Predicting the Presence and Cover of Management Relevant Invasive Plant Species on Protected Areas

Home , Brooker Creek Preserve, Bulow Creek State Park, Deer Lake State Park, Eden Gardens State Park, Grayton Beach State Park, Turkey Creek Sanctuary

Title: Predicting the presence and cover of management relevant invasive plant species on protected areas

Gwenllian Iaconaa∗ , [email protected] Franklin D. Priceb, [email protected] Paul R. Armswortha, [email protected] a Ecology and Evolutionary Biology The University of Tennessee 569 Dabney Hall Knoxville, TN 37996 U.S.A. b Florida Natural Areas Inventory 1018 Thomasville Road, Suite 200-C Tallahassee, Florida 32303 USA

∗ Corresponding author

Full contact details for corresponding author: Gwen Iacona Present address: Centre for Excellence in Environmental Decisions Room 523, Goddard Bldg 8, University Of Queensland, ST LUCIA, QLD, 4072 Australia Tel.: +61 (0) 424 578 995

Type of article: original research articles Number of words in abstract: 266 Number of words in article: 3690 Number of references: 52 Number of Figures: 2 Number of Tables: 3

1 1 Abstract

2 Invasive species are a management concern on protected areas worldwide. 3 Conservation managers need to predict infestations of invasive plants they aim to treat 4 if they want to plan for long term management. Many studies predict the presence of 5 invasive species, but predictions of cover are more relevant for management. Here we 6 examined how predictors of invasive plant presence and cover differ across species that 7 vary in their management priority. To do so, we used data on management effort and 8 cover of invasive plant species on central Florida protected areas. Using a zero-inflated 9 multiple regression framework, we showed that protected area features can predict the 10 presence and cover of the focal species but the same features rarely explain both. 11 There were several predictors of either presence or cover that were important across 12 multiple species. Protected areas with three days of frost per year or fewer were more 13 likely to have occurrences of four of the six focal species. When invasive plants were 14 present, their proportional cover was greater on small preserves for all species, and 15 varied with surrounding household density for three species. None of the predictive 16 features were clearly related to whether species were prioritized for management or 17 not. Our results suggest that predictors of cover and presence can differ both within 18 and across species but do not covary with management priority. We conclude that 19 conservation managers need to select predictors of invasion with care as species 20 identity can determine the relationship between predictors of presence and the more 21 management relevant predictors of cover.

22 Keywords:

23 conservation costs; Imperata cylindrica; Schinus terebinthifolius; Ludwigia peruviana; 24 Lygodium microphyllum; Urena lobata

25 1 Introduction

26 Invasive plant control is a management cost on protected areas (PAs) worldwide 27 (Frazee et al 2003; Goodman 2003; Reinhardt et al 2003; Pimentel et al 2005; 28 Pfennigwerth and Kuebbing 2012; Cleary 2013). Thus, predictions of invasion across 29 PAs are necessary to budget for long term management (Martin and Blossey 2012). 30 Such estimates contribute to regional scale conservation decisions (e.g. ranking sites for 31 potential acquisition) yet, many of the current predictive models assume all invasive 32 species are prioritized equally for management and predict species presence without 33 considering cover (Py˘seket al 2002a;b; Foxcroft et al 2011).

2 34 Conservation managers need estimates of both presence and cover. Predictions of 35 presence are useful for identifying possible invaders and planning for monitoring 36 (Catford et al 2012). Meanwhile, predictions of cover estimate the effort required to 37 manage an invasion. However, few studies consider cover (Alston and Richardson 2006; 38 Catford et al 2011; Polce et al 2011; Seabloom et al 2013), or how predictions of 39 presence relate to cover (but see Kuhman et al 2010). This shortcoming hinders the 40 application of existing studies to management. 41 Invasive plant species may co-occur on PAs, but they may not all be 42 management priorities (Allen et al 2009; Kuebbing et al 2013; Abella 2014). Cost can 43 constrain the management available at a PA (D’Antonio et al 2004; Tempel et al 2004), 44 inducing managers to prioritize species that respond better to treatment. 45 Alternatively, managers could aim to minimize invasive species impact and prioritize 46 treatment of species that negatively affect species of conservation interest. Finally, 47 managers may prioritize species that have historically been treated at the PA or whose 48 management is specifically mandated in management plans (Pullin and Knight 2005). 49 Regardless of why prioritization occurs, conservation planners need predictions of 50 presence and cover that are robust across species that are a management priority. 51 The potential variation across predictors of presence, cover, and management 52 priority, suggests that comparisons of invasion across species must consider three 53 contrasts (Figure 1). The first contrast is an examination of predictor variables across 54 species. The second contrast is the difference between predictors of presence and cover 55 for a single species. The final is a comparison of shared predictors for species that are a 56 high management priority (those that are likely to be treated) with shared predictors 57 of species that are low management priority (and potentially inflating estimates of 58 treatment need). 59 Previously, we related PA features to the proportional summed cover of the 60 dominant invasive species on the PA (Iacona et al 2014). Our predictors of invasion 61 included ecological attributes related to invasibility (Richardson 2011), and features 62 related to human disturbance. The best predictors of the invadedness of a PA were PA 63 size and the number of surrounding households (other variables tested include 64 minimum temperature, road density, and elevation). 65 We now examine the relationship between PA-level features and the presence and 66 cover of individual species that differ in management priority in central Florida. These 67 species are noxious invaders according to the Florida Exotic Pest Plant Council 68 (FLEPPC), and are documented as causing ecological harm (FLEPPC Category I 69 invaders) or increasing in abundance or frequency within the state (Category II 70 invaders). The FLEPPC categories roughly correspond with management priority in 71 Florida and we tailor our choice of study species by drawing on management records. 72 Variation in PA-level predictors of presence or cover is expected across species because 73 of different life histories and distributions. However, we seek to understand variation 74 within each species when predicting presence versus cover, and also whether predictors 75 across species relate to their prioritization for management. 76 We expect PA features that predict the presence of a species to differ from those 77 that predict cover. For instance, we expect the probability of presence to increase with

3 Presence Cover

PA area elevation no frost PA area houses roads no frost Schinus terebinthifolius + - + + Management Imperata cylindrica + + + - + + priority Lygodium microphyllum + + + - -

Ludwigia peruviana + + + - - Not a Urena lobata - management

priority Panicum maximum - +

Figure 1: Diagram outlining study design and results. We examine the site-level features that predict the presence and cover of six different invasive plant species in Florida. These species were chosen based on their abundance and management priority. The sign in the table indicates the relationship described by significant predictors in the multiple regression models relating invasive presence and cover to site-level features. The predictive site-level features are listed along the top of the table. See Table 2 for model coefficient values.

78 PA size because of species area effects (McKinney 2002), with surrounding households 79 because of human introductions (Gavier-Pizarro et al 2010), and with road density 80 because of propagule transport (Von Der Lippe and Kowarik 2007). We also expect 81 probability of presence to increase at lower latitudes (Marini et al 2009) because many 82 of the species that are invasive in central Florida are frost sensitive. Finally, habitat 83 type is likely to influence species presence (Chytrýet al 2008), so we include PA 84 elevation as a proxy for broad habitat classes such as floodplain forest (low elevation), 85 wet flatwood (medium elevation) or scrub (high elevation) (Myers and Ewel 1990). In 86 contrast, we expect proportional cover to decrease with PA size, increase with 87 household density, and potentially vary with winter temperature (Iacona et al 2014). 88 Predictive features may also be similar for species that are management 89 priorities. For instance, PA size is a predictor that could relate to management priority 90 if species that are more likely to be found on small PAs, because of disrupted ecological 91 processes, are also more likely to be managed because they are obvious or noxious. 92 Alternatively, species of management priority may be more likely on PAs close to 93 human development because species that are introduced as ornamentals may be more 94 likely to be invaders in natural areas than species that are agricultural pests 95 (Richardson and Rejmnek 2011; Zenni 2014). 96 Here we develop models that concurrently predict presence and cover for six 97 species that are prevalent on PAs in central Florida but differ in their priority for

4 98 management. We ask:

99 1. Are predictors of presence and cover similar for a species?

100 2. Do predictors of presence and cover vary across species in relation to priority for 101 management?

102 2 Methods

103 2.1 Study system

104 To address these questions we needed a comprehensive dataset of the relative 105 abundance and treatment eﬀort of invasive plants on PAs. There has been an extensive 106 recent eﬀort to quantify the extent of invasion on publicly-owned PAs within the state 107 of Florida, USA (Cleary 2007) which provides an ideal study system. Florida is heavily 108 impacted by invasion, and there are 159 species on the FLEPPC list of invasive plant 109 species (FLEPPC list 2015, http://www.fleppc.org/list/list.htm). The more 110 than 1800 publicly-owned PAs within the state range from temperate to tropical 111 climates, urban to rural locations, and small to large area (Florida Natural Areas 112 Inventory(FNAI), Florida Managed Areas (FLMA) database).

113 2.2 Study species and sites

114 For this study we selected six focal species, grouped according to priority for 115 management, from the 159 FLEPPC listed species in the state. To do so, we examined 116 the distribution of management effort across FLEPPC listed species using an 117 operations database from the Florida Fish and Wildlife Conservation Commission 118 (FWC) upland habitat management program (Cleary 2007). We calculated the 119 proportion of effort applied to each species (by acreage or by number of individuals 120 treated, depending on PA) to rank species by treatment effort (Supplementary Table 121 SI1). We then used the ranking to choose the six study species highlighted in Figure 122 2. Schinus terebinthifolius (Brazilian pepper, introduced 1840 (Hight et al 2002)), 123 Imperata cylindrica (cogon grass, introduced 1921 (Dozier et al 1998)), and Lygodium 124 microphyllum (Old World climbing fern, introduced 1958 (Langeland and Hutchinson 125 2013)) are prioritized for treatment. Meanwhile, Ludwigia peruviana (Peruvian 126 primrose-willow, introduced by 1877 (Kunzer, J., personal communication), Urena 127 lobata (Caesar’s weed, introduced 1897 (Austin 1999)), and Panicum maximum 128 (Guinea grass, introduced 1933 (FNAI 2014)) are widely prevalent on public 129 conservation lands but are a lower management priority.

5 250 5000 ● Schinus terebinthifolius Schinus terebinthifolius ● 200 4000

Urena lobata ● 150 3000 ● Ludwigia peruviana Imperata cylindrica ● ● Lygodium microphyllum Ludwigia peruviana Cover (HA) Cover ● Urena lobata ● Panicum maximum ● Number of PAs ● ●

100 ●

2000 ● ● ● ● ● Imperata cylindrica ● ● ● ● ● ● ● ● 6 ● ● ● ● ●

50 Lygodium microphyllum ● ● 1000 ● ● ● ● ● ● ● ● ● Panicum maximum● ● ● ● ● ● ● ● ●

● ● ● ● 0 0

−8 −6 −4 −2 0 −8 −6 −4 −2 0

log(Proportional effort) log(Proportional effort)

(a) (b)

Figure 2: Variation in management effort and cover on protected areas (PAs) across the 28 species in our occurrence database (Supplementary Table 2). For this study we chose six focal species (filled circles) that differ in the amount of effort that is allocated to their treatment on PAs across Florida. This figure displays effort for each species in relation to (a) cover on 365 PAs and (b) the number of PAs it occurs on. Schinus terebinthifolius, Lygodium microphyllum and Imperata cylindrica are high priority species for management and also have high levels of cover across the state. Ludwigia peruviana, and Urena lobata are low priority for management but have high cover. Panicum maximum is a low priority for management but is present on many PAs. 130 We obtained distribution data for these six species from the FLInv geodatabase. 131 This database was commissioned by the Florida Fish and Wildlife Conservation 132 Commission (FWC) to improve their prioritization of invasive species management 133 funds and is maintained by the Florida Natural Areas Inventory (FNAI; database 134 available at http://www.fnai.org/invasivespecies.cfm). Here we use the PA as the 135 analysis replicate to represent regional decision making. We chose to use only records 136 from PAs where data were collected by FNAI botanists between the years of 2008 and 137 2010 to standardize data collection protocols. We also chose PAs where all records for 138 that species were GPS point data with surveyor-estimated area of coverage to improve 139 the predictive capacity of our models (see Iacona et al 2014). The final dataset includes 140 presence and absence information for six focal species on 365 PAs across Florida 141 (Supplementary Information). Our criteria for species choice and data quality resulted 142 in a sample of species that are primarily observed on PAs in central Florida. These 143 PAs were slightly smaller (1st Q = 13.9 HA, Median = 62 HA, 3rd Q = 450 HA) than 144 all the PAs in Florida (from FLMA dataset: 1st Q = 15.42, Median = 78 HA, 3rd Q = 145 475 HA). While these records were a subset of the possible species and the entire 146 network of PAs within the state, it was still a large sample spanning gradients of PA 147 features. The limitations of this sample must be balanced against the desirability of 148 having all surveys conducted by one agency with standardized reporting protocols.

149 2.3 Protected area features

150 We used predictive features that relate to the presence and cover of invasive species 151 and that can be obtained from publicly available datasets. The first three factors relate 152 to ecological function and community composition. PA size information was obtained 153 from the FLMA dataset of PAs managed for conservation. We used elevation as a 154 proxy for plant community type because, in Florida, very small changes in elevation 155 correspond with obvious changes in vegetation (Myers and Ewel 1990), and it was a 156 solution to the statistical challenge of adding community type as a categorical variable 157 while still maintaining a balanced design. We derived PA average elevation from USGS 158 NED 1/3 arc second data layers at 1 m resolution. Winter PA temperature was 159 represented by a binary variable (frost-bin) indicating three days of frost per year or 160 not, as calculated from Daymet weather data for the state of Florida (Thornton et al 161 1997). The last two factors relate to anthropogenic disturbance at a PA. We estimated 162 the number of nearby households by weighting the number of households in nearby 163 year 2000 census-tracts by their overlap with a 25 km buffer around the PA. This 164 buffer distance was chosen to account for the impacts of low density human 165 development on invasion (Gavier-Pizarro et al 2010). We calculated road density as a 166 proxy of onsite disturbance by dividing area of roads by PA size for all roads that 167 intersected or were adjacent to the PA (USGS 24000:1 roads layer). We estimated 168 average road width to be 10 m by assuming that two lane roads likely abut PAs.

7 169 2.4 Modeling approach

170 Each species was present on only some of the 365 PAs (Supplementary Figure SI1) and 171 we were interested in both the probability of occurrence of a species on a PA and its 172 abundance if it was present. To model the two processes concurrently, we used a 173 zero-inﬂated negative binomial model (Zeileis et al 2008) to predict expected 174 proportional invasive cover at a PA in relation to PA features. We chose a negative 175 binomial model because our observed variance in invasive plant cover was greater than 176 the mean (overdispersion) and because AIC comparisons during the model-ﬁtting 177 process indicated that this error structure was the most appropriate. 178 Our modeling strategy considers the entire dataset to be binary data and models 179 the probability of the species presence at a given PA, assuming a binomial 180 distribution:

eα+β1X1i+...βnXni probability of presence = πi = 1 − ( ) (1 + eα+β1X1i+...βnXni

181 Here, α is the estimated model intercept and β is the ﬁtted coeﬃcient for the site 182 level factors Xi. Our model also uses the observed cover measurements and some of the 183 zero cover records to relate the mean cover of a species at a PA to site-level features, 184 assuming a negative binomial distribution:

α+β1X1i+...βnXni mean cover if present = µi = e

185 We can then calculate the expected value of the mean cover at a PA while 186 accounting for the zero-inﬂated process (mean cover = E(Yi) = µi ∗ πi). This 187 prediction of cover at a PA estimates expected cover weighted by the probability of the 188 species being present. 189 To construct our response variable, we binned proportional cover for each species 190 into thousandth of a percent bins to meet model assumptions of a discrete data 191 distribution (Table 1). We constructed separate models for each of the six species using 192 the pscl package in R (Zeileis et al 2008; R Development Core Team 2010; Jackman 193 2012). All models used a log link function to model the underlying proportional 194 invasive cover distribution, and a logit link function to model the excess zeros. We log 195 transformed all predictor variables, except frost-bin, to improve the model 196 appropriateness for the observed distributions of individual predictors. There was 197 moderate correlation between some predictor variables (e.g., negative correlation 198 between frost bin and elevation), but none that exceeded acceptable thresholds for 199 model assumptions (Dormann et al 2013). In addition, tolerance testing indicated that 200 the variation explained by each of the predictor variables was not more than 20%

8 201 dependent on variation in other predictor variables, ensuring that multicollinearity 202 requirements were adequate to proceed (Quinn and Keough 2001). Finally, 203 examination of semivariance plots of residuals from each model indicated that 204 remaining spatial autocorrelation was not a concern in this analysis.

205 This zero-inﬂated modeling technique considers records of zero cover to be one of 206 two types. If a species cover is zero it could be due to characteristics of the PA that 207 precluded the species presence (e.g., outside its range of temperature tolerance), or it 208 could be due to characteristics of the PA that relate to low amounts of cover (e.g., low 209 propagule pressure minimizing establishment). Hence, we present the results in two 210 sections even though they were produced in a single modeling process. The ﬁrst 211 describes the predictive features that relate to the probability that the species of 212 interest is present at the PA. The second describes the predictive features that relate to 213 the expected proportional cover of the species if it is present. For a prediction of 214 expected invasive plant proportional cover for a species at a PA, multiply the 215 estimated mean proportional cover (Table 2) by the probability of the PA having more 216 than zero cover present (Table 3).

217 3 Results

218 Features that proxy for ecological characteristics were most important for predicting 219 species presence in central Florida (Table 2). PAs with three frost days or fewer per 220 year were more likely to have occurrences of S. terebinthifolius, L. peruviana, I. 221 cylindrica, and L. microphyllum. In addition, the probability that I. cyclindrica, L. 222 peruviana, and L. microphyllum were present increased as PA mean elevation 223 increased. Meanwhile, PA features that relate to human disturbances were less 224 important in predicting the presence of a species. Although I. cylindrica and L. 225 microphyllum were more likely to be present as the PA size increased, households and 226 roads were not significant predictors of presence for any tested species. Finally, no PA 227 features predicted P. maximum or U. lobata presence. 228 In contrast, the PA-level features predicting proportional cover included both 229 those that related to ecological characteristics and those that related to human 230 disturbance (Table 3). The cover of all six species decreased as PA size increased. 231 However, the species differed in their relationship to household density. S. 232 terebinthifolius and I. cylindrica decreased in cover as the number of nearby 233 households decreased while the cover of L. peruviana increased. Finally, L. 234 microphyllum, P. maximum and U. lobata cover decreased as PA size increased but 235 was not related to household density. In addition, frost-bin was a significant predictor 236 of cover for S. terebinthifolius and P. maximum, with higher cover at PAs with three or 237 fewer frost days per year. Road density was related to increased cover for I. cylindrica 238 and decreased cover for L. microphyllum.

9 239 Outliers were present in the models of each species, but we think the information 240 provided by these highly invaded PAs was meaningful (very low proportional cover 241 outliers were absorbed by the zero-inﬂated process). Therefore we present the results 242 with all data included.

243 4 Discussion

244 This study examines how predictors of invasive plant cover and presence on PAs vary 245 with species identity and management priority in central Florida. This approach is 246 valuable from a conservation perspective because the fitted coefficients from models 247 such as ours permit regional planners to estimate invasion across PAs, and thus 248 management need, even if they only have information on PA features. For regional 249 level planning, such predictive models are useful for coarse decisions - such as refining a 250 list of potential acquisitions - and can postpone costly plant surveys. Here we find that 251 predictive features differ for presence and cover within and across species, suggesting 252 that predictions of presence and cover are not interchangeable. In addition, although 253 we identify predictors that are important across multiple species, no predictors related 254 to management priority species specifically. 255 There are many possible metrics of invasion (Richardson 2011), but we have 256 chosen to focus on presence and cover because of their management relevance. 257 Predictors of the presence of a species were often quite different than predictors of 258 cover in central Florida. For instance, in several models, the probability of presence of 259 a species increased with PA size, but the proportional cover decreased. This result 260 suggests that the two factors respond differently to site area and reinforces the call for 261 predictions of cover as well as presence for management applications (Catford et al 262 2012; Bradley 2013; Seabloom et al 2013). 263 The conservation implications of this discrepancy are relevant for management. 264 For instance, some species can be managed to minimize presence on PAs, such as when 265 the early detection and response to L. microphyllum is prioritized on the invasion front 266 in central Florida (Hutchinson et al 2006). Meanwhile other species are managed to 267 minimize cover, such as Melaleuca quinquenervia in the Florida Everglades 2014, 268 CISRERP. Notably, predictors of presence tended to be features related to ecological 269 characteristics. In contrast, the proportional cover of the species was more often 270 related to features that indicated human disturbance on a PA.

271 Speciﬁc PA-level predictors of presence and cover are important if a planner 272 wants to plan for management of one of our six study species. However, several 273 predictive features were common across many of the models, and we suggest that those 274 predictors could be appropriate for modeling invadedness of a PA regardless of species. 275 Across species in central Florida, the most important predictor of presence was 276 whether there were three or fewer frost days per year at the PA. Meanwhile, for

10 277 estimates of cover, these factors were PA size and the number of houses within 25 km 278 of the PA. When interpreting these results, it should be observed that our models were 279 parameterized on a subset of PAs that excluded the very largest sites. We suggest that 280 extrapolating our results to the very largest PAs (e.g. the 6 PAs in Florida that are 281 larger than 100,000 HA) should be done with caution. 282 Finally, we show that invasion by species of high management priority in central 283 Florida is not predicted by diﬀerent features than those that are low management 284 priority. Instead, both presence and cover of all six of our focal species were related to 285 slightly diﬀerent factors. These associations were reasonable based on the physiology 286 and life history of the individual species, as we discuss in section 4.1 below, but there 287 was no clear grouping of factors related to management priority.

288 4.1 Species speciﬁc results

289 4.1.1 High management priority species

290 Schinus terebinthifolius is a documented management priority in peninsular Florida 291 (Cuda et al 2006). It is found on the most PAs throughout the state, covers the most 292 area, and is allocated the most eﬀort and funding of species that are treated in Florida 293 (Supplementary Table SI1). It is range restricted in Florida and its presence at a PA 294 was almost entirely related to frost free days. However, on PAs where the species was 295 present, the proportional cover of S. terebinthifolius decreased as PA size increased, as 296 household density decreased, and when there were more than three days with frost. 297 Imperata cylindrica is also a management priority in Florida. Our model suggests 298 that its proportional cover is correlated with road density. This is not surprising 299 because I. cylindrica is a perennial rhizomatous grass that is primarily vegetatively 300 dispersed (Dozier et al 1998). The species is common along roadsides, as the rhizomes 301 are often transported in road ﬁll and by grading equipment (Jose et al 2002). In 302 addition, the positive relationship between the presence of I. cylindrica and elevation is 303 reasonable because this species is able to tolerate hot, dry conditions and is one of the 304 few species that will invade upland pine (Yager et al 2010) and scrub communities. 305 The presence of Lygodium microphyllum was positively related to elevation. This 306 may be related to its prevalence on large inland PAs in central Florida (Ferriter and 307 Pernas 2006). Its proportional cover also decreased with increased road cover. This 308 could be because, as a statewide management priority (Hutchinson et al 2006), it is 309 likely to be intensively treated in accessible areas. However, these results should be 310 interpreted with care because this species was present on only 18 PAs in our study and 311 our data selection criteria excluded several of the largest parks in south Florida where 312 this species is a primary pest.

11 313 4.1.2 Low management priority species

314 Our model suggests that the proportional cover of Ludwigia peruviana decreases with 315 proximity to human households. This wetland species is prevalent in large, shallow 316 wetlands produced by water control projects in south Florida (Toth 2010). These 317 wetlands may be less common in developed areas. In addition, the probability of 318 presence of L. peruviana is positively related to elevation. This seems counterintuitive 319 for a wetland species, but if it prefers the types of wetlands that are present in the 320 interior of the state, where there is higher elevation and greater distance from the 321 human development along the coast, both patterns would hold. 322 Finally, the presence of both Panicum maximum and Urena lobata was not 323 strongly related to PA features. In addition, the proportional cover of neither of these 324 species was well explained by the density of surrounding households, although they 325 both did decrease in cover as PA size increased. This is probably because both are 326 common ruderal species, with near ubiquitous presence on PAs in central Florida 327 (Austin 1999).

328 4.2 Extensions

329 We were particularly interested in features that predicted invasion across species of 330 management interest. However, it would be possible to refine the model to further 331 explore drivers behind species specific patterns, for instance PA area and access and L. 332 microphyllum. In addition, we did not find common site-level predictors across species 333 with similar management priority, but it is possible that other characteristics of these 334 species could meaningfully predict their presence and cover. For instance, time since 335 invasion is a species characteristic that could correlate with management priority. 336 Finally, one important potential predictor that we do not include in our models is a 337 measure of how much effort has been allocated to management at the site. There is 338 highly resolved cost data available for many of these PAs (Iacona et al 2014). However, 339 the invasion data is from a single sampling time and the relationship between observed 340 invasion and spend can go both ways (spending is higher at sites with more invasion, 341 but higher spending should reduce invasion). This endogeneity problem limits the 342 scope for including these data in our current models.

343 4.3 Conclusions

344 Invasive plant species management is often a priority for biodiversity conservation, and 345 we conﬁrm that species identity can be important from a planning perspective. Our 346 work suggests that certain PA features robustly predict presence and cover across 347 species, at least in central Florida, but that these predictors are not clearly related to

12 348 management priority. If such predictions are intended for conservation decision making 349 at a regional scale (e.g. if assessing the possible consequences of pursuing agency-wide 350 policies on minimum reserve sizes), it is important to understand the variation in 351 network wide trends of presence and proportional cover. In such cases, a model such as 352 ours, that uses cheap, easily obtainable coarse grain data to predict expected variation 353 in presence and cover at the scale of a PA would be appropriate.

354 4.4 Acknowledgments

355 We would like to thank R. Cleary at FWC for data, and T. Martin at CSIRO for 356 statistical assistance. GI was supported by the National Institute for Mathematical 357 and Biological Synthesis, an institute sponsored by the National Science Foundation, 358 the U.S. Department of Homeland Security, and the U.S. Department of Agriculture 359 through NSF Award #EF-0832858, with additional support from UTK department of 360 Ecology and Evolutionary Biology and the ARC Center of Excellence for 361 Environmental Decisions. We are also grateful to members of the Armsworth lab, S. 362 Wiggers, S. Kuebbing, D. Simberloﬀ, L. Gross, D. Hodges, and anonymous reviewers 363 for helpful comments on earlier versions of this manuscript.

13 364 References

365 Abella SR (2014) Eﬀectiveness of exotic plant treatments on national park service 366 lands in the United States. Invasive Plant Science and Management 7(1):147–163

367 Allen JA, Brown CS, Stohlgren TJ (2009) Non-native plant invasions of United States 368 National Parks. Biological Invasions 11(10):2195–2207

369 Alston KP, Richardson DM (2006) The roles of habitat features, disturbance, and 370 distance from putative source populations in structuring alien plant invasions at the 371 urban/wildland interface on the Cape Peninsula, South Africa. Biological 372 Conservation 132:183–198

373 Austin D (1999) Caesar’s weed (Urena lobata) - an invasive exotic, or a Florida native? 374 Wildland Weeds Winter:13–16

375 Bradley B (2013) Distribution models of invasive plants over-estimate potential 376 impact. Biological Invasions 15(7):1417–1429

377 Catford JA, Vesk PA, White MD, Wintle BA (2011) Hotspots of plant invasion 378 predicted by propagule pressure and ecosystem characteristics. Diversity and 379 Distributions 17(6):1099–1110

380 Catford JA, Vesk PA, Richardson DM, Pyek P (2012) Quantifying levels of biological 381 invasion: towards the objective classiﬁcation of invaded and invasible ecosystems. 382 Global Change Biology 18(1):44–62

383 ChytrýM, Jaros´ıkV, Py˘sekP, HájekO, KnollováI, TichýL, Danihelka J (2008) 384 Separating habitat invasibility by alien plants from the actual level of invasion. 385 Ecology 89(6):1541–1553

386 Cleary RL (2007) Controlling upland invasive exotic plants on public conservation 387 land: a strategic plan. Natural Areas Journal 27(3):218–225

388 Cleary RL (2013) A comparitive analysis of state programs and funding for terrestrial 389 invasive plant control. Tech. rep., Florida Fish and Wildlife Conservation 390 Commission, upland invasive plant management program

391 Committee on independent scientiﬁc review of Everglades restoration progress (2014) 392 Progress toward restoring the everglades: The ﬁfth biennial review, 2014. Tech. rep., 393 The National Academies Press

394 Cuda J, Ferriter A, Manrique V, Medal J (2006) Interagency Brazilian peppertree 395 (Schinus terebinthifolius) management plan for Florida. Tech. rep., Florida Exotic 396 Pest Plant Council

397 D’Antonio CM, Jackson NE, Horvitz CC, Hedberg R (2004) Invasive plants in 398 wildland ecosystems: merging the study of invasion processes with management 399 needs. Frontiers in Ecology and the Environment 2(10):513–521

14 400 Dormann CF, Elith J, Bacher S, Buchmann C, Carl G (2013) Collinearity: a review of 401 methods to deal with it and a simulation study evaluating their performance. 402 Ecography (Copenhagen) 36(1):27–46, DOI 10.1111/j.1600-0587.2012.07348.x

403 Dozier H, Gaﬀney JF, McDonald SK, Johnson ERRL, Shilling DG (1998) Cogongrass 404 in the United States: history, ecology, impacts, and management. Weed Technology 405 12(4):737–743

406 Ferriter A, Pernas T (2006) An explosion in slow motion: tracking the spread of 407 Lygodium microphyllum in Florida. Wildland Weeds Spring 2006:7–9

408 FNAI (2014) Panicum maximum. URL 409 http://fnai.org/invasives/panicum_maximum_fnai.pdf

410 Foxcroft LC, Jaro˘s´ık V, Py˘sekP, Richardson DM, Rouget M (2011) Protected-area 411 boundaries as ﬁlters of plant invasions. Conservation Biology 25(2):400–405

412 Frazee SR, Cowling RM, Pressey RL, Turpie JK, Lindenberg N (2003) Estimating the 413 costs of conserving a biodiversity hotspot: a case-study of the Cape Floristic Region, 414 South Africa. Biological Conservation 112(12):275–290

415 Gavier-Pizarro GI, Radeloﬀ VC, Stewart SI, Huebner CD, Keuler NS (2010) Housing is 416 positively associated with invasive exotic plant species richness in New England, 417 USA. Ecological Applications 20(7):1913–1925

418 Goodman PS (2003) Assessing management eﬀectiveness and setting priorities in 419 protected areas in KwaZulu-Natal. Bioscience 53(9):843–850

420 Hight S, Cuda J, Medal J (2002) Brazilian peppertree. in: Biological control of invasive 421 plants in the eastern united states. Report, United States Department of Agriculture 422 Forest Service. Forest Health Technology Enterprise Team

423 Hutchinson J, Ferriter A, Serbesoﬀ-King K, Langeland K, Rodgers L (2006) Old 424 weverorld climbing fern (lygodium microphyllum) management plan for Florida. 425 Tech. rep., Florida Exotic Pest Plant Council, Lygodium Task Force

426 Iacona GD, Price FD, Armsworth PR (2014) Predicting the invadedness of protected 427 areas. Diversity and Distributions 20(4):430–439

428 Jackman S (2012) pscl: Classes and methods for R developed in the political science 429 computational laboratory, stanford university

430 Jose S, Cox J, Miller DL, Schilling DG, Merrit S (2002) Alien plant invasions: the 431 story of cogongrass in southeastern forests. Journal of Forestry 100(1):41–44

432 Kuebbing SE, NuñezMA, Simberloff D (2013) Current mismatch between research and 433 conservation efforts: the need to study co-occurring invasive plant species. Biological 434 Conservation 160(0):121–129

15 435 Kuhman TR, Pearson SM, Turner MG (2010) Eﬀects of land-use history and the 436 contemporary landscape on non-native plant invasion at local and regional scales in 437 the forest-dominated southern appalachians. Landscape Ecology 25(9):1433–1445

438 Langeland K, Hutchinson J (2013) Natural area weeds: Old world climbing fern 439 (lygodium microphyllum). Report, University of Florida IFAS Extension

440 Marini L, Gaston KJ, Prosser P, Hulme PE (2009) Contrasting response of native and 441 alien plant species richness to environmental energy and human impact along alpine 442 elevation gradients. Global Ecology and Biogeography 18:652–661

443 Martin LJ, Blossey B (2012) Invasive plant cover impacts the desirability of lands for 444 conservation acquisition. Biodiversity and Conservation 21(8):1987–1996

445 McKinney ML (2002) Inﬂuence of settlement time, human population, park shape and 446 age, visitation and roads on the number of alien plant species in protected areas in 447 the usa. Diversity and Distributions 8:311–318

448 Myers RL, Ewel J John (1990) Ecosystems of Florida. University of Central Florida 449 Press, Orlando

450 Pfennigwerth AA, Kuebbing SE (2012) Direct costs associated with invasive non-native 451 plants in Tennessee. Wildland Weeds 15(3-4):4–6

452 Pimentel D, Zuniga R, Morrison D (2005) Update on the environmental and economic 453 costs associated with alien-invasive species in the United States. Ecological 454 Economics 52(3):273–288

455 Polce C, Kunin WE, Biesmeijer JC, Dauber J, Phillips OL, The Alarm Field Site 456 Network (2011) Alien and native plants show contrasting responses to climate and 457 land use in europe. Global Ecology and Biogeography 20(3):367–379

458 Pullin AS, Knight TM (2005) Assessing conservation management’s evidence base: a 459 survey of management-plan compilers in the United Kingdom and Australia. 460 Conservation Biology 19(6):1989–1996

461 Py˘sekP, Jaro˘s´ıkV, Ku˘ceraT (2002a) Patterns of invasion in temperate nature 462 reserves. Biological Conservation 104(1):13–24

463 Py˘sekP, Ku˘ceraT, Jaro˘s´ıkV (2002b) Plant species richness of nature reserves: the 464 interplay of area, climate and habitat in a central european landscape. Global 465 Ecology and Biogeography 11(4):279–289

466 Quinn G, Keough MJ (2001) Experimental design and data analysis for biologists. 467 Cambridge University Press, Cambridge

468 R Development Core Team (2010) R: a language and environment for statistical 469 computing

16 470 Reinhardt F, Herle M, Bastiansen F, Streit B (2003) Economic impact of the spread of 471 alien species in Germany. Tech. rep., J.W. Goethe-University Frankfurt/Main 472 Biological and Computer Sciences Division Department of Ecology and Evolution

473 Richardson DM (2011) Fifty years of invasion ecology: the legacy of Charles Elton. 474 Wiley-Blackwell, Chichester

475 Richardson DM, Rejmnek M (2011) Trees and shrubs as invasive alien species a global 476 review. Diversity and Distributions 17(5):788–809

477 Seabloom E, Borer ET, Buckley Y, Cleland WE, Davies K, Firn J, Harpole WS, 478 Hautier Y, Lind E, MacDougall A, Orrock JL, Prober SM, Adler P, Alberti J, 479 Anderson TM, Bakker JD, Biederman LA, Blumenthal D, Brown CS, Brudvig LA, 480 Caldeira M, Chu C, Crawley MJ, Daleo P, Damschen EI, D’Antonio CM, 481 DeCrappeo NM, Dickman CR, Du G, Fay PA, Frater P, Gruner DS, Hagenah N, 482 Hector A, Helm A, Hillebrand H, Hofmockel KS, Humphries HC, Iribarne O, Jin VL, 483 Kay A, Kirkman KP, Klein JA, Knops JMH, La Pierre KJ, Ladwig LM, Lambrinos 484 JG, Leakey ADB, Li Q, Li W, McCulley R, Melbourne B, Mitchell CE, Moore JL, 485 Morgan J, Mortensen B, O’Halloran LR, Prtel M, Pascual J, Pyke DA, Risch AC, 486 Salguero-Gmez R, Sankaran M, Schuetz M, Simonsen A, Smith M, Stevens C, 487 Sullivan L, Wardle GM, Wolkovich EM, Wragg PD, Wright J, Yang L (2013) 488 Predicting invasion in grassland ecosystems: is exotic dominance the real 489 embarrassment of richness? Global Change Biology 19(12):3677–3687

490 Tempel DJ, Gilimburg AB, Wright V (2004) The status and management of exotic and 491 invasive species in national wildlife refuge wilderness areas. Natural Areas Journal 492 24(4):300–306

493 Thornton P, Running S, White M (1997) Generating surfaces of daily meteorological 494 variables over large regions of complex terrain. Journal of Hydrology 190:214–251, 495 URL http://dx.doi.org/10.1016/S0022-1694(96)03128-9

496 Toth LA (2010) Unrealized expectations for restoration of a ﬂoodplain plant 497 community. Restoration Ecology 18(6):810–819

498 Von Der Lippe M, Kowarik I (2007) Long-distance dispersal of plants by vehicles as a 499 driver of plant invasions. Conservation Biology 21(4):986–996

500 Yager LY, Miller DL, Jones J (2010) Susceptibility of longleaf pine forest associations 501 in south Mississippi to invasion by cogongrass [imperata cylindrica (l.) beauv.]. 502 Natural areas journal 30(2):226–232

503 Zeileis A, Kleiber C, Jackman S (2008) Regression models for count data in R. Journal 504 of Statistical Software 27(8):URL http://www.jstatsoft.org/v27/i08/.

505 Zenni RD (2014) Analysis of introduction history of invasive plants in Brazil reveals 506 patterns of association between biogeographical origin and reason for introduction. 507 Austral Ecology 39(4):401–407

17 Table 1: Descriptive statistics. Our response variable is the proportional cover of the six study species. This value is the total cover for each species at a protected area (PA) divided by the size of the PA. The number of study PAs (n) with cover data varies with species. Total area (hectares =HA) invaded describes the sum of cover for each species across the 365 study PAs. Proportional eﬀort indicates the relative proportion of management activity allocated to each species between 1999 and 2009 (see Supplementary Table SI1 for more information). The predictor variables were site level features of PAs Variable Proportional Cover n HA invaded Proportion eﬀort 5th percentile Mean 95th percentile Schinus terebinthifolius 0 7.42E-03 1.57E-02 207 4644 0.25

18 Imperata cylindrica 0 1.16E-04 3.94E-04 312 1518 0.09 Lygodium microphyllum 0 3.88E-05 1.00E-07 342 2204 0.08 Ludwigia peruviana 0 4.94E-05 1.57E-04 326 2754 0.00 Urena lobata 0 2.96E-04 6.25E-04 277 2102 0.01 Panicum maximum 0 1.69E-04 4.58E-04 311 377 0.01 Protected Area Features 5th percentile Mean 95th percentile n Total HA 2 62 8600 365 Households within 25 km 9900 104 000 679 000 365 Frost-bin NA NA NA 365 Road Length (M) 45 2139 81 800 365 Mean Elevation (M) 1 4 32 365 Table 2: Parameter estimates and standard errors (Coefficient ± 1 SE) for the model component that predicts the presence of an invasive species on a protected area. The probability that a species is present at a protected area is calculated as 1 minus the logit of the linear combination of these coefficients multiplied by the predictor variable values for that protected area. See text in section 2.4 for the equation. Values in bold font are statistically significant at p ≤ 0.05. Intercept log HA log house density log road cover log elevation frost-bin Schinus terebinthifolius -11.65±8.52 -0.06±0.25 1.15±0.68 -0.18±0.11 0.76±0.58 -2.69±1.17 Imperata cylindrica -4.29±5.75 -0.70±0.19 0.26±0.28 0.45±0.29 -0.86±0.28 -1.46±0.65 Ludwigia peruviana 12.34±4.71 -0.68±0.17 -0.55±0.35 -0.05±0.07 -0.94±0.33 -1.60±0.72 Lygodium microphyllum 13.72±6.32 -1.13±0.33 0.00±0.43 0.10±0.09 -1.42±0.58 -7.25±2.75 Panicum maximum 6.33±4.54 -0.56±0.31 0.06±0.36 -0.32±0.21 0.69±0.39 -0.98±0.94 Urena lobata -14.51±48.45 -1.29±0.78 0.89±0.72 -1.10±1.00 -3.85±1.97 26.49±50.29 19 Table 3: Parameter estimates and standard errors (Coefficient ± 1 SE) for the model component that predicts proportion invasive cover of an invasive species at a protected area if the species is present. The mean proportion invasive cover at a protected area follows a binomial distribution and thus is calculated as exp(the linear combination of these coefficients multiplied by the predictor variable values for that protected area). See text in section 2.4 for the equation. Values in bold font are statistically significant at p ≤ 0.05. Intercept log HA log house density log road cover log elevation frost-bin dispersion Schinus terebinthifolius -5.16±3.26 -0.58±0.21 0.71±0.26 -0.11±0.08 0.00±0.44 4.68±0.97 5.34±1.17 Imperata cylindrica -11.17±3.67 -0.33±0.09 0.40±0.20 0.60±0.18 -0.07±0.16 0.42±0.38 1.26±0.41 Ludwigia peruviana 11.51±5.38 -0.51±0.10 -0.56±0.29 -0.19±0.19 -0.15±0.19 0.70±0.47 2.08±0.71 Lygodium microphyllum 10.77±5.66 -0.69±0.20 -0.22±0.41 -0.22±0.06 -0.48±0.37 -0.90±2.80 0.52±0.63 Panicum maximum 8.60±4.67 -0.73±0.17 -0.41±0.35 -0.15±0.29 0.28±0.46 2.32±0.86 3.75±2.36 Urena lobata -1.86±2.82 -0.46±0.09 0.41±0.21 -0.09±0.08 0.22±0.21 0.95±0.56 6.41±1.21 20 508 Supplementary Information

509 Table SI1 The effort values are sums of proportional effort per protected area (PA). 510 For instance, if all of the treatment effort at a PA was focused on a single species it 511 would receive a value of 1 for that PA. We summed these proportions of effort per 512 species across all PAs. If only one species was managed across all PAs it would receive 513 an effort ranking of 365. In our dataset, Schinus terebinthifolius has an effort ranking 514 of 225 which would be equivalent to Schinus terebinthifolius being the only species 515 treated on 225 of our 365 PAs. Case study species are in bold font. Florida Exotic Pest 516 Plant Council (FLEPPC) category I species are those documented as impacting native 517 plant communities. FLEPPC category II species have been observed to have increased 518 in abundance or frequency in natural areas.

Effort Prop. Common name Scientific name Cat. 1 224.99 0.24 Brazilian pepper Schinus terebinthifolius I 2 86.84 0.09 cogon grass Imperata cylindrica I 3 73.57 0.08 Old World climbing fern Lygodium microphyllum I 4 55.68 0.06 melaleuca, paper bark Melaleuca quinquenervia I 5 52.49 0.06 Australian pine Casuarina species I 6 49.60 0.05 Japanese climbing fern Lygodium japonicum I 7 42.89 0.05 Chinese tallow, popcorn tree Triadica sebifera I 8 39.31 0.04 air-potato Dioscorea bulbifera I 9 31.14 0.03 Unidentified species NA NA 10 24.76 0.03 coral ardisia Ardisia crenata I 11 16.03 0.02 Chinaberry Melia azedarach I 12 15.15 0.02 skunk vine Paederia foetida I 13 14.71 0.02 camphor tree Cinnamomum camphora I 14 14.27 0.02 lead tree Leucaena leucocephala II 15 13.94 0.02 Bay Biscayne creeping-oxeye Sphagneticola trilobata II 16 12.91 0.01 Caesar’s weed Urena lobata II 17 12.54 0.01 mimosa, silk tree Albizia julibrissin I 18 11.58 0.01 tropical soda apple Solanum viarum I 19 9.29 0.01 lantana, shrub verbena Lantana camara I 20 8.80 0.01 bowstring hemp Sansevieria hyacinthoides II 21 7.85 0.01 rosary pea Abrus precatorius I 22 6.61 0.01 shoebutton ardisia Ardisia elliptica I 23 5.85 0.01 torpedo grass Panicum repens I 24 5.63 0.01 Guinea grass Panicum maximum II 25 5.11 0.01 lather leaf Colubrina asiatica I 26 4.98 0.01 Japanese honeysuckle Lonicera japonica I 27 4.87 0.01 downy rose-myrtle Rhodomyrtus tomentosa I 28 4.74 0.01 wild taro Colocasia esculenta I 29 4.64 0.01 seaside mahoe Thespesia populnea I 30 4.03 0 guava Psidium guajava I 31 3.88 0 scaevola, beach naupaka Scaevola sericea I 32 3.88 0 aquatic soda apple Solanum tampicense I 33 3.84 0 paper mulberry Broussonetia papyrifera II 34 3.78 0 nandina, heavenly bamboo Nandina domestica I

21 Table 4: (continued) Effort Prop. Common name Scientific name Cat. 35 3.55 0 kudzu Pueraria montana I 36 3.20 0 Par grass Urochloa mutica I 37 2.75 0 Burma reed, cane grass Neyraudia reynaudiana I 38 2.00 0 arrowhead vine Syngonium podophyllum I 39 1.93 0 tung oil tree Aleurites fordii II 40 1.91 0 white-flowered wandering Jew Tradescantia fluminensis I 41 1.72 0 cat’s claw vine Macfadyena unguis-cati I 42 1.64 0 wedelia Wedelia trilobata II 43 1.49 0 Chinese brake fern Pteris vittata II 44 1.48 0 mahoe, sea hibiscus Hibiscus tiliaceus II 45 1.48 0 strawberry guava Psidium cattleianum I 46 1.25 0 bischofia Bischofia javanica I 47 1.24 0 life plant Kalanchoe pinnata II 48 1.09 0 sapodilla Manilkara zapota I 49 1.06 0 purple sesban, rattlebox Sesbania punicea II 50 0.90 0 carrotwood Cupaniopsis anacardioides I 51 0.80 0 Natal grass Rhynchelytrum repens II 52 0.76 0 woman’s tongue Albizia lebbeck I 53 0.76 0 Chinese or hedge privet Ligustrum sinense I 54 0.76 0 castor bean Ricinus communis II 55 0.75 0 winged yam Dioscorea alata I 56 0.71 0 confederate jasmine Trachelospermum jasminoides NA 57 0.51 0 sweet autumn virginsbower Clematis terniflora NA 58 0.48 0 thorny eleagnus Elaeagnus pungens II 59 0.46 0 twin-flowered passion vine Passiflora biflora II 60 0.44 0 earleaf acacia Acacia auriculiformis I 61 0.42 0 West Indian marsh grass Hymenachne amplexicaulis I 62 0.41 0 Peruvian primrose-willow Ludwigia peruviana I 63 0.38 0 arrow bamboo Pseudosasa japonica NA 64 0.34 0 schefflera, umbrella tree Schefflera actinophylla I 65 0.29 0 oyster plant Tradescantia spathacea I 66 0.25 0 flamegold tree Koelreuteria elegans II 67 0.24 0 Surinam cherry Eugenia uniflora I 68 0.24 0 pothos Epipremnum pinnatum II 69 0.20 0 malanga, elephant ear Xanthosoma sagittifolium II 70 0.19 0 rose Natal grass Melinis repens I 71 0.18 0 climbing or Christmas cassia Senna pendula I 72 0.14 0 alligator weed Alternanthera philoxeroides II 73 0.13 0 coconut palm Cocos nucifera NA 74 0.13 0 flamevine Pyrostegia venusta NA 75 0.12 0 rubber vine Cryptostegia madagascariensis II 76 0.09 0 bush morning-glory Ipomoea fistulosa II 77 0.08 0 common asparagus fern Asparagus setaceus NA 78 0.07 0 umbrella plant Cyperus involucratus II 79 0.07 0 glossy privet Ligustrum lucidum I 80 0.06 0 Senegal date palm Phoenix reclinata II 81 0.05 0 Napier grass Pennisetum purpureum I 82 0.05 0 sisal hemp Agave sisalana II 83 0.04 0 rose-apple Syzygium jambos II 84 0.04 0 Egyptian grass Dactyloctenium aegyptium NA

22 Table 4: (continued) Effort Prop. Common name Scientific name Cat. 85 0.04 0 Mexican petunia Ruellia brittoniana I 86 0.04 0 tropical almond Terminalia catappa II 87 0.04 0 coral vine Antigonon leptopus II 88 0.04 0 red sandalwood Adenanthera pavonina II 89 0.03 0 jambolan, Java plum Syzygium cumini I 90 0.03 0 solitary palm Ptychosperma elegans II 91 0.03 0 orchid tree Bauhinia variegata I 92 0.03 0 Taiwanese cheesewood Pittosporum pentandrum II 93 0.02 0 Puerto Rico silver palm Coccothrinax barbadensis NA 94 0.02 0 queen palm Syagrus romanzoffiana II 95 0.01 0 Arabian jasmine Jasminum sambac II 96 0.01 0 laurel fig Ficus microcarpa I 97 0.01 0 asparagus-fern Asparagus densiflorus I 98 0.01 0 limpo grass Hemarthria altissima II 99 0.00 0 Washington fan palm Washingtonia robusta II 100 0.00 0 puncture vine, bur-nut Tribulus cistoides II 101 0.00 0 santa maria, mast wood Calophyllum antillanum I 102 0.00 0 white cypress-pine Callitris glaucophylla NA 103 0.00 0 Indian rosewood, sissoo Dalbergia sissoo II 104 0.00 0 governor’s plum Flacourtia indica II 105 0.00 0 wood-rose Merremia tuberose II 106 0.00 0 susumber, turkey berry Solanum torvum II 107 0.00 0 simpleleaf chastetree Vitex trifolia II

519

23 (a) (b) (c) 24

(d) (e) (f) Fig. SI1 Distribution of study species across sample protected areas. Circles indicate a protected area with the focal species present but the analysis was performed across all 365 protected areas for each species. Size of the circle corresponds with size of the protected area (HA) 520 Table SI2 Invasive species presence and cover across study protected areas

Plant name HA cover Number of protected areas Schinus terebinthifolius 4644 211 Ludwigia peruviana 2754 101 Lygodium microphyllum 2204 41 Urena lobata 2102 154 Imperata cylindrica 1518 120 Colocasia esculenta 1362 56 Lygodium japonicum 1099 77 Solanum viarum 1034 28 Panicum repens 876 94 Melaleuca quinquenervia 787 52 Casuarina equisetifolia 687 60 Leucaena leucocephala 642 56 Urochloa mutica 526 53 Dioscorea bulbifera 406 96 Panicum maximum 377 98 Rhynchelytrum repens 310 64 Cinnamomum camphora 305 87 Ricinus communis 291 44 Nephrolepis cordifolia 263 90 Sphagneticola trilobata 245 72 Sapium sebiferum 240 92 Lantana camara 236 97 Abrus precatorius 235 72 Melia azedarach 223 79 Syngonium podophyllum 52 48 Ardisia crenata 31 25 Lonicera japonica 7 34 Albizia julibrissin 1 51

25 521 Table SI3 Study protected areas

26 Protected Area name Hectares County 1 Morningstar Parcel 0.29 STLU 2 Trailhead Park 0.37 BROW 3 Vilano Oceanfront Park 0.73 STJO 4 North Fork Riverwalk 0.8 BROW 5 Sea Oats Park 0.9 BREV 6 Sailboat Bend Preserve 1.02 BROW 7 Coontie Hatchee Landing 1.05 BROW 8 Rodney Kroegel Homestead 1.06 INDI 9 Sand Hill Trailhead 1.11 BREV 10 Windswept Acres Park 1.21 STJO 11 Carlton Village Park 1.25 LAKE 12 Lucas Tract 1.44 SEMI 13 Lake Tarpon West Management Area 1.5 PINE 14 Hallandale City Beach 1.56 BROW 15 Overlook Park 1.56 SEMI 16 Timer Powers Park Conservation Area 1.62 MART 17 Bear Creek Nature Trail 1.66 SEMI 18 Mullock Creek Preserve 1.74 LEEX 19 Spessard Holland North Beach Park 1.83 BREV 20 Lake Thomas Cove Park 1.86 LAKE 21 Whispering Pines Hammock Preserve 2.1 DADE 22 Warbler Wetland Natural Area 2.46 BROW 23 Orange River Parcel 2.58 LEEX 24 Spessard Holland South Beach Park 2.63 BREV 25 Punta Rassa Preserve 2.72 LEEX 26 Baycliff Preserve 2.76 SANT 27 Mallory Heights Park #3 2.83 ESCA 28 Northwest 39th Avenue Park 2.87 BROW 29 West Creek Pineland Natural Area 3.01 BROW 30 Ryall Parcel 3.13 INDI 31 Long Branch Management Area 3.16 PINE 32 Ozona Management Area 3.24 PINE 33 Cypress Creek Sand Pine Preserve 3.3 BROW 34 Varn Parcel 3.46 STLU 35 Blackwater Hammock 3.64 HILL 36 Columbus G. MacLeod Preserve 3.64 LEEX 37 Vinkemulder LAPC 3.78 BROW 38 Ingram Pineland 4.01 DADE 39 Helwig (456) 4.03 BROW 40 Sebastian Scrub Conservation Area 4.05 INDI 41 Silver Palm Hammock 4.05 DADE 42 Ludlam Pineland 4.14 DADE 43 John Williams Park 4.44 BROW 44 Saw Palmetto Natural Area 4.53 BROW 45 Harden Hammock 5 DADE 46 River Tower Restoration Site 5.18 HILL 47 Bay Bluffs Park 5.26 ESCA 48 Sheridan Oak Forest 5.33 BROW 49 South Lake Jesup Property 5.48 SEMI 50 John Mahon Park 5.67 ALAC 51 Gopher Tortoise Preserve (Broward County) 5.72 BROW 52 Blackburn Point Park and Addition 6.07 SARA 53 County Line Scrub 6.08 DADE 54 Golden Sands Park 6.23 INDI 55 Joe’s River Park 6.23 MART 56 Green Salt Marsh 6.48 INDI 57 Manatee Park 6.48 LEEX 58 Cross Bayou North 6.54 PINE 59 William J. Kelly Rookery 6.79 BROW 60 Bowditch Point Park 7.28 LEEX 61 Windmiller Parcel 7.54 INDI 62 Herman and Dorothy Shooster Preserve 8 BROW 63 Military Trail Natural Area 8.06 BROW 64 Lucille Hammock 8.4 DADE 65 Woodmont Natural Area 8.43 BROW 66 Charlie’s Marsh Preserve 8.44 LEEX 67 Tivoli Sand Pine Park 9.13 BROW 68 Anclote Gulf Park 9.29 PASC 69 Palatlakaha River Park 9.31 LAKE 70 East Lake Management Area 9.71 PINE 71 Holland Park 9.89 BROW 72 King Islands Management Area 10.12 PINE 73 Prange Islands Conservation Area 10.72 INDI 74 Seminole Wayside Park 11.21 DADE 75 Ansin Tract 11.38 INDI 76 Escambia Bay Bluffs 11.49 ESCA 77 Shadowbrook Tract 12.04 INDI 78 Lake Seminole Management Area 12.42 PINE 79 Rotunda Community Park and Preserve 12.54 CHAR 80 Golden Aster Preserve 12.56 LEON 81 Allen’s Creek Management Area 13.36 PINE 82 Highland Scrub Natural Area 13.87 BROW 83 Twin Rivers 2 Preserve 13.88 SEMI 84 Grassy Point Preserve 14.16 MANA 85 Jim Wingate Park 14.16 DUVA

27 Protected Area name Hectares County 86 Marsh Park and Boat Ramp 14.16 LAKE 87 Palmetto Estuary Preservation Project 14.42 MANA 88 Marianna Greenway 14.46 JACK 89 Haynes Creek Park 14.61 LAKE 90 Flinn Tract Conservation Area 15.18 INDI 91 Westmoreland 15.38 STLU 92 Oak Island Nature Preserve 15.42 OSCE 93 Hathaway Park 15.78 CHAR 94 Imperial River Preserve 15.78 LEEX 95 Lake Tarpon Management Area 15.78 PINE 96 Ribault River Preserve 15.78 DUVA 97 Eagle Lake Preserve 16.19 LEEX 98 Goulds Pineland and Addition 16.19 DADE 99 Salinas Park 16.19 GULF 100 Atlantic Ridge Parcels 16.37 MART 101 Oak Hammock Park 16.59 STLU 102 Ollie’s Pond Park 16.74 CHAR 103 Cypress Point Park 17.16 HILL 104 Lake Runnymede Conservation Area 17.4 OSCE 105 Dicerandra Scrub Sanctuary 17.81 BREV 106 Pendarvis Cove Park 17.81 MART 107 Madison Blue Spring 17.83 MADI 108 Black Hammock Trail Head 18.03 SEMI 109 Robert K. Rees Memorial Park 18.21 PASC 110 University of Central Florida Riparian Area 18.53 ORAN 111 Travatine Island Management Area 18.82 PINE 112 Easterlin Regional Park 18.86 BROW 113 Russell Grove 19.12 INDI 114 Falling Creek Park 20.81 COLU 115 Secret Woods Buffer and Nature Center 21.68 BROW 116 Alligator Lake Management Area 22.19 PINE 117 John David Patton Wildlife Park 22.26 FRAN 118 Mills Pond Park 22.5 BROW 119 Matanzas Pass Preserve 23.88 LEEX 120 Orange River Preserve 23.88 LEEX 121 Round Island South Conservation Area 23.88 INDI 122 Price Park 24.5 BROW 123 Hollywood North Beach Regional Park 24.8 BROW 124 University of Central Florida East Parcel 25.14 ORAN 125 Chain of Lakes Stormwater Park 25.27 BREV 126 Natural Bridge Battlefield Historic State Park 25.86 LEON 127 Shell Bluff 25.9 FLAG 128 Hillsboro Pineland Natural Area 26.85 BROW 129 Tall Cypress Natural Area 26.87 BROW 130 Watson Island Parcel 26.9 STJO 131 Hogtown Creek Headwaters Nature Park 29.73 ALAC 132 Hackberry Hammock 30.18 STLU 133 Pine Island Preserve 30.35 MANA 134 Flatwoods Conservation Area 30.74 ALAC 135 Mariner’s Point Management Area 30.76 PINE 136 Alafia Scrub Preserve 31.49 HILL 137 Bivens Arm Nature Park 32.78 ALAC 138 Jerry Lake 32.78 PINE 139 Dead Lakes Park 33.78 GULF 140 Clam Bayou 34.01 PINE 141 Cooper’s Point 34.03 PINE 142 Brohard Beach and Paw Park 34.12 SARA 143 Cow Branch Management Area 34.72 PINE 144 The Hammock 35.61 PINE 145 James E. Grey Preserve 36.38 PASC 146 Harmony Oaks Conservation Area 36.42 INDI 147 Hallstrom Farmstead 37.64 INDI 148 Oyster Bar Salt Marsh 38.85 INDI 149 Temple Terrace Riverfront Park 38.85 HILL 150 Marjorie Kinnan Rawlings Historic State Park 40.07 ALAC 151 Mikes Donation 40.34 SEMI 152 Cameron Preserve 40.47 BREV 153 Reddie Point Preserve 41.2 DUVA 154 Pine Island Ridge Natural Area 41.33 BROW 155 Key Vista Nature Park 41.81 PASC 156 Wilson’s Landing 42.4 SEMI 157 Kendall Indian Hammocks Park 42.49 DADE 158 Shoreline Park South 42.49 SANT 159 Murdock Point Cayo Costa 42.57 LEEX 160 Wabasso Scrub Conservation Area 44.92 INDI 161 Captain Forster Hammock Preserve 45.33 INDI 162 Peck Sink Preserve 45.33 HERN 163 Charles Lee Property 45.58 SEMI 164 Hamilton Reserve 46.14 OSCE 165 Wildcat Cove 46.14 STLU 166 Chinsegut Hill Conference Center 46.35 HERN 167 Pasco Palms Preserve 46.9 PASC 168 St. James Creek Preserve 47.75 LEEX 169 Verdie Forest 47.82 DUVA 170 River City Nature Park 48.56 VOLU 171 Upper Pithlachascotee River Preserve 48.56 PASC 172 New Tampa Nature Park 49.25 HILL 173 Lake Lotus Park 50.59 SEMI 174 Thomas Creek Preserve 50.77 DUVA 175 Cayo Pelau Preserve28 51.08 LEEX Protected Area name Hectares County 176 Navarre Beach Park 52.68 SANT 177 Turkey Creek Sanctuary 53.83 BREV 178 University of Central Florida McKay Tract 54.55 ORAN 179 Edwards Bottomland 55.69 BRAD 180 Crews Lake Wilderness Park 56.66 PASC 181 Gamble Place 58.28 VOLU 182 Fickett Hammock Preserve 60.7 HERN 183 Ravine Gardens State Park 61.79 PUTN 184 Miramar Pineland Natural Area 63.52 BROW 185 Anclote Islands Management Area 64.75 PINE, PASC 186 Eden Gardens State Park 65.55 WALT 187 Sebastian Harbor Preserve 65.97 INDI 188 Fox Creek 66.14 SARA 189 University of Central Florida Pond Pine 66.2 ORAN 190 North Buck Lake Scrub Sanctuary 71.4 BREV 191 Brooksville Plant Materials Center 73.66 HERN 192 East 417 Property 74.33 SEMI 193 Lathrop Bayou Tract 75.68 BAYX 194 Letchworth-Love Mounds Archaeological State Park 76.16 JEFF 195 Carver Preserve 76.49 LEEX 196 Curry Island 78.96 GLAD 197 Ferndale Preserve 79.32 LAKE 198 Bocilla Preserve 79.73 LEEX 199 Fanning Springs State Park 80.28 LEVY 200 Lake Jackson Mounds Archaeological State Park 82.94 LEON 201 West Marsh Preserve 83.37 LEEX 202 Alligator Creek Conservation Area 85.39 SARA 203 Morsani Ranch 86.77 PASC 204 Inland Groves 88.63 LAKE 205 Half Moon Island Preserve 92.27 DUVA 206 Econ River Wilderness Area 97.15 SEMI 207 Daniels Preserve at Spanish Creek 98.34 LEEX 208 Pond Apple Slough 98.35 BROW 209 Estero Marsh Preserve 98.48 LEEX 210 Lake Stone Fish Management Area 100.74 ESCA 211 Big Hickory Island Preserve 109.12 LEEX 212 Lakes Regional Park 112.91 LEEX 213 South Marianna Trail and Canoe Launch 112.95 JACK 214 Chipola River Greenway 118.06 JACK 215 Fort Matanzas National Monument 121.45 STJO 216 Palatlakaha Environmental and Agricultural Reserve Park 128.69 LAKE 217 Cypress Lakes Preserve 130.31 HERN 218 Cabbage Key Management Area 134.98 PINE 219 Walsingham Park 141.65 PINE 220 Tippecanoe Environmental Park 143.26 CHAR 221 Lyonia Preserve 144.94 VOLU 222 Four Mile Cove Ecological Preserve 147.72 LEEX 223 Gum Root Park 149.33 ALAC 224 Lake Harney Wilderness Area 149.74 SEMI 225 Shell Creek Preserve 154.6 CHAR 226 Morsani Conservation Easement 155.42 PASC 227 Tradewinds Regional Park 156.42 BROW 228 Ponce de Leon Springs State Park 156.59 HOLM, WALT 229 Greenbriar Swamp Preserve 159.54 LEEX 230 Mangrove Preserve 161.88 DADE 231 Mobbly Bayou Preserve 162.69 PINE 232 Sal Taylor Creek Preserve 164.48 DUVA 233 Tampa Bay Estuarine Ecosystem - Terra Ceia 167.65 MANA 234 Joe’s Creek Management Area 168.33 PINE 235 Washington Oaks Gardens State Park 172.21 FLAG 236 Oslo Riverfront Conservation Area 178.35 INDI 237 Cutler Wetlands 181.31 DADE 238 Mashes Sands Park 195.47 WAKU 239 Cedar Point 196.28 DUVA 240 Brooker Creek Buffer Preserve 198.3 HILL 241 Lost Tree Islands Conservation Area 205.59 INDI 242 North Sebastian Conservation Area 219 INDI 243 Indian River Lagoon Preserve State Park 220.19 BREV 244 Lake Talquin State Park 223.2 LEON, GADS, LIBE 245 North Peninsula State Park 225.64 VOLU 246 Fowlers Bluff Conservation Area 226.08 LEVY 247 Lake Griffin State Park 226.16 LAKE 248 Hidden Lake Project 238.33 PASC 249 Fern Prairie Preserve 241.52 LAKE 250 De Leon Springs State Park 245.3 VOLU 251 Prairie/Shell Creek 246.87 CHAR 252 Palatka-Lake Butler State Trail 248.09 BRAD, CLAY, PUTN, UNIO 253 Matheson Hammock Park 254.56 DADE 254 Elinor Klapp-Phipps Park 274.27 LEON 255 Eagle Point Park 274.4 PASC 256 Black Hammock Wilderness Area 286.39 SEMI 257 San Carlos Bay - Bunche Beach Preserve 289.77 LEEX 258 Ward Creek West 291.1 BAYX 259 Lower Alapaha Conservation Area 295.33 HAMI 260 St. Marks Headwaters 305.14 LEON 261 Withlacoochee State Trail 307.46 CITR, HERN, PASC 262 Caloosahatchee Regional Park 310.81 LEEX 263 Boyd Hill Nature Park 323.76 PINE 264 Fort Desoto Park 364.23 PINE 265 Crandon Park29 365.81 DADE Protected Area name Hectares County 266 Valkaria Scrub Sanctuary 368.37 BREV 267 Pine Island Conservation Area 370.39 BREV 268 Sebastian Inlet State Park 392.97 BREV, INDI 269 Flint Pen Strand 399.9 LEEX 270 Haw Creek Preserve 406.72 FLAG 271 Wacissa Conservation Area 429.14 JEFF 272 Little Manatee River (SWFWMD) 433.43 MANA 273 North/Walk-in-Water Creek 446.99 POLK 274 Brooker Creek Headwaters 449.78 HILL 275 Yellow Jacket Conservation Area 453.07 DIXI 276 Yellow River Wildlife Management Area - Escribano Point 471.88 SANT 277 Honey Creek Research Natural Area 485.64 VOLU 278 Grissom Parkway 518.53 BREV 279 Jack Creek 520.22 HIGH 280 Upper Little Manatee River 558.24 HILL 281 Little Manatee River 572.05 HILL 282 Valkaria Expansion 580.51 BREV 283 Clear Springs 582.36 POLK 284 Princess Place Preserve 608.26 FLAG 285 Prairie Creek Preserve (Charlotte County) 648.73 CHAR 286 Tomoka State Park 655.65 VOLU 287 Shell Key Preserve 709.92 PINE 288 Deep Creek Conservation Area (SRWMD) 726.45 COLU 289 Homeland 778.24 POLK 290 Deer Lake State Park 807.27 WALT 291 Austin Cary Memorial Forest 809.4 ALAC 292 Lower Peace River Corridor 829.63 DESO, CHAR 293 Hixtown Swamp Conservation Area 844.4 MADI 294 Grayton Beach State Park 875.09 WALT 295 Upper Alapaha Conservation Area 914.17 HAMI 296 Annutteliga Hammock 941.88 HERN 297 Fox Lake Sanctuary 954.28 BREV 298 Flat Island Preserve 957.48 SUMT, LAKE 299 Six Mile Cypress Slough Preserve 979.05 LEEX 300 Lonesome Camp Ranch Conservation Area 988.6 OSCE 301 Tampa Bay Estuarine Ecosystem 1009.32 HILL 302 Pepper Ranch Preserve 1016.61 COLL 303 Holton Creek Conservation Area 1028.66 HAMI 304 St. Marks River State Park 1048.04 LEON, JEFF 305 Weedon Island Preserve 1092.69 PINE 306 T. H. Stone Memorial St. Joseph Peninsula State Park 1099.1 GULF 307 Olustee Experimental Forest 1138.42 BAKE 308 Upper Aucilla Conservation Area 1159.06 JEFF, MADI 309 Tate’s Hell Wildlife Management Area 1175.65 FRAN 310 Haw Creek Preserve State Park 1238.96 VOLU, FLAG, PUTN 311 Wild Turkey Strand Preserve 1270.77 LEEX 312 Graham Swamp Conservation Area 1294.64 FLAG 313 Garcon Point Water Management Area 1313.25 SANT 314 Headwaters at Duette Park 1364.9 MANA 315 Devil’s Hammock 1465.01 LEVY 316 Subtropical Agricultural Research Station 1558.1 HERN 317 Econfina River State Park 1832.59 TAYL 318 Colt Creek State Park 2050.61 POLK 319 Cecil Field Conservation Corridor 2171.62 DUVA, CLAY 320 Lake Jesup Conservation Area 2245.12 SEMI, VOLU 321 Bulow Creek State Park 2260.59 FLAG, VOLU 322 Faver-Dykes State Park 2446.81 STJO, FLAG 323 Edward Ball Wakulla Springs State Park 2450.55 WAKU 324 T. M. Goodwin Waterfowl Management Area 2537.47 BREV 325 Pine Glades Natural Area 2688.01 PALM 326 San Felasco Hammock Preserve State Park 2893.44 ALAC 327 Santa Fe Swamp Conservation Area 2971.79 ALAC, BRAD 328 Tenoroc Fish Management Area 2973.61 POLK 329 Allen David Broussard Catfish Creek Preserve State Park 3301.22 POLK 330 Brooker Creek Preserve 3539.55 PINE 331 Upper Chipola River Water Management Area 3680.23 JACK, CALH 332 Ordway-Swisher Biological Station 3682.77 PUTN 333 Buck Lake Conservation Area 3880.66 BREV, VOLU 334 Alafia River Corridor 3997.01 HILL 335 Fellsmere Water Management Area 4122.82 INDI 336 Lochloosa Wildlife Conservation Area 4189.62 ALAC 337 Canaveral Marshes Conservation Area 4427.64 BREV 338 Frog Pond/L-31 N Transition Lands 4579.3 DADE 339 Snipe Island Unit 4729.73 TAYL 340 St. Vincent National Wildlife Refuge 5054.67 FRAN, GULF 341 Grassy Waters Preserve 5180.16 PALM 342 Cone Ranch 5665.8 HILL 343 Model Lands Basin 6894.47 DADE 344 Lower Wekiva River Preserve State Park 7043.64 LAKE, SEMI, VOLU 345 Yellow River Water Management Area 7180.2 SANT, OKAL 346 Paynes Prairie Preserve State Park 8570.45 ALAC 347 St. Sebastian River Preserve State Park 8642.54 BREV, INDI 348 Lake Woodruff National Wildlife Refuge 8724.94 VOLU, LAKE 349 Dupuis Reserve 8859.53 PALM, MART 350 Bull Creek Wildlife Management Area 9569.54 OSCE 351 Crystal River Preserve State Park 10261.19 CITR 352 Tyndall Air Force Base 11331.6 BAYX 353 Kissimmee Chain of Lakes 11844.02 OSCE, POLK 354 Seminole Ranch Conservation Area 11918.2 BREV, VOLU, ORAN, SEMI 355 Chassahowitzka National Wildlife Refuge30 12482.13 CITR, HERN Protected Area name Hectares County 356 Southern Glades 13122.56 DADE 357 Lower Escambia River Water Management Area 14331.8 SANT, ESCA 358 Apalachicola River Water Management Area 14371.71 LIBE, GULF 359 River Lakes Conservation Area 15933.53 BREV, OSCE 360 Kissimmee River 17556.04 HIGH, OSCE, POLK, OKEE 361 Blue Cypress Conservation Area 20204.41 INDI 362 Kissimmee Prairie Preserve State Park 21758.97 OKEE, OSCE 363 Choctawhatchee River Water Management Area 24625.74 WASH, WALT, BAYX, HOLM 364 Apalachicola River Wildlife and Environmental Area 26091.41 GULF, FRAN 365 Gulf Islands National Seashore 26932.48 ESCA, OKAL, SANT