Assessing potential invasiveness of woody horticultural plant species using seedling growth rate traits


Correspondence author. E-mail:


1. The ornamental plant trade, forestry, and agriculture have been responsible for the initial introduction of over 60% of invasive alien plant species. Screening tools to test potentially new horticultural species should help curtail the continued introduction of new invaders.

2. Using two methods for analysing phylogenetically independent contrasts (PICs) of known invasive and non-invasive, exotic woody horticultural species, we tested the potential of relative growth rate (RGR) and related traits including net assimilation rate (NAR), leaf area ratio (LAR), and specific leaf area (SLA) as predictors of invasiveness. These 29 PICs include 65 species and broadly cover angiosperms.

3. Without accounting for phylogeny, no significant differences were found in seedling RGR or related traits between invasive and non-invasive woody species. Using PICs, invasive species’ RGRs were significantly higher. RGR was considerably more significant using our extensive dataset than in previous smaller studies, while SLA and LAR remained marginally significant. NAR was significantly higher for invasives for the 10–20 days interval.

4. Analysis of this broad data set confirms that RGR is significantly higher for invasive woody species than their non-invasive counterparts, and may serve as a useful biological predictor of invasiveness for woody angiosperms. This expanded study shows that plant species use different physiological and biomass allocation patterns to achieve higher RGR; therefore individual components of RGR, such as SLA, do not consistently predict potential invasiveness of species.

5.Synthesis and applications. Comparative seedling RGR studies show that this measure has potential as a screening tool for new exotic plant species. Unfortunately, more easily measurable components of RGR do not consistently predict invasiveness, as previously thought. Using seedling RGR analysis as an invasive species’ screening tool requires growing a species proposed for introduction with related invasive and non-invasive species. If the tested species’ RGR is higher or not significantly different from its known invasive counterpart, it should be considered highly likely to become invasive, and excluded from further consideration as a potential horticultural species. Seedling RGR could potentially produce a useful, straightforward screening tool when phylogenetically related species or cultivars are available.


Woody plant invaders have been introduced through the ornamental plant trade, forestry, agriculture, and/or as accidental introductions (Pemberton & Liu 2009). Escaped ornamental and other horticultural species account for 52% of naturalised species in Europe (Lambdon et al. 2008), more than half of wildland invasive plants globally, and more than 80% of woody invaders in the USA (Reichard in press).

Horticultural organisations, governmental agencies, and academic institutions recognise the link between the ornamental plant trade and invasion and have introduced a number of voluntary and mandatory regulations to curtail the continued introduction of invasive species (Pheloung, Williams & Halloy 1999; Daehler et al. 2004; Burt et al. 2007). While engagement and outreach with industry members and the public can help decrease the planting of readily available invasive species, screening tools are needed to help prevent further introductions of potential invaders. Biologists have been researching fundamental differences between invasive and non-invasive species since Baker (1965). Biological traits common in invasive species, many of which are also favoured by horticulture and gardeners, include longer flowering and fruiting periods (Lloret et al. 2005; Pyšek & Richardson 2007), faster germination (Schlaepfer et al. 2010), shorter minimum generation times (Rejmánek & Richardson 1996; Widrlechner et al. 2004), higher specific leaf area (SLA) and relative growth rate (RGR; Grotkopp, Rejmánek & Rost 2002; Burns 2004; Hamilton et al. 2005), smaller seeds (Rejmánek & Richardson 1996; Hamilton et al. 2005), greater seed production (Mason et al. 2008), animal dispersal (Widrlechner et al. 2004; Lloret et al. 2005), vegetative growth (Lloret et al. 2005), and self-fertilisation (van Kleunen et al. 2008).

Risk assessment tools have been used to screen potentially invasive species, but they require a large amount of biological and geographical information for each species tested, some of which is difficult to obtain (Pyšek & Richardson 2007). Rejmánek & Richardson (1996) developed successful predictive models for woody plant invasiveness, and their work has been elaborated upon by others (Grotkopp, Rejmánek & Rost 2002; Simberloff, Relva & Nuñez 2002; Jaryan et al. 2007). Minimum generation time, an integral trait in the model, is very difficult to obtain for non-forestry species, making their work inapplicable for general screening purposes. Additionally, complexity of trait interactions (Küster et al. 2008; Bucharova & van Kleunen 2009) and confounding effects such as phylogeny, residence time, and propagule pressure (Pyšek & Richardson 2007; Pemberton & Liu 2009) make invasion predictions difficult. Several new frameworks have been developed to predict species’ potential invasiveness into a particular habitat using biological traits as well as climate information, native range, and plant community traits (Widrlechner et al. 2004; Herron et al. 2007; Moles, Gruber & Bonser 2008).

Among the more promising invasive screening tools are comparative growth rate studies that capitalise on the ability of invasive species to exploit excess resources for growth and reproduction in high resource environments (Burns 2004; Blumenthal 2006; Grotkopp & Rejmánek 2007; Leishman et al. 2007). High-resource environments, often the initial sites of naturalisation and the first front for invasion of new species (Blumenthal 2006), can be simulated in a greenhouse. The RGR of a species determined under such conditions often is comparable in rank to that performed in field studies despite large differences in resources, environmental conditions, and competition (Vile, Shipley & Garnier 2006). Although seedling RGR generally decreases under lower light and/or nutrient levels, woody species’ RGR rankings across different light and nutrient conditions are strongly correlated with those obtained under more optimal conditions (reviewed by Cornelissen, Castro-Díez & Carnelli 1998). Additionally, species’ RGR rankings are consistent over time; seedling RGR is strongly correlated with both RGR of older seedlings as well as G, the long-term growth constant of mature trees (Cornelissen, Castro-Díez & Carnelli 1998). Therefore the results from comparative seedling growth studies reflect, at least in relative terms, the potential growth patterns of species in many natural habitats. Exceptions are those habitats with temperature, light, and/or moisture extremes, and those with strong resident competitors (Rejmánek in press, and references therein).

RGR is a robust tool for understanding life history traits because it integrates a species’ anatomy, physiology, and morphology (Grotkopp, Rejmánek & Rost 2002). RGR can be broken into two components: net assimilation rate (NAR) and leaf area ratio (LAR). Furthermore, LAR can be decomposed into leaf mass ratio (LMR) and SLA (Causton & Venus 1981; Hunt 1982):


These growth traits are influential at the physiological and ecological levels. High SLA is physiologically related to lower leaf construction costs, higher leaf nitrogen levels, higher nitrogen allocation to photosynthesis, and therefore higher photosynthetic nitrogen use efficiency (Meziane & Shipley 2001; Feng, Fu & Zheng 2008). These traits may allow invasive species to exploit resources opportunistically for fast growth and early reproduction (Davis, Grime & Thompson 2000; Blumenthal 2006; Leishman et al. 2007), especially when exotics are released from their natural enemies (Herms & Mattson 1992). Invasive species tend to have smaller seed mass (but see Pyšek & Richardson 2007), which is often correlated with larger seed production per unit canopy per year (Moles & Westoby 2006) and with higher SLA and RGR (Grotkopp, Rejmánek & Rost 2002; Hamilton et al. 2005). Non-invasive species tend to be those with lower SLA (leaves that are more expensive, better defended, and have longer leaf lifespans), lower LAR (more investment into storage and support tissues), and therefore lower RGR (more conservative growth strategies), all strategies well suited for long-term survival under less favourable conditions (Herms & Mattson 1992; Westoby et al. 2002). Understanding these fundamental differences in strategies in terms of the trade-offs in plant growth and biomass allocation patterns can potentially form the basis of successful screening procedures.

Since many traits, including RGR and SLA, are dependent on evolutionary history, differences in these traits between groups of plants often are only apparent when treated as phylogenetically independent contrasts (PICs) and not across species (Saverimuttu & Westoby 1996). PICs account for shared phylogeny by analysing attributes of closely phylogenetically related species that differ in one important trait, in this case invasiveness. Several studies using PICs have found that under optimal conditions and no competition, invasive species have higher RGR, LAR, and/or SLA than their phylogenetically related non-invasive congeners (Grotkopp, Rejmánek & Rost 2002; Burns 2004; Hamilton et al. 2005; Grotkopp & Rejmánek 2007; Feng, Fu & Zheng 2008).

Our goal was to assess the use of RGR and its components as predictors of invasiveness by testing a broad range of phylogenetically related sets of known invasive and non-invasive, horticultural, non-native, woody species. Our species selection avoids the comparison of invasive exotics with natives that may be invasive in other geographic locations (Rejmánek 1999) and reduces the differences in human introduction effort that can confound analyses of invasive and non-invasive species (Pyšek & Richardson 2007). We chose to perform this study under near optimal resources since Cornelissen, Castro-Díez & Carnelli (1998) found that RGR rankings for woody seedlings were similar under higher- and lower-resource levels and Burns (2004) found that differences between invasive and non-invasive species were more apparent at higher nutrient levels. Seedlings grow faster under more optimal conditions, which also better represent our generally resource-rich disturbed habitats. We expand on earlier work (10 intrafamilial contrasts from Grotkopp & Rejmánek 2007) by adding data from 43 species to form a total of 29 PICs that cover a broad spectrum of the angiosperm phylogeny (Table 1).

Table 1.   Contrasts, species, year grown, and relative growth rate (RGR) for the 10–20 and 10–30 days intervals. All growth analysis components are means ± SE (standard error). Invasive species within a contrast are in bold
ContrastSpeciesYearRGR 10–20 days (mg g−1 day−1)RGR 10–30 days (mg g−1 day−1)
  1. aSeeds were commercially obtained.

  2. bSeeds were collected.

  3. cCommonly sold in California and South Africa as Eucalyptus lehmannii (Schauer) Benth.

Acacia 1Acacia dealbata Linka2004114·1 (13·0)79·3 (6·6)
Acacia pendula A. Cunn ex G. Dona200452·5 (9·5)80·2 (6·6)
AcerAcer tataricum subsp. ginnala (Maxim.) Wesm.a200498·9 (18·3)103·4 (7·4)
Acer truncatum Bungeaa2004128·6 (13·7)94·7 (5·7)
Broom 1Cytisus scoparius (L.) Linka200485·7 (11·7)69·5 (6·4)
Genista monspessulana (L.) L.A.S. Johnsonb200489·9 (16·1)65·3 (8·0)
Genista aetnensis (Raf. Ex Biv.) DC.b200485·3 (10·3)76·7 (4·9)
Eucalyptus 1Eucalyptus camaldulensis Dehnh.a2004281·4 (14·4)214·9 (9·2)
Eucalyptus leucoxylon F. Muell.a2004163·6 (13·4)142·6 (7·7)
Fabaceae 1Albizia julibrissin Durazz.b200479·7 (11·8)78·4 (5·2)
Ceratonia siliqua L.b200426·0 (6·0)38·7 (3·3)
Fabaceae 2Robinia pseudoacacia L.b2004176·1 (46·9)101·2 (15·6)
Sesbania punicea (Cav.) Benth.a2004102·9 (24·7)93·2 (14·4)
Cercis canadensis L.a2004107·5 (5·9)82·3 (3·4)
MoraceaeFicus carica L.b2004183·9 (13·0)154·8 (7·1)
Maclura pomifera (Raf.) C.K. Schneid.b2004126·9 (7·6)84·4 (5·4)
OleaceaeFraxinus velutina Torr.a2004120·4 (22·5)87·8 (10·4)
Syringa vulgaris L.a200458·3 (13·9)51·4 (7·3)
RosaceaeCotoneaster lacteus W.W. Sm.a2004133·3 (6·1)125·5 (3·1)
Photinia serratifolia (Desf.) Kalkmanb200485·9 (14·9)101·8 (5·8)
RubusRubus armeniacus Fockeb2004176·4 (17·8)160·4 (9·9)
Rubus idaeus L. a2004207·8 (28·9)186·6 (13·5)
Acacia 2Acacia cyclops A. Cunn. Ex G. Dona200656·1 (13·7)
Acacia melanoxylon R. Br.a2006101·2 (11·1)
Acacia pendula A. Cunn. Ex G. Dona200671·4 (10·3)
Acacia 3Acacia saligna (Labill.) H. L. Wendl.a200697·9 (12·4)73·0 (6·5)
Acacia cultriformis A. Cunn. Ex G. Dona200670·7 (16·8)74·6 (7·9)
ApocynaceaeNerium oleander L.b2006110·5 (9·4)109·9 (4·3)
Thevetia peruviana (Pers.) K. Schum.a200639·8 (8·9)42·6 (3·7)
Broom 2Genista monspessulana (L.) L.A.S. Johnsonb200695·9 (8·2)91·7 (3·0)
Genista tinctoria L.a2006108·2 (15·3)106·9 (6·2)
Broom 3Retama monosperma (L.) Boiss.a200663·8 (13·9)61·5 (6·9)
Genista aetnensis (Raf. Ex Biv.) DC.b200664·4 (11·3)61·2 (7·0)
Broom 4Spartium junceum L.a200684·5 (7·7)58·0 (4·3)
Ulex europaeus L.a200688·3 (9·7)86·4 (5·6)
Genista hispanica L.a200685·7 (6·5)63·9 (4·2)
BuddlejaBuddleja davidii Franch.a2006214·3 (13·9)216·1 (11·4)
Buddleja globosa Hopea2006258·7 (24·0)
Eucalyptus 2Eucalyptus camaldulensis Dehnh.a2006147·5 (18·1)126·2 (11·1)
Eucalyptus pulverulenta Simsa2006138·9 (14·5)118·7 (7·4)
Eucalyptus 3Eucalyptus cladocalyx F. Muell.a2006149·0 (13·2)
Eucalyptus conferruminata D. Carr & S. Carra,c2006101·9 (15·5)103·3 (8·2)
Eucalyptus nicholii Maiden & Blakelya2006111·6 (14·9)106·0 (7·8)
MorusMorus alba L.a2006148·7 (9·8)129·5 (4·4)
Morus rubra L.a2006115·4 (17·3)120·7 (5·1)
AnacardiaceaeSchinus molle L.a2007206·1 (17·6)167·4 (8·6)
Schinus terebinthifolia Raddia200793·4 (16·2)113·4 (6·8)
Searsia lancea (L. f.) F.A. Barkleya2007102·7 (22·5)113·8 (8·7)
BerberisBerberis thunbergii DC.a2007125·1 (19·5)108·5 (10·2)
Berberis koreana Palib.a2007125·5 (23·7)74·4 (1·3)
CaesalpiniaCaesalpinia gilliesii (Hook.) D. Dietr.a200780·3 (6·3)68·8 (5·3)
Caesalpinia cacalaco Humb. & Bonpl.a200742·6 (16·7)78·8 (4·1)
EriobotryaEriobotrya japonica (Thunb.) Lindl.a200798·5 (8·2)
Eriobotrya deflexa (Hemsl.) Nakaia200762·0 (8·5)
ErythrinaErythrina crista-galli L.a200788·1 (16·3)71·2 (7·2)
Erythrina coralloides DC.a200757·6 (17·7)55·3 (5·7)
Eucalyptus 4Eucalyptus globulus Labill.a2007199·9 (15·1)159·3 (6·2)
Eucalyptus pauciflora Sieber ex Spreng.a2007220·6 (50·3)196·5 (14·8)
Eucalyptus 5Eucalyptus conferruminata D. Carr & S. Carra,c2007217·9 (20·3)150·8 (8·0)
Eucalyptus baueriana Schauera2007336·7 (28·7)221·7 (18·5)
Eucalyptus macrocarpa Hook.a2007136·3 (16·3)138·5 (11·3)
LavandulaLavandula stoechas L.a2007159·3 (9·7)
Lavandula angustifolia Mill.a2007168·2 (13·6)149·5 (13·0)
LeptospermumLeptospermum laevigatum (Gaertn.) F. Muell.a200783·6 (28·3)105·3 (14·0)
Leptospermum lanigerum (Sol. ex Aiton) Sm.a200755·3 (45·8)135·4 (22·8)

Specifically, we asked the following questions:

  • 1 Do invasive species have higher seedling RGR, LAR, SLA, and other growth-related variables and/or smaller seed mass, than phylogenetically related non-invasive species?
  • 2 Are comparative seedling RGR studies useful as invasive species screening tools?

Materials and methods

Species selection and invasiveness status

We paired species into contrasts (PICs) that differ in invasiveness but are phylogenetically related and similar in growth habit, leaf habit (evergreen/deciduous), seed dispersal mode, flower colour, and aesthetics (Table 1). We consulted with horticultural and invasive species experts (W. Roberts, University of California-Davis Arboretum, and J. Randall, The Nature Conservancy, personal communications) to identify sets of similar species. The species used were limited to woody horticultural plants exotic to California (except Fraxinus velutina, native in parts of California) that have been commonly planted in California for at least 100 years (see Grotkopp & Rejmánek 2007; Appendix S1, Supporting Information).

We categorised species’ invasiveness as described in Grotkopp & Rejmánek (2007)sensuPyšek et al. (2004). For the purpose of our analyses, invasive woody species are those that are invasive in California or have been reported as clearly invasive in other states or regions of the world (Randall 2007; Appendix S2, Supporting information). The non-invasive (or much less-invasive) species are those that, despite widespread planting, have not been reported as invasive anywhere, or that have only limited, local spread.

Species selection was based on theoretical as well as practical considerations. For final selection, we used species with available seeds, either commercially or from local specimens. Several species were grown in multiple years as members of different contrasts (Table 1). In all cases, the species forming a PIC were grown together, at the same time. Therefore, each PIC is considered independent for data analysis purposes. In total, we planted 65 species for 29 sets of PICs to test hypotheses that invasive species have higher RGR and higher SLA than non-invasive species (Table 1).

Planting and harvesting

Mean seed mass was determined for most species by weighing 100 air-dried seeds together. Seeds were collected, weighed, prepared, planted (late February/early March each year), recorded, and harvested in a similar manner as those grown in 2004 (Grotkopp & Rejmánek 2007; Appendices S1, S2, Supporting information). Species within a contrast were planted in 164-ml SC-10 Super Cells (Stuewe & Sons, Corvallis, OR, USA) that were placed together in racks so that species within a contrast experienced the same greenhouse environment. The racks were moved to different assigned areas of two benches every five days using a random numbers table to account for potential environmental inconsistencies within the greenhouse. Plants were randomly harvested at their appropriate individual ages after emergence. The average number of plants per species harvested at 10 and 30 days was nine plants (range: 3–15 plants). For the 20 days harvest, the average number of plants harvested per species was 8·5 (range: 3–15 plants). In total, we harvested 1712 seedlings.

Growth conditions

All pots were allowed to dry out at the surface and then fully saturated with alternating distilled water and nutrient solution (Appendix S1, Supporting information). For 2004, mean minimum and maximum temperatures were 15·0 °C and 28·6 °C, respectively, while those for 2006 were 18·8 °C and 27·1 °C, respectively. For 2007, plants of different contrasts were grown at three different times due to stratification timing. All members of a contrast were grown at the same time. The minimum and maximum temperatures respectively for these three groups were 19·6 °C and 29·4 °C, 21·2 °C and 33·7 °C, and 17·9 °C and 35·6 °C.

Growth analysis

NAR, RGR, and their variances were calculated for the intervals 10–20 and 10–30 days after emergence according to the formulae from Causton & Venus (1981) for ungraded and unpaired harvests. The interval from 10–30 days was used for most analyses because we had complete data for all species. The contrasts grown in 2004 were analysed previously in a slightly different manner (Grotkopp & Rejmánek 2007) and are here re-analysed together with contrasts from all years. Contrary to Grotkopp & Rejmánek (2007), LMR, LAR, and SLA were quantified at the end of the growth interval, 30 days, rather than the midpoint of the growth interval, 20 days. This was due to insufficient germination and harvesting of 20 days plants for some species. For the Eriobotrya contrast, we did not include the mass of the cotyledons for growth analyses because they were large storage cotyledons that inconsistently fell from the plants.

Statistical analysis

Species’ means were used in all analyses. For cross-species analyses, growth data were compared by pooling all invasive species vs. all non-invasive species using one-tailed t-tests. We used two methods for analysing PICs. We used one-tailed paired t-tests, pairing species based on phylogeny using StatView 5·0·1 (SAS Institute Inc. 1998). If more than one species was used for half of a PIC, then the values for those species were averaged. We also used the program Comparative Analysis by Independent Contrasts (CAIC version 2·6·9; Purvis & Rambaut (1995); available at to construct the contrasts within a phylogeny (Appendix S3, Supporting information) and analyse the data. Branch lengths were set to be equal since we did not have data for a completely resolved phylogeny. This option has been shown to be the most reliable in such cases (Purvis & Rambaut 1995). We used the more conservative ‘Brunch’ option for all analyses.


Cross-species analyses

Without taking phylogeny into account, no significant differences were observed between all invasive species and all non-invasive species for RGR, NAR, LAR, SLA, LMR, log(seed mass), total biomass, or total leaf area, over either time interval (one-tailed t-tests, all P > 0·136). RGR was significantly positively correlated with its components, NAR (P < 0·001) and LAR (P < 0·01) (Table 2). Similarly, LAR was significantly positively correlated with its components, SLA (P < 0·0001) and LMR (P < 0·05) (Table 2). RGR, LAR, and SLA were all significantly negatively correlated with log(seed mass) (all P < 0·05) (Table 2).

Table 2.   Coefficients of determination (r) and corresponding one-tailed P values (in parentheses) of growth variables and log(seed mass) based on cross-species correlations and on phylogenetically controlled correlations for both the 10–20 and the 10–30 days intervals. P values < 0·05 are in bold
Dependent and independent variablesCross-species correlationsCAIC PIC correlations
10–20 days intervala10–30 days intervalb10–20 days intervalc10–30 days intervald
  1. an = 55 species.

  2. bn = 65 species.

  3. cn = 26 contrasts.

  4. dn = 27 contrasts.

RGR (mg·g−1·day−1):
NAR (mg·cm−2·day−1)+0·717 (<0·001)+0·411 (0·001) +0·632 (0·001)+0·424 (0·023)
LAR (cm2·g−1)+0·368 (0·006)+0·616 (<0·0001)+0·346 (0·074)+0·332 (0·068)
log(seed mass)−0·501 (<0·0001)−0·712 (<0·0001)−0·400 (0·036)−0·500 (0·004)
LAR (cm2·g−1)
SLA (cm2·g−1)+0·934 (<0·0001)+0·916 (<0·0001)+0·933 (<0·0001)+0·917 (<0·0001)
LMR (g·g−1)+0·302 (0·025)+0·514 (<0·0001)+0·374 (0·061)+0·424 (0·019)
log(seed mass)−0·354 (0·008)−0·580 (<0·0001)−0·624 (0·0005)−0·640 (0·0001)
NAR (mg·cm−2·day−1):
log(seed mass)−0·200 (0·143)−0·014 (0·909)+0·141 (0·473)+0·283 (0·114)
SLA (cm2·g−1):
log(seed mass)−0·321 (0·017)−0·487 (<0·0001)−0·707 (<0·0001)−0·663 (<0·0001)

Phylogenetically independent analyses

Results from the paired t-tests and the phylogenetic program CAIC agreed (Table 3). Contrary to our results with cross-species analyses, we found that invasive species had significantly higher seedling RGR than phylogenetically related non-invasive species for the 10–20 and 10–30 days growth intervals (< 0·01 and P < 0·05, respectively; Fig. 1 and Table 3). With PICs, invasive species grew faster than related non-invasive species by an average of 42·0% ± 11·9% (mean 23·8 ± 7·3 mg g−1 day−1; n = 25) for the 10–20 days interval and by 20·9% ± 7·7% (mean 10·6 ± 5·6 mg g−1 day−1; n = 29) for the 10–30 days interval. NAR, the physiological component of RGR, was significantly higher for invasive species during the shorter growth interval, 10–20 days. Beyond this early time frame, NAR no longer was significantly higher for invasive species (Table 3). Both LAR and SLA were marginally higher (0·05 ≤ ≤ 0·10) for invasive species over both growth intervals (Table 3).

Table 3.   One-tailed P values of plant and growth traits examined with PICs (paired t-tests and CAIC) between invasive and non-invasive species. Invasive species were hypothesised to have higher means than non-invasive species for all growth variables and smaller means for log(seed mass). RGR and NAR are from 10 to 20 days (n = 25) and 10 to 30 days (n = 29) after emergence, while data for LAR, SLA, LMR, total biomass, and total leaf area are shown for the end of the growth intervals, 20 and 30 days respectively. P values < 0·05 are in bold
VariablePaired t test PIC correlationsCAIC PIC correlations
10–20 days10–30 days10–20 days10–30 days
RGR (mg·g−1·day−1)0·0020·0350·0020·044
NAR (mg·cm−2·day−1)0·0280·1940·0360·206
LAR (cm2·g−1)0·0960·0530·1040·100
SLA (cm2·g−1)0·0730·0660·0930·075
LMR (g·g−1)0·1650·1350·2550·197
log(seed mass)0·2610·2910·3020·328
Total biomass (g)0·4400·3770·4250·394
Total leaf area (cm2)0·4430·3450·4640·362
Figure 1.

 RGR of invasive species vs. non-invasive species for the (a) 10–20 days interval and (b) 10–30 days interval. Each point represents a contrast. When more than one species represent part of a contrast, the average RGR is used. Points above the line represent contrasts with invasive species having higher RGRs than their non-invasive counterparts.

No significant differences were found between invasive and non-invasive species using PICs for LMR, total biomass, or total leaf area at either 20 or 30 days after seedling emergence (Table 3). No difference between invasive and non-invasive species was observed for log(seed mass) (Table 3).

For phylogenetically corrected analyses with only continuous traits we found that RGR was significantly positively correlated with NAR (P < 0·05) and marginally positively correlated with LAR (P < 0·10) (Table 2). LAR was strongly positively correlated with SLA for both time intervals (P < 0·0001). LAR and LMR were marginally positively correlated for the 10–20 days interval (P < 0·10), but were significantly correlated for the 10–30 days interval (P < 0·05) (Table 2). RGR, LAR, and SLA were each significantly negatively correlated to log(seed mass) (P < 0·05) (Table 2).


Invasive species have higher seedling RGRs

Using PICs, RGR emerged as the most significant growth rate trait separating invasive and related non-invasive species (Fig. 1 and Table 3). RGRs from both intervals were strongly correlated (P < 0·0001), with RGR from 10 to 20 days about 11·2% greater than RGR from 10 to 30 days across all species. The 10–20 days growth interval was likely more significant because it is closer to the maximum growth rate exhibited by woody angiosperm seedlings (Hunt 1982). This probably explains the greater standard errors associated with the 10–20 days interval (Table 1) since individual plants may be just below, at, or slightly past the maximum RGR. NAR was also significantly higher for invasive species, but only for the 10–20 days interval (Table 3). This is probably because for short time periods, environmental fluctuations caused by varying light, temperature, and moisture are most quickly acclimated to by physiological traits such as NAR. Over longer time periods, it is the more stable and species-specific morphological traits such as LAR and SLA that influence species’ growth rates (Villar et al. 2005).

Overall, these results confirm previous conclusions for a smaller subset of data (Grotkopp & Rejmánek 2007) and a recent meta-analysis (van Kleunen, Weber & Fischer 2010) that RGR is significantly correlated with invasiveness of woody angiosperms when compared within PICs. With an increase in sample size and more extensive coverage of the angiosperm phylogeny, we found a more highly significant difference in RGR between invasive and related non-invasive species in this study than in a previous study (Grotkopp & Rejmánek 2007) for the 10–20 days interval (P < 0·01, n = 25 and P < 0·05, n = 10; respectively). The main difference is that higher SLA appeared to drive the higher RGR of invasive species in the smaller studies of angiosperms (P < 0·05, n = 10 in Grotkopp & Rejmánek 2007), pines (Grotkopp, Rejmánek & Rost 2002), and woody species (reviewed by Cornelissen, Castro-Díez & Carnelli 1998). In this expanded study, SLA was found to be only marginally significant (Table 3). We expected, with an increased sample size, that SLA would have been much more significant if it truly was the main driver of differences in RGR. For this larger pool of contrasts, we covered a wider spectrum of invasive/non-invasive PICs from a substantial proportion of the angiosperm phylogeny (Appendix S3, Supporting information) and therefore revealed that a larger variety of pathways determine growth rate patterns in different species/clades.

RGR and correlated traits leading to invasiveness

Seedling RGR is the strongest variable linked to invasiveness across the many contrasts we studied (Fig. 1 and Table 3). The ability to opportunistically capture resources, often through high RGR and SLA, is important in many plant invasions (Davis, Grime & Thompson 2000; Blumenthal 2006). The high RGR of invasive species is advantageous initially when seedlings experience little or no competition and/or herbivory (Herms & Mattson 1992). Even though seedlings from large-seeded species are initially larger after germination, smaller-seeded species with higher RGRs can soon grow taller (Marañon & Grubb 1993; Chacón & Muñoz 2007). In this way, high-RGR species can compete more successfully for light and space. The relationship between RGR and competitiveness is undoubtedly highly context- and stage-dependent (Chacón & Muñoz 2007; Turnbull et al. 2008). However, even if high RGR is not necessarily advantageous in all scenarios, high seedling RGR is correlated with other attributes that are causally responsible for invasiveness in disturbed, resource-rich environments.

While LAR, SLA (via LAR), and NAR contribute to the strength of RGR’s ability to predict invasiveness, these factors also contribute to other traits important for invasiveness. We found significant negative correlations of RGR, LAR, and SLA with log(seed mass) using both cross-species and PIC analyses (Table 2). Because of this strong association with seedling RGR, we expected invasive species to have significantly smaller seeds than non-invasive species, as was found with pines (Grotkopp, Rejmánek & Rost 2002). We did not find a significant difference in seed mass between invasive and non-invasive species, either across species or when corrected for phylogeny. Although small seed mass is not significantly associated with invasiveness in woody angiosperms, it is correlated with other traits that are important for invasiveness. Species with smaller seeds tend to disperse further than larger seeded species (excluding those with vertebrate dispersers) and produce a significantly greater number of seeds on an area basis (Rejmánek & Richardson 1996; Moles & Westoby 2006). Smaller seeds are also associated with higher SLA and higher seedling RGR (Table 2) (Marañon & Grubb 1993; Grotkopp, Rejmánek & Rost 2002).

Minimum generation time also has a key role in invasiveness and is positively correlated with seed mass, and negatively with RGR and SLA (Brzeziecki & Kienast 1994; Rejmánek & Richardson 1996; Grotkopp, Rejmánek & Rost 2002). Species with smaller seeds have both shorter minimum and average generation times, while species with larger seeds reproduce later (Moles & Westoby 2006). Shorter minimum generation time is strongly associated with invasiveness in woody species (Rejmánek & Richardson 1996). Ideally, we would have included minimum generation time in our study, but such data were unavailable for most of our species.

Is seedling RGR suitable for invasive species risk assessment?

Comparative seedling RGR studies have potential as screening tools of new exotic plant species. Using seedling growth analysis as an invasive species’ screening tool requires growing a species proposed for introduction with related invasive and non-invasive species under standardized greenhouse conditions. If the tested species’ RGR is higher or not significantly different from its known invasive counterpart, it should be considered highly likely to become invasive, and excluded from further consideration as a potential horticultural species. If it grows significantly slower than its related invasive species and/or not significantly faster than its non-invasive relative, then the potential new horticultural species should be considered for further testing. The use of comparative growth analysis for invasive screening purposes will not be ideal for all new potential horticultural species because related species of known invasiveness are sometimes difficult to find, but there are many species that will meet such conditions, as well as new cultivars of known invasive species.

While our results show that high RGR is a major biological trait associated with invasiveness, only two-thirds of the invasive species within our PICs had higher RGR than their non-invasive counterparts (17/25 and 19/29 for the 10–20 and 10–30 days intervals, respectively; Fig. 1 and Table 1). The question arises, why do some species classified as non-invasive have higher RGR than their invasive counterparts? One hypothesis is that they could be still in a lag-time phase that follows their introduction (Crooks & Soulé 1996). We believe this is plausible and that these presumptive non-invasive species should be treated with caution. One such case could be our ‘non-invasive’, but fast growing Buddleja globosa that recently has been reported as naturalising in Australia and New Zealand ( On the other hand, one renowned invasive species, Acacia cyclops, has an extremely low RGR compared to its non-invasive counterparts (Table 1). Clearly, Acacia cyclops has other characteristics that make it extremely invasive, and it should not be used for screening procedures. This contrast also reveals that comparative seedling RGR is not a foolproof screening tool for invasiveness on its own. Further testing of known invasive and non-invasive woody angiosperm species comparisons that include seedling RGR with other easily attainable traits, such as climatic and geographic data, may produce a straightforward tool for screening potential horticultural introductions.


Taking a multivariate problem, such as the invasive potential of a plant, and whittling it down to a univariate trait, is a difficult, if not impossible task. We conclude that, overall, RGR is a strong single biological predictor of invasiveness in woody species using PICs. It is unfortunate that some more easily measurable and less labour- and time-intensive attribute, such as SLA, did not emerge as the best trait for separating invasive species from their non-invasive counterparts. This probably reflects the fact that different species or clades have different ways of increasing RGR, whether by increasing NAR or by increasing allocation to leaf area (SLA) or overall leaf mass (LAR). RGR is a good candidate for inclusion in screening procedures because it encompasses many traits related to morphology, physiology, and ecology, but it should not be used on its own for risk assessment purposes.

The use of seedling RGR studies as invasive screening tools still needs to be fine-tuned including standardising greenhouse conditions (including resource levels) and determining the appropriate growth intervals for varying growth forms of species. From our study, it appears that the 10–20 days interval is better suited to woody angiosperms than the 10–30 days interval, although the 10–20 days interval tends to have a higher SE because of its proximity to maximum RGR. Minimally, related species should be within family (Grotkopp & Rejmánek 2007), but for large families such as Fabaceae, they should be within closely related genera. If combined with easily obtainable climatic and geographic information, seedling RGR could produce a useful, straightforward screening tool when phylogenetically related species or cultivars are available.

While the initial escape and naturalisation of a species may occur following disturbance or other increases in nutrients/resources, the survival and further spread of a species in the wild will more likely occur under less than ideal conditions, such as drought. To gain a better understanding of the underlying mechanisms that influence both RGR and the invasive potential of a species, further basic research should include examining how invasive and non-invasive species respond to stresses and increases in resources.


Thanks to P. Riley, M. Bower, T. Metcalfe, and the many undergraduate assistants for their help in the greenhouse, and to R. Drenovsky and four reviewers for useful comments. This research was funded by USDA-CSREES UC-IPM EPDRP Projects #01XN020 and #05XN027.