Native to North America, Acer negundo L. is the most widely distributed of all North American maple. A. negundo was intentionally introduced in Europe during the seventeenth century (in France around 1749 [64, 65]). It is a small to medium sized tree with pinnately compound leaves that usually have five leaflets. First planted in parks, this species is now widely used in South of Europe as an urban tree for avenues for ornamental purposes. The actual distribution area of A. negundo in Europe now extends from southern France to Lithuania and from Italy to Germany . In France, its ongoing invasion takes place in the southern two-thirds , mainly in riparian habitats. This species is of limited commercial importance and is considered an ecological pest inducing biodiversity losses and river banks instability .
At the interface between aquatic and terrestrial ecosystems, riparian forests constitute a key ecosystem that shapes many species' habitats  and are particularly vulnerable to invasions . Acer negundo mostly invades riparian zones at the ecotone between native softwood and hardwood communities [43, 44, 70]. In these habitats, five native species can commonly be found in France and thus are likely to compete one or two at a time with A. negundo: Populus nigra, Salix alba and Alnus glutinosa are early-successionnal species highly tolerant to disturbances; Fraxinus excelsior and Fraxinus angustifolia are late-successional and more shade-tolerant species.
Greenhouse experiment design
The objective was to compare the invasive tree species, A. negundo, to the four native tree species: F. excelsior, F. angustifolia, S. alba and P. nigra. During fall 2003 seeds of A. negundo and both Fraxinus species were collected in situ on populations located along the Garonne River and were sown after vernalization, in spring 2004 at the nursery of the INRA Pierroton research station (44°44'N 0°46'W, west of Bordeaux, Gironde, France). In February 2005, one-year-old seedlings of S. alba and P. nigra were bought. In March 2005, seedlings of all five species were transplanted in 4 L pots filled with a commercial sphagnum soil mixture (organic matter 80 % of dry matter, pH = 6; Le terreau du producteur, HTA, Saint Cyr en Val, France) and placed in a greenhouse under natural air relative humidity and controlled temperature (day T° 25°C and night T° 15°C). Plants were watered daily to field capacity. The experiment was arranged in a split-split-plot design with complete random blocks (3). The treatments were applied to mimic riparian habitat conditions: shade (3 levels, main plot), nutrient availability (2 levels, sub-plot) and mechanical disturbance (2 levels, sub-sub-plot). Treatments were applied from April 1st 2005, 15 days after leaf unfolding, till June 14th. The shade treatments consisted in a control full light (C, 100% of the ambient radiation), shade (S, 25% of full light) and deep shade (SS, 7% of full light). It was obtained combining thermal cloths over the plants. The nutrition treatment was obtained by providing a complete fertiliser (N+, 4 mg of fertilizer Compo Floranid Permanent, 16% N; 7% P2O5; 22.5% SO3; + metal elements) versus no fertiliser (N-). The fertiliser was applied three times on the 3rd, 14th and 53rd day after the start of the experiment. The fertiliser treatment corresponded to a nutrient level equivalent to that of riparian forest soils in South-West France [71, 72]. Finally, disturbance (D) by river bank flooding was simulated by applying a hand-made partial defoliation (25%, on the 21st and 48th day after the start of the experiment) and compared to non-disturbed (ND) plants. Four individuals per species were randomly assigned to each of the 12 treatments, leading to a total of 720 individuals.
Growth and biomass measurements
At the beginning and at the end of the experiment, total height (cm, ruler, nearest mm, H1 and H2 respectively) was measured on each seedling. The relative height growth rate (RGRh
, mm. mm-1
) was calculated for each individual as the difference between the logarithms of final and initial height divided by the number of days between the beginning of the experiment and the harvest:
where ln (H1) and ln (H2) are the ln-transformed plant heights at the initial (t1) and final (t2) time of the experiments respectively .
At the end of the experiment, all seedlings were harvested to measure above- and below-ground biomasses (oven-dried at 65°C until constant dry weight) which were used to calculate the root/shoot ratio (RSR, g.g-1). Within each treatment and block, 180 plants out of the 720 were sampled randomly but equally amongst the treatments and species to undertake detailed biomass measurements: leaves, stems (branches + stem) and roots were separated. All the leaves were immediately set in distilled water for a minimum of 12 h to reach full hydration  and total leaf area per individual (TLA, m2) was determined then with a planimeter (Light box, Gatehouse, Scientific Instruments LTD, Norfolk, UK). Stem, root and leaf dry weights (oven-dried at 65°C until constant weight) were measured. For each species, specific leaf area (SLA, m2.kg-1) was calculated as the ratio of TLA to leaf dry weight; the leaf weight ratio (LWR, g.g-1) as the ratio of leaf dry weight to total individual biomass (stems + leaves + roots).
Photosynthesis and nitrogen content measurements
Gas exchange measurements were carried out in early June, between 8.00 am and 12.00 am, with a steady state through flow chamber (PLC4, PP-Systems, Hitchin, UK) coupled with an infra-red gas analyzer (CIRAS II, PP-Systems, Hitchin, UK). During the measurements, air CO2 concentration, air temperature and relative humidity (RH) in the chamber were controlled to match ambient air values: 375 ± 3 ppm of CO2, 25 ± 1°C and 70 ± 10% of RH. All the measurements were made at saturated light (PPFD = 1500 μmol.m-2.s-1) in order to obtain a light-saturated photosynthetic assimilation rate (Amax, μmol CO2.m-2.s-1) at ambient CO2. No gas exchange measurements were conducted under the deep shade treatment due to the very low number of leaves per individual. For Salix alba, no measurements could be performed either, whatever the treatment, due to a too small leaf size compared to the leaf chamber surface. Three repetitions were made per species and per treatment, leading to a total number of 96 photosynthesis measurements. Light-saturated photosynthetic assimilation rate per unit leaf dry weight (Amaxw, μmol CO2.kg-1.s-1) was calculated as the ratio of Amax to SLA.
Leaf nitrogen content was analysed from the leaf samples used for photosynthetic rate measurements (n = 96). Leaf samples were crushed to powder with a ball mill (MM 200, Fisher Bioblock Scientific, France), then nitrogen content (Nm, %) was measured with an elementary analyser Eager 300 CHONS (FlashEA 1112, ThermoElectron Corporation, Waltham, MA, USA). Nitrogen content per leaf area (Na, g N.m-2) was calculated as Nm divided by SLA and the photosynthetic nitrogen use efficiency (PNUE, μmol CO2.g N-1.s-1) as Amax/ Na.
In situ measurements
In situ measurements were conducted in May 2006 in four invaded riparian habitats of South-West France. Two sites were located in Cestas along the Eau Bourde River (44°45'20.37''N, 0°40'49.95''W and 44°44'47.00''N, 0°41'17.93''W), one in Bruges along The Jalles River (44°54'12.45''N, 0°36'16.40''W) and one in Saint-Denis-de-Pile along the Isle River (44°59'35.66''N, 0°12'28.45''W). In each site, ten adult individuals from the upper canopy were selected for each species (the invasive species A. negundo and the co-occurring native species late-successional F. excelsior and early-successional Alnus glutinosa). Light-saturated photosynthetic assimilation rate measurements were carried out following the same protocol as for the greenhouse experiment. Leaves used for photosynthesis measurements were collected and their leaf area, dry weight, SLA and nitrogen contents were determined as indicated previously.
Statistical analyses were conducted using the SAS software package (SAS 9.1, SAS Institute Inc., Cary, NC). For the controlled conditions experiment, a split-split-plot analysis of variance was performed (proc GLM) and mean differences assessed with SNK and Tukey multiple comparison tests (α < 5%). Main plot (shade) and block effects were tested using shade*block as an error term, the sub-plot effects (fertilisation, fertilisation*shade) were tested using block*fertilisation(shade) as an error term and sub-sub-plot effects (disturbance, disturbance*shade, disturbance*fertilisation, disturbance*shade*fertilisation) were tested using the regular error term according to Federer and King [75, 75]. Analysis of variance (proc GLM) and SNK multiple comparison tests (α < 5%) were used to test species differences in situ.