The short-term benefit obtained from ve-getation control is consistent with our expectation that reduced com-petition results in greater seedling growth and earlier grass stageemergence (Knapp et al., 2006, 2008). However, competition controlfrom herbicide or mechanical treatments should diminish through timefollowing vegetation recovery. For example, Freeman and Jose (2009)found that shrub cover was similar between treated and untreated areasfour years after herbicide application on flatwoods sites in Florida.Boyer (1990) suggested that duration of the grass stage has a criticalimpact on subsequent longleaf pine growth and stand development,with delayed emergence resulting in lower long-term growth rates(Boyer, 1983). Twenty years after a single herbicide release treatment,Michael (1980) reported volume differences that represented an 8-yeargrowth advantage for longleaf pine on treated sites. Although this couldbe due to more years of growth from earlier grass stage emergence, ourresults suggest that early competition control results in long-term,sustained increases in growth rate. In contrast to the recovery of ve-getation following a single mechanical or chemical application, soilmanipulation treatments may have persistent effects for several decades(Brockway et al., 2015). Although mounds and beds are likely to settleover time (Londo and Mroz, 2001), these treatments may have lastingimpacts on hydrology. Given that the majority of the root system oflongleaf pine is within the upper 30 cm of soil (Boyer, 1990; Heyward,1933), it is possible that the drier conditions at the soil surface continueto directly benefit tree growth through time. Drier soil might also in-directly benefit longleaf pine by producing less optimal conditions forwoody competitors that thrive in wetter soils.

Our findings suggest that the growth rate after grass stage emer-gence is more important than the timing of emergence for determiningpatterns of stand development through year 15. For example, the un-treated control (F) did not differ from any treatment other than CF intiming of grass stage emergence, yet resulted in the fewest trees/ha andsmallest mean tree size. One explanation for this pattern may be thesurvivorship bias implicit in our sampling approach. By studying arandom sample of trees 15 years after planting, we were limited tostudying only those trees that were still alive at sampling, thus basingconclusions on successful trees. On the untreated control, which ex-perienced a high level of mortality during the study period, the stemanalysis results suggest that for trees to survive to age 15, they mustemerge from the grass stage early. If they fail to do so, competitors willpresumably overtake the grass stage seedlings, resulting in suppressionand eventual mortality. Because competition and poor drainage werenot improved on the control, post-emergence growth remained low fortrees that did manage to survive.

Several of our findings suggest that the more intensive site pre-paration treatments used in this study resulted in a Type A response(Morris and Lowery, 1988; Nilsson and Allen, 2003) through 15 yearsafter treatment. For example, survival did not differ among treatmentsduring early seedling establishment, but by year 15, survival rangedfrom 11.7% on F to 44.7% on HB. The results for growth differenceswere much more pronounced; differences in relative diameter amongtreatments were significant in year two and persisted at the samemagnitude through year 15. Finally, stem analysis indicated that annualheight growth remained consistently higher through time on treatedsites compared to F and CF. Long-term responses of longleaf pine to sitepreparation have not been well studied, but the Type A response ob-served in our study is consistent with the growth response of loblollypine through 18 years after intense site preparation reported by Nilssonand Allen (2003). In contrast, slash pine has generally exhibited a TypeC response after bedding on flatwoods sites (Outcalt, 1984; Zhao et al.,2009). Differences in annual growth rates observed in our study

Fig. 7. Photos to contrast the stand structure of (A) an untreated control, (B) aCB treatment, and (C) an HB treatment.

C.D. Crouch, et al. Forest Ecology and Management 461 (2020) 117928

9

magnify the cumulative differences in tree size among treatments,suggesting that differences among treatments are likely to persist orincrease.

5. Management implications

Longleaf pine management may encompass a range of specific ob-jectives and structural or compositional conditions. Site preparation iscommonly used in conjunction with timber production of othersouthern pines, and our results show that intensive site preparationresulted in the greatest growth and survival of longleaf pine. However,management objectives for longleaf pine are often different within thecontext of ecosystem restoration. Consequently, maintaining or en-hancing other attributes, such as ground flora communities, appropriatefire regimes, and soil properties, are also targeted during restoration.The natural community of our study area has been described as longleafpine wet savanna (Frost, 2001), which has relatively open foreststructure and low stand density. Our study shows that planting ap-proximately 1,050 trees/ha with no site preparation results in ap-proximately 125 longleaf pine trees/ha at year 15. Although the treeswere small (mean height = 1.2 m), the resulting stand may meet someobjectives of savanna restoration if the trees continue to grow andsurvive to maturity. Although the stand development phase would beprolonged, slow tree growth could be considered beneficial becauseslower growth is correlated with increased longevity (Black et al.,2008). An alternative restoration scenario could be optimizing longleafpine stand density and growth rates to minimize time required to satisfyred-cockaded woodpecker (Leuconotopicus borealis) habitat guidelines(Shaw and Long, 2007). In this scenario, site preparation may be usefulto improve the establishment and long-term growth of planted seed-lings on hydric sites.

Site preparation treatments are applied to address site-specificchallenges, and the effectiveness of the treatments used in our studywould be expected to vary across the wide range of site types on whichlongleaf pine is managed. On mesic and xeric sites, soil manipulationtreatments are likely unnecessary because these site types lack the highwater table that inhibits seedling growth and survival on hydric sites(Fox et al., 2007), although bedding has been shown to increasegrowth, but not survival, of longleaf pine seedlings on well-drained sites(Loveless et al., 1989). In contrast, herbicide is commonly used for sitepreparation on mesic and xeric sites because of longleaf pine’s intol-erance to competition (Brockway et al., 2006). Our findings supportprevious studies and suggest that herbicide site preparation improveslongleaf pine seedling growth regardless of site type (Addington et al.,2012; Boyer, 1988; Loveless et al., 1989).

Although longleaf pine would have naturally regenerated on hydricflatwoods sites under a regime of frequent fire (Frost, 2001; Peet,2006), loss of overstory longleaf pine and legacies of infrequent firehave disrupted ecosystem processes and created barriers to restorationsuccess. Feedbacks among longleaf pine dominance, fuel character-istics, and frequent fire contribute to ecosystem function. The disrup-tion of these feedbacks presents a potential ecological threshold forlongleaf pine restoration because conditions suitable for regenerationsuccess are limited within current, pre-restoration community states(Martin and Kirkman, 2009; Suding and Hobbs, 2009). Site preparation,as an example of intensive intervention during restoration, may benecessary to promote longleaf pine establishment and overcome thethreshold between the current and desired community states. Oncelongleaf pine is successfully established, the desired community statemay be maintained by re-establishing ecosystem function through lessintense management (e.g., frequent prescribed fire) (Martin andKirkman, 2009). Thus, site preparation treatments for longleaf pinerestoration must (1) result in adequate longleaf pine regeneration todevelop a future stand of desired structure and composition and (2)sustain a frequent fire regime that is continuous in both time and space(Mitchell et al., 2009) to maintain ecosystem function. An evaluation of

the study treatments’ impacts on the ecology of fire and fuels is beyondthe scope of this study; however, our study does demonstrate that sitepreparation significantly improves establishment and growth of long-leaf pine on hydric sites. We also found that survival of planted seed-lings is still possible without site preparation, although this approachresults in high mortality and slow growth.

Longleaf pine ecosystem restoration seeks to not only establish anoverstory of longleaf pine but also maintain or improve the groundlayer plant community (Walker and Silletti, 2006). Previous studieshave documented reductions in native understory plants following sitepreparation, which raises concerns about using such treatments forrestoration (Knapp et al., 2008; Litt et al., 2001; Pritchett, 1979; Schultzand Wilhite, 1974). For example, mechanical treatments reduce wire-grass populations, with slow recovery following disruption of the soil(Clewell, 1989). Deleterious impacts would be expected to be mostsevere in remnant, high-quality sites. However, the impact of site pre-paration may be less of a concern for restoration scenarios in which theunderstory community has already been degraded by past land use ormanagement activity. Our future work will address these issues byexploring long-term understory responses to site preparation withinthese plantations.

CRediT authorship contribution statement

Connor D. Crouch: Conceptualization, Formal analysis,Investigation, Writing - original draft, Writing - review & editing.Benjamin O. Knapp: Conceptualization, Methodology, Writing - re-view & editing, Project administration, Supervision, Funding acquisi-tion. Susan A. Cohen: Conceptualization, Writing - review & editing,Project administration. Michael C. Stambaugh: Methodology, Writing- review & editing. Joan L. Walker: Conceptualization, Methodology,Writing - review & editing. G. Geoff Wang: Conceptualization,Methodology, Writing - review & editing.

Declaration of Competing Interest

The authors declare that they have no known competing financialinterests or personal relationships that could have appeared to influ-ence the work reported in this paper.

Acknowledgments

Funding for this work was provided by the Strategic EnvironmentalResearch and Development Program (SERDP) for both the initial studyestablishment (Project SI-1303) and long-term remeasurement (ProjectRC-2706). We are immensely grateful for the support provided by CampLejeune’s Environmental Management Division throughout the project’sduration. We would also like to thank Kristy McAndrew, Max Farver,Joseph Gray, Jacob Hart, Bennett Wickenhauser, and Joe Marschall forthe field, technical, and lab support they provided. This paper is theTechnical Contribution No. 6819 of the Clemson University ExperimentStation.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.foreco.2020.117928.

References

Addington, R.N., Greene, T.A., Elmore, M.L., Prior, C.E., Harrison, W.C., 2012. Influenceof herbicide site preparation on longleaf pine ecosystem development and firemanagement. South. J. Appl. For. 36, 173–180. https://doi.org/10.5849/sjaf.11-012.

Barnhill, W.L., 1992. Soil Survey of Onslow County. North Carolina. U.S, Department ofAgriculture, Soil Conservation Service, Washington, DC.

Bates, D., Maechler, M., Bolker, B., Walker, S., 2015. Fitting linear mixed-effects modelsusing lme4. J. Stat. Softw. 67, 1–48. https://doi.org/10.18637/jss.v067.i01.

C.D. Crouch, et al. Forest Ecology and Management 461 (2020) 117928

10

Black, B.A., Colbert, J.J., Pederson, N., 2008. Relationships between radial growth ratesand lifespan within North American tree species. Ecoscience 15, 349–357. https://doi.org/10.2980/15-3-3149.

Boyer, W.D., 1990. Pinus palustris Mill. longleaf pine, in: Burns, R.M., Honkala, B.H. (Eds.), Silvics of North America, Vol. 1, Conifers. USDA Handbook 654. pp. 405–412.

Boyer, W.D., 1988. Effects of site preparation and release on the survival and growth ofplanted bare-root and container-grown longleaf pine (Georgia Forest Research PaperNo. 76). Research Division, Georgia Forestry Commission.

Boyer, W.D., 1983. Variations in height-over-age curves for young longleaf pine planta-tions. Forest Sci. 29, 15–27.

Brockway, D.G., Outcalt, K.W., Boyer, W.D., 2006. Longleaf pine regeneration ecologyand methods. In: Jose, S., Jokela, E.J., Miller, D.L. (Eds.), The Longleaf PineEcosystem: Ecology, Silviculture, and Restoration. Springer, New York, pp. 95–133.

Brockway, D.G., Outcalt, K.W., Tomczak, D.J., Johnson, E.E., 2015. Restoring longleafpine forest ecosystems in the southern U.S. In: Stanturf, J.A., Madsen, P. (Eds.),Restoration of Boreal and Temperate Forests. CRC Press, Boca Raton, pp. 501–519.

Carmean, W.H., 1972. Site index curves for upland oaks in the Central States. Forest Sci.18, 109–120. https://doi.org/10.1093/forestscience/18.2.109.

Clewell, A.F., 1989. Natural history of wiregrass (Aristida stricta Michx., Gramineae).Natural Areas J. 9, 223–233.

De Rosario-Martinez, H., 2015. phia. Post-Hoc Interaction Analysis.Dyer, M.E., Bailey, R.L., 1987. A test of six methods for estimating true heights from stem

analysis data. Forest Sci. 33, 3–13. https://doi.org/10.1093/forestscience/33.1.3.Fox, T.R., Jokela, E.J., Allen, H.L., 2007. The development of pine plantation silviculture

in the southern United States. J. Forest. 105, 337–347. https://doi.org/10.1093/jof/105.7.337.

Freeman, J.E., Jose, S., 2009. The role of herbicide in savanna restoration: Effects of shrubreduction treatments on the understory and overstory of a longleaf pine flatwoods.For. Ecol. Manage. 257, 978–986. https://doi.org/10.1016/j.foreco.2008.10.041.

Frost, C., 2006. History and future of the longleaf pine ecosystem. In: Jose, S., Jokela, E.J.,Miller, D.L. (Eds.), The Longleaf Pine Ecosystem: Ecology, Silviculture, andRestoration. Springer, New York, pp. 9–42.

Frost, C., 2001. Presettlement vegetation and natural fire regimes of Camp Lejeune. NorthCarolina Department of Agriculture, Plant Conservation Program, Raleigh, NC.

Glitzenstein, J.S., Streng, D.R., Wade, D.D., 2003. Fire frequency effects on longleaf pine(Pinus palustris P. Miller) vegetation in South Carolina and Northeast Florida, USA.Natural Areas Journal 23, 22–37.

Glitzenstein, J.S., Streng, D.R., Wade, D.D., Brubaker, J., 2001. Starting new populationsof longleaf pine ground-layer plants in the Outer Coastal Plain of South Carolina,USA. Natural Areas Journal 21, 89–110.

Graves, S., Piepho, H.-P., Selzer, L., 2015. multcompView: Visualizations. of PairedComparisons.

Haywood, J.D., 1987. Response of slash pine planted on mounds in central and south-western Louisiana. New Forest. 1, 291–300. https://doi.org/10.1007/BF00031739.

Heyward, F., 1933. The root system of longleaf pine on the deep sands of western Florida.Ecology 14, 136–148.

Hiers, J.K., Mitchell, R.J., Barnett, A., Walters, J.R., Mack, M., Williams, B., Sutter, R.,2012. The dynamic reference concept: measuring restoration success in a rapidlychanging no-analogue future. Ecological Restoration 30, 27–36. https://doi.org/10.3368/er.30.1.27.

Johnson, R., Gjerstad, D., 2006. Restoring the overstory of longleaf pine ecosystems. In:Jose, S., Jokela, E.J., Miller, D.L. (Eds.), The Longleaf Pine Ecosystem: Ecology,Silviculture, and Restoration. Springer, New York, pp. 271–295.

Jose, S., Ranasinghe, S., Ramsey, C.L., 2010. Longleaf pine (Pinus palustris P. Mill.) re-sotration using herbicides: overstory and understory vegetation responses on aCoastal Plain flatwoods site in Florida, U.S.A. Restor. Ecol. 18, 244–251. https://doi.org/10.1111/j.1526-100X.2008.00440.x.

Kassambara, A., Kosinski, M., 2018. survminer: Drawing Survival Curves using ggplot2.Kirkman, L.K., Jack, S.B., McIntyre, R.K., 2017. The fire forest of the past and present. In:

Kirkman, L.K., Jack, S.B. (Eds.), Ecological Restoration and Management of LongleafPine Forests. CRC Press, Boca Raton, pp. 3–15.

Knapp, B.O., Wang, G.G., Walker, J.L., 2008. Relating the survival and growth of plantedlongleaf pine seedlings to microsite conditions altered by site preparation treatments.For. Ecol. Manage. 255, 3768–3777. https://doi.org/10.1016/j.foreco.2008.03.013.

Knapp, B.O., Wang, G.G., Walker, J.L., Cohen, S., 2006. Effects of site preparationtreatments on early growth and survival of planted longleaf pine (Pinus palustrisMill.) seedlings in North Carolina. For. Ecol. Manage. 226, 122–128. https://doi.org/10.1016/j.foreco.2006.01.029.

Landers, J.L., Van Lear, D.H., Boyer, W.D., 1995. The longleaf pine forests of theSoutheast: Requiem or renaissance? J. Forest. 93, 39–44. https://doi.org/10.1093/jof/93.11.38.

Lenth, R., 2019. emmeans: Estimated Marginal Means, aka. Least-Squares Means.Litt, A.R., Herring, B.J., Provencher, L., 2001. Herbicide effects on ground-layer vegeta-

tion in southern pinelands, USA: A review. Natural Areas Journal 21, 177–188.Londo, A.J., Mroz, G.D., 2001. Bucket mounding as a mechanical site preparation tech-

nique in wetlands. North. J. Appl. For. 18, 7–13. https://doi.org/10.1093/njaf/18.1.7.

Loveless, R.W., Pait III, J.A., McElwain, T., 1989. Response of longleaf pine to varyingintensity of silvicultural treatments, in: Miller, J.H. (Ed.), Proceedings of the FifthBiennial Southern Silvicultural Research Conference. Gen. Tech. Rep. SO-74. U.S.Department of Agriculture, Forest Service, Southern Research Station, New Orleans,LA, pp. 159–164.

Lowery, R.F., Gjerstad, D.H., 1991. Chemical and mechanical site preparation. In: Duryea,M.L., Dougherty, P.M. (Eds.), Forest Regeneration Manual, Forestry Sciences.

Springer, Dordrecht, pp. 251–261.Martin, K.L., Kirkman, L.K., 2009. Management of ecological thresholds to re-establish

disturbance-maintained herbaceous wetlands of the south-eastern USA. J. Appl. Ecol.46, 906–914. https://doi.org/10.1111/j.1365-2664.2009.01659.x.

MCBCL, 2015. 2015–2020 Integrated Natural Resource Management Plan. US MarineCorps, Camp Lejeune, NC.

McCaskill, G.L., Jose, S., 2012. The ecological classification of coastal wet longleaf pine(Pinus palustris) of Florida from reference conditions. Am. J. Plant Sci. 3, 1205–1218.https://doi.org/10.4236/ajps.2012.39146.

W.H. McNab D.T. Cleland J.A. Freeouf J.E. Keys Jr. G.J. Nowacki C.A. CarpenterDescription of ecological subregions: Sections of the conterminous United States(General Technical Report No. WO-76B). United States Department of AgricultureForest Service 2007 Washington, DC.

Michael, J.L., 1980. Long-term impact of aerial application of 2,4,5-T to longleaf pine(Pinus palustris). Weed Sci. 28, 255–257.

Miller, J.H., 1980. Competition after windrowing or single-roller chopping for site pre-paration in the southern piedmont. Proc. South Weed Sci. Soc. 33, 139–245.

Mitchell, R.J., Hiers, J.K., O’Brien, J., Starr, G., 2009. Ecological forestry in the Southeast:understanding the ecology of fuels. J. Forest. 107, 391–397. https://doi.org/10.1093/jof/107.8.391.

Mitchell, R.J., Hiers, J.K., O’Brien, J.J., Jack, S.B., Engstrom, R.T., 2006. Silviculture thatsustains: The nexus between silviculture, frequent prescribed fire, and conservationof biodiversity in longleaf pine forests of the southeastern United States. Can. J. For.Res. 36, 2724–2736. https://doi.org/10.1139/X06-100.

Morris, L.A., Lowery, R.F., 1988. Influence of site preparation on soil conditions affectingstand establishment and tree growth. South. J. Appl. For. 12, 170–178. https://doi.org/10.1093/sjaf/12.3.170.

Nilsson, U., Allen, H.L., 2003. Short- and long-term effects of site preparation, fertiliza-tion, and vegetation control on growth and stand development of planted loblollypine. For. Ecol. Manage. 175, 367–377. https://doi.org/10.1016/S0378-1127(02)00140-8.

NRCS, 2014. Official Soil Description-Leon Series. Natural Resources ConservationService, USDA.

Oswalt, C.M., Cooper, J.A., Brockway, D.G., Brooks, H.W., Walker, J.L., Connor, K.F.,Oswalt, S.N., Conner, R.C., 2012. History and current condition of longleaf pine in thesouthern United States (General Technical Report No. SRS-166). USDA ForestService, Southern Research Station, Asheville, NC.

Outcalt, K.W., 1984. Influence of bed height on the growth of slash and loblolly pine on aLeon fine sand in northeast Florida. South. J. Appl. For. 8, 29–31. https://doi.org/10.1093/sjaf/8.1.29.

Peet, R.K., 2006. Ecological classification of longleaf pine woodlands. In: Jose, S., Jokela,E.J., Miller, D.L. (Eds.), The Longleaf Pine Ecosystem: Ecology, Silviculture, andRestoration. Springer, New York, pp. 51–93.

Pinherio, J., Bates, D., DebRoy, S., Sarkar, D., Core Team, R., 2018. nlme: Linear andNonlinear. Mixed Effects Models.

Pritchett, W.L., 1979. Site preparation and fertilization of slash pine on a wet savannasoil. South. J. Appl. For. 3, 86–90. https://doi.org/10.1093/sjaf/3.3.86.

Core Team, R., 2018. R: A Language and Environment for Statistical Computing. RFoundation for Statistical Computing, Vienna, Austria.

Schultz, R.P., Wilhite, L.P., 1974. Changes in a flatwoods site following intensive pre-paration. Forest Science 20, 230–237.

Shaw, J.D., Long, J.N., 2007. A density management diagram for longleaf pine standswith application to red-cockaded woodpecker habitat. South. J. Appl. For. 31, 28–38.https://doi.org/10.1093/sjaf/31.1.28.

Suding, K.N., Hobbs, R.J., 2009. Threshold models in restoration and conservation: adeveloping framework. Trends Ecol. Evol. 24, 271–279.

Walker, J., 1998. Ground layer vegetation in longleaf pine landscapes: An overview forrestoration management, in: Proceedings of the Longleaf Pine Ecosystem RestorationSymposium. Longleaf Alliance Report No. 3. pp. 2–13.

Walker, J., 1993. Rare vascular plant taxa associated with the longleaf pine ecosystems:patterns in taxonomy and ecology, in: Hermann, S.M. (Ed.), Proceedings of the TallTimbers Fire Ecology Conference, No. 18. The Longleaf Pine Ecosystem: Ecology,Restoration and Management. Tall Timbers Research Station, Tallahassee, FL, pp.105–125.

Walker, J., Peet, R.K., 1983. Composition and species diversity of pine-wiregrass savannasof the Green Swamp, North Carolina. Vegetatio 55, 163–179. https://doi.org/10.1007/BF00045019.

Walker, J.L., Silletti, A.M., 2006. Restoring the ground layer of longleaf pine ecosystems.In: Jose, S., Jokela, E.J., Miller, D.L. (Eds.), The Longleaf Pine Ecosystem: Ecology,Silviculture, and Restoration. Springer, New York, pp. 297–325.

Wang, G.G., Pile, L.S., Knapp, B.O., Hu, H., 2016. Longleaf pine adaptation to fire: Is earlyheight growth pattern critical to fire survival?, in: Schweitzer, C.J., Clatterbuck, W.K.,Oswalt, C.M. (Eds.), Proceedings of the 18th Biennial Southern Silvicultural ResearchConference. Gen. Tech. Rep. SRS-212. U.S. Department of Agriculture, Forest Service,Southern Research Station, Asheville, NC, pp. 214–218.

Wickham, H., 2016. ggplot2: Elegant Graphics for Data Analysis. Springer-Verlag, NewYork.

Wickham, H., Francois, R., Henry, L., Muller, K., 2019. dplyr: A Grammar. of DataManipulation.

Zhao, D., Kane, M., Borders, B., Harrison, M., 2009. Long-term effects of site preparationtreatments, complete competition control, and repeated fertilization on growth ofslash pine plantations in the flatwoods of the southeastern United States. ForestScience 55, 403–410.

C.D. Crouch, et al. Forest Ecology and Management 461 (2020) 117928

11