Southeastern Naturalist K.E. Tisshaw and E.S. Menges 2018 Vol. 17, No. 4 654 2018 SOUTHEASTERN NATURALIST 17(4):654–670 The Influence of Management Regimes and Habitat Characteristics on the Persistence and Current Occupancy of the Non-native Melinis repens (Natalgrass) Kathryn E. Tisshaw1,* and Eric S. Menges2 Abstract - Although prescribed fires and pre-treatments (e.g., roller chopping and mowing) are used by public and private landowners to manage natural habitats in Florida, they can influence the invasion and spread of non-native plants in natural areas. Firelanes and roads used to access habitat for management practices create corridors for invasive grasses. Melinis repens (Natalgrass) is an invasive plant found throughout Florida. Fire regimes and roads acting as corridors may affect invasion and persistence of Natalgrass, but these topics have not been well-studied. Following up on distribution data originally collected in 2002 at Archbold Biological Station, we explored how fire regimes, distance to road, habitat type, and microhabitat factors influenced Natalgrass persistence through 2016, and current Natalgrass occupancy. Persistence from 2002–2016 was not influenced by distance to road. However, Natalgrass was currently more likely to occupy habitat closest to roads and was more likely to persist in areas burned within 16 y. Although Natalgrass was most likely to persist in human modified habitat, it still persisted in and occupied interior scrub habitat. Natalgrass was more likely to occupy areas with lower litter, shrub, and palmetto cover, which are characteristics of many habitats, including sandy roadsides and recently burned scrub habitat. These results suggest Natalgrass is able to persist in habitats other than roads, and distance to road did not influence its persistence; thus, land managers should treat interior habitat where Natalgrass is persisting. At the same time, searches for new populations of Natalgrass should be focused largely in areas close to corridors, such as in roads and firelanes. Introduction Prescribed fires are used as a management practice in a wide variety of habitats and regions to restore vegetation structure, reduce fuels, and enhance wildlife habitat where certain endangered or threatened species require fire (D’Antonio and Vitousek 1992, Elgersma et al. 2017, Long et al. 2004, Wade et al. 1980). Fire severity, fire frequency, and time-since-fire can vary among vegetation types and within the same fire or vegetation type, and can influence post-burn vegetation patterns (D’Antonio and Vitousek 1992, Flannigan et al. 2013, Long et al. 2004). Prescribed fire has been used extensively in Florida for many years (Boughton et al. 2016, Mitchell et al. 2006, Wade et al. 1980), but urbanization and concerns about control and smoke have led land managers to use roller-chopping and mowing, which allow for more effective carry and easier control during a prescribed burn 1Department of Environmental and Life Sciences, Trent University, 1600 West Bank Drive, Peterborough, Ontario K9J 0G2, Canada. 2Plant Ecology Program, Archbold Biological Station, 123 Main Drive, Venus, FL 33960. *Corresponding author - ktisshaw@trentu.ca. Manuscript Editor: Matthew Heard Southeastern Naturalist 655 K.E. Tisshaw and E.S. Menges 2018 Vol. 17, No. 4 (Menges and Gordon 2010), as pre-treatments or an alternative to prescribed fires. Although these practices are employed to manage natural habitat, they can act as disturbances that increase the risk of non-native plant invasions into natural areas (Hobbs 1991, Hobbs and Huenneke 1992, Long et al. 2004). In order to manage pyrogenic habitats using prescribed fires and pre-treatments, roads or firelanes are used to access protected habitat and provide firebreaks. Roads and firelanes act as corridors for non-native plants to colonize open areas and gaps of natural habitats (Wace 1977). These corridors may increase the risk of non-native invasive plants entering natural habitat (Gelbard and Belnap 2003, Hansen and Clevenger 2005, Yates et al. 2004), and lead to a loss of native species (Tomimatsu and Ohara 2004, Turner 1996). In general, non-native grasses are often common along roads because of conditions such as increased nutrient availability, reduced competitive pressure, stimulation of seed germination, and increased seed dispersal (D’Antonio 1993, Forman and Alexander 1998). Consequently, active management can have unintended consequences on the distribution of non-native invasive plant species in natural areas. Prescribed fires and pre-treatments can increase the risk of non-native plant invasions through soil disturbance and reduction of competition (Catling et al. 2002, Dodson and Fiedler 2006, Hobbs and Atkins 2006, Lonsdale 1999). A disadvantage to using pre-treatments is that they create more soil disturbance than using prescribed fires alone (Menges and Gordon 2010). Some non-native plants have been observed to increase in abundance following fire (D’Antonio and Vitousek 1992, Just et al. 2017, Lippincott 2000) and following soil disturbances (Catling et al. 2002, Dodson and Fiedler 2006). For instance, Imperata cylindrica (L.) P. Beauv. (Cogongrass) is a non-native grass found throughout Florida that is more likely to invade natural habitat when soil disturbance has occurred (Lippincott 2000, Schmalzer and Foster 2016). Cogongrass is able to colonize areas after prescribed fire and, once established, increases fuel load, fire severity, and Pinus (pine) mortality (Lippincott 2000, Matlack 2002). Melinis repens (Willd.) Zizka (Natalgrass) is a grass native to Africa that has become a problematic invasive species in many areas around the world, including Florida (Stokes et al. 2011). Its approximate date of arrival in Florida was in 1893 (Scott 1913). Natalgrass is a Category I invasive plant which can displace native plants and alter community structure (FLEPPC 2015). It is a short-lived perennial C4 grass which is primarily wind-dispersed (Possley and Maschinski 2006). Natalgrass is one of the few grasses observed to invade scrub habitat (Gordon et al. 2005, Hutchinson and Menges 2006). The optimal habitat of Natalgrass is along roads and in disturbed areas (David and Menges 2011, Gordon et al. 2005); however, it is unknown if these sites can serve as continuous source populations to disperse seed into adjacent scrub habitat that has been recently disturbed or burned. In some Florida habitats, Natalgrass has been observed to increase in density after a prescribed burn, especially when accompanied with soil disturbance (Hutchinson and Menges 2006, Williges et al. 2006). David and Menges (2011) speculated that fire may influence Natalgrass distribution through creating open Southeastern Naturalist K.E. Tisshaw and E.S. Menges 2018 Vol. 17, No. 4 656 habitat and decreasing competitive effects of shrubs. Despite this research, no one has quantified the influence of fire regimes on Natalgrass distribution, the conditions under which Natalgrass can persist, and whether roads and firelanes act as corridors for Natalgrass invasion or persistence. In this paper, we examine how effects of land management, such as distance to roads or firelanes, time-since-fire, fire severity, and historical fire frequency influenced the persistence of Natalgrass from 2002 to 2016, and how persistence differs from current Natalgrass occupancy (hereafter current occupancy). We also examine how habitat type, vegetation cover, and gap cover influence current occupancy. Undisturbed Florida scrub seems resistant to most plant invasions (Greenberg et al. 1997); thus, we predicted that Natalgrass would persist only in road habitats and that it would have the greatest persistence and current occupancy closest to roads. Since some non-native invasive grass species are able to colonize areas after prescribed fires, we predicted that Natalgrass would have greater persistence and current occupancy in recently burned management units, those that have been burned with the highest severity, and those that have been burned most frequently. We predicted that Natalgrass would persist more frequently in human-modified areas, pastures, and disturbed scrub habitat than in other interior scrub habitats. Certain non-native invasive species can take advantage of areas where competition is reduced, so we predicted that Natalgrass would occupy open areas with characteristically low litter-cover, high bare-sand cover, and low shrub and palmetto cover. In order to test these predictions, in 2016, we conducted a resurvey of the distribution of Natalgrass, previously surveyed in 2002, to determine how its persistence was affected by management regimes. In addition, we surveyed random road transects in different habitats in 2016 to determine how current Natalgrass occupancy is affected by management regimes and microhabitat characteristics. Field Site Description We conducted this study at the Archbold Biological Station (ABS; Swain 1998) in Lake Placid, Highlands County, FL (27º11'N, 81º21'W). ABS is one of the largest remaining tracts of the Lake Wales Ridge ecosystems (Weekley et al. 2008), and includes over 2100 ha of southern ridge sandhill, Florida scrub, flatwoods, and seasonal-pond habitats (Abrahamson et al. 1984). For the purpose of this study, we categorized the habitat types at ABS into xeric yellow sands, flatwoods, scrubby flatwoods/Ceratiola ericoides Michx. (Florida Rosemary) scrub, human-modified areas, and roads. Xeric yellow sands combines the Abrahamson et al. (1984) categories of Quercus (oak)-Carya (hickory) scrub and sandhill. Scrubby flatwoods/rosemary combines scrubby flatwoods, Pinus clausa (Chapman ex Engelmann) Vasey ex Sargent (Sand Pine) scrub, and rosemary scrub (Abrahamson et al. 1984). Human-modified areas includes old-field habitats. Roads includes sandy roadsides adjacent to interior habitat. We also sampled invasive species in the Archbold Reserve (a former ranchland; hereafter the Southeastern Naturalist 657 K.E. Tisshaw and E.S. Menges 2018 Vol. 17, No. 4 Reserve). Reserve habitat types were categorized into either pasture or disturbed scrub with a history of mechanical disturbance. Methods Data collection (persistence) To evaluate persistence, we resurveyed 220 plots where Natalgrass was found within ABS and the Reserve in 2002 and 2003 (Hutchinson and Menges 2006). These plots included interior habitats, >2.5 m from the vegetation edge along roads. This distance is beyond the range of edge effects in vegetation (E. Menges, unpubl. data). These plots were in management units with various habitat types, fire histories (Menges et al. 2017), and distances to roads. We navigated to these plots using the 2002 coordinates and a Trimble GeoXT GPS unit with submeter accuracy. We recorded whether Natalgrass was present or absent in 2.5-m–radius circular plots centered on the GPS point. We did not record data at 20 of these plots, which were ponded or within new development. Data collection (current occupancy) We used a stratified random-transect design (Holly and Hampton 1990) to evaluate current occupancy of Natalgrass in interior habitats. We randomly generated 32 transects among 10 different management units with varying fire histories at ABS, beginning at the vegetation edge along unpaved, sandy roads or firelanes, and running perpendicular to the road into adjacent interior habitat. Using a randomtransect generator tool in ArcGIS, we created 30-m transects with one 5-m plot randomly established within each of three 10-m intervals: within 10 m from the road, 10–20 m from the road, and 20–30 m from the road (resulting in 3 plots per transect, and 96 plots total). Using a Trimble GeoXT GPS unit, we navigated to the coordinates of the start of the transects and placed 30-m transects by hand using a transect tape and the predetermined coordinates for plot intervals as reference to ensure a perpendicular transect from the road into interior habitat. Following establishment of the 5-m circular plots, we recorded whether Natalgrass was present or absent within each plot, and collected microhabitat data including relative percent cover of shrubs (Serenoa repens (W. Bartram) Small [Saw Palmetto] and Sabal etonia Swingle ex Nash [Scrub Palmetto]) and gaps using the line-intercept method (Canfield 1941). We defined gaps as areas without woody or herbaceous plant-cover, but included ground lichens, bare sand, and litter, and which intercepted more than 10 cm along the transect line. Statistical analysis We obtained data on time since last fire (years), fire frequency (number of times an area burned since 1967), fire severity of the most recent fire: never burned, patchy (low severity), scorched (intermediate severity), or consumed (high severity), and habitat type (defined above under Field Site Description from a 5 x 5-m–grid database [Menges et al. 2017]). In ArcGIS 10.3.1, we measured distances to the nearest road or firelane in meters using shapefiles of habitat types and road Southeastern Naturalist K.E. Tisshaw and E.S. Menges 2018 Vol. 17, No. 4 658 boundaries. We measured from the vegetation edge of the nearest road or firelane to the edge of the plot. We employed generalized linear-model logistic regressions for model formation with multiple predictor variables to ascertain the effects of distance to road, fire-regime variables, and habitat type on the persistence and current occupancy of Natalgrass, and microhabitat data on the current occupancy of Natalgrass. We used 2 models to analyze (1) persistence, and (2) occupancy as binary response-variables against the predictors: distance to road, time-since-fire, fire severity, fire frequency, and habitat type (Table 1: m1 and m2, respectively). Using backwards stepwise procedures, we determined which predictors were the most import ant in determining Natalgrass persistence and occupancy. We used a binomial-regression model to test the effects of microhabitat variables on the current occupancy of Natalgrass in interior native habitat (Table 1: m3). We removed from the models observations that did not have fire-history data. We used Bartlett tests to assure homogeneity of variances among predictors within models. In m2, distance to road did not meet the assumption of homogeneity of variance; therefore we used a non-parametric function for this predictor within a generalized additive model using the gam package in R v 3.4.4 (Hastie 2018). There were no predictors with correlations exceeding 0.5 (Pearson correlation coefficient), therefore all predictors stayed in the model before stepwise procedures. We employed the Wald statistic to test the significance of the regression variables, wherein levels of factors were compared to 1 level assigned as the reference category. We ran the chi-square goodness-of-fit test (χ2) to assess the model goodness- of-fit. Details on the model diagnostics are outlined in Table 1. We conducted all analyses in R v.3.4.4 within R studio v.1.1.383 (R Core Team 2017; RStudio Team 2016). Results Distance to road and fire-regime effects on Natalgrass The effects of distance to roads on persistence and current occupancy of Natalgrass were variable. Although persistence was not affected by distance to roads (P = 0.811; Table 2; Fig. 1A), current occupancy of Natalgrass decreased significantly with distance to roads (P = 0.002; Table 3; Fig. 1B). Natalgrass persistence and current occupancy were affected by the fire regime. Persistence of Natalgrass was significantly influenced by time-since-fire (P = 0.036; Table 1. Model diagnostics including: the chi-square goodness-of-fit test statistic (χ2), degrees of freedom (df), and the significance of the fitted models (P > 0.5 indicates the fitted model does not differ significantly from our observed values). Model name Model (y ~ x1 + x2 + x3 + x4) χ2 df P m1 Persistence ~ time-since-fire + fire severity + habitat type 68.418 13 0.968 m2 Current occupancy ~ distance to road + fire frequency 9.405 8 1.000 m3 Current occupancy ~ bare sand + litter + shrub + palmetto 84.448 4 0.617 Southeastern Naturalist 659 K.E. Tisshaw and E.S. Menges 2018 Vol. 17, No. 4 Figure 1. The (A) persistence or (B) current occupancy of Melinis repens (Natalgrass) in relation to distance to road in meters. Table 2. Binary logistic regression predicting the likelihood of persistence of Natalgrass from 2002 to 2016 based on distance to roads, time-since-fire, fire severity, fire frequency, and habitat type (n = 105). Stepwise procedures resulted in the best model (m1) including the predictor’s time-since-fire, fire severity, and habitat type. Details on the influence of distance to road and fire frequency before being removed from the full model with all explanatory variables are also included. Factor levels in parentheses were reference conditions in regressions. * denotes significance at P < 0.05; ** denotes significance at P < 0.01. 95% CI for odds ratio Variables (m1) n z-value SE Wald df P Odds ratio Lower Upper Time-since-fire (3–4 y) 18 2.680 4 0.0360* 0–2 y 58 -1.724 0.519 1 0.0847 0.166 0.017 1.174 5–8 y 10 -0.570 0.411 1 0.5690 0.452 0.024 6.580 9–16 y 7 -0.003 0.372 1 0.0144* 0.026 0.001 0.467 17–34 y 12 -1.300 0.415 1 0.1940 0.185 0.012 2.211 Fire severity High (consumed) 51 3.815 2 0.0256* Low (patchy) 44 1.712 0.354 1 0.0870 3.394 0.869 14.967 Intermediate (scorched) 10 2.589 0.384 1 0.0096** 29.126 3.046 726.940 Habitat type Flatwoods 27 2.748 5 0.0232* Human modified 31 3.094 0.362 1 0.0020** 11.475 2.615 60.046 Pasture 7 0.005 995.235 1 0.9960 1.79e+9 0.000 . Xeric yellow sands 16 -0.659 0.364 1 0.5090 0.514 0.053 3.243 Scrubby flatwoods/ 15 -0.008 890.556 1 0.9940 0.000 . 2.45e+65 rosemary Disturbed scrub 9 1.626 0.271 1 0.1040 4.798 0.730 34.635 Variables excluded from m1 Distance to road 105 0.002 0.007 0.057 1 0.8110 137.938 3.624 5250.464 Fire frequency 105 1.308 1.358 0.927 1 0.3360 3.698 0.258 52.940 Southeastern Naturalist K.E. Tisshaw and E.S. Menges 2018 Vol. 17, No. 4 660 Table 2). Compared to the 3–4 y-since-fire category, Natalgrass persistence was significantly less likely in areas burned 9–16 y ago (P = 0.014), and marginally significantly less likely in areas burned 0–2 y ago (P = 0.085). Natalgrass persistence in areas that were burned 5–8 y ago and 17–34 y ago were not significantly different than areas that were burned 3–4 y ago (P = 0.569 and P = 0.194, respectively). Persistence of Natalgrass was also affected by fire severity (P = 0.026; Table 2; Fig. 2a). Compared to high-severity patches where Natalgrass persisted only 41% of the time, Natalgrass persistence in intermediate-intensity patches was significantly higher (80%; P = 0.0096). Natalgrass persistence in low-severity patches was marginally higher (~48%) than in high-severity patches (P = 0.087). On the other hand, current occupancy of Natalgrass was not influenced by time-since-fire (P = 0.373; Table 3) or fire severity (P = 0.938; Table 3; Fig. 2B). Fire frequency since 1967 was a significant predictor of Natalgrass occupancy (P = 0.038; Table 3) but not a significant predictor of Natalgrass persistence (P = 0.336; Table 2). Natalgrass occupied plots burned twice 42% of the time, while it only occupied plots burned 3 times or more 11% of the time (P = 0.035). Habitat type effects on Natalgrass Habitat type affected persistence (P = 0.0232; Table 2; Fig. 3A) but not current occupancy (P = 0.979, Table 3; Fig. 3B) of Natalgrass. Compared to flatwoods habitats where Natalgrass persisted only 26% of the time, Natalgrass persisted in human-modified habitats 84% of the time. According to the model, Natalgrass persistence in pasture, xeric yellow sands, scrubby flatwoods/rosemary, and disturbed scrub was not significantly different from flatwoods habitats (P = 0.996, P = 0.509, P = 0.994, and P = 0.104, respectively). However, in pastures and scrubby flatwoods/ rosemary plots, Natalgrass persisted 100% and 0%, respectively (Fig. 3A). Therefore, there was no variation in which the model could prov ide an estimate. Table 3. Binary logistic regression predicting the likelihood of current occupancy of Natalgrass based on distance to roads, time-since-fire, fire severity, fire frequency, and habitat type (n = 79). Stepwise procedures resulted in the best model (m2) including the predictors distance to road and fire frequency. Details on the influence of time-since-fire, fire severity, and habitat type before being removed from the full model with all explanatory variables are also included. Factor levels in parentheses were reference conditions in regressions. * denotes significance at P < 0.05; ** denotes significance at P < 0.01. 95% CI for odds ratio Variables (m2) n SE Wald df P Odds ratio Lower Upper Distance to road 79 0.085 10.109 1 0.002** 0.763 0.617 0.873 Fire frequency (burned twice) 43 4.436 1 0.038* Burned ≥3 37 1.449 1 0.035* 0.047 0.001 0.525 Variables excluded from m2 Time-since-fire 79 5.365 5 0.373 Fire severity 79 0.128 2 0.938 Habitat type 79 0.194 3 0.979 Southeastern Naturalist 661 K.E. Tisshaw and E.S. Menges 2018 Vol. 17, No. 4 Figure 2. Bar graphs depicting the (A) persistence or (B) current occupancy of Melinis repens (Natalgrass) in relation to categorical fire severity (never burned, low, intermediate or high severity). Proportions are based on averages of Natalgrass either being present (value of 1) or absent (value of 0) among plots with varying fire severity categories. An average value of 1 would signify that Natalgrass is always present, and an average value of 0 would signify that Natalgrass is always absent. Reference categories are dark grey, and significant differences from reference categories are denoted by an asterisk ( *) for P < 0.05. Southeastern Naturalist K.E. Tisshaw and E.S. Menges 2018 Vol. 17, No. 4 662 Figure 3. Bar graphs depicting the (a) persistence or (b) current occupancy of Melinis repens (Natalgrass) in relation to habitat type categories (either pasture, human modified, roads, disturbed scrub, xeric yellow sands, flatwoods, and scrubby flatwoods/rosemary). Proportions are based on averages of Natalgrass either being present (value of 1) or absent (value of 0) among plots with varying habitat types). An average value of 1 would signify that Natalgrass is always present, and an average value of 0 would signify that Natalgrass is always absent. Natalgrass was always absent from occurring in flatwoods. Reference categories are dark grey, and significant differences from reference categories are denoted by an asterisk (*) for P < 0.05. Southeastern Naturalist 663 K.E. Tisshaw and E.S. Menges 2018 Vol. 17, No. 4 Table 4. Binary logistic regression predicting the likelihood of current occupancy of Natalgrass based on relative percent cover of bare sand, litter, shrub, and palmetto. We used absent as the reference category for the regressions (n = 94). There were no variables taken out of m3. Average % cover where Natalgrass 95% CI Variables was absent or present for odds ratio (m3) Absent Present SE Wald df P Odds ratio Lower Upper Bare sand 5.078 10.356 0.017 1.663 1 0.197 0.978 0.945 1.012 Litter 18.245 11.669 0.018 8.846 1 0.003** 0.947 0.913 0.982 Shrub 31.205 19.921 0.017 10.213 1 0.001** 0.948 0.918 0.980 Palmetto 18.997 12.626 0.024 8.028 1 0.005** 0.933 0.889 0.979 Microhabitat effects on Natalgrass Microhabitat affected Natalgrass current occupancy and cover. Natalgrass presence was associated with significantly lower litter-cover than plots where it was absent (12% vs. 18%; P = 0.003; Table 4). Natalgrass presence was 5% higher in areas with bare sand cover, although this difference was not significant (P = 0.197; Table 4). Plots where Natalgrass was present had 11% less shrub cover and 6% less palmetto cover than plots where it was absent (P = 0.001, P = 0.005, respectively; Table 4). Discussion In this study, we examined how the persistence and current occupancy of Natalgrass were affected by management regimes, habitat types, and microhabitat factors. Distance to roads did not influence the persistence of Natalgrass; however, current occupancy of Natalgrass was lower farther from roads. We found that Natalgrass persisted more often in areas that were recently burned with intermediate severity and in human-modified habitats. Natalgrass was able to persist with lower probabilities in xeric yellow sands, flatwoods, and disturbed scrub habitats. We also found new populations of Natalgrass occupying xeric yellow sands, scrubby flatwoods, and rosemary scrub habitats. Natalgrass preferred microhabitats with low litter-, shrub-, and palmetto-cover. Influence of distance to roads on persistence and current occupancy Distance to roads had dissimilar effects on the persistence and current occupancy of Natalgrass. Persistence of Natalgrass was not influenced by distance to roads, implying that Natalgrass is able to persist once it has established no matter what distance it is to the roadside source-population. In contrast, current occupancy of Natalgrass was greatest at shorter distances to road, a result consistent with the findings of David and Menges (2011). Although Natalgrass persists no matter the distance to roads, it still exhibits patterns similar to those of other non-native invasive grasses that are more likely to invade and occupy habitat closest to roads (Barbosa et al. 2010, Christen and Matlack 2009, Spellerberg 1998). Natalgrass is able to persist in interior habitats; thus, land managers should treat Natalgrass in all Southeastern Naturalist K.E. Tisshaw and E.S. Menges 2018 Vol. 17, No. 4 664 interior habitats (>2.5 m from vegetation edge). Managers should prioritize greater search effort for new interior populations of Natalgrass near roads because the species had greater current occupancy closer to roads. Influence of fire regimes on Natalgrass Natalgrass persisted more frequently in areas that were last burned 0–16 y ago, consistent with our prediction that it would persist more frequently in recently burned units. The 17–34 y-since-fire category, where persistence was lower, is the only category that included areas that had not been burned since the 2002 survey was conducted (~15 y ago). These results suggest that Natalgrass could persist in an area for up to 16 y in the absence of fire. Other studies have shown that certain grasses persist for many years after fire (Lippincott et al. 2001). For example, cover of non-native invasive Urochloa maxima (Jacq.) R. (Guineagrass) was 6 times higher in burned plots compared to unburned plots 3 years after a fire (Veldman et al. 2009). Alien grasses in Hawaii increased in abundance following a fire compared to pre-burn levels, and they maintained high cover-levels even 18 y after the fire (Hughes et al. 1991). Although, in our study, current Natalgrass occupancy did not depend on time-since-fire, and Natalgrass did not occur more frequently in recently burned areas, Natalgrass sometimes invades areas shortly after fire (Hutchinson and Menges 2006, Williges et al. 2006). Duration of typical Natalgrass persistence after a recent invasion is not known, but our results suggest that once established, Natalgrass patches are capable of persisting for many years. This finding again suggests that interior populations of Natalgrass need to be treated aggr essively. In other circumstances, prescribed burning has been effective in managing invasive plants. Prescribed burning to reduce Poa pratensis L. (Kentucky Bluegrass) invasions has been effective (Abrams 1988, Becker 1989, McMurphy and Anderson 1965) and has even stimulated populations of native grasses such as Schizachyrium scoparium Michx. (Little Bluestem) and Andropogon gerardii Vitman (Big Bluestem) (Sverdarsky et al. 1986, Towne and Owensby 1984). Populations of the invasive grass Bromus inermus Leyss. (Smooth Brome) were reduced after repeated prescribed burns (Willson and Stubbendieck 1996). These contrasting findings on the effects of fire on non-native grass distribution emphasize the need for land managers to understand which grasses in the ecosystem are controlled by or benefit from the use of prescribed fires. Eradicating or reducing invasive grass populations before prescribed burns to prevent subsequent post-fire invasion would be an efficient management strategy. Our results showed the highest persistence of Natalgrass in intermediate burnseverity patches, while other studies have found mixed results varying from no effects of burn severity on the persistence of non-native species (Kuenzi et al. 2008), to high proportions of non-native species compared to native species associated with high-severity burns (Crawford et al. 2001, Fornwalt et al. 2010, Griffis et al. 2001, Turner et al. 1997). Studies that found an association between high occurrences of non-native species and high-severity burns attributed their findings to non-native species that resprouted after fire (Turner et al. 1997), dispersed into recently burned areas (Fornwalt et al. 2010), or were intentionally Southeastern Naturalist 665 K.E. Tisshaw and E.S. Menges 2018 Vol. 17, No. 4 seeded into the area (Kuenzi et al. 2008). Collectively, the findings from these studies are inconclusive, suggesting that post-burn vegetation patterns following fires of varying fire severity are influenced by the interplay of past management practices, species composition pre-burn, dispersal abilities of neighboring individuals, and resprouting ability of individuals within burn patches. Our observations suggest Natalgrass is often killed by fire, in contrast with grasses that can reproduce asexually from belowground rhizomes that are not destroyed by fire (e.g., Cogongrass; Dozier et al. 1998). Once Natalgrass has been consumed by a high-severity fire, it would need to recolonize from neighboring Natalgrass seed sources or surviving seeds in the soil. It is unlikely that Natalgrass would resprout once consumed by a high-intensity fire, which may contribute to its greater persistence in patches burned with intermediate severity. Habitat preferences of Natalgrass Our prediction that Natalgrass would persist more often in human-modified areas than in interior scrub habitats was supported. However, Natalgrass was still able to persist in and occupy interior scrub habitats. Propagule pressure along roads may help explain the presence of Natalgrass in interior scrub habitats (Colautti et al. 2006). Future studies should focus on measuring the density of road populations and monitoring dispersal of Natalgrass into adjacent interior habitat. Researchers should investigate how long populations of Natalgrass persist once established in natural habitats, and whether the road populations can act as continual sources for dispersing seeds into interior habitats. Land managers should focus on eradicating large patches of Natalgrass along or near roads and disturbed areas. Microhabitat preferences of Natalgrass Our results supported most of our predictions for microhabitat preferences of Natalgrass. As predicted, Natalgrass was less likely to occupy areas with high litter-, shrub-, or palmetto- cover. David and Menges (2011) studied microhabitat preferences of Natalgrass and found that Natalgrass biomass decreased with increasing litter volume, which is analogous to the results of our study in that Natalgrass was more frequently present in areas with lower litter-cover. They also found that Natalgrass presence significantly increased with greater distance to shrubs (David and Menges 2011), which is comparable to our results in that Natalgrass is more likely found in plots with low shrub-cover. Bare sand had no significant effect on current occupancy of Natalgrass in our study, although the pattern of more Natalgrass occupancy with bare sand was consistent with our predictions. David and Menges (2011) found that Natalgrass was less microhabitat-limited on roads than interior scrubby flatwoods habitats, and attributed this effect to the openness of sandy roads. In our study, bare sand did not have a significant effect on Natalgrass presence, although Natalgrass occurred more frequently on roads. Our contrasting results might be explained by Natalgrass persisting in a range of interior habitat types with variable bare sand cover in our study, since David and Menges (2011) only studied scrubby flatwoods interior habitats, which have little bare sand (Dee and Menges 2014, Young and Menges 1999). Southeastern Naturalist K.E. Tisshaw and E.S. Menges 2018 Vol. 17, No. 4 666 Management and research implications To protect native habitat from the invasion of Natalgrass, land managers should reduce the total area of firelanes and roads along borders of management units when possible. These areas may be constant sources of invasion for Natalgrass and other invasive species, and can affect invasions into adjacent closed habitats. Land managers should focus on eradicating persistent populations of Natalgrass on both roads and interior habitats. Herbicide application (Stokes 2010) and careful hand pulling (ideally before fruit maturation) can be used to reduce Natalgrass abundance. Elimination of seed input for several months can be effective in reducing Natalgrass because this species does not have a long-lived seedbank (Stokes 2010). Both road and interior patches should be treated, with a priority of eradicating large patches close to roads that may act as sources. Road populations should be treated before a prescribed burn so that post-fire invasion into interior habitats is discouraged. In addition, land managers need to be vigilant in surveying for other invasive species that could take advantage of areas cleared of Natalgrass. Finally, the use of mechanical pre-treatments or surrogates should be approached with caution (Menges and Gordon 2010), as Natalgrass is known to increase in areas where soil has been disturbed. Our work demonstrates contrasting results regarding where an in vasive species persists and occupies a landscape. Resurveys are important to understand the persistence of invasive species. When resurveys show substantial persistence, as was the case for Natalgrass even in interior areas, land managers do not have the luxury of allowing natural processes to eliminate problem species. Instead, invasive species persistence demonstrated by resurveys make a case for aggressive treatment of invasive species. Acknowledgments We thank Vivienne Sclater and Rebecca Waskovich for help with ArcGIS and SQL database queries. Thanks to Stephanie Koontz for her statistical expertise, as well as Stacy Smith, Brian Josey, and Kevin Main for their input and support. This study was conducted as part of Kathryn Tisshaw’s internship at Archbold Biological Station, supported by a grant from the Vaughn-Jordan Foundation. Literature Cited Abrahamson, W.G., A.F. Johnson, J.N. Layne, and P. Peroni. 1984. Vegetation of the Archbold Biological Station, Florida: An example of the southern Lake Wales Ridge. Florida Scientist 47. Abrams, M.D. 1988. Effects of burning regime on buried seed banks and canopy coverage in a Kansas tallgrass prairie. Southwestern Naturalist 33:65–70 . Barbosa, N.P.U., G.W. Fernandes, M.A.A. Carneiro, and L.A.C. Junior. 2010. Distribution of non-native invasive species and soil properties in proximity to paved roads and unpaved roads in a quartzitic mountainous grassland of southeastern Brazil (rupestrian fields). Biological Invasions 12:3745–3755. Southeastern Naturalist 667 K.E. Tisshaw and E.S. Menges 2018 Vol. 17, No. 4 Becker, D.A. 1989. Five years of annual prairie burns. Pp. 163–168, In T. B. Bragg and J. Stubbendieck (Eds.). Proceedings, 11th North American Prairie Conference—Prairie Pioneers: Ecology, History, and Culture. University of Nebraska, Lincoln, NE. Boughton, E.H., P.F. Quintana-Ascencio, P.J. Bohlen, J.E. Fauth, and D.G. Jenkins. 2016. Interactive effects of pasture-management intensity, release from grazing, and prescribed fire on forty subtropical wetland-plant assemblages. Journal of Applied Ecology 53:159–170. Canfield, R.H. 1941. Application of the line-interception method in sampling range vegetation. Journal of Forestry 39:388–394. Catling, P.M., A. Sinclair, and D. Cuddy. 2002. Plant community composition and relationships of disturbed and undisturbed alvar woodland. The Canadian Field-Naturalist 116:571–579. Christen, D.C., and G.R. Matlack. 2009. The habitat and conduit functions of roads in the spread of three invasive plant species. Biological Invasions 1 1:453–465. Colautti, R.I., I.A. Grigorovich, and H.J. MacIsaac. 2006. Propagule pressure: A null model for biological invasions. Biological Invasions 8:1023–103 7. Crawford, J.S., C.H.A. Wahren, S. Kyle, and W.H. Moir. 2001. Responses of exotic plant species to fires in Pinus ponderosa forests in northern Arizona. Journal of Vegetation Science 12:261–268. D’Antonio, C.M. 1993. Mechanisms controlling invasion of coastal plant communities by the alien succulent Carpobrotus edulis. Ecology 74:83–95. D’Antonio, C.M., and P.M. Vitousek. 1992. Biological invasions by exotic grasses, the grass/fire cycle, and global change. Annual Review of Ecology, Evolution and Systematics 23:63–87. David, A., and E. Menges. 2011. Microhabitat preference constrains invasive spread of nonnative Natalgrass (Melinis repens). Biological Invasions 13:2309–2322. Dee, J.R., and E.S. Menges. 2014. Gap ecology in Florida scrubby flatwoods: Effects of time-since-fire, gap size, gap aggregation, and microhabitat on species occurrence and composition. Journal of Vegetation Science 25:1235–1246. Dodson, E.K., and C.E. Fiedler. 2006. Impacts of restoration treatments on alien plant invasion in Pinus ponderosa forests, Montana, USA. Journal of Applied Ecology 43:887–897. Dozier, H., J.F. Gaffney, S.K. McDonald, E.R.R.L. Johnson, and D.G. Shilling. 1998. Congongrass in the United States: History, ecology, impacts, and management. Weed Technology 12:737–743. Elgersma, K.J., J.P. Martina, D.E. Goldberg, and W.S. Currie. 2017. Effectiveness of cattail (Typha spp.) management techniques depend on exogenous nitrogen inputs. Elementa: Science of the Anthropocene 5:19. DOI:10.1525/elementa.147. Flannigan, M., A.S. Cantin, W.J. DeGroot, M. Wotton, A. Newberry, and L.M. Gowman. 2013. Global wildland fire-season severity in the 21st century. Forest Ecology and Management 294:54–61. Florida Exotic Plant Pest Council (FLEPPC). 2015. Available online at http://www.fleppc. org/list/2015FLEPPCLIST-LARGEFORMAT-FINAL.pdf. Accessed 15 October 2017. Forman, R.T.T., and L.E. Alexander. 1998. Roads and their major ecological effects. Annual Review of Ecology, Evolution, and Systematics 29:207–231. Fornwalt, P.J., M.R. Kaufmann, and T.J. Stohlgren. 2010. Impacts of mixed-severity wildfire on exotic plants in a Colorado Ponderosa Pine–Douglas-fir forest. Biological Invasions 12:2683–2695. Southeastern Naturalist K.E. Tisshaw and E.S. Menges 2018 Vol. 17, No. 4 668 Gelbard, J.L., and J. Belnap. 2003. Roads as conduits for exotic plant invasions in a semiarid landscape. Conservation Biology 17:420–432. Gordon, D.R., C.H. Greenberg, S.H. Crownover, and J.L. Slapcinsky. 2005. Effects of unpaved road soils on persistence of three non-native grass species. Natural Areas Journal 25:257–262. Greenberg, C.H., S.H. Crownover, and D.R. Gordon. 1997. Road soils: A corridor for invasion of xeric scrub by nonindigenous plants. Natural Areas Journal 17:99–109. Griffis, K.L., J.A. Crawford, M.R. Wagner, and W.H. Moir. 2001. Understory response to management treatments in northern Arizona Ponderosa Pine forests. Forest Ecology and Management 146:239–245. Hansen, M.J., and A.P. Clevenger. 2005. The influence of disturbance and habitat on the presence of non-native plant species along transport corridors. Biological Conservation 125:249–259. Hastie, T. 2018. gam: Generalized Additive Models. R package version 1.15. Available online at https://CRAN.R-project.org/package=gam. Accessed 12 April 2018. Hobbs, R.J. 1991. Disturbance as a precursor to weed invasion in native vegetation. Plant Protection Quarterly 6:99–104. Hobbs, R.J., and L. Atkins. 2006. Effect of disturbance and nutrient addition on native and introduced annuals in plant-communities in the western Australian wheat-belt. Australian Ecology 13:171–179. Hobbs, R.J., and L.F. Huenneke. 1992. Disturbance, diversity, and invasion: Implications for conservation. Conservation Biology 6:324–337. Holly, G.M., and I. Hampton. 1990. A stratified random-transect design for acoustic surveys of fish stocks. Canadian Journal of Fisheries and Aquatic Sciences 47:1282–1291. Hughes, F., P.M. Vitousek, and T. Tunison. 1991. Alien grass invasion and fire in the seasonal submontane zone of Hawai’i. Ecology 72:743–746. Hutchinson, J.T., and E.S. Menges. 2006. Evaluation of the invasiveness of non-native plants at the Archbold Biological Station. Florida Scientist 69:62–68. Just, M.G., M.G. Hohmann, and W.A. Hoffmann. 2017. Invasibility of a fire-maintained savanna–wetland gradient by non-native, woody plant species. Forest Ecology and Management 405:229–237. Kuenzi, A.M., P.Z. Fulé, and C.H. Sieg. 2008. Effects of fire severity and pre-fire stand treatment on plant community recovery after a large wildfire. Forest Ecology and Management 255:855–865. Lippincott, C.L. 2000. Effects of Imperata cylindrica (L.) Beauv. (Cogongrass) invasion on fire regime in Florida sandhill (USA). Natural Areas Journal 20:140–149. Long, A.J., D.D. Wade, and F.C. Beall. 2004. Managing for fire in the interface: Challenges and opportunities. Pp. 201–223, In S.W Vince, M.L. Duryea, E.A. Macie, and L.A. Hermansen (Eds.). Forests at the Wildland–Urban Interface. CRC Press, Boca Raton, FL. Lonsdale, W.M. 1999. Global patterns of plant invasions and the concept of invisibility. Ecology 80:1522–1536. Matlack, G.R. 2002. Exotic plant species in Mississippi, USA: Critical issues in management and research. Natural Areas Journal 22:241–247. McMurphy, W.E., and K.L. Anderson. 1965. Burning Flint Hills range. Journal of Range Management 18:265–269. Menges, E.S., and D.R. Gordon. 2010. Should mechanical treatments and herbicides be used as fire surrogates to manage Florida’s uplands? A review. Florida Scientist 73:147–174. Southeastern Naturalist 669 K.E. Tisshaw and E.S. Menges 2018 Vol. 17, No. 4 Menges, E.S., K.N. Main, R.L. Pickert, and K. Ewing. 2017. Evaluation of a firemanagement plan for fire-regime goals in a Florida landscape. Natural Areas Journal 37:212–227. Mitchell, R.J., J.K. Hiers, J.J. O’Brien, S.B. Jack, and R.T. Engstrom. 2006. Silviculture that sustains: The nexus between silviculture, frequent prescribed fire, and conservation of biodiversity in Longleaf Pine forests of the southeastern United States. Canadian Journal of Forest Research 36:2724–2736. Possley, J., and J. Maschinski. 2006. Competitive effects of the invasive grass Rhynchelytrum repens (Willd.) C.E. Hubb. on pine rockland vegetation. Natural Areas Journal 26:391–395. R Core Team. 2017. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria URL. Available online at https://www.Rproject. org/. Accessed 12 April 2018.. RStudio Team. 2016. RStudio: Integrated development for R. RStudio, Inc., Boston, MA. Available online at http://www.rstudio.com/. Accessed 12 April 2018.. Schmalzer, P.A., and T.E. Foster. 2016. Flora and threatened and endangered plants of Canaveral National Seashore, Florida. Castanea 81:91–127. Scott, J.M. 1913. Natalgrass. Press Bulletin 208. University of Florida Agricultural Experiment Station, Gainesville, FL. Spellerberg, I. 1998. Ecological effects of roads and traffic: A literature review. Global Ecology and Biogeography Letters 7:317–333. Stokes, C.A. 2010. Biology, ecology, and management of Natalgrass (Melinis repens). M.Sc. Thesis. University of Florida, Gainesvile, FL. 108 pp. Stokes, C.A., G.E. MacDonald, C. Reinhardt Adams, K.A. Langeland, and D.L. Miller. 2011. Seed biology and ecology of Natalgrass (Melinis repens). Weed Science 59:527–532. Sverdarsky, W.D., P.E. Buckley, and T.A. Feiro. 1986. The effect of 13 years of annual burning on an aspen–prairie ecotone in northwestern Minnesota. Pp. 118–122, In G.K. Clambey and R.H. Pemble (Eds.). Proceedings, 9th North American Prairie Conference— The Prairie: Past, Present, and Future. Tri-College University, Moorhead, MN, and Fargo, ND. Swain, H. 1998. Archbold Biological Station and the MacArthur Agro-Ecology Research Center. Bulletin of the Ecological Society of America 79:114–120. Tomimatsu, H., and M. Ohara. 2004. Edge effects on recruitment of Trillium camschatcense in small forest fragments. Biological Conservation 117:509–519. Towne, G., and C. Owensby. 1984. Long-term effects of annual burning at different dates in ungrazed Kansas tallgrass prairie. Journal of Range Management 37:392–397. Turner, I.M. 1996. Species loss in fragments of tropical rain forest: A review of the evidence. Journal of Applied Ecology 33:200–209. Turner, M.G., W.H. Roime, R.H. Gardner, and W.W. Hargrove. 1997. Effects of fire size and pattern on early succession in Yellowstone National Park. Ecological Monographs 67:411–433. Veldman, J.W., B. Mostacedo, M. Pena-Claros, and F.E. Putz. 2009. Selective logging and fire as drivers of alien grass invasion in a Bolivian tropical dry forest. Forest Ecology and Management 258:1643–1649. Wace, N. 1977. Assessment of dispersal of plant species: The carborne flora in Canberra. Proceedings of the Ecological Society of Australia 10:167–186. Wade, D., J. Ewel, and R. Hofstetter. 1980. Fire in South Florida ecosystems. US Department of Agriculture. Forest Service General Technical Report SE-17. Asheville, NC. Southeastern Naturalist K.E. Tisshaw and E.S. Menges 2018 Vol. 17, No. 4 670 Weekley, C.W., E.S. Menges, and R.L. Pickert. 2008. An ecological map of Florida’s Lake Wales Ridge: A new boundary delineation and an assessment of post-Columbian habitat loss. Florida Scientist 71:45–64. Williges, K., P. Kubilis, J. Ellenberger, P. Hansen, and M. McMillan. 2006. Effects of management regimes on restoring long-unburned scrub to an early-successional structure favored by obligate scrub species. Florida Fish and Wildlife Research Institute, Upland Habitat Research and Monitoring Subsection, Gainesville, FL. Willson, G.D., and J. Stubbendieck. 1996. Suppression of Smooth Brome by atrazine, mowing, and fire. The Prairie Naturalist 28:13–20. Yates, E.D., D.F. Levia, and C.L. Williams. 2004. Recruitment of three non-native invasive plants into a fragmented forest in southern Illinois. Forest Ecology and Management 190:119–130. Young, C.C., and E.S. Menges. 1999. Gap-phase dynamics and microhabitat specialization: Post-fire regeneration of scrubby flatwoods communities on the Lake Wales Ridge. Florida Scientist 62:1–10.