Chapitre 3 : Influence des propriétés initiales du sol sur l’impact des plantes envahissantes

Chapitre 3 : Influence des propriétés initiales du sol sur l’impact des plantes envahissantes

Dans le premier chapitre, nous avons identifié des patrons similaires dans l’impact de cinq plantes envahissantes à écologie et forme de vie contrastées : augmentation de la disponibilité de nombreux éléments dans le sol sous les envahissantes par rapport aux sols non envahis.

Dans le deuxième chapitre, nous avons nuancé ces conclusions en mettant en évidence que l’impact de Fallopia japonica dépendait du site et plus particulièrement des conditions de fertilité initiales de celui-ci avant l’invasion. En effet, dans les sites où la concentration en éléments minéraux était la plus élevée avant invasion, F.

japonica avait tendance à réduire la disponibilité des éléments.

Dans ce troisième chapitre, nous avons considérablement élargi l’échantillonnage en ajoutant deux espèces (Impatiens glandulifera et Senecio inaequidens) et surtout en augmentant le nombre de sites étudiés par espèce (4 à 6 sites par espèce). Cet échantillonnage important nous a permis de tester si le patron d’impact identifié chez F. japonica pouvait être appliqué à l'ensemble des plantes exotiques envahissantes étudiées. Il nous a également permis de déterminer l’influence relative de l’espèce et du site sur la direction et l’intensité de l’impact de l’invasion sur les propriétés du sol, la productivité et le stock d’éléments minéraux dans la phytocénose.

Ces résultats ont été soumis à Oecologia et sont actuellement en révision (accepté

sous réserve de modifications).

49 Dassonville N. ¹ , Vanderhoeven S. ² , Vanparys V. ³ , Hayez M., Gruber W. ¹ , Meerts P. ¹

Predictable site-specific impacts of alien invasive plants on soil nutrients in Western Europe

1 Laboratoire de génétique et écologie végétales. Université Libre de Bruxelles.

2 Laboratoire d'écologie. Faculté universitaire des sciences agronomiques de Gembloux.

3 Unité d'écologie et de biogéographie. Université catholique de Louvain.

Abstract

Alien invasive plants are capable of modifying ecosystem function. However, generalisations are difficult because impacts often appear to be species- and site- specific. In this study, we examine impacts on nutrient pools in the topsoil and the standing biomass of 7 highly invasive plant species in NW Europe (Fallopia japonica, Heracleum mantegazzianum, Impatiens glandulifera, Prunus serotina, Rosa rugosa, Senecio inaequidens, Solidago gigantea). We test if impacts follow predictable patterns, across species and sites or, alternatively, if they are entirely idiosyncratic.

To that end, we compare invaded and adjacent uninvaded plots in a total of 36 sites with widely divergent soil chemistry and vegetation composition. For all species, invaded plots had increased aboveground biomass and nutrient stocks in standing biomass compared to uninvaded vegetation. This suggests that enhanced nutrient use may be a key trait of highly invasive plant species. There was no significant difference among species in their impacts on topsoil. In contrast, impacts magnitude and direction were strongly site-specific. A striking finding is that the direction of change in soil properties followed a predictable pattern. Thus, strong positive impacts (i.e. increased nutrient concentrations) were most often found in sites with low nutrient concentrations in the topsoil while negative impacts were generally found in the opposite conditions. This pattern was significant for K, Mg, P, Mn and N. The particular site-specific pattern in the impacts provides the first evidence that alien invasive species contribute to homogenization of soil conditions in invaded landscapes.

Key words

Homogenization, mineral nutrients, nutrient cycling, species effects, topsoil, biological

invasions.

Introduction

During the last decade, the impacts of alien invasive plants on ecosystems have received growing attention. These species alter ecosystems because they differ from natives in growth and allocation patterns (Wilsey and Polley 2006). Many alien invasive species have been shown to affect plant and animal communities (Braithwaite et al. 1989; Alvarez and Cushman 2002), ecosystem functioning (D'Antonio 1992; Belnap and Philips 2001), soil properties and nutrient fluxes (Asner and Beatty 1996, Blank and Young 2001) in the sites they invade. Not surprisingly, the documented impacts on soil properties are diverse. While most published studies report increased soil nutrient availability under invasive plant species (Musil 1993;

Scott et al. 2001; Duda et al. 2003; Vanderhoeven et al. 2005; Chapuis-Lardy et al.

2006), some studies show the opposite pattern (Christian and Wilson 1999; Leary et al. 2006). In addition, the same species may have different impacts depending on local conditions (Belnap and Philips 2001; Scott et al. 2001). Most studies investigating impacts of alien invasive plants consider only a single species in one or very few sites and with tens of different methods, thus making generalisation difficult.

Based on a literature review, Ehrenfeld (2003) stated that soil C and nutrients pools were often modified by invasions and that the direction and magnitude of the impacts were probably determined by the composition of the invaded community and soil properties. One of the aims of this article is to examine the relationship, if any, between initial soil conditions and direction and magnitude of impacts, based on a large sample of species and sites and with a common methodology.

In Europe, many alien invasive plant species have dramatically increased their range during the last decades (Muller 2004; Williamson et al. 2005; Pyšek and Prach 1995;

Pyšek 1991). It is important to test if invaded ecosystems are modified in a systematic and predictable way. This might be the case because invasive species often share common functional traits which promote invasiveness (Grotkopp et al.

2002). Invaded ecosystems also share common attributes making them susceptible to invasion (Alpert et al. 2000; Davis et al. 2000) and which might result in similar impacts.

In this study, we examine impacts on soil nutrients of seven highly invasive plant species in Belgium, in a total of 36 sites. In addition to nutrients in soil, we examine nutrient pools in aboveground biomass.

The specific questions addressed are:

- Are there general patterns in the impacts of these highly successful invasive

species on soil nutrients?

51 - Alternatively, do impacts vary depending on site conditions? If so, do impacts follow some predictable pattern or are they completely idiosyncratic?

- Are impacts on soil nutrients explained by differences in nutrient use between native and alien species?

Materials and methods

Species selection

Seven species have been selected: Impatiens glandulifera (annual), Heracleum mantegazzianum (hemicryptophyte), Senecio inaequidens (chamaephyte), Solidago gigantea, Fallopia japonica (perennial rhizomatous geophytes), Rosa rugosa (woody shrub) and Prunus serotina (tree). All these species can form monospecific stands in invaded communities and are recognized among the most highly invasive plant species in Belgium and elsewhere in Europe (Pyšek 1991; Pyšek and Prach 1995;

Verloove 2002; Verloove 2006; Muller 2004).

Site selection

For each species, five or six sites were selected in ecosystems with contrasting resident vegetation structure and composition, in an attempt to sample the range of habitats colonised by the study species in Belgium. In all selected sites, vegetation structure is profoundly affected by invasion, with alien species generally forming pure stands. The sites fulfilled the following conditions: 1) having well-established, and still expanding populations of the target species surrounded by native uninvaded vegetation, 2) having sufficiently homogeneous soil. Site selection aimed at minimizing the probability of differences existing prior to the invasion event. To that end, invaded and control uninvaded plots within a site were in the same topographic situation and had the same soil texture (See Vanderhoeven et al. 2005 for methods).

Moreover, the uninvaded control plots were located as close as possible to the front of expansion of the invader. Therefore systematic differences between invaded and uninvaded soils, if any, can reasonably be ascribed to the contrasted plant cover.

The sample comprises sites with widely divergent plant cover (from open grassland

in sand dunes to alluvial eutrophic forest), in line with the contrasted ecological

niches of the seven alien invasive species. In most sites, invasion typically dates

back to 2 or 3 decades.

Soil sampling

At each site, six 1-m² plots were located in invaded patches and six 1-m² plots were located in adjacent, uninvaded vegetation. Soil was sampled from February to April 2004. In each plot, five soil cores (0-10 cm, litter discarded) were collected with a soil borer (4 cm in diameter, one core at each corner of the square and one core at the centre of the square). These five cores were mixed up into a single bulk sample for each plot. Soil samples were air-dried and sieved (< 2 mm).

Soil analysis

The following parameters were assessed on each sample: Soil pH, exchangeable Mg, K, Zn, Mn and P (1M CH 3 COONH 4 pH 4.65 extraction and ICP-OES determination). Exchangeable Ca was not determined because a substantial proportion of soil samples contained free CaCO 3 . Organic C and N content were also assessed with a dry combustion C-N analyser (NC-2100, Carlo Erba Instruments) and C/N was then computed. For further methodological details, see Vanderhoeven et al. (2005).

Nutrients in standing biomass

Aboveground biomass was harvested in invaded and control plots at the peak of biomass (between June and August) for 4 of the 7 species (H. mantegazzianum, Prunus serotina and Rosa rugosa were not sampled for practical reasons) on the same plots as for soil sampling. Standing dead stems were discarded. Thus, the harvested biomass can be used as a proxy for the annual production. Species were not sorted before weighing. However, in the invaded plots, the contribution of native species was always negligible. The plant samples were dried at 70°C to constant weight. A representative subsample was finely ground (0.12 mm) and analysed for Mg, K, P, Mn and Zn concentration (550°C calcinatio n, HCl dissolution of the ashes and ICP-OES determination). C and N content were assessed using a dry combustion C-N elemental analyser (NC-2100, Carlo Erba Instruments). The aboveground nutrient stock (mg m ^-2 ) was then calculated as the product of mineral nutrient concentration and biomass. For further technical details, see Vanderhoeven et al. (2005).

Statistical analyses

For soil and plant parameters, a three-way nested ANOVA was applied on all sites

and species pooled, with species (fixed factor), site (random factor, nested in

species) and invasion (fixed factor) as main effects. In this analysis, significant

53 species X invasion and site (species) X invasion interactions will point to differences in the impacts of invasion among species and sites, respectively.

Log transformation was applied as necessary to meet assumptions of ANOVA.

Relationships between magnitude of impacts and initial site conditions were examined in two different ways. First, impact magnitude expressed as the difference between invaded stands (I) and uninvaded stands (U), was plotted against values in uninvaded stands (U). Correlation between impact magnitude and values in uninvaded stands were calculated. When necessary, the correlation was computed on log transformed data (Mg, P, Mn, Zn) and the magnitude of impact was calculated as difference between log(invaded value) and log(uninvaded value). Secondly, a regression analysis was conducted. To avoid autocorrelation, impacts (I-U) were not regressed directly against values in uninvaded stands (U). Instead, values in invaded stands (I) were regressed on values in uninvaded stands (U). Under the null hypothesis of no impact, all data points align on the line I = U. Deviation from this line, in the form of I= b + aU, indicates departure from the null hypothesis. Intercept (b) values >1 and <1 indicate positive and negative impacts, respectively. Slope (a) values >1 and <1 indicate that impacts increase and decrease with initial conditions, respectively. Conformity tests of intercept and slope were performed. Regression analyses were performed on log-transformed values as necessary to meet normality assumptions. Analyses were performed with Statistica 7.1 software (StatSoft Inc.

2005).

Results

The seven species considered strongly modified plant community structure with formation of (quasi)monospecific stands. Sampled sites spanned a very broad range of vegetation and soil types (Table 1).

Table 1: Sites localization and characterization: vegetation and soil type (FAO, ISRIC, IUSS, 1998)

Species Localization Invaded vegetation Soil type

H. mantegazzianum Crombach Wet grassland with Phalaris arundinacea fluvisol Ganshoren Wet wasteland with Urtica dioica and Petasites

hybridus

gleysol Koersel Dry grassland with Festuca ovina arenosol

Boitsfort Willow scrub gleysol

Boitsfort Willow scrub gleysol

Table 1 (continuation)

Species Localization Invaded vegetation Soil type

F. japonica Boitsfort Forest pond bank with Petasites hybridus and Chrysosplenium oppositifolium

gleysol Gembloux Cultivated field margin with Urtica dioica and Rubus

sp.

anthrosol Haren Oldfield with Dactylis glomerata, Urtica dioica and

Cirsium arvense

anthrosol

Boitsfort Beech forest luvisol

Saint- Ghislain

Mesic rough grassland with Calamagrostis epigejos, Eupathorium cannabinum, Achillea millefolium…

luvisol Boitsfort Wasteland with Urtica dioica anthrosol S. gigantea Saint-

Ghislain

Mesic rough grassland with Calamagrostis epigejos, Arrhenatherum elatius…

luvisol Boitsfort Wasteland with Urtica dioica anthrosol Kraainem Rough grassland with Agrostis sp., Holcus lanatus,

Leucanthemum vulgare, Senecio jacobea...

gleysol Woluwé Rough grassland with Dactylis glomerata and

Arrhenatherum elatius

luvisol Marche-les-

dames

Wasteland with Tanacetum vulgare, Rubus sp., Origanum vulgare

luvisol S. inaequidens Koersel Dry grassland with Festuca ovina arenosol

Harchies Open dry vegetation with Hypericum perforatum and Fragaria vesca

arenosol Antwerpen Wasteland with Senecio jacobea and Melilotus albus Cambisol De Panne Coastal sand dune with Ammophila arenaria arenosol Antwerpen Open dry vegetation with Sedum acris arenosol I. glandulifera Uccle Small valley in a beech forest gleysol

Temse Phragmites australis gleysol

Kessel Riverbank with Phalaris arundinacea and Urtica dioica

gleysol Muizen Riverbank with Phalaris arundinacea and Urtica

dioica

gleysol Ways Poplar plantation clearcut with Galeopsis tetrahit and

Urtica dioica

luvisol

K

-150 -100 -50 0 50 100 150

0 100 200 300 400

In v a s io n i m p a c t (I -U )

r = -0.3385*

pH

-1,5 -1,0 -0,5 0,0 0,5 1,0

2 4 6 8

In v a s io n i m p a c t (I -U )

r = -0.1655

N

-0,6 -0,5 -0,4 -0,3 -0,2 -0,1 0,0 0,1 0,2 0,3 0,4

0,0 0,2 0,4 0,6 0,8

Uninvaded soil properties (U)

In v a s io n i m p a c t (I -U )

r = -0.6515***

C org

-6 -4 -2 0 2 4 6

0 5 10 15

In v a s io n i m p a c t (I -U )

r = -0.5818*

P

-0,6 -0,4 -0,2 0,0 0,2 0,4 0,6 0,8

0 0,5 1 1,5 2 2,5

In v a s io n i m p a c t (L o g I - L o g U )

r = -0.6781**

Mn

-0,3 -0,2 -0,1 0,0 0,1 0,2 0,3 0,4

0,0 0,5 1,0 1,5 2,0 2,5

In v a s io n i m p a c t (L o g I - L o g U )

r = -0.5690***

Zn

-0,4 -0,3 -0,2 -0,1 0,0 0,1 0,2 0,3 0,4 0,5

-0,5 0,5 1,5 2,5

Uninvaded soil properties (log U)

In v a s io n i m p a c t (L o g I - L o g U )

r = -0.2249 Mg

-0,3 -0,2 -0,1 0 0,1 0,2 0,3 0,4

1 1,5 2 2,5 3

In v a s io n i m p a c t (L o g I - L o g U )

r = -0.4598**

Fig. 1 Impact of invasive species as a function of soil chemical parameters in uninvaded plots (U).

Impact is calculated as the difference between invaded and uninvaded plots, I-U. Positive values of I-U indicate increased element concentration in the invaded soil. Pearson correlation coefficient (r) between U and I-U. Significance levels (*: p<0.05, **: p<0.01, ***: p<0.001). Values are in mg kg

except for C and N (%). Log-transformed data for Mg, P, Mn, Zn. ♦ S. gigantea, ■ F. japonica, ▲ H.

mantegazzianum , X I. glandulifera, ○ R. rugosa, + S. inaequidens, □ P. serotina.

Table 1 (continuation)

Species Localization Invaded vegetation Soil type

P. serotina Uccle Rough grassland with Dactylis glomerata and Arrhenatherum elatius

luvisol La Hulpe Young forest with Betula pendula and Sorbus

aucuparia

podzol Louvain-La-

Neuve

Beech forest luvisol

Kessel Mixed forest with Quercus robur, Acer pseudoplatanus, Betula pendula…

luvisol Westerlo Pine plantation with Betula pendula and Sorbus

aucuparia

podzol R. rugosa Oostduinkerke Coastal sand dune with Hippophae rhamnoides and

Ammophila arenaria

arenosol Nieuwpoort Coastal sand dune with Hippophae rhamnoides and

Ammophila arenaria

arenosol Bredene Coastal sand dune with Hippophae rhamnoides and

Ammophila arenaria

arenosol Wenduine Coastal sand dune with Hippophae rhamnoides and

Ammophila arenaria

arenosol Zeebrugge Wet depression in a coastal sand dune with

Phragmites australis, Arhhenatherum elatius…

arenosol

Soil

The 36 sites spanned a very broad range of soil conditions (Fig. 1) (45-fold variation for K, 40-fold for P, 248-fold for Zn, 64-fold for organic C, 44-fold for Mn, 35-fold for Mg, 32-fold for N). The results of the three-way ANOVA are shown in Table 2.

Species effects

The species factor was significant for all elements except P, Zn (Table 2). This indicates that the seven species tend to occupy distinct soil niches. The most striking species differences were observed for pH, C and N. Thus, sites with Prunus serotina and Rosa rugosa occupied the lowest and the highest end of the pH range, respectively (Fig. 1). R. rugosa was restricted to sites with the lowermost C and N content in our sample (maritime sand dunes).

Site effect

The site (within species) effect was highly significant for all parameters (Table 2).

This underscores the broad soil niche occupied by most species in our sample. 10-

fold variation in nutrient concentrations across sites within species was not rare. For

instance, 51-fold variation was observed for K in sites invaded by Prunus serotina,

56 40-fold variation for P in sites invaded by Solidago gigantea, 19-fold variation for Mg in sites invaded by Impatiens glandulifera.

Invasion and species X invasion effect

The invasion effect and the interaction invasion X species were mostly not significant (Table 2). K stands out as an exception, with a significant invasion effect. In 26 of the 36 sites, soil showed increased K availability under invasive plants compared to uninvaded plots (Fig. 1).

Site X invasion interaction

The site X invasion interaction was highly significant for all elements (Table 2). This indicates that the impact of invasive plants on soil chemistry varied depending on local site conditions. Actually, for a single species, impacts were either positive (i.e.

increased values in invaded plots), negative (decreased values) or not significant. For instance, K increased (+ 135 mg kg ^-1 ) in one site invaded by Heracleum mantegazzianum while it decreased (-86 mg kg ^-1 ) in another site invaded by the same species (Fig. 1). The same was true for Mg in sites invaded by F. japonica (+128 mg kg ^-1 in one site vs. -85 mg kg ^-1 in another), for P in sites invaded by S.

gigantea (+13 mg kg ^-1 vs. -48 mg kg ^-1 ) or for Mn in sites invaded by I. glandulifera (+10 mg kg ^-1 vs. -54 mg kg ^-1 ) (Fig. 1).

Table 2: Three way ANOVA for soil parameters: F values and significance level (*:

p<0.05, **: p<0.01, ***: p<0.001). Species and invasion are fixed factors and site within species is a random factor.

d.f. pH K Mg P Mn Zn Corg N C/N

Species 6 8.2*** 6.2*** 2.9*** 0.9 3.1* 2.2 3.6** 3.3* 3.7**

Site (species) 29 38.3*** 13.0*** 17.8*** 11.6*** 12.1*** 80.8*** 17.0*** 30.6*** 19.3***

Invasion 1 2.6 10.5** 4.7* 1.5 3.4 0.3 0.4 0.5 1.3

Species x Invasion 6 1.1 0.7 1.2 0.7 1.8 0.2 1.1 1.9 1.1

Site (sp) x Invasion 29 2.9*** 4.6*** 4.8*** 5.7*** 2.8*** 2.9*** 2.4*** 3.3*** 3.9***

Interestingly, the magnitude of impacts, expressed as the difference between values in invaded stands (I) and uninvaded stands (U), shows a remarkable pattern (Fig. 1).

Thus, positive impacts (i.e. increased values) were most often observed in sites with low values of initial conditions. On the contrary, sites with relatively high concentration values in the soil prior to invasion tend to show negative impacts, i.e.

decreased concentrations. Correlation between impact magnitude and initial

conditions was negative for all parameters (Fig. 1).

This pattern was confirmed by regression analysis. In this analysis, values in invaded stands (I) were regressed against values in uninvaded stands (U). The equation of the regression line was calculated (I= b + aU). Conformity tests of the slope (a) and intercept (b) were performed. There was a highly significant, positive correlation between I and U, for all parameters (Table 3). The slope of the regression line was systematically <1 and intercepts were systematically >0. This was significant for all variables except zinc, carbon and pH. Thus, impacts varied with initial conditions. In sites with low nutrient concentrations in the soil, invasion tended to increase nutrient concentrations. In sites with high nutrient concentrations, the opposite pattern i.e.

decreased concentrations in invaded stands, was most often observed.

Table 3: Conformity tests for intercept (H 0 : b=0) and slope (H 0 : a=1) of the linear regression line between values in invaded (I) and uninvaded (U) plots for soil parameters (I= b + aU). Correlation coefficients (r), t-values and significance levels (NS: not significant, *: p<0.05, **: p<0.01, ***: p<0.001). Degree of freedom = 34.

r a ± SE t b ± SE t

pH 0.98*** 0.11 ± 0.21 0.5 NS 0.96 ± 0.036 0.9 NS

K 0.82*** 41 ± 14 2.9** 0.80 ± 0.09 2.1*

C 0.89*** 0.86 ± 0.51 1.7 NS 0.82 ± 0.07 2.4*

N 0.69*** 0.12 ± 0.04 5.2 *** 0.53 ± 0.10 4.8***

Log Mg 0.93*** 0.42 ± 0.12 3.3** 0.82 ± 0.06 3.0**

Log P 0.86*** 0.32 ± 0.09 3.6** 0.74 ± 0.08 3.3**

Log Mn 0.90*** 0.37 ± 0.08 4.5 *** 0.75 ± 0.06 4.0***

Log Zn 0.96*** 0.079 ± 0.048 1.6 NS 0.94 ± 0.04 1.3 NS Biomass

There was a 77-fold variation in aboveground biomass in uninvaded plots, in line with the large variation in vegetation structure in our sample. S. inaequidens sites generally had very low biomass in the resident vegetation (80 g m ^-2 on average). In all sites, biomass was always higher in invaded compared to uninvaded vegetation (on average, F. japonica: 2.95 vs. 0.56 kg m ^-2 , S. gigantea: 0.99 vs. 0.47 kg m ^-2 , I.

glandulifera: 1.0 vs. 0.68 kg m ^-2 , S. inaequidens: 0.24 vs. 0.08 kg m ^-2 ). The biomass increase was highly significant for all species (tests not shown).

Nutrient concentrations and stocks in plants

The 4 species had widely different nutrient concentrations in shoots, with I.

glandulifera and F. japonica having the highest and the lowest nutrient

concentrations, respectively. Compared to the uninvaded vegetation, F. japonica had

Mn

-100 -50 0 50 100 150 200 250

0 50 100

K

0 10000 20000 30000 40000

0 5000 10000 15000 20000 25000

In v a s io n i m p a c t (I -U ) Mg

0 1000 2000 3000 4000

0 1000 2000

P

-1000 0 1000 2000 3000

0 1000 2000 3000

In v a s io n i m p a c t (I -U )

Zn

-40 -20 0 20 40 60 80 100

0 50 100 150 200 250

In v a s io n i m p a c t (I -U ) C

-200 0 200 400 600 800 1000 1200 1400 1600 1800

0 100 200 300 400 500

Uninvaded vegetation nutrient stock

N

-20 -15 -10 -5 0 5 10 15 20 25 30 35

0 10 20 30 40

Uninvaded vegetation nutrient stock

In v a s io n i m p a c t (I -U )

Fig. 2 Nutrient stocks in aboveground biomass in uninvaded plots (U) and difference between invaded

and uninvaded plots (I-U). Values are in mg m

for K, Mg, P, Mn, Zn and g m

for C and N. The line

denotes equal value for invaded and control plots. Value above the line indicates higher value in

invaded plot. ♦ S. gigantea, ■ F. japonica, X I. glandulifera, + S. inaequidens

significantly lower tissue concentrations of all nutrients (K: 12200 vs. 20700 mg kg ^-1 , Mg: 1150 vs. 1880 mg kg ^-1 , P: 1280 vs. 2710 mg kg ^-1 , Mn: 31 vs. 35 mg kg ^-1 , Zn:

29.1 vs. 45.5 mg kg ^-1 , N: 1.10 vs. 1.58 %). Similarly, S. gigantea had significantly lower tissue concentration of most nutrients (Mg: 1380 vs. 1660 mg kg ^-1 , P: 1820 vs.

2480 mg kg ^-1 , Mn: 41 vs. 54 mg kg ^-1 , N: 0.87 vs. 1.16 mg kg ^-1 ). I. glandulifera had significantly higher values for K (31100 vs. 17000 mg kg ^-1 ), Mg (2320 vs. 1390 mg kg ^-

1 ) and P (2810 vs. 2210 mg kg ^-1 ) but significantly lower concentration of Zn (53.3 vs.

87.4 mg kg ^-1 ) and N (1.32 vs. 1.66 %). S. inaequidens had a significantly higher concentration of K (11500 vs. 8230 mg kg ^-1 ) and Zn (50.0 vs. 35.8 mg kg ^-1 ) but a significantly lower concentration of P (1310 vs. 1770 mg kg ^-1 ) than the uninvaded vegetation.

Despite contrasting patterns for nutrient concentrations, nutrient stocks were consistently higher in invaded compared to uninvaded plots, for nearly all elements in all species with only few exceptions (Fig. 2). In the three-way ANOVA, the invasion effect was significant for all elements (results not shown).

Discussion

AIS increase biomass and nutrient stocks in standing biomass

Without exceptions, all studied species increase aboveground biomass and nutrient stocks in standing biomass. Higher net primary productivity and faster growth rates have often been reported in alien invasive species compared to native vegetation (Wilsey and Polley 2006; Ehrenfeld 2003; Ehrenfeld et al. 2001; Blank and Young 2002). Two distinct mechanisms can be identified. First, three species have higher shoots and faster growth rate compared to the resident vegetation (F. japonica, I.

glandulifera, S. gigantea). Secondly, S. inaequidens invades native communities with a sparse ground layer and achieves higher population density and ground cover compared to the resident vegetation.

In all cases, higher biomass translates into enhanced mineral nutrient fluxes as reflected in steadily increased mineral nutrient stocks. Successful invasive species thus appear as enhancers of nutrient use in invaded ecosystems. However, the strategy of mineral nutrient use varies strikingly between species. For some elements and especially N, concentrations in invasive species are always lower compared to the resident vegetation. This is apparently at odds with published results showing higher concentrations of nutrients in invasive species (Baruch and Goldstein 1999;

Ehrenfeld et al. 2001; Blank and Young 2002). This apparent discrepancy may be

explained by the fact that our data refer to whole shoots, while most published

papers have analysed foliar concentrations only. Low concentrations of all nutrients

in whole shoots of F. japonica are related to a high biomass allocation to nutrient-

59 poor stems (Dassonville et al. 2007) and this may hold true for S. gigantea. In contrast, I. glandulifera has very high cation concentrations in shoots, in line with tissue turgescence being the main mechanism ensuring shoot rigidity in this species with low dry matter content (Andrews et al. 2005).

In I. glandulifera, high nutrient standing stocks certainly result in increased nutrient returns in litterfall since this species is annual. In the other species, which are perennial, this most likely holds true, except if invasive had considerably higher nutrient resorption from senescing shoots, which seems unlikely. It thus appears that invasive species steadily enhance specific components of nutrient fluxes, including net uptake rates from soil and annual returns in dead organic matter.

These results support the view that highly successful invasive species share common functional attributes (e.g. high growth rate, large size …) that result in similar impacts on invaded ecosystems. The non-significant species X invasion interaction in the three-way ANOVA of soil parameters may indeed indicate convergence of impact across species.

Impacts on soil are site-specific and partially predictable

Our results confirm earlier reports of site-specific impacts of alien invasive species (Scott et al. 2001; Ehrenfeld 2003). However, the picture emerging from our study is not one of idiosyncratic impacts. A striking finding is that the direction of change in soil properties follows a partially predictable pattern. Strong positive impacts are most often found in sites when the initial state is poor and negative impacts are generally found in the opposite conditions. This pattern holds true for all elements and is significant for K, Mg, P, Mn and N.

In our sample, sites with negative impacts of invasion were relatively less frequent than sites with positive impacts. This may be due to a sampling bias. Thus, highly eutrophicated sites, in which invasive species are most successful, were often discarded because they had heterogeneous substrate due to topsoil disturbance.

Mechanisms of site-specific impacts on soil properties

Impact on nutrient concentrations in the topsoil may result from alterations of the

balance of inputs to and output from topsoil to other pools of nutrients in the soil-plant

system. The relevant pools are deep soil (>10 cm), living biomass (roots and shoots)

and dead organic matter on the soil. Invasive plants can alter those pools in several

ways.

Increased pools of nutrients in nutrient-poor soil

Soils with low organic matter content and sparse vegetation cover which are invaded by plants with a high biomass may accumulate C through higher litter return. This is true of coastal sand dunes invaded by R. rugosa. Succession from open herbaceous vegetation to closed shrub vegetation is accompanied by an increase of soil organic matter content. In this particular case, the closed shrub layer may also fix the soil and limit soil C loss through erosion. Such an effect may not be specific of alien invasive species also being observed during natural succession in dunes (Kovel et al. 2000).

Increased nitrogen pool in the topsoil can result from enhanced asymbiotic atmospheric nitrogen fixation in invaded plots (Ley and D'Antonio 1998). Increased plant cover may also decrease nitrogen losses through leaching.

Nutrient uplift from deep soil layers is a well known mechanism by which deep-rooted species can move nutrients up to the topsoil (Blank and Young 2002; Jobbagy and Jackson 2004). Several invasive species in our sample do root at deeper depth compared to the native vegetation (F. japonica, H. mantegazzianum and possibly also P. serotina and R. rugosa). Dijkstra and Smits (2002) showed that a small amount of Ca uptake in the deep soil under Acer saccharum was sufficient to sustain high amounts of available Ca in the surface soil.

Finally, increased inorganic phosphorus concentrations in the topsoil under Solidago gigantea has been explained by decreased soil pH and enhanced production of phosphomonoesterase in the topsoil (Chapuis-Lardy et al. 2006).

Decreased pools of nutrients in the topsoil in nutrient-rich soil

In sites with high initial nutrient pools in the topsoil, depletion of topsoil nutrient pools may result from different mechanisms depending on species. F. japonica and S.

gigantea immobilise large amounts of nutrients in perennial organs (rhizomes and roots). In addition, highly productive invasive species with recalcitrant litter may accumulate large amounts of nutrients in dead organic matter on the floor. Thus, both F. japonica and S. gigantea have very high C/N ratio in shoots (42 and 54 respectively) which may slow down litter decomposition. In these species, standing dead shoots actually persist for a long time.

Finally, invasive species that leave soil uncovered in winter, like F. japonica, S.

gigantea, I. glandulifera and H. mantegazzianum, may stimulate nutrient export by

erosion and leaching. Some invasive species accelerate litter decomposition rate and

N release (Allison and Vitousek 2004). This could be the case of H. mantegazzianum

and I. glandulifera, which leave little litter on soil surface in early spring. These

species do not cover the soil surface during winter in contrast to the resident

vegetation mostly composed of perennial grass species. Bare soil is more prone to

erosion and N loss through nitrate leaching.

61 Even though nutrient loss through leaching and nutrient storage in dead or living biomass have the same apparent effect (i.e. decreased availability in the topsoil), these two processes may have completely different implications in the long term.

Indeed, nutrient leaching results in irreversible nutrient loss in contrast to nutrient immobilisation in plant biomass. In the latter case, availability of nutrients may be restored by removal of the invasive species.

Homogenization effects

The particular site-specific pattern in the impacts of alien invasive species may result in convergence of invaded plots towards similar values of soil parameters. There is indeed some support for this prediction in our results. Thus, variation range of element concentrations across sites was narrower in invaded plots than in uninvaded ones, most strikingly so for Mn (44-fold variation among resident plots vs. 15-fold in invaded plots), K (45-fold vs. 22-fold), P (40-fold vs. 22-fold), Mg (35-fold vs. 23-fold), N (32-fold vs. 21-fold), Zn (248-fold vs. 174-fold). It is remarkable that this pattern has been detected in spite of the variation in age of invasion which should have increase the noise in our data and blur other effects.

It has been suggested that alien invasive species might result in biotic homogenization (Vitousek et al. 1996; Olden and Poff 2003). Homogenization of plant community composition in invaded sites has been demonstrated in North America (McKinney 2004). Up to now, such homogenization effect on soil properties had only been demonstrated for Fallopia japonica (Dassonville et al. 2007). Our results provide the first indication that highly successful alien invasive species may promote homogenization of soil conditions in invaded landscapes. This conclusion should be tested in other edaphic contexts, using invaded stands of known age.

Conclusion

Alien invasive species have contrasting functional attributes compared to the resident vegetation they invade. This study suggests that enhanced nutrient use may be a common feature of the most successful alien invasive species in Western Europe.

These species have the potential to alter soil conditions in a partially predictable

pattern. Convergence of topsoil properties due to invasion by those species may

result in ecosystem homogenization. The implications of this for management and

restoration have yet to be investigated.