1 2 Characteristics of the Psidium cattleianum invasion of secondary rainforests 3 DAVID Y. P. TNG1, MIRIAM W. GOOSEM1, CLAUDIA P. PAZ1, NOEL D. 4 PREECE1,2, STEPHEN GOOSEM3, RODERICK J. FENSHAM4,5, and SUSAN G. W. 5 LAURANCE1* 6 1 7 and Environmental Sciences, James Cook University, Cairns, Qld 4870, Australia (*Email: 8 Susan.Laurance@jcu.edu.au), 2Research Institute for Environment and Livelihoods, Charles 9 Darwin University, Darwin, NT, 0801, Australia, 3Wet Tropics Management Authority, 1 Centre for Tropical Environmental and Sustainability Science (TESS) and College of Marine 10 Grafton Street & Hartley St, Cairns Qld. 4870, Australia, 4School of Biological Sciences, 11 University of Queensland, St Lucia, Qld. 4072, Australia, 5Queensland Herbarium, Mt Coot- 12 tha Road, Toowong, Qld. 4066, Australia. 13 14 15 2 16 Abstract Strawberry guava (Psidium cattleianum) is a shade-tolerant shrub or small tree 17 invader in tropical and subtropical regions and is considered among the world’s top 100 worst 18 invasive species. Studies from affected regions report deleterious effects of strawberry guava 19 invasion on native vegetation. Here we examine the life history demographics and 20 environmental determinants of strawberry guava invasions to inform effective weed 21 management in affected rainforest regions. We surveyed the vegetation of eight mature 22 rainforest and thirty-three successional sites at various stages of regeneration in the 23 Australian Wet Tropics and found that strawberry guava invasion was largely restricted to 24 successional forests. Strawberry guava exhibited high stem and seedling densities, 25 represented approximately 8% of all individual stems recorded and 20% of all seedlings 26 recorded. The species also had the highest basal area among all the non-native woody species 27 measured. We compared environmental and community level effects between strawberry 28 guava-invaded and non-invaded sites, and modelled how the species basal area and 29 recruitment patterns respond to these effects. Invaded sites differed from non-invaded sites in 30 several environmental features such as aspect, distance from intact forest blocks, as well as 31 supported higher grass and herb stem densities. Our analysis showed that invasion is 32 currently ongoing in secondary forests, and also that strawberry guava is able to establish and 33 persist under closed canopies. If left unchecked, strawberry guava invasion will have 34 deleterious consequences for native regenerating forest in the Australian Wet Tropics. 35 Keywords: community species diversity, biological invasions, Psidium cattleianum, 36 secondary rainforests, shade-tolerant invaders, strawberry guava 37 INTRODUCTION 38 Tropical forests are of great importance due to their immense contribution to global 39 biodiversity and carbon budgets, but are experiencing major changes in species composition 3 40 and richness at local and global scales due to environmental changes caused by 41 anthropogenic activities. A major symptom of this change is the establishment of invasive 42 species in natural environments (Ortega & Pearson 2005; Ditham et al. 2007). Once they 43 achieve a certain level of abundance introduced species may displace native species by 44 competing for resources, such as space, water, nutrients, and light (Levine et al. 2003; Vila & 45 Weiner 2004). Invasive species can also cause community species shifts by impeding the 46 colonization success of native species (Hager 2004; Yurkonis & Meiners 2004). Invasive 47 species typical impose considerable propagule pressure and can saturate suitable microsites 48 (Brown & Fridley 2003) reducing the rate of establishment by native species. Where 49 competitive interactions shape community structure, the invasion process may more strongly 50 inhibit colonization of species within the same functional group as the invader (Symstad 51 2000; Fargione et al. 2003). Furthermore, some invasive species can behave like ‘ecosystem 52 engineers’ and modify important ecological processes such as disturbance regimes and 53 nutrient cycling (Yurkonis et al. 2005). 54 Heywood (1989) argued that invasion of tropical forests follows widespread disruption or 55 conversion of the primary forest to secondary successional communities, and invasive plant 56 species are thought to establish mostly on forest edges and in disturbed closed forest with 57 high light and nutrient availability (Milbau & Nijs 2004; Gilbert & Lechowizc 2005). As a 58 result, closed-canopied vegetation has long been regarded as highly resistant to invasion 59 (Cavers & Harper 1967; Rejmánek 1989; Von Holle et al. 2003). However, evidence is 60 mounting, that shade-tolerant invasive plants can invade forests with relatively closed 61 canopies (Murphy et al. 2008; Martin et al. 2009). There is a need for detailed studies on the 62 effects of such invasive plant species in tropical forests. The susceptibility of a community to 63 invasion of new species can be assessed quantitatively to determine community invasibility 64 (Burke & Grime 1996) – an approach that takes into account ecosystem and native species 4 65 properties (Lonsdale 1999; Hejda et al. 2009). But given the difficulties associated with 66 collecting quantitative data in species-rich tropical forest, such studies are few (e.g. Mullah et 67 al. 2014). 68 While high species richness has been hypothesized to confer resistance to the invasion of a 69 community (e.g. Elton 1958; Hooper et al. 2005), many studies present contrasting results 70 (e.g. Levine & D’Antonio 1999; Shea & Chesson 2002; Hejda et al. 2009). It is assumed that 71 higher species richness may repel invasion because species-rich communities exploit 72 available resources more completely and thus leave fewer niches open for colonization 73 (Levine & D’Antonio 1999). However, community vulnerability to invasion may depend on 74 many factors, such as species composition, interactions and successional stage (Hejda et al. 75 2009; Shea & Chesson 2002). Community attributes associated with a healthy ecosystem 76 such as high indigenous species richness and evenness may also potentially facilitate species 77 invasion by conferring protection from pests or predators (Bruno et al. 2003; Dunstan & 78 Johnson 2006). 79 Comprehensive reviews of invasive plant impacts have covered the ecological effects of 80 invaders (Pyšek et al. 2012), the modification of nutrient cycles (Ehrenfeld 2003), 81 mechanisms of plant invasion (Levine et al. 2003) and competition (Vila et al. 2004). 82 Synthesizing accurate predictions of the invasive potential of specific plant taxa has proven 83 difficult and there is no universal trait that can be applied to predict invasiveness (Rejmanek 84 & Richardson, 1996; Hayes & Barry, 2008; Thompson & Davis, 2011; Morin et al. 2013). 85 The few studies that have examined the relationship between invasive species and community 86 properties have shown that impacts on species diversity and composition depend on the 87 individual invader (Hulme & Bremner 2006; Hejda & Pysék 2006). Similarly, native species 88 also differ in their relationship with the invader, as some are excluded from a community 89 more easily than others (Standish et al. 2001). A quantitative way to assess invasibility of a 5 90 non-native species on a community is to examine how the adults, saplings and seedlings of an 91 invader affect community attributes of the community being invaded. 92 In Australia the presence of strawberry guava, Psidium cattleianum Sabine (Myrtaceae), a 93 shade-tolerant invasive shrub or small tree species, was first documented in the 1940s (Atlas 94 of Living Australia 2014), and has since been recorded from several tropical and subtropical 95 rainforest habitats. P. cattleianum is able to to form monospecific stands, and has been 96 implicated in studies from numerous tropical regions to alter habitats (Motley 2005), modify 97 successional trajectories and impede native plant regeneration (Lorence & Sussman 1986; 98 Fleischmann 1997), pose a threat to endangered plant species (Meyer 2004), and interact with 99 other invasive species hence causing further ecological damage (i.e. pigs: Huenneke & 100 Vitousek 1990). The presence of P. cattleianum in tropical and subtropical regions in 101 Australia is therefore a matter of concern. Having been gazetted as a World Heritage Area in 102 1988, and having relatively long-term and comprehensive records of land use, the Australian 103 Wet Tropics provides an important setting for examining the invasibility of P. cattleianum. 104 We (i) characterise the demographics of P. cattleianum populations across rainforest sites of 105 different stages of successional development; (ii) compare various community (e.g. species 106 diversity indices and abundance of other species) and site (e.g. soil properties, aspect and 107 slope) attributes between P. cattleianum-invaded and non-invaded sites; and (iii) model P. 108 cattleianum invasibility in relation to these attributes. 109 METHODS 110 Study species and sites 111 Psidium cattleianum is an evergreen shrub or small tree (2-4 m, occasionally taller) native to 112 the Atlantic forests of Brazil, extending from the Ceará state of Northeast Brazil to Uruguay 113 (Reitz et al. 1983; Sobral et al. 2013). The species has been cultivated, to a small extent, in 6 114 various parts of the world for its edible fruits and ornamental value (Patel 2012). The species 115 embodies a wide range of traits that could facilitate its invasibility and affect ecosystems, 116 including: a high relative growth rate (Pattison et al. 1998), extremely abundant fruit and seed 117 set (Huenneke & Vitousek 1990); high coppicing and resprouting ability (Huenneke 1989), 118 an ability to form dense thickets (Uowolo & Denslow 2008), and high stemflow that funnels 119 water, to increase water available for transpiration (Safeeq & Fares 2014). As a result of 120 human introductions and interactions with birds or feral mammals (i.e. feral pigs; Diong 121 1982), P. cattleianum has become invasive in at least 31 countries, representing major 122 biogeographical regions in tropical to subtropical zones (Ellshoff et al. 1995; Richardson & 123 Rajmanek 2011). Hence, P. cattleianum is listed in the Global Invasive Species Database as 124 being among the world’s top 100 worst invasive alien species (Global Invasive Species 125 Database 2014), with the islands of Hawaii (Huenneke & Vitousek 1990), Mauritius (Lorence 126 & Sussman 1986), Réunion (Tassin et al. 2006) and Seychelles (Fleischmann 1997; Dietz et 127 al. 2004) being especially affected. 128 The form of P. cattleianum afflicting Australia is red-fruited, and the earliest known 129 collection in 1945 was from Koah, north Queensland (Atlas of Living Australia 2014). The 130 species was probably introduced for its edible fruits, and is presently classed as a potential 131 environmental weed (Csurhes & Edwards 1998). It is now known to be invasive in various 132 tropical and subtropical locations stretching from north Queensland to northern New South 133 Wales (Downey et al. 2010) and also on some offshore islands like Lord Howe (Auld & 134 Hutton 2004) and Norfolk (Mills 2012) island. 135 Our study sites are on the Atherton Tablelands (17° 21' S, 145° 35' E) in north-eastern 136 Queensland, and ranged in altitude from 700-830 m asl (Fig. 1). Annual rainfall in the study 137 area ranges from 1,700-2,600 mm with a distinct dry season (where mean monthly rainfall is 138 less than 100 mm) from July to September. Mean monthly temperatures range from a 7 139 minimum of 10°C to 29°C. The vegetation of the study sites comprises upland tropical 140 evergreen rainforest in various stages of recovery from anthropogenic disturbance. Primary 141 rainforests are confined to ranging from 1-600 ha (Laurance & Laurance 1999), and 142 secondary rainforests comprises > 11,000 ha and distributed into almost 2,500 patches (Sloan 143 et al., in review). 144 [INSERT FIG. 1 HERE] 145 Survey Methods 146 Thirty-three regrowth rainforest sites on soils derived from granite and basalt were selected 147 along a successional chronosequence which ranged in age of abandonment from 3 to 69 148 years. An additional eight sites representing primary rainforest were sampled for comparison. 149 Psidium cattleianum was present at 27 sites, all of which were successional. Adult P. 150 cattleianum was present in 19 of these sites. We determined site ages was using a range of 151 Queensland State Government digital or hardcopy aerial photography (1943 -2011), and 152 satellite imagery from Google Earth from 2002-2014 (© 2014 Google Image, ©2014 153 DigitalGlobe), and Queensland Globe from 2011 and 2014 (©State of Queensland 2013, 154 ©CNES 2012, Spot Image S.A. France, ©2013 Pitney Bowes). Each image was examined for 155 vegetation cover, using stereo pairs of images where available (1943-1997). Otherwise, aerial 156 photographs were scanned at high resolution and successive pairs of digital images were 157 compared side-by-side on-screen. We determined the age since abandonment to be the mid- 158 point between successive images where pasture had been replaced by another vegetation type 159 (e.g. shrubby weeds, scramblers, shrubs and scattered tree saplings). 160 At each site, we recorded plant community structure and composition along 50 m transects, 161 using survey methods described in Preece et al. (2012). All stems >2.5 m in height and <10 162 cm dbh were recorded in 3 m belt transects and trees ≥10 cm dbh in 10 m belt transects. At 5 8 163 metre intervals along the transect we counted seedlings in 1 m x 1 m plots, and estimated 164 canopy cover using a spherical crown densiometer, canopy height and slope using a 165 clinometer, and aspect with a compass. 166 Distances to continuous rainforest blocks, to remnant primary rainforest, to waterbodies and 167 to the nearest anthropogenic land-use feature (roads, pastures, abandoned fallows, human 168 residences, etc.) and elevation were derived from aerial photography and GIS layers. For each 169 site, we obtained 10 soil samples – one every 5 m along the transect and a single sample 5 m 170 perpendicular to transect. At each sample site, the top 30 cm of the soil layer (after removing 171 the leaf litter) was collected with a hand auger. The 10 soil samples for each site were pooled 172 and sent to a commercial laboratory (Nutrient Advantage - Incitec Pivot ®, Victoria, 173 Australia) for analysis. Soil particle analysis (clay, silt and sand percentages) was performed 174 using the hydrometer method, pH was determined using a digital pH meter in a 1:5 soil-water 175 suspension, and cation exchange capacity (CEC = sum of exchangeable cations) was obtained 176 using standard protocols by Rayment & Lyons (2011). 177 Data Analysis 178 First, we described the population structure of P. cattleianum across our study sites. From the 179 19 sites with P. cattleianum adults we grouped individuals (each individual being the sum of 180 all stems or coppice shoots) into six dbh size class categories ranging from 2.5 cm to 17.5 cm 181 dbh. To further determine if duration since site abandonment had an effect on P. cattleianum 182 demographics, we plotted P. cattleianum size classes segregated into three categories of 183 forest succession (years since abandonment: <15; 15-29 and; >30 years). A chi-squared test 184 was used to test for homogeneity of size class distribution in the different forest succession 185 categories. 9 186 Second, we compared the density of P. cattleianum (mean number of individual stems) with 187 that of four common shrub species that occupy the same ecological niche in the study area. 188 This allowed us to examine the relative proportions of the shrub niche occupied by these 189 species in the understorey. We restricted this demographic comparison to individuals within 190 the stem size range of 2.5 - 10 cm dbh for the shrub species: Guioa lasioneura, Neolitsea 191 dealbata, Rhodamnia sessiliflora and Rhodomyrtus pervagata. We used Kruskal-Wallis H 192 test to test for differences in the mean percentage of individuals that formed multi-stemmed 193 plants among the five shrub species. 194 Third, we examine what environmental parameters best account for P. cattleianum basal area 195 and seedling (stems < 2.5 cm) density. We used two sets of environmental variables, one 196 pertaining to vegetation community attributes and another to site attributes. Community 197 attributes included Shannon Weiner diversity index, evenness, and the densities of grass 198 clumps and tree, shrub, vine, herb, ferns, and exotic species seedlings. We also computed and 199 compared the basal area of G. lasioneura, N. dealbata, R. sessiliflora and R. pervagata 200 between invaded and non-invaded sites, to examine if these species occupy more niche space 201 in non-invaded secondary forest sites. We calculated the basal area (m2) of each species as 202 (dbh/200)2 × 3.14, and in the case of multiple-stemmed individuals, the basal area was the 203 sum of all stems. Site attributes included canopy height, slope, aspect, soil pH and cation 204 exchange capacity (CEC), and fractions of sand and clay. We used Mann-Whitney U-tests (P 205 < 0.05) to compare these attributes between P. cattleianum-invaded (n=27) and non-invaded 206 secondary rainforest sites (n=6). As the differences between primary and secondary forests 207 were largely floristic (see later), we restricted these univariate comparisons between 27 208 Psidium cattleianum-invaded and the six non-invaded secondary forest sites. 209 Fourth, we examined how basal areas and seedling densities of P. cattleianum varied among 210 communities and correlated with site attributes, using all 41 sites, including sites with and 10 211 without P. cattleianum. Site attributes and community attributes were examined in separate 212 models. For the community attribute models we included gradients in community 213 composition as an additional explanatory variable. To achieve this, we performed Non-metric 214 Multidimensional Scaling (NMDS) ordinations on species presence/absence data (excluding 215 the presence of P. cattleianum) using Bray-Curtis similarity. NMDS ordinations were 216 performed using the vegan package (Oksanen et al. 2014) in R 2.10.0 (R Development Core 217 Team 2009). The NMDS axes reflect floristic gradients, which are associated with the 218 distribution of mature-phase rainforest species. NMDS Axis 1 increases with greater numbers 219 of mature-phase species (Fig. 2). A standard protocol of data exploration was used to 220 determine significantly correlated variables which were excluded from the models. In the 221 final generalised linear models (GLM) we included nine community attributes (NMDS axis 222 1, NMDS axis 2, grass clump density, and seedling densities of vines, shrubs, trees and other 223 exotic species, and canopy height) and five site attributes (cation exchange capacity, distance 224 to remnant forest, pH and sand fraction). Because the response variables were zero-inflated, 225 we fitted our GLM models using a Tweedie distribution and log link (Dunn et al. 2009). 226 GLM models were fitted in SPSS (IBM Corp 2011). 227 [INSERT FIG. 2 HERE] 228 RESULTS 229 Psidium cattleianum demographics 230 Psidium cattleianum was recorded in 27 of the 41 sampled sites, comprising 326 established 231 stems (including all coppice stems) and accounting for 7.9% of all individual stems recorded 232 in the study. We counted 1324 seedlings of P. cattleianum which represents 19.5% of all 233 seedlings recorded and also the highest number of seedlings per species. Psidium 234 cattleianum was also the most abundant non-native species encountered across all of our 11 235 study sites. Among four other woody non-native species with basal area measurements 236 (Cinnamomum camphora, Lantana camara, Ligustrum sinense, Michelia champaca), P. 237 cattleianum comprised 88.3% of the stems and 49.9% of their total basal area. Among non- 238 native seedlings and herbs (<2.5 cm dbh), P. cattleianum comprised 59% of the stems. 239 The size class distribution of P. cattleianum individuals exhibited a consistent reverse J- 240 shaped distribution regardless of the time since forest abandonment (Fig. 3). However, 241 individuals with a dbh size class above 12.5 cm were found only in sites from the recently 242 abandoned category (<15 years) (χ2 = 4.492, d.f. = 1, p = 0.034) – a result that can be 243 attributed to one particularly infested site. 244 Among the shrub species compared across the 19 sites with established P. cattleianum 245 individuals, P. cattleianum was the most common shrub species encountered and exhibited 246 the highest mean number of individual stems within the 2.5 to 10 cm dbh range (Fig. 4a). 247 Whilst all the four native shrub species formed multi-stemmed plants, none exhibited as high 248 a percentage of multi-stemmed individuals as P. cattleianum (Kruskal-Wallis H test: χ2 = 249 10.8, d.f. = 4, p = 0.009; Fig. 4b). Psidium cattleianum also achieved almost four times the 250 stem number of the most abundant native species Rhodomyrtus pervagata (Fig. 4c) 251 (INSERT FIG. 3 HERE) 252 (INSERT FIG. 4 HERE) 253 Environmental correlates of Psidium cattleianum invasion 254 Relative to non-invaded sites, invaded sites tended to be more west-facing (higher aspect 255 degrees; p = 0.011) and were further away from intact forest blocks (p = 0.009). However, 256 invaded and non-invaded sites did not differ in canopy cover (p > 0.05). Invaded sites also 257 had higher grass clump densities, higher densities of seedlings of tree species and native 12 258 groundcover species, but lower densities of seedlings of other exotic species (Table 1). 259 Importantly, the diversity and the basal areas of the four understorey shrub species (Guioa 260 lasioneura, Neolitsea dealbata, Rhodomyrtus pervagata, Rhodamnia sessiliflora) did not 261 differ between P. cattleianum-invaded and non-invaded sites (all p > 0.05). 262 (INSERT TABLE 1 HERE) 263 Community and environmental correlates of Psidium cattleianum invasion 264 For each of the two response variables, P. cattleianum basal area and seedling density, we 265 fitted two sets of GLMs – one using community attributes and the other using site attributes. 266 Across the 41 sites, P. cattleianum basal area and seedling density increased as forest canopy 267 height declined (Table 2). Psidium cattleianum seedling density further exhibited negative 268 relationships with NMDS Axis 1 and the seedling densities of other exotic species. No 269 significant relationship with the other predictive variables was detected. Both P. cattleianum 270 basal area and seedling density increased with distance to remnant forest, soil pH, and a 271 declined with soil cation exchange capacity. Finally, P. cattleianum seedling density 272 associated with soil sand fraction, suggesting seedling recruitment is higher in sandy soils 273 (Table 3). 274 [INSERT TABLE 2 HERE] 275 [INSERT TABLE 3 HERE] 276 DISCUSSION 277 Our study of secondary forest communities revealed that Psidium cattleianum is now well- 278 established in upland successional forests of the Australian Wet Tropics. Adult stems 279 occurred in 65% of our study sites and seedlings comprised 20% of the 6,800 individuals that 280 we identified. Other shade-tolerant understorey weeds in rainforest in the current study 13 281 include Ardisia crenata and Cinnamomum camphora but none of these achieved stem 282 densities or basal areas as high as P. cattleianum. Contrary to expectation the age of 283 secondary forest did not influence the number of P. cattleianum individuals or their size, with 284 almost equal numbers found in young and older forests, and some of the largest stems were 285 found at the youngest sites. However P. cattleianum was strongly associated with sites that 286 had low forest canopies and abundant in grass, herbs, which suggests that this species may 287 spread into abandoned lands and secondary rainforest in this region. 288 Little invasion of P. cattleianum into primary rainforest was detected, suggesting that primary 289 rainforest may have some resilience to invasion. Significantly, P. cattleianum-invaded sites 290 were further from intact forest blocks, which could be a result of a longer history of 291 disturbance at these invaded sites. Similarly, invasions were more prominent further from 292 primary rainforest fragments. It is unlikely that the lack of P. cattleianum in the eight 293 primary rainforest remnants was due to seed limitation as the species produces abundant seed 294 and exerts considerable propagule pressure wherever found. Taking into account the well- 295 established shade-tolerance of the species (Huenneke & Vitousek 1990; Fleischmann 1997), 296 we can only hypothesize that some other undetermined factor reduces seedling recruitment. 297 Soil pathogens, seed or seedling predators, competition with or allelopathy from established 298 tree species or combinations of these factors can all potentially limit recruitment. 299 Psidium cattleianum was the most common stem (<10 cm dbh) encountered in secondary 300 rainforests and occurred in 82% of these sites. Relative to non-invaded sites, invaded sites 301 were generally abandoned more recently, and associated with a higher abundance of grass 302 and herbs, which suggests that if there is no seed limitation in the area then recruitment is 303 higher in younger rather than older secondary rainforests. While P. cattleianum invasion 304 appeared to be favoured by a number of community and site attributes, in particular, soil 14 305 structural and chemical attributes, the persistence and demographic structure of P. 306 cattleianum in these forests suggests that further spread is likely. 307 Site and community correlates of Psidium cattleianum invasibility 308 Communities with high species richness are thought to be resistant to invasion (Hooper et al. 309 2005; Martin et al. 2009) because local niches are filled by representatives from different 310 functional groups (Zavaleta & Hulvey 2007). We found no direct evidence of this in our 311 analyses. However, we found a negative relationship between P. cattleianum seedling density 312 and the gradient of community composition determined in our NMDS ordination, which in 313 general segregated secondary rainforest from primary rainforest (Fig. 2). The primary 314 rainforest in turn had a higher Shannon-Weiner diversity than all secondary forest sites as a 315 group (Mann-Whitney U, p < 0.001), so it is plausible that the floristic composition and 316 higher diversity of the primary rainforests either separately or in combination suppress the 317 establishment of P. cattleianum seedlings. Likewise, the negative association of canopy 318 height with density of P. cattleianum seedlings is likely to be a function of the generally taller 319 canopy found in primary rainforests. However, it is notable that canopy cover did not have a 320 significant effect on P. cattleianum basal areas or seedling densities, reinforcing the broad 321 range of light conditions the species can tolerate. 322 The invasibility of a community is thought to decrease when the native species matrix 323 includes species of similar functional groups or with traits similar to the invader because such 324 species will fill a greater proportion of potentially available niches (Elton 1958; Gilbert & 325 Lechowizc 2005). Several native understorey shrubs, Guioa lasioneura, Neolitsea dealbata, 326 Rhodomyrtus pervagata and Rhodamnia sessiliflora that are shade-tolerant, coppice like P. 327 cattleianum. These species could be interpreted as belonging to the same functional group but 328 the mean basal areas of these species did not differ in P. cattleianum-invaded and non- 15 329 invaded secondary forest sites. Follow-up studies using functional niche-modelling (e.g. 330 Moles et al. 2008) may provide insight on whether P. cattleianum is occupying empty niches 331 in these secondary rainforests. 332 333 Empirical observations and implications of Psidium cattleianum invasion 334 The probability plant invasiveness increases if a species reproduces vegetatively and has a 335 history of invasion elsewhere (Kolar & Lodge 2001). Psidium cattleianum meets both these 336 criteria – it has the highest number of coppice stems of any woody species examined in the 337 study, and has a significant history of invasion in Hawaii and many other tropical regions 338 dating back to the early- to mid-1800s (Lorence & Sussman 1986; Huenneke & Vitousek 339 1990; Kueffer et al. 2008). With prolific seeds and seedling establishment and ability to 340 survive under a broad range of light conditions (Loh & Daehler 2007), Psidium cattleianum 341 fits the profile of a ‘‘super invader’’ (Daehler 2003). Given that Psidium cattleianum has 342 established self-sustaining populations in tropical north Queensland, the invasion of more 343 areas is imminent. 344 The continuing establishment and spread of P. cattleianum in Australian rainforest could also 345 be favoured by the apparent lack of natural herbivore enemies. In our study, P. cattleianum 346 leaves were observed to be always in a healthy state and with no signs of herbivory. This is 347 consistent with the enemy-release hypothesis of Keane & Crawley (2002), which posits that 348 there is greater impact of natural enemies on natives than on a given exotic species in its 349 introduced range. In Hawaii, Shields et al. (2014) reported experimental findings of P. 350 cattleianum seedlings being less susceptible to herbivory than native plant seedlings. 351 Similarly, myrtle rust, which is an exotic rust disease affecting members of the myrtle family 352 (Morin et al. 2011), has been observed to affect several native members of the myrtle family 16 353 in the study sites. In contrast, we have observed no certifiable cases of infection in P. 354 cattleianum, even when growing adjacent to myrtle rust-infected native Rhodomyrtus 355 pervagata and Rhodamnia sessiliflora. 356 Another set of features that could facilitate the success of P. cattleianum is the rapidity with 357 which it attains reproductive maturity, and its dispersal mechanism. We have observed stems 358 below 30cm height and less than 2.5cm stem diameter in flower. The seedling density of P. 359 cattleianum was also the highest among all species encountered, reflecting the copious fruit 360 set from preceding fruiting seasons. Psidium cattleianum fruits are eaten by native birds and 361 spectacled flying foxes (Cooper & Cooper 2004), which undoubtedly aids in the spread of the 362 species. 363 The ability of P. cattleianum to persist under shade and to attain high basal areas and stems 364 densities can have serious ecological consequences. Psidium cattleianum comprised about 365 60% of woody basal area and 85% of the seedlings at an extreme site that had been 366 abandoned for 12.5 years. At this stage there is insufficient evidence to conclude whether the 367 species will self-thin and be replaced as the forest matures, although this is unlikely in light 368 of the shade tolerance of the species. For instance, Zimmerman et al. (2008) found in lowland 369 Hawaii that the functional and compositional integrity of forests were increasingly 370 compromised by P. cattleianum invasion, even though these forests remained at least 371 partially intact in several locations. Likewise, P. cattleianum invasion has been implicated in 372 modifying successional trajectories in Mauritius (Lorence & Sussman 1986) and on the 373 Seychelles (Fleischmann 1997). How severely P. cattleianum invasion will arrest the 374 succession of rainforest in the region will require monitoring. However, we may speculate 375 that in the absence of P. cattleianum invasion, successional trajectories involving the native 376 woody genera Acacia, Alphitonia, or Rhodomyrtus (Yeo & Fensham 2015; Goosem et al. in 377 review) would proceed. 17 378 While it is arguable that P. cattleianum may at least provide some ecosystem services such as 379 habitat and seasonal food resources for native animals, the species has potential to alter 380 ecosystem structure and function ways that are difficult or impossible to reverse (Gaertner et 381 al. 2014). We therefore advocate that P. cattleianum be prioritized for control in Australia. 382 The current national and state listings are inadequate for P. cattleianum, and a first step could 383 be to list the species as an environmental weed under the National Environmental Alert List 384 (Department of the Environment 2014). With increased and sufficient public awareness, and 385 given the recent occurrence in Australia relative to other infested tropical regions, controlling 386 P. cattleianum invasion may be achievable. However decisive action is required, and control 387 efforts need to be sustained and monitored for efficacy. 388 ACKNOWLEDGEMENTS 389 We thank Steve McKenna, Rigel Jensen, Andrew Hunter, Ana Palma and Martha Karafir for 390 their help with fieldwork. This research was supported by a Linkage Grant LP110201093 391 from the Australian Research Council and the Queensland Herbarium and an Australian 392 Research Council Future Fellowship to SL. We are deeply appreciative for the access 393 provided by the landholders who supported this project. We also thank our two anonymous 394 reviewers for their comments. The authors declare no conflicting interests. 395 REFERENCES 396 Atlas of Living Australia. (2014) [Cited 29 October 2014.] Available from URL: 397 http://www.ala.org.au 398 Auld T. D. & Hutton I. (2004) Conservation issues for the vascular flora of Lord Howe 399 Island. Cunninghamia 8, 490–500. 18 400 Brown R. L. & Fridley J. D. (2003) Control of plant species diversity and community 401 invasibility by species immigration: seed richness versus seed density. Oikos 102, 15–24. 402 Bruno J. F., Stachowicz J. J. & Bertness M. D. (2003) Inclusion of facilitation into ecological 403 theory. Trends Ecol. Evol. 18, 119–25. 404 Burke M. J. & Grime W. J. P. (1996) An experimental study of plant community invasibility. 405 Ecology 77, 776–90. 406 Cavers P. B. & Harper J. L. (1967) Studies in the dynamics of plant populations. I. The fate 407 of seed and transplants introduced into various habitats. J. Ecol. 55, 59–71. 408 Cooper W. T. & Cooper W. (2004) Fruits of the Australian Tropical Rainforest. Nokomis 409 Editions, Melbourne, Australia. 410 Csurhes S. & Edwards R. (1998) Potential environmental weeds in Australia. Queensland 411 Department of Natural Resources, Queensland, Australia. 412 Daehler C. C. (2003) Performance comparisons of co-occurring native and alien invasive 413 plants: implications for conservation and restoration. Ann. Rev. Ecol. Syst. 34, 183–211. 414 Department of the Environment. (2014) National Environment Alert List. [Cited 8 December 415 2014.] Available at URL: 416 http://www.environment.gov.au/biodiversity/invasive/weeds/weeds/lists/alert.html 417 Dietz H., Wirth L. R. & Buschmann H. (2004) Variation in herbivore damage to invasive and 418 native woody plant species in open forest vegetation on Mahé, Seychelles. Biol. Invasions 6, 419 511–21. 420 Diong C. H. (1982) Population biology and management of the feral pig (Sus scrofa) in 421 Kipahulu Valley, Maui. PhD dissertation. University of Hawaii Manoa, Hawaii. 19 422 Ditham R. K., Tylianakis J. M., Gemmel N. J., Rand T. A. & Ewers R .M. (2007) Interactive 423 effects of habitat modification and species invasion on native species decline. Trends Ecol. 424 Evol. 22, 489–96. 425 Downey P. O., Scanlon T. J. & Hosking J. R. (2010) Prioritizing weed species based on their 426 threat and ability to impact on biodiversity: a case study from New South Wales. Plant 427 Protect. Quart. 25, 111–26. 428 Dunn P. K., Smyth G. K., Walker N., Saveliev A. A. & Smith G. M. (2009) Series evaluation 429 of Tweedie exponential dispersion model densities. Stat. Comput. 15, 267–80. 430 Dunstan P. K. & Johnson C. R. (2006) Linking richness, community variability, and invasion 431 resistance with patch size. Ecology 87, 2842–50. 432 Ehrenfeld J. G. (2003) Effects of exotic plant invasions on soil nutrient cycling processes. 433 Ecosystems 6, 503–23. 434 Ellshoff, Z. E., Gardner D. E., Wikler C. & Smith C. W. (1995) Annotated bibliography of 435 the genus Psidium with emphasis on P. cattleianum (Strawberry guava) and P. guajava 436 (common guava), forest weeds in Hawai'i. Technical Report 95. Department of Botany, 437 Honolulu, Hawai'i. 438 Elton C. S. (1958) The ecology of invasions by animals and plants. Methuen, London, UK. 439 Fargione J., Brown C. S. & Tilman D. (2003) Community assembly and invasion: An 440 experimental test of neutral versus niche processes. PNAS 100, 8916–20. 441 Fleischmann K. (1997) Invasion of alien woody plants on the islands of Mahé and Silhouette, 442 Seychelles. J. Veg. Sci. 8, 5–12. 20 443 Gaertner M., Biggs R., Te Beest M., Hui C., Molofsky J. & Richardson D. M. (2014) 444 Invasive plants as drivers of regime shifts: identifying high-priority invaders that alter 445 feedback relationships. Divers. Distri. 20, 733–44. 446 Gilbert B. & Lechowizc M. J. (2005) Invasibility and abiotic gradients: the positive 447 correlation between native and exotics plant diversity. Ecology 86, 1848–55. 448 Global Invasive Species Database. (2014) Psidium cattleianum. [Cited 27 September 2014.] 449 Available at URL: 450 http://www.issg.org/database/species/ecology.asp?si=59&fr=1&sts=sss&lang=EN 451 Goosem MW, Paz CP, Fensham RJ, Preece ND, Goosem S, Laurance SGW. (in review) 452 Delayed recovery of species diversity and community composition in secondary forests in 453 tropical Australia. J. Veg. Sci. 454 Hager H. A. (2004) Competitive effect versus competitive response of invasive and native 455 wetland plant species. Oecologia 139, 140–9. 456 Hayes K. R. & Barry S. C. (2008) Are there any consistent predictors of invasion success? 457 Biol. Invasions 10, 483–506. 458 Heip C. H. R., Herman P. M. J. & Soetaert K. (1998) Indices of diversity and evenness. 459 Océanis 24, 61–87. 460 Hejda M. & Pyšek P. (2006) What is the impact of Impatiens glandulifera on species 461 diversity of invaded riparian vegetation? Biol. Conserv. 132, 143–52. 462 Hejda M., Pyšek P. & Jarošík V. (2009) Impact of invasive plants on the species richness, 463 diversity and composition of invaded communities. J. Ecol. 97, 393–403. 21 464 Heywood V. H. (1989) Patterns, extents and modes of invasions by terrestrial plants. In: 465 Biological Invasions: a Global Perspective, SCOPE 37. (eds J. A. Drake, H. A. Mooney, F. 466 diCastri, R. H. Groves, F. J. Kruger, M. Rejmanek, & M. Williamson) pp 31–60. John Wiley, 467 New York, US. 468 Hooper D. U., Chapin III F. S., Ewel J. J., et al. (2005) Effects of biodiversity on ecosystem 469 functioning: a consensus of current knowledge. Ecol. Monographs 75, 3–35. 470 Huenneke L. F. (1989) Contribution of sprouting behavior to population persistence in woody 471 species. Amer. J. Bot. (Suppl.) 76, 84–5. 472 Huenneke L. F. & Vitousek P. M. (1990) Seedling and clonal recruitment of the invasive tree 473 Psidium cattleianum: Implications for management of native Hawaiian forests. Biol. Conserv. 474 53, 199–211. 475 Hughes R. F. & Denslow J. S. (2005) Invasion by a N2-fixing tree alters function and 476 structure in wet lowland forests of Hawaii. Ecol. Appl. 15, 1615–28. 477 Hulme P. E. & Bremner E. T. (2006) Assessing the impact of Impatiens glandulifera on 478 riparian habitats: partitioning diversity components following species removal. J. Appl. Ecol. 479 43, 43–50. 480 IBM Corp. (2011) IBM SPSS Statistics for Windows, Version 20.0. IBM Corp, Armonk, NY. 481 Keane R. M. & Crawley M. J. (2002) Exotic plant invasions and the enemy release 482 hypothesis. Trends Ecol. Evol. 17, 164–70. 483 Kolar C. S. & Lodge D. M. (2001) Progress in invasion biology: predicting invaders. Trends 484 Ecol. Evol. 16, 199–204. 22 485 Krebs C. J. (2009) Ecology: The Experimental Analysis of Distribution and Abundance, 6th 486 edn. Benjamin Cummings, San Francisco, US. 487 Kueffer C., Klinger G., Zirfass K., Schumacher E., Edwards P. J. & Güsewell S. (2008) 488 Invasive trees show only weak potential to impact nutrient dynamics in phosphorus-poor 489 tropical forests in the Seychelles. Funct. Ecol. 22, 359–66. 490 Laurance S. G. W. & Laurance W. F. (1999) Tropical wildlife corridors: use of linear 491 rainforest remnants by arboreal mammals. Biol. Conserv. 91, 231–239.Levine J. M. & 492 D'Antonio C. M. (1999) Elton Revisited: A review of evidence linking diversity and 493 invasibility. Oikos 87, 15–26. 494 Levine J. M., Vila M., D’Antonio C. M., Dukes J. S., Grigulis K. & Lavorel S. (2003) 495 Mechanisms underlying the impacts of exotic plant invasions. Proc. Biol. Sci. 270, 775–81. 496 Loh R. K. & Daehler C. C. (2007) Influence of invasive tree kill rates on native and invasive 497 plant establishment in a Hawaiian forest. Restoration Ecol. 15, 199–211. 498 Lonsdale W. M. (1999) Global patterns of plant invasions and the concept of invasibility. 499 Ecology 80, 1522–36. 500 Lorence D. H. & Sussman R. W. (1986) Exotic species invasion into Mauritius wet forest 501 remnants. J. Trop. Ecol. 2, 147–62. 502 503 Mackey B. G. (1993) A spatial-analysis of the environmental relations of rain-forest 504 structural types. J. Biogeogr. 20, 303–36. 23 505 Martin P. H., Canham C. D. & Marks P. L. (2009) Why forests appear resistant to exotic 506 plant invasions: intentional introductions stand dynamics and the role of shade tolerance. 507 Front. Ecol. Environ. 7, 142–9. 508 Meyer, J.-Y. (2004) Threat of invasive alien plants to native flora and forest vegetation of 509 eastern Polynesia. Pac. Sci. 58, 357–75. 510 Milbau A. & Nijs I. (2004) The role of species traits (invasiveness) and ecosystem 511 characteristics (invasibility) in grassland invasions: a framework. Weed Technology 18, 512 1301–4. 513 Mills K. (2012) The endemic flora of Norfolk Island: Conservation challenges on a remote 514 oceanic island. Aust. Plant Conserv. 21, 19–21. 515 Moles A. T., Gruber M. A. M. & Bonser S. P. (2008) A new framework for predicting 516 invasive plant species. J. Ecol. 96, 13–7. 517 Motley, T.J. (2005) Tetraplasandra lydgatei (Araliaceae): Taxonomic recognition of a rare, 518 endemic species from O‘ahu, Hawaiian Islands, Pac. Sci. 59, 105–10. 519 Morin L., Paini D. R. & Randall R. P. (2013) Can global weed assemblages be used to 520 predict future weeds? PLoS ONE 8, e55547. 521 Morin L., Aveyard R. & Lidbetter J. (2011) Myrtle rust: host testing under controlled 522 conditions. NSW Department of Primary Industries, West Pennant Hills, NSW, Australia. 523 Mullah C. J. A., Klanderud K., Totland Ø. & Odee D. (2014) Community invasibility and 524 invasion by non-native Fraxinus pennsylvanica trees in a degraded tropical forest. Biol. 525 Invasions 16, 2747–55. 24 526 Murphy H. T., Hardesty B. D., Fletcher C. S., Metcalfe D. J., Westcott D. A. & Brooks S. J. 527 (2008) Predicting dispersal and recruitment of Miconia calvescens (Melastomataceae) in 528 Australian tropical rainforests. Biol. Invasions 10, 925–36. 529 Oksanen J., Blanchet F. G., Kindt R. et al. (2014) Package vegan: Community ecology 530 package. http://CRAN.R-project.org/package=vegan. Accessed 21 February 2014. 531 Ortega Y. K. & Pearson D. E. (2005) Strong versus weak invaders of natural plant 532 communities: distinguishing invasibility from impact. Ecol. Appl. 15, 651–61. 533 Patel S. (2012) Exotic tropical plant Psidium cattleianum: a review on prospects and threats. 534 Rev. Environ. Sci. Biotech. 11, 243–8. 535 Pattison R. R., Goldstein G. & Ares A. (1998) Growth, biomass allocation and 536 photosynthesis of invasive and native Hawaiian rainforest species. Oecologia 117, 449–59. 537 Preece N. D., Crowley G. M., Lawes M. J. & Van Oosterzee P. (2012) Comparing above- 538 ground biomass among forest types in the Wet Tropics: Small stems and plantation types 539 matter in carbon accounting. Forest Ecol. Manag. 264, 228–37. 540 Pyšek P., Jarošík V., Hulme P. E. et al. (2012) A global assessment of invasive plant impacts 541 on resident species, communities and ecosystems: the interaction of impact measures, 542 invading species’ traits and environment. Glob. Change Biol. 18, 1725–37. 543 R Development Core Team. (2009) R: a language and environment for statistical computing. 544 R Foundation for Statistical Computing, Vienna. [Cited 21 February 2014.] Available at 545 URL: http://www.R-Project.org 546 Rayment G. E. & Lyons D. J. (2011) Soil chemical methods: Australasia (Vol. 3). CSIRO 547 publishing, Victoria, Australia. 25 548 Reitz P. R., Klein R. M. & Reis A. (1983) Flora Catarinense (Psidium). Sellowia 35, 684– 549 715. 550 Rejmánek M. (1989) Invasibility of plant communities. In: Biological Invasions. A Global 551 Perspective. (eds J. A. Drake, H.A. Mooney, F. di Castri et al.) pp. 369–388. John Wiley & 552 Sons, Chichester, UK. 553 Rejmanek M. & Richardson D. M. (1996) What attributes make some plant species more 554 invasive? Ecology 77, 1655–61. 555 Richardson D. M. & Rejmánek M. (2011) Trees and shrubs as invasive alien species – a 556 global review. Divers. Distri. 17, 788–809. 557 Safeeq M. & Fares A. (2014) Interception losses in three non-native Hawaiian forest stands. 558 Hydrol Processes 28, 237–54. 559 Shea K. & Chesson P. (2002) Community ecology theory as a framework for biological 560 invasions. Trends Ecol. Evol. 17, 170–6. 561 Shiels A. B., Ennis M. K. & Shiels L. (2014) Trait-based plant mortality and preference for 562 native versus non-native seedlings by invasive slug and snail herbivores in Hawaii. Biol. 563 Invasions 16, 1929–40. 564 Sloan S., Goosem M. W. & Laurance S. G. W. (in review) Tropical forest recovery following 565 land abandonment is driven by primary rainforest distribution in an old pastoral region. 566 Landscape Ecol. 567 Sobral M., Proença C., Souza M., Mazine F. & Lucas E. (2013) Myrtaceae in Lista de 568 Espécies da Flora do Brasil. Jardim Botânico do Rio de Janeiro. [Cited 30 January 2015.] 569 Available from URL: http://floradobrasil.jbrj.gov.br/jabot/floradobrasil/FB10858 26 570 Standish R. J., Robertson A. W. & Williams P. A. (2001) The impact of an invasive weed 571 Tradescantia fluminensis on native forest regeneration. J. Appl. Ecol. 38, 1253–63. 572 Symstad A. J. (2000) A test of the effects of functional group richness and composition on 573 grassland invasibility. Ecology 81, 99–109. 574 Tassin J., Riviere J-N., Cazanove M. & Bruzzese E. (2006) Ranking of invasive woody plant 575 species for management on Réunion Island. Weed Research 46, 388–403. 576 Thompson K. & Davis M. A. (2011) Why research on traits of invasive plants tells us very 577 little. Trends Ecol. Evol. 26, 155–6. 578 Uowolo A. L. & Denslow J. S. (2008) Characteristics of the Psidium cattleianum 579 (Myrtaceae) seed bank in Hawaiian lowland wet forests. Pacific Sci. 62, 129–35. 580 Vila M. & Weiner J. (2004) Are invasive species better competitors than native plant species? 581 Evidence from pair-wise experiments. Oikos 105, 229–38. 582 Vila M., Williamson M. & Lonsdale M. (2004) Competition experiments on alien weeds with 583 crops: lessons for measuring plant invasion impact? Biol. Invasions 6, 59–69. 584 Von Holle B., Delcourt H. R. & Simberloff D. (2003) The importance of biological inertia in 585 plant community resistance to invasion. J. Veg. Sci. 14, 425–32. 586 Yurkonis K. A. & Meiners S. J. (2004) Invasion impacts local species turnover in a 587 successional system. Ecol. Letters 7, 764–9. 588 Yurkonis K. A., Meiners S. J. & Wachholder B. E. (2005) Invasion impacts diversity through 589 altered community dynamics. J. Ecol. 93, 1053–1061.Zavaleta E. S. & Hulvey K. B. (2007) 590 Realistic variation in species composition affects grassland production, resource use and 591 invasion resistance. Plant Ecol. 188, 39–51. 27 592 Zimmerman N., Hughes R. F., Cordell S. et al. (2008) Patterns of primary succession of 593 native and introduced plants in lowland wet forests in eastern Hawai‘i. Biotropica 40, 277– 594 84. 595 28 596 Table 1. Means (±1SD) of significantly different site and community attributes between 597 Psidium cattleianum-invaded (n=27) and non-invaded (n=6) secondary rainforest plots (50 x 598 10 m). Seedling averages are of 10 quadrats (1 m x 1 m) per plot extrapolated to 500 m2. 599 Significance was determined by Mann-Whitney U-tests (p < 0.05) Descriptor P. cattleianum-invaded Non-invaded sites sites (n=27) (n=6) Mann-Whitney U-tests Mean ±SD Mean ±SD U-statistic p 183.41 107.52 159.00 66.02 26 0.011 2.63 1.21 2.48 0.98 25 0.009 687.04 1095.85 33.33 60.55 6 <0.001 1411.11 2625.59 1533.33 3537.47 27.5 0.013 3642.59 9037.77 1758.33 1225.73 36.5 0.040 5925.93 9772.97 3700.00 2230.47 27.5 0.013 7338.89 9962.43 5233.33 3867.51 33 0.026 Site attributes Aspect (˚) Distance to intact forest block (m) Community attributes Grass clump density Other exotics seedling density Tree species seedling density Total seedling density of native groundcover Total seedling density (excluding Psidium) 600 601 29 602 Table 2. Results of generalized linear models (GLM) fitted with community attributes to 603 estimates of Psidium cattleianum basal area and seedling density across 41 rainforest sites (8 604 primary and 33 secondary). NMDS axis 1 is an ordination axis reflecting a floristic gradient 605 that increases with mature-phase rainforest species (See Fig. 2) Attribute Psidium cattleianum basal area Estim SE ate Walds Psidium cattleianum seedling density p Estimate SE Walds Chi- Chi- Square Square p (Intercept) -1.815 1.614 1.264 0.261ns 9.591 1.393 47.426 <0.001 NMDS Axis 1 -1.038 0.732 2.011 0.156ns -1.764 0.631 7.830 0.005** NMDS Axis 2 -0.335 0.755 0.197 0.657ns -.538 0.667 0.650 0.420ns Grass clump density -0.003 0.019 0.029 0.864ns 0.016 0.015 1.129 0.288ns Vine seedling density -0.050 0.054 0.867 0.352ns -0.068 0.046 2.215 0.137ns Shrub seedling density 0.013 0.029 0.197 0.657ns 0.026 0.025 1.100 0.294ns Tree seedling density -0.002 0.008 0.048 0.826ns -0.001 0.006 0.008 0.928ns Seedling density of other -0.013 0.012 1.244 0.265ns -0.023 0.011 4.299 0.038* -0.204 0.071 8.219 0.004** -0.207 0.061 11.603 0.001** exotic species Canopy height (m) 606 607 608 609 Significance levels: p<0.05*, p<0.01**, p<0.001***; ns, not significant 30 610 Table 3. Results of generalized linear model (GLM) fitted with site attributes to estimates of 611 Psidium cattleianum basal area and seedling density across 41 rainforest sites (8 primary and 612 33 secondary) Attribute Psidium cattleianum basal area Estimate (Intercept) Cation exchange SE Walds Psidium cattleianum seedling density p Estimate SE Walds Chi- Chi- Square Square p -57.045 13.272 18.474 <0.001*** -62.955 11.900 27.986 <0.001*** -1.119 0.307 13.315 <0.001*** -0.996 0.299 11.122 0.001** 1.891 0.598 10.015 0.002** 2.589 0.632 16.788 <0.001*** 73.028 19.767 13.649 <0.001*** 83.323 17.471 22.745 <0.001*** 2.160 2.446 0.780 0.377ns 7.216 2.552 7.997 0.005** capacity Distance to remnant forests (m) pH Sand fraction 613 614 615 Significance levels: p<0.05*, p<0.01**, p<0.001***; ns, not significant 31 616 617 Fig. 1. Locations of sample sites on the Atherton Tablelands, Queensland, Australia. 32 618 619 Fig. 2. Ordinations of the floristic composition of species presence-absence (with the site 620 presences of Psidium cattleianum omitted) of 41 tropical rainforest sites in the Atherton 621 Tablelands, Australia, using Non-metric multidimensional scaling (NMDS). Closed and open 622 symbols respectively represent P. cattleianum-invaded and non-invaded sites. Open squares 623 represent primary rainforest sites, and open and closed triangles represent secondary forest 624 sites. The gradients represented are largely floristic. NMDS axis 1 shows no significant 625 correlation with any of the site and community variables measured in the study, and NMDS 626 axis 2 correlated positively only with soil cation exchange capacity (r = 0.424, p = 0.006) 627 33 628 629 630 631 Fig. 3. Size class distributions of Psidium cattleianum individuals (i.e. all multistems 632 regardless of number considered part of that individual) from the sites divided into three age 633 classes based on their number of years since abandonment (Black: <15 years; Grey: 15-29 634 years; White: >30 years) 635 34 636 637 Fig. 4. Comparisons of the mean (± standard errors) of (a) number of stems of the top five 638 most abundant stems-species within the 2.5 to 10 cm diameter at breast height range, and (b) 639 percentage of individuals with multiple stems. The means of stems ha-1 were extrapolated 640 from nineteen 50 x 3 m transects with Psidium cattleianum invasion. Across all woody 641 species, the greatest capacity for forming multiple-stemmed individuals was observed in P. 642 cattleianum (c), with up to 21 stems >2.5 cm dbh