The experiment was conducted at the Phytozone, a plant growth research facility at the Montréal Botanical Garden (45°33′43.2″N; 73°34′18.0″W), in a polyethylene grow tunnel. Dormant cuttings (20-cm) of four willow species, S. miyabeana ‘SX64’, obtained from Ramo Inc., a Quebec willow producer, as well as three Canadian indigenous species, S. amygdaloides, S. bebbiana and S. nigra, kindly provided by Verbinnen’s Nursery Ltd. (Ontario, Canada), were planted in plastic bags filled with a mix (1:1) of mature compost and BM6 all-purpose high porosity potting substrate (80% peat moss, 20% perlite; Berger, Quebec, Canada) to allow them to root. After 35 days, 120 willows (30 of each species) were transplanted into 12 L pots filled with high quality plantation soil (pH: 6.4, organic matter: 5.8%, CEC: 23 meq.100 g^-1, P: 326 kg ha^-1, K: 224 kg ha^-1, Ca: 3,849 kg ha^-1, Mg: 478 kg ha^-1, Al: 579 ppm, Mn: 16.1 ppm, Cu: 3.68 ppm, Zn: 13.25 ppm, B: 1.11 ppm, Fe: 190 ppm; Pépinière Mucci, Quebec, Canada). A bag of green tea (Tetley) and a bag of rooibos tea (Twinings) were weighed and buried in each of the pots at plantation to assess soil microbial activity following the protocol for the tea bag index (TBI) method (Keuskamp et al., 2013).

Willows were exposed to five treatments: (1) adequate watering and fertilization regime (control), (2) nitrogen overfertilization, (3) flooding, (4) combination of flooding and nitrogen overfertilization, and (5) a cyclic regime. A complete randomized block design including the five treatments and the four willow species was set up and replicated in six blocks for a total of 120 pots (5 × 4 × 6 experimental units). Treatments began on 18 June 2020, a week after transplantation, and lasted for 6 weeks.

For the flooding treatments, the pots were placed in pairs in larger containers (40 × 60 × 36 cm) which were filled up to 10 cm below the soil surface with either tap water (flooding treatment) or a fertilizing solution (flooding and nitrogen overfertilization treatment). Each willow receiving the nitrogen overfertilization treatment was exposed to a dose equivalent to 800 kg of N.hectare^-1.season⁻¹ in the form of an ammonium sulphate (NH₄)₂ SO₄) solution (250 mg of N.l^-1) without any other form of fertilization. ‘Cyclic treatment’ was applied over a 2-week cycle. During the first week, the willows were flooded and fertilized with nitrogen following the same procedure as for the ‘flooding and nitrogen overfertilization’ treatment. The second week, the liquid was emptied from the containers and no nitrogen was applied. The removed water was stored in a cold room and reused to refill the containers the following week.

Plant height increment was measured weekly throughout the experiment. The presence or absence of upward growing roots, adventitious roots and hypertrophied lenticels, which are morphological characteristics associated to flooding (Shuwen et al., 2006), was noted during the last week of the experiment while teabags, roots, stems and leaves were harvested.

Total leaf area was measured using a LI-3100C Area Meter. Biomass weight was measured after the entire root systems and aerial parts of all 120 plants were washed and oven-dried (70 °C, 72 h) (Hamilton Cabinet Dryer). One leaf per plant was weighed individually before being dried to determine the fresh weight. This leaf was then reweighed and dried to determine the dry mass and thus calculate the leaf water content (WC). For nutrient content analysis, sub-samples were sent to a certified laboratory (Agroenvirolab, LaPocatière, Québec, Canada, accredited by CEAEQ and ISO-CEI 17025). The foliar content of total P, K, Ca, Mg, and S was analyzed by ICP-OES (Carter and Gregorich, 2007). Total nitrogen was determined by combustion analysis (CHNOS Elemental Analyzer Vario Micro Select; Elementar Analyse systeme GmbH, Hanau, Germany). Samples of the water contained in the tanks (leachate) were taken at the very end of the experiment and sent to an accredited laboratory (H2Lab, Quebec, Canada) to measure total nitrogen. An approximation of the total amount of foliar nitrogen removal (kg.ha^-1) envisioned for the four willow species according to the different treatments was calculated by multiplying the leaf dry weight by the percentage of leaf N. We then cautiously multiplied the result by three, since the experiment lasted only 6 weeks (whereas full growing season lasts >6 months in Québec, eastern Canada) and by 16,000, the usual number of trees planted per hectare in conventional SRC willow plantation. Such a calculation, although imperfect, is intended as a proxy for the first-year potential of cuttings at maximum growth rate (4 months of spring and summer) and therefore when fertigation could be first planned in situ. The potential for total above-ground filtration and root nitrogen sequestration were not considered in this study since the stem and root nitrogen contents were not measured.

Statistical analyses were performed using R version 3.6.3. The significance level of the analyses was set at α = 0.05. Linear mixed models (nlme library) were built for each of the components under study. The “block” variable was used as a random intercept in all models. The “species” and “treatment” variables were used as fixed variables in all models. The significance of the interaction was verified using ANOVA followed by a Tukey’s test. For all tests, differences were considered to be significant at P ≤ 0.05.

Throughout the experiment, plant growth patterns were stable without reaching a plateau regardless of willow species (Figure 1A) but ‘nitrogen overfertilization’ quickly induced the greatest growth with Salix miyabeana trees reaching almost 2 m after only 6-week (dark blue lines, p < 0.001). When treatment included ‘flooding’ (different shade of green), the height of S. miyabeana, S. amygdaloides and S. nigra with or without nitrogen application did not significantly differ from ‘control’ (purple lines) whether overfertilization occurred. The growth of S. bebbiana was much more severely impacted by inundation with its height being 50% lower (i.e., < 0.4 m; p < 0.05) than that of ‘control’. Individuals of all species in the cyclic treatment (lightest green lines) had equal or greater growth than individuals in other treatments with flooding.

Total dry biomass results differed from those reported above for growth increment. Under all treatment conditions except control, S. nigra performed much better than other species and S. bebbiana was the least productive (Figure 1B); p < 0.05, different capital letters indicate significant difference between species exposed to the same treatment). All trees benefited from nitrogen overfertilization. ‘Flooding + N’ treatment’ conditions were also beneficial to S. nigra, S. miyabeana and S. amygdaloides but the root biomass of S. miyabeana was particularly impacted by fertigation including the ‘cyclic’ regime that may better simulate an operational vegetative filter in which treatment could be intermittent (Figure 1B, left panel).

In addition to plant growth parameters mentioned above, a few morphometrics, anatomical, and nutritional variables were studied to evaluate the potential of the different willow species to adapt to the artificially created environment of a filtering plantation.

In comparison to ‘control’, root-to-shoot ratio decreased in the presence of excessive nitrogen for all plant species, although the ‘cyclic’ treatment slightly alleviated this tendency for S. amygdaloides and S. nigra (Figure 2A). Regarding more specific physiological manifestations of plasticity at the leaf level, no comparisons were made between species but their respective response profiles to the different fertigation treatments were recorded. Neither the results for total leaf area nor those for leaf water content were like those of biomass (Figure 2B) and the specific behavior of S. nigra was highlighted. For total leaf area (left panel), a maximum was observed in the ‘overfertilization’ conditions for all species. While S. miyabeana and S. amygdaloides exhibited rather similar patterns (i.e., more leaves in ‘nitrogen overfertilisation’ condition but not when flooded), S. nigra leaf area remained high in the mesocosms that copied vegetative filters: ‘flooding + overfertilization’ and ‘cyclic’ treatments. On the contrary, both leaf area and leaf water content parameters were indicative of S. bebbiana impairment when flooded (leaf water content was as low as 44% in cyclic conditions) while, for the later, that of S. nigra remained invariable between treatments including ‘control’ (see Figure 2B right panel, RWC remained slightly above 70% in all conditions while S. miyabeana and S. amygdaloides was much more plastic in presence of ‘nitrogen’ but not when flooded). Leaf nutrient content was also impacted by the different treatments in a species-dependant manner (Table 1). While foliar potassium and magnesium contents were not altered for any species, the nutrient status of the plants was logically modified by the different treatments. While potassium and magnesium content varied very little regardless treatment or plant species, phosphorus and calcium monitoring more specifically revealed some disorders, especially for the leaves of S. bebbiana and S. miyabeana, which were respectively much more affected by flooding and/or the combination of flooding and nitrogen overfertigation. Calcium content dropped by about half in ‘flooding’ and ‘overfertilization + flooding’ for S. bebbiana and that number was even worst for S. miyabeana under ‘overfertilization + flooding’. On the other hand, the data collected for S. amygdaloides and S. nigra shows much greater stability, with values even much closer to the ones found in control condition for the cyclic treatment.

At the root and stem level (Figure 3), the presence of adventitious and upward growing roots as well as lenticels was recorded throughout the experiment. S. bebbiana was the species for which the least number of individuals exhibited all these characteristics (Figure 3, see lightest grey bars on left panel). In general, adventitious roots were the feature least observed (Figure 3 right panel, top block), while lenticels were the most common (Figure 3 right panel, bottom block), especially under flooded conditions (with or without overfertilization). Only some individuals of S. miyabeana and S. nigra developed adventitious roots in the cyclic treatment. Specimens of all species developed vertical roots, especially in ‘flooding’ and ‘cyclic’ conditions. Vertical roots were found in all individuals of S. amygdaloides in all treatments except ‘flooding + nitrogen’. Lenticels were almost systematically present on stems of S. miyabeana regardless of treatment conditions, but not in ‘control’ conditions for the other three species.

We investigated different components of the treatment potential of the envisioned vegetative filter which are summarized in Figure 4. We used the Tea Bag Index method to evaluate the carbon storage potential of willows (Figure 4A). While plant species did not significantly influence the stabilization factor, treatment did, with a very low level of stabilization under ‘control’ and ‘overfertilization’ regimes. A contrario, ‘flooding’ (with or without addition of nitrogen) showed the highest potential, and ‘cyclic’ conditions yielded intermediate values. For all species treated with an excess of nitrogen except S. bebbiana, N concentration in leaves was high, above 3% under such conditions (<2% under ‘control’ condition). Together with biomass data. These measurements enabled us to estimate the amount of nitrogen taken up by the willows (Figure 4B), i.e., >30 kg/ha for S. bebbiana, >75 kg/ha for S. miyabeana, S. amygdaloides and S. nigra under ‘N overfertilization’ conditions. The removal potential of S. nigra was comparable in all treatments that included nitrogen. No such level was obtained for the other species. For S. miyabeana, the data calculated for the ‘cyclic’ treatment was higher than for the other two treatments including flooding and ‘control’. For S. amygdaloides and S. bebbiana, no increase in nitrogen uptake under flooding with or without overfertilization was calculated.

At first glance, all species benefited from nitrogen overfertilization with clear biomass gain, especially aboveground. Reduced root-to-shoot ratios were thus observed, which concurs with findings in the literature (Jerbi et al., 2015; Jerbi et al., 2020; Liu and Dickmann, 1992). In flooded conditions, however, opposite profiles were observed. Results for Salix bebbiana, a species that grows naturally in more xeric habitats (Isebrands and Richardson, 2013), contrast with those of the other tested species, showing both reduced growth and almost no adaptive traits such as upward growing roots, lenticels or longer stomata (Cameron et al., 2010; Jackson and Attwood, 1996; Jerbi et al., 2020; Kozlowski, 1997; Kuzovkina et al., 2004; Li et al., 2006; Pereira and Kozlowski, 1977; Wang et al., 2017). The biomass of more wetland-adapted willow species was only slightly affected, and apart from a nutrient imbalance at leaf level of S. miyabeana that could be indicative of stress response and alter long-term growth if no amendment is considered (Barak et al., 1997; Kozlowski, 1997), biomass distribution remained constant which is promising under the conditions herein reproduced.

A general pattern of response to leachate irrigation (the combined ‘flooding + overfertilization’ treatments) emerged for all three species adapted to riparian environments that clearly benefit from this treatment. Salix miyabeana, S. amygdaloides and S. nigra, performed equally well, or even better when irrigated, than in the ‘control’ conditions. These more flood-tolerant species had greater aboveground biomass production and a higher leaf surface area to maximize transpiration, thus resulting in increased foliar concentration of the applied nitrogen. Similar results were observed in Weih and Nordh (2002) when they tested different irrigation and fertilization regimes on fourteen willow species.

One of the objectives of the present study was to compare the tested species native to Canada to S. miyabeana ‘SX64’, which was expected to be the best performing species under leachate irrigation as it has been used frequently in nitrogen-rich water treatment with a high success rate (Guidi Nissim et al., 2014; Guidi Nissim et al., 2015; Jerbi et al., 2020; Lachapelle-T et al., 2019; Mirck and Volk, 2010; Sas et al., 2021). However, a similar or higher level of productivity was observed for well-adapted indigenous species in the present study.

In particular, S. nigra, recognized as a species well adapted to wetlands (Mosseler et al., 2017), displayed an interesting plasticity in physiological and morphological characteristics, as described above, and the most conservative isohydric behavior with no change in leaf water content (Sade et al., 2012). It was the most productive when nitrogen supply was maximized under flooded conditions. Furthermore, changes in its nutritional status also remained relatively buffered compared to others, although of course there was an overall increase in foliar N content, a good indicator of the plant’s ability to take up nitrogen from the substrate (Simon et al., 1990). The differences highlighted here show a clear distinction in response among willow species. S. nigra shows a wide range of adaptive traits that allow the conception of more versatile leachate phytofiltration components for engineered ecological landscaping.

The results from soil and leachate monitoring in this study highlight the need for a large-scale testing trial of these indigenous willow species using an intermittent ‘cyclic’ fertigation regime. In this experiment we tested two leachate application procedures and found that the ‘cyclic’ treatment was as promising as a conventional permanent fertigation regime in regard to nitrogen filtration potential. Of course, a significant amount of nitrogen-loaded water was leached at the end of the experiment (eq. > 200 kg ha^-1 more in ‘flooding + overfertilization’ than in ‘flooding’ conditions), but plants were still in a juvenile phase with no mature root system to fully colonize the substrate and capture the available nutrient stock (Wang et al., 2017). Furthermore, the nitrogen concentration reported in leaves was indicative of an optimal nutrient supply (Kopinga and van den Burg, 1995) and consistent with the values reported in the literature (Aronsson and Perttu, 2001; Hofmann-Schielle et al., 1999), as was the projected nitrogen removal rate (Fabio and Smart, 2018), which could rank close to the >400 kg ha^-1 estimates by Labrecque and Teodorescu (2001) when aboveground biomass is more realistically harvested at 2–3 years of age after being adequately fertilized.