Herbicide control of swamp titi (Cyrilla racemiflora) in restoration of ephemeral wetlands

Swamp titi (Cyrilla racemiflora L.) is a vigorous native shrub or small tree found in poorly drained areas of the southeastern U.S. Coastal Plain. In the absence of fire, it creates dense thickets that prevent herbaceous groundcover in ephemeral wetlands, degrading habitat for uncommon and endangered amphibians and other genera. In our study, aquatic herbicides and methods of application were examined at two forests in north Florida for restoration of ephemeral wetland habitat to benefit wetland‐breeding amphibians. Glyphosate, imazapyr, and combinations of glyphosate and imazapyr were tested in four study protocols that included an operational standard (triclopyr), treatment method control, and untreated control where appropriate. Results of foliar protocols (targeting rootstocks less than 1.5 m tall) demonstrated that glyphosate provided better titi control than imazapyr (average of 79% vs. 12% mortality), there was no strong improvement with the addition of imazapyr to glyphosate (84% mortality), glyphosate treatments had titi mortality as high or higher than the triclopyr operational treatment (57% mortality), and that methylated seed oil surfactant improved glyphosate efficacy but might have been antagonistic for combination treatments. Glyphosate and imazapyr cut stump treatments did not provide higher efficacy than the triclopyr standard treatment (60% mortality). Cut stem treatments (targeting rootstocks greater than 1.5 m tall) provided the highest level of control averaging 97, 90, 97, and 80% mortality for glyphosate, imazapyr, combination treatments, and triclopyr standard, respectively. Only directed foliar sprays reduced total groundcover and graminoid cover, with triclopyr treatment giving the least reduction of desired herbaceous groundcover.


Introduction
Titi is a vigorous native shrub or small tree widely dispersed on somewhat poorly drained and poorly drained sites of the U.S. southeastern Coastal Plain.Three species are found in the region, swamp titi (Cyrilla racemiflora L.), black titi (Cliftonia monophylla (Lam) Britton ex Sarg), and littleleaf titi (Cyrilla parvifoia Rav) (USDA-NRCS 2022a).Swamp titi, the subject of our study, is most widespread and is a challenge to many restoration and reforestation efforts.In the absence of fire or grazing titi can form dense thickets (Petrick & Pertrick 1987) which limit herbaceous cover preferred by rare or imperiled amphibians such as the flatwoods salamanders (Ambystoma cingulatum and bishopi), and Florida bog frog (Lithobates okaloosae) (Sekerak et al. 1996;Gorman 2009;Gorman et al. 2014).Dense thickets of titi produce a thick duff layer of leaf litter which can modify wetland hydrology and/or prevent associated species from reproducing.Consequently, titi thickets can alter structure and composition of wetlands and their ecosystem function over time (Facelli 1991;Lamont 1995).
Direct control measures using hand cutting, triclopyr amine herbicide, and prescribed fire have generally not been effective in reducing titi in wetlands in a single application.Repeated reapplication of fire and/or herbicide over an extended period is often required to effectively reduce titi density.Even when aboveground portions die-back, re-sprouting occurs from the root collar of stumps and the extensive and interconnected root system.In addition, titi has rapid seed spread rate (USDA-NRCS 2022a).Herbicide control of sprouting species is most effective with herbicides such as glyphosate and imazapyr which are preferentially translocated to the roots and meristems (Shaner 2014).
However, aquatic and terrestrial amphibians are sensitive to herbicides (Relyea 2005) and herbicides may be especially lethal to larvae (Allran & Karasov 2001;Wojtaszek et al. 2004;Wagner et al. 2013), whereas juveniles and adults may be more tolerant (Greenberg et al. 2016;McFall et al. 2023).Therefore, breeding ponds should not be disturbed during critical habitat periods, which are between October 1 and May 1 for both species of flatwood salamanders, and between April 15 and August 15 for Florida bog frogs (Sekerak et al. 1996;Gorman 2009;Gorman et al. 2014).To avoid these critical periods herbicides are applied in early September.
Control of titi in reforestation activities in the region commonly utilizes broadcast applications of imazapyr alone or in combination with triclopyr or glyphosate (Lauer et al. 2006).Titi control is not complete but adequate to establish pine plantations, often followed by prescribed burning when pines grow above the titi canopy.We found no published research regarding selective herbicides for titi control in restoration activities.
The objectives of most restoration activities are not consistent with broadcast herbicide applications which may damage or kill desired vegetation, though in some cases the high degree of disturbance may foster desired pioneer species or plants tolerant to a particular herbicide (Harrington et al. 2021).Another concern is that imazapyr is a persistent, broad-spectrum herbicide; however, conifers, legumes, Rubus spp., and other desirable species are tolerant (Harrington et al. 1998;Minogue & Quicke 1999).Glyphosate and triclopyr are strongly adsorbed by soil such that plant uptake by roots is negligible, which provides a measure of plant selectivity.However, glyphosate is broad spectrum and selectivity is obtained by carefully applying the herbicide only to target plants.Triclopyr amine (Garlon 3A) is recommended for shrub control in many restoration activities because of selectivity to nontarget grasses (Klaus & Klaus 2011;DiAllesandro 2012;GeFellers et al. 2020).In addition, triclopyr applications to control midstory forest vegetation had no effect on salamander capture rate (Greenberg et al. 2016) or juvenile frog dispersal traits following short-term exposure (McFall et al. 2023).
Titi has a thick, waxy cuticle which limits herbicide absorption.Glyphosate is lipophobic and movement though the cuticle is very limited in woody plants, especially shrubs with a waxy cuticle such as titi (Green et al. 1992).Addition of the adjuvant methylated seed oil (MSO) improves the efficacy of glyphosate and imazapyr due to improved foliar uptake in the control of difficult species such as cogongrass (Imperata cylindrica (L.) P. Beauv.) (Ramsey et al. 2012) and MSO is recommended for titi applications, but we found no research supporting its use for woody plants.Triclopyr is the most lipophilic among these herbicides (Shaner 2014) and is commonly used for titi control or other waxy leaf shrubs in the region (Cristan et al. 2019).
Three approaches to individual stem treatments are currently used in forestry and restoration efforts, directed foliar spray, cut stump, and cut stem (Zedaker et al. 1987;Miller et al. 2010;Harrington et al. 2021).Foliar sprays are chosen when crowns are less than 1.5 m tall and can be sprayed effectively by backpack crews.One disadvantage with this approach is that foliar sprays are likely to contact nontarget plants, and the risk of damaging desirable vegetation increases with stem height, so this method is commonly limited to small rootstocks.For rootstocks with stems greater than 1.5 m tall herbicides are applied by cut stem (hack and squirt) or cut stump methods.In cut stem treatment a hatchet or similar tool is used to cut though the cambium at a downward angle near the tree base and a measured amount of aqueous herbicide solution is placed into the cut reservoir.In cut stump treatment concentrated or diluted herbicide solution or emulsion is applied to the cambium of freshly cut stumps.Cut stump applications are used when high titi stem density necessitates stem removal.Cut stump treatments may be used to control stems of any height but are labor intensive.Cut stem and cut stump treatments are very selective since herbicides are applied only to a limited surface of the target vegetation.
Here, our objective was to identify aquatic herbicide treatments that provide swamp titi control while minimizing adverse impacts to rare amphibian habitat.Our approach was to use individual stem treatments using relatively high rates of the nonresidual herbicide glyphosate with low rates of the persistent herbicide imazapyr in comparison to standard triclopyr amine treatments.A secondary objective was to assess the benefit for inclusion of MSO for foliar applications of these herbicides.We evaluated many treatments to determine effective herbicides, rates, and methods of application, leaving a need for further research to examine direct effects on amphibians and other taxa when using the most effective approaches.
Uniform studies were established in ephemeral wetlands under active amphibian restoration activities at the Apalachicola National Forest (ANF) and The Nature Conservancy (TNC) Flint Rock Wildlife Management Area in north Florida.The ANF study (lat 30 N, long 85 W) in Liberty County was on Rains and Bladen soil series, typically found in drainageways on fluviomarine terraces and flats.Both series are characterized by poorly drained, hydric, sandy loam A and Eq overlying a sandy clay loam Btg horizon.The ecological site is East Central Sandy Flat-F152AY320FL (USDA-NRCS 2022b).The TNC study (lat 30 N, long 84 W) in Jefferson County was on Nutall-Tooles complex soils, found in flats on marine terraces.These soils are characterized by poorly drained, hydric, fine sand A and E over sandy clay loam Btg horizons.The ecological site is East Karst Flat-F152AY500FL.

Uniform Experimental Design
In all studies at both locations the treatments were replicated 10 times in a completely randomized design with single rootstocks serving as the experimental unit.A minimum 3 m buffer was maintained between treatment rootstocks in the cut stump and cut stem studies, whereas a 2 m minimum buffer was used in the directed foliar spray studies in which small rootstocks were treated.In all studies, we selected treatment rootstocks that appeared to be individuals, though root connectivity is common in titi.The relevant initial titi rootstock characteristics are listed in Table S1.

Directed Foliar Studies
At both study locations, two directed spray studies examined applications to shrub form titi (less than 1.5 m tall).Directed foliar spray Study 1 tested all herbicides in MSO emulsion (1.0%Premium MSO product, Helena Agri-Enterprises, Collierville, TN, U.S.A.) to compare rate response for glyphosate (5, 7.5, 10% Rodeo product, Corteva Agriscience, Indianapolis, IN, U.S.A.) and imazapyr (1, 2, 3% Habitat product, BASF, Research Park, NC, U.S.A.) applied alone, or when applied in varying proportions of these two herbicides (7.5% Rodeo plus 0.5, 1, or 2% Habitat) as compared to standard triclopyr amine reference treatment (4% Garlon 3A product, Corteva Agriscience, Indianapolis, IN, U.S.A.) and a untreated control.Directed foliar spray Study 2 tested the efficacy of the middle rates of glyphosate (7.5% Rodeo) and imazapyr (2% Habitat) with or without MSO surfactant (1% Premium MSO), for these two herbicides applied alone or in combination (7.5% Rodeo plus 1% Habitat), as compared to MSO only and untreated control.Individual titi rootstocks were tagged with numbered metal tags and height of tallest stem was measured to 1 cm as a pretreatment baseline.Herbicides were applied using backpack sprayers equipped with adjustable cone nozzles to direct the spray to individual titi crowns in early September 2018.Titi rootstocks were remeasured 1 and 2 years after treatment (YAT) in late August and September of 2019 and 2020 to assess phytotoxicity code (0 none to 5 mortality) and live height (cm) of the tallest stem.

Cut Stump Study
At both locations, cut stump applications tested rate response of glyphosate (50, 75, 100% Rodeo), imazapyr (25, 50, 75% Habitat), or when applied in varying proportions of these two herbicides (75% Rodeo plus 25% Habitat, 50% Rodeo plus 50% Habitat, 25% Rodeo plus 75% Habitat) as compared to standard triclopyr amine standard treatment (50% Garlon 3A), cut only method control, and a untreated control.Prior to treatment, individual titi rootstocks were tagged with numbered metal tags and baseline live height (cm) of tallest stem and all stem diameters (0.1 cm) at 15 cm height, excluding those stems less than 0.5 cm diameter.In September 2018, both shrub and tree form titi study rootstocks were cut at 15 cm height and herbicide treatments were applied to the freshly cut stump surface using a paint brush to treat the cambium and cut edge of bark.In late August and early September of 2019 and 2020 we remeasured titi rootstocks to obtain number of live stems per rootstock and all rootstock live stem diameters at 15 cm height, excluding those stems less than 0.5 cm diameter.

Cut Stem Study
The cut stem (hack and squirt) study was done only at the ANF location where tree-form titi rootstocks were abundant.This study tested the same treatments as in the cut stump study comparing various concentrations of glyphosate, imazapyr, and their combination, standard triclopyr amine treatment, overlapping stem cuts without herbicide (cut only method control), and an untreated control.Herbicides were applied in September 2018 using a calibrated squirt bottle to apply 1 mL herbicide solution to cuts (hacks) made with a hatchet at a downward angle on standing tree form titi stems, leaving a small reservoir in which herbicide solutions were applied.In accordance with product herbicide labels, the Rodeo and Rodeo plus Habitat treatments were placed in cuts 5-7.6 cm apart (2-3 in) at 15 cm stem height and Habitat treatments were placed with one cut per 7.6 cm stem diameter at 15 cm stem height.The cut stem study followed the same titi rootstock measurement procedures as the cut stump study.

Impacts to Groundcover
Impacts to groundcover were evaluated only at the ANF location because areas of the TNC location were inundated at time of scheduled herbicide application.Groundcover at ANF was discontinuous and predominantly perennial grasses.We proposed to assess herbicide impacts to ground vegetation in all studies but herbicide phytotoxicity was only observed in the directed foliar spray studies.Results were similar in the two foliar studies, and we report groundcover results for the more robust directed foliar Study 1. Groundcover was assessed prior to treatment and at 1 and 2 YAT in plots defined as the area below the extent of each rootstock crown (within the drip line).Assessments included the percentage of plots with groundcover, number of plots with no damage, percent cover reduction of graminoids, forbs, and total cover, and number of critical species, that is, plants important to ephemeral wetland-breeding amphibian habitat (Sekerak et al. 1996;Gorman et al. 2014; P. Hill 2019, Florida Wildlife Commission, Fish and Wildlife Research Institute, Joe Budd Field Office, Gadsden County, FL, personal communication).The common critical species observed included bunch grasses such as Aristida spp., Rhynchospora spp., Dichanthelium spp., and Scleria sp., and forb species including Eriocaulon sp., Xyris spp., Lobelia floridana Chapm., and Helenium pinnatifidum (Schwein.ex Nutt.)Rydb.

Statistical Analysis
Inclusion of location as a random effect and measures of efficacy differed by study.The directed foliar Study 1, directed foliar Study 2, and cut stump study were performed at two locations with location included in the analysis of variance (ANOVA) as a random effect.The cut stem study was only performed at one location.Evaluation of treatment efficacy on swamp titi differed for the foliar, cut stump, and cut stem studies as related to the relevant posttreatment measurements.Treatment efficacy for titi control 2 YAT was evaluated in terms of rootstock mortality for all studies, live height control (change from baseline) for the directed foliar studies, live rootstock basal area for both cut stump and cut stem studies, and resprout number and height for the cut stump study.
The ANOVA for mortality was performed using the logistic distribution with live/dead treated as a binary random variable for all studies.Mortality models included initial rootstock height as a significant covariate for the directed foliar with MSO Study 1 (p = 0.023) but not for the directed foliar MSO comparison Study 2 (p = 0.874), and initial stem count (p = 0.049) as a covariate for the cut stump study.The analysis was modified to exclude the untreated control in the directed spray MSO Study 2 and the cut-only treatment in the cut stump study because these reference treatments had no mortality.Only a main effects analysis of herbicide active ingredients was possible for the cut stem study due to the high level of control for these treatments.
The analysis of live height control (directed foliar studies) was performed using proportion height control calculated as PH 2YAT = rootstock height 2YAT/pre-treatment rootstock height.The square root transformation (1 + PH 2YAT ) 0.5 was used to improve residual homogeneity (PH 2YAT can be greater than 1).An R-sided covariance structure (Littell et al. 2006) with treatments grouped into 2 groups based on Akaike's Information Criteria (AIC) was included because variation was low for efficacious treatments even after transformation.Percent height control means were calculated as PC 2YAT = 100 Â (1 À PH 2YAT ).
The analysis of rootstock basal area for the cut stump and cut stem studies accounted for heterogeneity of variance using treatment groups based on graphical examination of residuals and AIC.Rootstock basal area was computed as the sum of crosssectional area of measured stems greater than 0.5 cm and crosssectional area of resprouts (<0.5 cm) assuming resprouts had a diameter of 0.25 cm.Quadratic mean rootstock diameter (Dq) was computed as the diameter for a stem with equivalent rootstock basal area to test as a covariate and for use as a size reference that is easier to visualize.Initial stems/rootstock was a significant covariate (p < 0.001) for the cut stump study.Mean comparisons were performed using rootstock basal area but this was also presented in the results for the cut stump study in terms of percent control calculated as 100 Â (1 À treatment basal area/ cut untreated control basal area) for a given level of the covariate.
The cut stump study ANOVA for resprout count used a negative binomial model with log link and log of initial stem count as an offset.This compares treatments with respect to rate of resprouting per stem in a rootstock.A square root transformation was used for cut stump study resprout height to account for increasing variation with height.Grouping treatments with respect to variation based on AIC was used to address heterogeneity of treatment variance for both variables.
In the analysis of impacts to groundcover, the reduction in forb cover was not analyzed because forbs were mostly absent.The analysis of percent reduction was performed using the arcsine-square root transformation for normality.The analysis of species counts in the plots used a negative binomial model.The analysis of percent of plots with given species or no damage used a logistic model with presence of species or absence of damage considered a binomial random variable.
Statistical analysis was performed using SAS/STAT software version 9.4 of the SAS System for Windows (SAS Institute Inc. 2013).The SAS GLIMMIX procedure was used to perform analysis of variance due to the ease of producing the necessary output as well as the need to fit logistic models and specify R-sided covariance structures to address heterogeneity of variance (Littell et al. 2006).Orthogonal contrast statements were used to partition treatment effects (Milliken & Johnson 1992) and to partition effects to test hypotheses as appropriate for each study.Several nonorthogonal comparisons were included because they were of interest, such as comparisons to the untreated experimental control or Garlon 3A reference treatments.Transformed means are reported using back transformation and approximate standard errors as described in Jørgensen and Pedersen (1998).Treatment means and averages of treatment groups were compared using Tukey's or Bonferroni's adjustments for multiplicity as noted with a significance level of 5%.Multiple comparisons of individual treatment to reference treatments used the Dunnett-Hsu adjustment for multiplicity.

Directed Foliar Spray Studies
Directed foliar spray Study 1 indicated no significant Rodeo or Habitat rate effects for swamp titi rootstock mortality or height control (ANOVA not shown).Averaged over the three Rodeo rates, titi mortality was 79% compared to 4% for the untreated control and height control was 91% compared to À17% for untreated control (negative height control indicates height growth) (Table 1).Averaged over the three Habitat rates, titi mortality (12%) and height control (20%) were significantly less than averages for the other herbicide treatments.Mortality and height control for the three rates of Habitat did not differ from the untreated control.There were no significant Habitat rate effects when combined with 7.5% Rodeo.Rodeo or Rodeo plus Habitat treatments achieved 68-91% mortality and 87-94% height control, but did not differ from 57 and 66%, respectively, observed for the Garlon 3A reference standard.Adding Habitat to Rodeo did not improve titi mortality or height control over Rodeo.There was also no evidence of antagonism.The Garlon 3A standard had greater height control than the untreated control, but titi mortality did not differ.
Results from directed foliar spray Study 2 to determine the effects of MSO surfactant addition indicated significant interactions ( p < 0.005) for both swamp titi mortality and height control including herbicide treatment Â MSO levels and herbicide treatments with Rodeo Â MSO levels (ANOVA not shown).These interactions identify that the low efficacy of Habitat was not improved by MSO, that Rodeo efficacy was improved by MSO, and that MSO was antagonistic for the Rodeo plus Habitat combination (Table 2).This antagonism is inconsistent with foliar spray Study 1 which had 84% mortality and 93% height control for the Rodeo 7.5% plus Habitat 1% with MSO treatment compared to 55% mortality and 81% height control for this treatment in Study 2. Field records were verified to be certain they were correct, and we cannot explain why MSO was antagonistic with the combination treatment in Study 2, but this was observed at both study locations.Rodeo plus MSO and Rodeo Table 1.Comparison of mean swamp titi rootstock mortality and height control with approximate standard errors (SE) 2 years after treatment for directed foliar Study 1 testing directed foliar spray applications that included 1% methylated seed oil (MSO) at two forest sites in north Florida.Mortality means are predicted mortality at the covariate average initial rootstock height of 95 cm (initial heights ranged from 28 to 150 cm).Effect means followed by the same letter within each grouping are not significantly different at the 5% level using Bonferroni's adjustment for multiplicity (Bon).Treatment means followed by the same letter within a column are not significantly different at the 5% level using Tukey's adjustment for multiplicity.Asterisk (*) indicates treatment mean is significantly different than the untreated control using Dunnett-Hsu adjustment at the 5% level.

Mortality (%)
Height Highest mortality (60%) was achieved for the Garlon 3A standard treatment.Titi rootstock basal area was significantly less than the untreated control for all herbicide treatments.Basal area was significantly less than the cut only method control for all but the low-rate Rodeo treatment.Percent control (PC) in Table 3 was calculated using the cut only method control as PC = 100 (1 À treatment basal area/cut only method control basal area).In terms of percent control relative to cut only, Garlon 3A had 87% control compared to 64-79% control for the other significant herbicide treatments.The Garlon 3A standard treatment was the only herbicide treatment that significantly reduced resprout count compared to cut only control (Table 4).There were no significant treatment differences for the highly variable early resprout height.

Cut Stem
The cut stem study was evaluated in terms of swamp titi rootstock mortality and surviving rootstock basal area 2 YAT (Table S2).
The high level of mortality required a reduced logistic model in which treatments were grouped by herbicide active ingredient with no significant differences among the groups (80-97% mortality) but all differed from the untreated control (0%) and cut-only method control (20%).Rootstock basal area did not differ for herbicide treatments (0-0.1 cm 2 ) and all herbicide treatments differed from the untreated control (84.52 cm 2 ) and cut-only method control (31.32 cm 2 ).Rootstock height responses were not analyzed because of the overall high mortality from these treatments.

Impacts to Groundcover
Treatments were compared with respect to the percent of plots with groundcover present and percent of plots with no damage to groundcover after application for foliar Study 1 at both 1 and 2 YAT assessments.Percent of plots with groundcover ranged from 20 to 70% and treatments or treatment effects did not differ (data not shown).The percentage of plots with no damage to groundcover did not change in any treatment from 1YAT to 2YAT.However, the percentage of plots with groundcover that showed no damage significantly differed (p = 0.015) for the average of Rodeo treatments (32%) when compared to untreated control (90%) and (p = 0.025) the Garlon 3A standard (80%) at both Table 3. Swamp titi mortality and rootstock basal area 2 years after cut stump treatment compared at study average of 3.4 stems per rootstock (significant covariate).Means followed by the same letter are not significantly different and estimates followed by asterisk (*) differ from the untreated control for mortality and differ from the cut only method control for rootstock basal area at the 5% level using the Bonferroni adjustment (Bon).Means followed by the same letter are not significantly different at the 5% level using Tukey's adjustment and means followed by asterisk (*) are significantly different from the untreated control for mortality and significantly different from the cut-only method control for rootstock basal area using Dunnett's at the 5% level.showing number of resprouts (stems + resprouts for untreated control) per initial stem and resprout height 2 years after treatment with approximate standard errors (SE).Means followed by the same letter are not significantly different and estimates followed by asterisk (*) differ from the cut only method control at the 5% level using the Bonferroni adjustment (Bon).Means followed by the same letter are not significantly different at the 5% level using Tukey's adjustment and means followed by asterisk (*) are significantly different from the cut only method control using Dunnett's at the 5% level.assessments.The percentage of plots that showed no damage was marginally less (p = 0.086) for Habitat (54%) when compared to the untreated control (90%) at both assessments.There were no significant Rodeo rate effects for reduction of total ground vegetation cover or graminoid cover (ANOVA not shown).Percent reduction of total ground vegetation cover (55-58%) and graminoid cover (28-31%), averaged over the three Rodeo rates, differed (p ≤ 0.016) from the untreated control (0-2%) and the Garlon 3A standard (4-7%) at 1 YAT and 2 YAT (Tables 5, S3 & S4).There were no Habitat rate effects on reduction of total ground vegetation cover or graminoid cover when applied alone or mixed with 7.5% Rodeo.Averaged over rates of Habitat applied alone, reduction of total ground vegetation (28-31%) was greater (p ≤ 0.042) than the untreated control (2%) at 1 and 2 YAT.Habitat applied alone reduced total groundcover less than Rodeo applied alone or the combinations of Rodeo plus Habitat, using Bonferroni's test (Table 5).Among the herbicide treatments, Rodeo alone at 5-7.5% or Rodeo 7.5% plus Habitat at 1-2% resulted in the greatest reduction in total groundcover (66-58%) and graminoid cover (38-31%), whereas the Garlon 3A standard caused least reduction in total groundcover (7%) or graminoid cover (4%) at 1 and 2 YAT.

Resprout count
Critical plant species were identified within all treatments with the number of species means ranging from 0.6 to 1.4 and no changes were observed between 1 and 2 YAT (Table S5).Treatments and treatment effects did not differ in number of critical species at each annual assessment.

Directed Foliar Spray Studies
Rodeo and combinations with Habitat gave the greatest swamp titi mortality and height control but the addition of Habitat did not improve control over Rodeo and introduces a persistent herbicide which can damage groundcover.Although the Rodeo rate effect for rootstock mortality or height control was not significant, the partitioned linear rate effect for titi mortality was p = 0.090.The importance of this trend depends on operational objectives.If canopy suppression is adequate, perhaps when followed by planned prescribed burning, then 68% titi mortality for the low rate versus 91% for the high rate may suggest that a low rate of Rodeo is sufficient.However, if application costs are high relative to herbicide or burning is restricted, and long-term control of rootstocks is preferred, then a high-rate Rodeo treatment might be warranted.Addition of MSO to Rodeo improved titi control but was antagonistic for Rodeo plus Habitat combinations in Study 2. The potential antagonism may not be of great concern because there is no compelling reason to use Habitat as these studies identify Rodeo the effective herbicide.In Study 1, Habitat averaged 12% titi mortality and 20% height control, significantly less than the 57% and 66%, respectively, obtained with the Garlon 3A standard which provided the best tolerance to nontarget vegetation.

Cut Stump
The Garlon 3A 50% standard reference treatment, Rodeo 100%, Habitat 100%, and Rodeo + Habitat tank mixes were the only treatments resulting in greater swamp titi rootstock mortality than the untreated control (4%).No herbicide treatment performed better than the Garlon 3A reference with 60% rootstock mortality, 87% rootstock basal area control, and 1.7 sprouts per rootstock.Garlon 3A is recommended for restoration activities because it does not pose a concern for root uptake by nontarget vegetation.Rodeo is also tightly absorbed by soil and thus not prone to uptake by roots.The 100% Rodeo rate gave comparable rootstock mortality (48%) as Garlon 3A 50% and may be considered as an alternative to the standard Garlon 3A treatment.Rootstock basal area was greatly reduced by cutting only, 50.35 cm 2 for the untreated control versus 3.40 cm 2 for the cut only method control, but cutting promoted sprouting, 1.1 stem + sprouts/rootstock for untreated control versus 12.1 sprouts/ rootstock for cut only.Cutting had the most immediate effect on rootstock basal area reduction, but the concern here is longterm control of rootstocks utilizing the comparisons of the cut only method control to herbicide treatments.

Cut Stem
Comparatively, cut stem herbicide treatments provided the greatest control of swamp titi rootstocks and did not stimulate rootstock sprouting as did the cut stump treatments.Although titi rootstock mortality and basal area did not differ among herbicides, averages for Rodeo, Habitat, and their combination provided 90-97% mortality compared to 80% for the Garlon 3A standard reference treatment.There was no rate response for Rodeo or Habitat, suggesting the use of the lowest rate (50%) would be adequate, and there was no advantage for using the combination of these herbicides.

Impacts to Groundcover
Cut stem treatments are thought to provide the best safety to desired groundcover if the herbicides used have limited soil activity.Cut stem treatments are quick, generally use less herbicide than the other methods, but are limited to stems greater than 3-4 in. in diameter.Following cut stem treatments eventual tree fall can shade out desired groundcover if tree density is great and prescribed burning is not employed to reduce the debris.Cut stump treatments are labor intensive, especially if titi tops are piled and removed, but quickly clear the site.Future studies might investigate the introduction of prescribed burning to reduce cut tops in situ for additional control of titi rootstocks and potential beneficial fire disturbance to foster desired habitat.
Among directed foliar treatments, Rodeo or Rodeo plus Habitat gave the greatest titi control but had the most detrimental effect to graminoids, including several critical grass species.At 1 year after treatment, we observed reduction of the critical bunch grass species Aristida spp., Rhynchospora spp., Dichanthelium spp., and Scleria sp.At that assessment reduction of critical forb cover including Eriocolon sp., Xyris spp., Lobelia floridana, and Helenium pinnatifidum was only noted for the low rate of the Rodeo plus Habitat tank mix.At 2 YAT there was recovery of the critical grass species Aristida spp., Rhynchospora sp., and Scleria sp. in 7.5% and 10% Rodeo treatments, 1% Habitat, and the combination of 7.5% Rodeo plus 3% Habitat.Unlike the impact to critical grasses, the critical forb species assessment remained unchanged from the 1 YAT to 2 YAT assessment, with little herbicide damage observed.At both annual assessments, the Garlon 3A 4% standard resulted in the least total cover reduction (mostly grasses) and critical grass and forb species, which is consistent with reported tolerance to this active ingredient (Klaus & Klaus 2011;DiAllesandro 2012;GeFellers et al. 2020).
Over time the disturbance from foliar sprays may open groundcover and foster colonization of desired species.In a study of hardwood control methods for longleaf pine restoration significant herbicide damage to nontarget groundcover with glyphosate plus imazapyr directed sprays was observed but 2 years after treatment herbaceous species diversity was greater than the untreated control and there was no difference in number of desired remnant woodland species (Harrington et al. 2021).
To have the least detrimental effect when treating shrub form titi, standard Garlon 3A 50% cut stump treatment could be followed, perhaps a year later, by directed foliar sprays of standard Garlon 3A 4% for retreatment of small titi stump and root sprouts where the spray can be better targeted to titi crowns.
We identified effective treatment approaches to control swamp titi rootstocks of various heights while minimizing adverse impacts to desired herbaceous groundcover habitat.Further research is needed focusing on these methods to determine direct impacts to amphibians and other taxa, including imperiled plants.

Table 2 .
Comparison of swamp titi rootstock mortality and height control with approximate standard errors (SE) 2 years after treatment (YAT) for directed foliar spray Study 2 testing applications with and without methylated seed oil (MSO) at two forest sites in north Florida.Effect means followed by the same letter within each grouping are not significantly different at the 5% level using Bonferroni's adjustment for multiplicity (Bon).Treatment means followed by the same letter within a column are not significantly different at the 5% level using Tukey's adjustment for multiplicity.Asterisk (*) indicates that mortality is significantly different than MSO only treatment or that height control is significantly different than untreated control at the 5% level using Dunnett-Hsu adjustment.

Table 4 .
Swamp titi cut stump study results

Table 5 .
Comparison of percent reduction of total groundcover 1 and 2 YAT for directed foliar Study 1 treatments that included 1% methylated seed oil (MSO) at the Apalachicola National Forest study location.Effect means followed by the same letter within each grouping are not significantly different at the 5% level using Bonferroni's adjustment for multiplicity (Bon).Treatment means followed by the same letter within a column are not significantly different at the 5% level using Tukey's adjustment for multiplicity.Asterisk (*) indicates mean is significantly different than the untreated control using Dunnett-Hsu adjustment at the 5% level.