Background ¹⁵N abundance in test species was determined from the literature and from plants collected in the Tidal Linkage (Figure 1; Supplementary Text S1). On December 12, 2002, we collected shoot and root samples from one individual of each species. Plants were refrigerated, taken to the laboratory, and rinsed with deionized (DI) water over a 500-μm sieve to eliminate soil and salts while preserving the root mass. We separated roots and shoots for each plant. In the case of Frankenia salina, Batis maritima, and S. pacifica, any stem fragments with green tissue were classified as shoot throughout the experiment. Tissue N content (%N) and atoms % ¹⁵N in shoot and roots were determined directly on dried (60°C for 48 h) and ground tissue samples, after milling through a 40 mesh (420-μm openings). Analysis was by a Carla Erba 1,110 Elemental analyzer coupled to a Finnigan MAT 252 Isotope Ratio Mass Spectrometer at the Stable Isotope Ratio Facility for Environmental Research in the University of Utah. Precision (based on internal standards) was ±0.07‰ for ¹⁵N.

We applied the ¹⁵N enrichment approach due to the overlap between Triglochin’s isotope natural abundance values (δ¹⁵N) and values from other common salt marsh species (Table 2). Additionally, Tijuana Estuary receives sewage overflow during storm events, which is enriched in ¹⁵N and results in higher δ¹⁵N across the food web relative to natural marshes, which could confound natural abundance levels (Moseman et al., 2004).

Seedlings for the greenhouse and field experiments were grown in 2001 in our greenhouse at San Diego State University, CA. Seeds were collected from the Tidal Linkage in fall–winter 2000, air dried, and stored at 5°C until planting. On November 1, 2001, seeds of all marsh plain species (Table 1), except the annual Salicornia bigelovii, were germinated in flats filled with topsoil covered in a ~ 2-cm layer of potting soil. Seedlings were then transplanted individually into sterile 5×5-cm plastic pots and grown for 8 months, watering with freshwater two to three times a week as needed. We germinated S. bigelovii seeds 2 months prior to the start of each experiment.

To label Triglochin plants, we followed the method proposed by Hart (1990). On August 20, 2002, we transported the 60 largest Triglochin plants from San Diego to UW-Madison. Soil was washed off the roots with DI water using a pressure hose. We standardized plant size by removing all but three randomly chosen shoots. Roots were cut to 7-cm length, then 10 large roots were separated and remaining root mass was cut to 2-cm length. Plants were rinsed again in DI water, blotted with paper, weighed, and measured as the length of the tallest shoot.

Triglochin propagules were planted in sterile square plastic pots 9.6-cm tall (112.36 cm² area), filled with sterilized clay soil (35.7% clay, 56.4% silt, and 7.9% sand) similar to marsh plains in southern California (Zedler, 1982). The soil had: 5.6 pH, 5 ppm NO₃, 0.24 ppm P, and 2 ppm K (UW-Madison Soil and Plant Analysis Laboratory). We randomly arranged the pots in plastic trays in the greenhouse (Supplementary Figure S1). We supplemented natural light with 1,000-watt high-pressure sodium lamps in a 14:10-h photoperiod. Trays were filled with 2 g L⁻¹ Proline™ synthetic sea water, to 2 cm below the top of the pots, to maintain soil saturation. We added DI water to offset evaporation and transpiration.

We fertilized Triglochin pots with Technigro™. ¹⁵N natural abundance was 0.368 ± 0.002 atoms % ¹⁵N, so we aimed to enrich plant tissue to 1.102 atoms % ¹⁵N. The water-soluble Technigro™ fertilizer has a 20 N: 9 P: 20 K ratio, trace elements, and 8.5% N (as CO(NH₂)₂) and 11.5% NO₃ (as KNO₃). This fertilizer was enriched with 99% ¹⁵N ammonium nitrate (NH₄NO₃), which was 35% N (Isotec, Inc. Miamisburg OH). Salt marsh sediments release mineralized nitrogen predominantly as ammonium, which is the main source of nitrogen uptake by marsh plants (Gardner et al., 1991; Chambers et al., 1992). We applied the enriched fertilizer to each pot at a rate of 5 g N m⁻² by dissolving the fertilizer in 240 ml of water. The fertilizer was added in two 2.5 g N m⁻² applications on September 9 and on September 23, 2002. Following fertilization and algal proliferation, we treated each pot with ZeroTol^ZT, a broad spectrum algicide containing 27% H₂O₂, in a 1:100 concentration, on September 25, then repeated on September 27 with a 1:300 concentration. On October 7 the plants were transplanted to larger plastic pots, 14-tall × 15.2-cm diameter (153.4-cm⁻² area), filled with sterilized clay soil. We covered each pot’s soil with Perlite© to limit algal growth. After transplanting, we applied another dose of ¹⁵N enriched fertilizer (15 g m⁻², mixed as described above); each pot was irrigated with 2 ml of the liquid mixture on October 25 and on November 1. In total, each plant received 0.133 g N ¹⁵N enriched.

On November 22, we began to salt-harden plants and eliminate freshwater algae. Over 2 weeks, we applied dilute seawater in increasing concentrations: 4, 6, 15, and 35 g L⁻¹ every 3 days. In between seawater additions, we added DI water and monitored soil surface salinities as interstitial water squeezed through filter paper onto a refractometer. Once soil surface salinities were > 35 g L⁻¹ salt, pots were watered with DI water only. On December 3, plants were removed from the pots and we carefully washed each plant over a 350-μm sieve with DI water to clean the root mass.

We blotted the labeled 25 Triglochin plants, obtained wet weight, and measured the height of the tallest shoot to analyze ¹⁵N label uptake. of these, 20 Triglochin plants were analyzed whole and 5 were separated into green shoots, senesced shoots, inflorescences, and roots. Then, they were then dried at 60°C for at least 48 h and weighed. We determined Tissue N content and atoms % ¹⁵N as with the background samples (as above). We determined recovery of the ¹⁵N label following Cabrera and Kissel (1989) as:

The remaining labeled Triglochin plants were held in DI water at 5°C for use in the field and greenhouse experiments.

To test for N uptake and N transfer from Triglochin to seven halophytes, we set up a complete block experiment on December 11–12, 2002 on the unvegetated marsh plain of the Model Marsh surface ~20 m away from the edge of the S. foliosa plantings (Figure 1). The 5 blocks were ~ 80 m apart, and each block encompassed four 0.04-m⁻² plots. In each plot, we planted one individual each of S. pacifica, F. salina, Jaumea carnosa, S. esteroa, L. californicum, B. maritima, and 4 S. bigelovii seedlings in a circle, with position assigned randomly. Four S. bigelovii seedlings were planted to compensate for their small size. In each block, we established 4 plots ±Triglochin (n = 2), i.e., 2 control plots with no Triglochin (−Triglochin; extrapolated density = 273 plants m⁻²) and 2 plots +Triglochin, planted with ¹⁵N-labeled Triglochin plants (extrapolated density = 328 plants m⁻²). The ±Triglochin plots were 6 m apart. Each +Triglochin plot held a pair of plants with approximately the same wet weight (36.49 ± 0.66 g). We assessed mortality on January 6 and replaced dead plants. Strong winter storms in 2003 during a mild El Niño/Southern Oscillation delivered >16 cm of heavy rainfall (NWS 2004), which dislodged S. bigelovii seedlings. After storms ceased, we replaced missing seedlings on January 30 and again on April 11, 2003. No further seedlings were replaced, due to transpiration stress as temperatures rose to 27°C in April (NOAA, 2003).

We determined soil salinity from 5-cm deep cores collected next to the treatment plots on December 12, 2002, May 19, 2003, and March 23, 2004; interstitial water was minimal, so we mixed soil pastes by drying soil samples at 60°C, mixed with DI to make a saturated soil paste (Richards, 1954), and expressed the filtrate onto a refractometer. The experiment was terminated March 23–26, 2004.

Prior to harvesting, we counted survivors (with green tissue). We collected live and dead shoots on the surface of the plot. We extracted roots using a corer 22 × 25 cm (diameter × length). Cores were taken to the lab and washed over a series of four stacked 500-μm mesh sieves to collect roots. Shoots and roots were rinsed with DI water to remove soil and salts and then stored at 5°C. Individual species’ rhizomes were separated from the root mass with dissection needles; fine roots were combined for each plot. Shoot and root samples were oven-dried at 60°C for 2 days and weighed to obtain dry weight. We analyzed whole plant %N and atoms % ¹⁵N (shoot and root tissue combined) from all plants in +Triglochin plots as described in the previous section. The N pool for each plot was calculated based on root, shoot, algal turf, and fine root biomass and tissue N content. Root:shoot ratio was calculated per plot and by species.

We repeated the design of the field experiment except that we had to divide treatment replicates between two UW greenhouses to fit available space. Shading differed somewhat and one greenhouse averaged 7°C higher in ambient temperature. Evapotranspiration and soil saturation were not measured but appeared to vary with temperature.

We grew seedlings for 20+ pots in San Diego and flew them to UW-Madison, then transplanted them to sterilized plastic pots 22 × 22 cm (diameter x length) filled with sterilized clay soil. Pairs of Triglochin in each pot weighed on average 59.85 ± 1.28 g. To introduce microflora, we inoculated each pot with a small amount (~10 g) of marsh soil taken from a natural marsh in Tijuana Estuary. We assessed mortality after 2 weeks, when only S. bigelovii plants had died and were replaced. Three subsequent replants all died within a week, so we stopped replanting after March 16, 2003, when transpiration stress made S. bigelovii seedling survival unlikely.

To salt-harden seedlings, starting January 2, 2003 we watered pots with artificial Proline™ seawater (2, 4, 6, 15, and 35 g L⁻¹ salt). In between seawater additions, we watered with DI water. Once soil surface interstitial salinities were > 35 g L⁻¹ salt, pots were watered with DI water only. On April 8, plants were placed inside 11-L plastic containers and assigned to each of two greenhouses. Starting May 11, we alternated between dry and wet soil by filling containers with either 3 or 1.5 l of DI water every 2 weeks. We used 1,000-watt high pressure sodium lamps to supplement natural light in a 14-h photoperiod.

Triglochin produced seeds that germinated continually, so we maintained pot density by pulling up seedlings on October 17, December 4, December 18, 2003, and February 1, 2004. Seedlings were left on the soil surface to maintain each pot’s N pool. On October 17, Triglochin plants were pruned of standing dead biomass and the canopy of the other species was shaken to remove dead tissue. We fragmented this plant biomass manually and returned it to the pots. We measured interstitial soil surface salinity on June 26, August 8, September 19, 2003, and February 20, 2004. As possible, soil was returned to each pot after measuring salinity. Salinity in the bottom 5 cm of each pot was measured on February 20, 2004 while harvesting the experiment. Litter and Triglochin seedlings present at the end of the experiment were collected separately for each pot. We processed plant material as with the field experiment (see above), to determine whole plant biomass (shoots + roots), %N and atoms % ¹⁵N for +Triglochin treatments and for six plants from −Triglochin pots (three from each UW greenhouse). We pooled Triglochin seedlings to provide enough material for %N and ¹⁵N analysis. The N pool for each pot was calculated based on root, shoot, litter, fine root biomass, and tissue N content. Root:shoot ratio was calculated per plot and by species.

We quantified the isotopic composition relative to the ¹⁵N/¹⁴N atmospheric ratio. Results are expressed as parts per thousand (‰) differences from the corresponding standard:

higher δ values indicate a greater proportion of the heavy ¹⁵N isotope.

Differences among ± Triglochin treatments in shoot and root biomass, root:shoot ratio, %N, and δ¹⁵N values were analyzed using a two-way ANOVA with interactions. Total shoot biomass excluded litter and total root biomass excluded fine root weight; we also eliminated Triglochin from +Triglochin treatments to assess its effect on the 7-species assemblage. We specified block and greenhouse effects as factors. We analyzed individual species performance between plots and pots ± Triglochin using a multistratum ANOVA with the dependent variable (shoot and root biomass, root:shoot ratio, %N, and δ¹⁵N) nested within plots or pots.

Dependent variables that did not meet the assumptions of ANOVA were transformed: biomass was transformed using x’ = √x + ⅜ to stabilize the variance, and root:shoot data were log-transformed (Zar, 1999). Tukey’s Honest Significant Difference (HSD) was computed for post-hoc pairwise multiple comparisons. Statistical analyses were performed using R statistical framework (R Development Core Team, 2012) and the package MASS (Venables and Ripley, 2002). Means ±1 SE are reported in the text.

Delta values from shoot samples were low and ranged from −0.2 to 6.7‰ (Table 3). Triglochin shoots had the highest δ¹⁵N value and B. maritima was depleted. Root samples of F. salina, J. carnosa, L. californicum, and S. pacifica were depleted relative to shoots. The δ¹⁵N values found for F. salina and S. pacifica were lower than records from other sites (Table 3); we found no published information for other species. Tissue N contents in shoots (Table 3) were highest for Triglochin (4.5%), and ranged from 0.8 to 1.6% for other species. N content in Triglochin root tissue was also higher (2.9%) than for other species (~0.7%).

Biomass of Triglochin plants increased 20.3 ± 1.0 g wet weight by the end of the labeling process. The mean δ value in labeled tissue was 656 ± 13.2‰ and N tissue content was 1.9 ± 0.3%. There was no difference in δ¹⁵N values between tissue types but N content in senesced shoots was 0.9 ± 0.01 compared to 2.2 ± 0.05% in live tissue. The recovery rate of the ¹⁵N label was on average 14.4 ± 0.3%.

Two years after it was opened to tidal flows, the soil surface salinity did not vary between blocks and it averaged 56.4 ± 2.5 g L⁻¹ in December 2002, 72.4 ± 6.5 g L⁻¹ in May 2003, and 70.1 ± 4.9 g L⁻¹ in March 2004, reflecting seasonal differences.

Plant survival at the end of the experiment, in 2004, was high: 100% for B. maritima, F. salina, and S. pacifica; 65% in −Triglochin plots and 74% in +Triglochin plots. 75% for J. carnosa and S. esteroa, but only 14% for S. bigelovii. Four of the S. bigelovii plants collected were natural recruits. At the end of the experiment, only one −Triglochin plot and three +Triglochin plots still contained the original species assemblage.

Unless specified, total plot biomass for +Triglochin treatments considers all plants, including Triglochin; also, “roots” includes rhizomes throughout. Biomass did not differ between ± Triglochin treatments, despite different planting density (n = 10 in −Triglochin vs. n = 12 in +Triglochin plots). Scaling from 0.04 m⁻² plots, mean shoot biomass was 363.6 ± 41.3 g m⁻² in −Triglochin and 352.44 ± 58.1 g m⁻² in +Triglochin plots. Shoot biomass varied among species (Figure 2; F = 77.42, p < 0.001; Supplementary Table S1). S. pacifica was the dominant; with 57.6 ± 3.7% of all shoot biomass per plot. Although Triglochin plants were still alive at the end of the experiment, most had little shoot biomass (0.96 ± 0.5 g m⁻²) and showed evidence of herbivory.

Root biomass was 60.6 ± 5.5 in −Triglochin and 56.4 ± 7.2 g m⁻² in +Triglochin plots. S. pacifica and B. maritima contributed 31.7 ± 2.6 and 36.7 ± 3.1%, respectively, of total root biomass within plots (Figure 2; Supplementary Table S1). Triglochin root biomass was 4.2 ± 0.9 g m⁻². Fine root biomass was high in both −Triglochin, 33.6 ± 4.4, and + Triglochin, 51.4 ± 8.9 g m⁻² treatments. Root:shoot ratio did not vary either among ± Triglochin treatments. Average root:shoot ratios were 0.4 ± 0.04 in −Triglochin and 0.5 ± 0.1 in +Triglochin plots. Root:shoot ratio differed among species (Supplementary Table S2; F = 28, p < 0.0001). S. pacifica had the lowest root:shoot ratio (0.09) and Triglochin the highest (4.3), largely due to reduced shoot biomass. No litter was found in the plots.

Labeled Triglochin plants retained the ¹⁵N label after 14 months in the Model Marsh (December 2002–March 2004). Delta values were significantly different between ± Triglochin plots (F = 6.582, p = 0.0116) and between species (F = 143.8, p < 0.01). Delta values for Triglochin were 343.1 ± 39.9‰, on average 50% lower than δ¹⁵N values in plants immediately after labeling (656 ± 13.2‰). Delta values for other species ranged from 5 to 26‰ (Figure 3) and were consistently higher than background levels for plants of similar age collected in the Tidal Linkage (Table 3). Seedlings that germinated in field plots had high δ15N values (25‰) and 0.4%N. Average tissue N content was highest for Triglochin, 1.7 ± 0.1% N (Figure 3; F = 5.23, p = 0.01) and was not significantly different from background levels (Table 3). The N pool in +Triglochin plots was 5.9 ± 0.9 g m⁻². The contribution of individual species to the N pool followed biomass patterns; S. pacifica accumulated 44 ± 2.0% of total N.

Growing conditions in the greenhouse differed from the field, in part because pots confined plant roots. Soil surface salinity (5-cm depth) did not vary between pots in the two greenhouses; the average was 66.5 ± 5.3 g L⁻¹ in June, 98.5 ± 2.4 g L⁻¹ in August, 74.5 ± 5.3 g L⁻¹ in September 2003, 53.5 ± 5.9 in February 2004. Salinity in the bottom 5 cm of pots was 13.1 ± 2.3 in the lower vs. 16.4 ± 2.1 g L⁻¹ in the less-shaded greenhouse. Still the greenhouse results supported those of the field test.

As in the field, only S. bigelovii had a low survival rate (10%); all other plants survived. In both ± Triglochin pots, collected S. bigelovii plants were small (< 5 cm tall) and none was branching. At the end of the experiment, only five −Triglochin pots still contained the complete species assemblage. One −Triglochin pot was contaminated by a Triglochin seed that germinated and grew to ~10 cm before being removed at the end of the experiment.

As found in the field, there was a significant overall effect of ± Triglochin treatments on shoot biomass (Figure 4). Mean shoot biomass, scaling from 0.04-m⁻² pots, was significantly higher in −Triglochin pots (373.4 ± 22.9) than +Triglochin (225.8 ± 21.3 g m⁻²; F = 22.16, p < 0.001). This difference can be attributed to S. pacifica, whose shoot biomass was 44% lower in +Triglochin than in −Triglochin pots (52.5 ± 7.4 vs. 120.4 ± 18.6 g m⁻²; F = 5.44, p < 0.0001); F. salina shoot biomass followed a similar pattern (Figure 4). Sarcocornia pacifica showed a significant difference in shoot biomass between ± Triglochin treatments. Triglochin shoot biomass was 17.64 ± 2.2 g m⁻². Litter collected in +Triglochin pots was mostly Triglochin and J. carnosa senesced shoots (180.7 ± 23.3 g m⁻²); in −Triglochin pots litter was mostly F. salina and J. carnosa senesced leaves (95 ± 13.2 g m⁻²).

Neither shoot nor root biomass differed between greenhouses. Root biomass did not vary between ± Triglochin treatments (96.9 ± 10.5 − Triglochin vs. 102.8 ± 6.0 g m⁻² + Triglochin). F. salina root biomass was significantly higher than other species (54.73 ± 6.74 g m⁻²; F = 96.45, p < 0.001; Supplementary Table S1), including Triglochin root biomass (35.6 ± 5.1 g m⁻²). Fine root biomass was similar in ± Triglochin treatments (121.1 ± 11.7 in −Triglochin and 147.1 ± 10.3 g m⁻² in +Triglochin). Root:shoot ratio was significantly higher in +Triglochin treatments, average ratio was 0.5 ± 0.04 in −Triglochin and 0.7 ± 0.1 in +Triglochin pots (F = 6.14, p = 0.02; Supplementary Table S2). Triglochin had the highest root:shoot ratio (2.0) and J. carnosa the lowest (0.08). Litter production was similar between greenhouses, although litter collected in the cooler greenhouse was moist and appeared more decomposed than litter from the warmer greenhouse.

Delta values were significantly higher for species in +Triglochin pots (Figure 5; F = 13.49, p < 0.0001; Supplementary Table S3). Delta values in pots −Triglochin ranged between −2.6 and 6.8‰ and between 231.1 ± 23.2‰ for +Triglochin pots. These values for Triglochin were on average 65% lower than δ¹⁵N values found in plants right after labeling (656 ± 13.2‰). Triglochin seedlings that germinated from seeds produced during the experiment had high δ¹⁵N values, 158.8 ± 21.9‰. Delta values for other species in +Triglochin pots ranged between 4–80‰. Delta values were highest for L. californicum (89.5 ± 13‰) and lowest for B. maritima (15 ± 23.2‰). B. maritima had a high root:shoot ratio, with little growth aboveground. Delta values for plants grown in +Triglochin pots were consistently higher than background levels and values for other studies (Table 3; Supplementary Table S3). Delta values of litter in +Triglochin pots were lower than for live shoots (29.1 ± 34.8‰); δ¹⁵N values in litter from −Triglochin pots were 1.5 ± 0.4‰.

There were no differences in whole plant tissue N content between ± Triglochin treatments (Figure 5). There was a significant difference between the two greenhouses (F = 2.6, p < 0.001); % N was lower in the greenhouse that was warmer and received more light (0.6 ± 0.03 vs. 0.7 ± 0.1% N). As in the field experiment, tissue N content was highest in Triglochin seedlings that germinated during the experiment 1.9 ± 0.2% (F = 59.7, p < 0.001); followed by Triglochin and L. californicum (Figure 5). Percent N in litter was similar in −Triglochin, 0.8 ± 0.1, and + Triglochin, 0.73 ± 0.09% N. Tissue N content was positively related to root:shoot ratio (r = 0.39, F = 4.9, p < 0.001); suggesting higher N content in roots than shoots.

Overall, the N pool was lower in −Triglochin than in +Triglochin pots (F = 8.09, p = 0.01); this difference can be attributed to the addition of Triglochin plants. Triglochin plants and seedlings accumulated 0.6 ± 0.03 g N m⁻² (Figure 6). We combined N-pools for the remaining species and we found N accumulation in tissue of the 7-species assemblage was higher in −Triglochin than in +Triglochin pots (2.17 ± 0.08 vs. 1.37 ± 0.11 g N m⁻²; F = 10.89, p < 0.001). This difference can be attributed to the increase in F. salina and S. pacifica shoot biomass in −Triglochin pots. There was a significant effect of greenhouse on the N pools among pots; on average, the N pool was on average higher in pots in the cooler greenhouse (4.66 ± 0.12 vs. 3.53 ± 0.05 g N m⁻²; F = 11.91, p = 0.003).

A reservoir of N is important where N is limiting, as in newly excavated, formerly buried substrates (Valiela and Teal, 1974; Boyer and Zedler, 1998). We found that Triglochin had high uptake of the ¹⁵N label, and high N content and retention (Figures 3, 5, 6), likely through recycling or internal translocation. These findings are consistent with previous results showing that Triglochin can limit growth and N accumulation of other plants in the 7-species assemblage (Table 1), particularly the biomass dominants S. pacifica and F. salina (Sullivan and Zedler, 1999; Callaway et al., 2003; Sullivan et al., 2007; Morzaria-Luna and Zedler, 2014). Our results support the hypothesis that Triglochin is a superior competitor for N and a functionally important halophyte that dominates N cycling, as proposed by (Zedler and Callaway, 1999; Zedler, 2005).

We demonstrated that a subordinate, low-biomass, species can take up more N than other marsh halophytes in a regional model restoration site. The whole plant N content of Triglochin was 1–5 times as high as that of other species in greenhouse and field experiments. Sullivan and Zedler (1999) and Callaway et al. (2003) also found high N in Triglochin shoots (1.7%–5.2%). Although waterlogged soils lose N through denitrification, Triglochin appears to have a competitive advantage in such soil, because it was able to grow substantial root biomass deep into the 10–12 cm of waterlogged soil at the base of of 38-cm tall greenhouse pots (Sullivan et al., 2007). Plant nitrogen uptake also responds to drought, flooding, salt stress and nitrogen fertilization (Morzaria-Luna and Zedler, 2014; Bai et al., 2017).

Nitrogen has important influences in salt marshes, some of which can be attributed to succulent Triglochin spp. with leaves having abundant free amino acids and low cell-wall content, preferred by several bird species as food (Popp and Albert, 1980; Naidoo, 1994). At the same time, herbivory is deterred by cyanogenic glucosides taxiphyllin and triglochinin that Triglochin spp. synthesizes (Majak et al., 1980). These compounds contain N (Nielsen and Moller, 1999), and feeding causes the cyanogenic glucosides to degrade rapidly into an aldehyde or ketone, cyanide, and sugar (Jones et al., 2000). Anecdotally, crushing the shoots of this Triglochin releases a strong smell that suggests cyanide (Alonso-Amelot and Oliveros, 2000).

Our field and greenhouse experiments showed that seven halophyte species and labeled Triglochin had higher δ¹⁵N values (Figures 3, 5) than plants grown without labeled Triglochin. Values also exceeded previously reported natural abundances (Table 3). Thus, the labeled δ¹⁵N from Triglochin tissue was transferred to the plant tissue of the other common marsh halophytes. The ¹⁵N content in vascular plant tissue is a reflection of the N sources assimilated during the growing season (Shearer and Kohl, 1986; Johannisson and Högberg, 1994). Multiple mechanisms could account for the observed transfer of δ¹⁵N label from Triglochin to other plants. Although litter collected from +Triglochin pots was predominantly Triglochin senesced shoots, litter N content was similar in ± Triglochin pots (~0.8%); litter δ¹⁵N values in +Triglochin pots were also lower (60%) than live Triglochin shoots. This is consistent with death, decomposition, and remineralization of Triglochin shoots in the sediments and subsequent plant uptake (White and Howes, 1994a,b).

Labeled N could also reach other plant species by root exudates, leaching, or bacterial remineralization (Paynel et al., 2001; Levin et al., 2006), or by simple mass flow or diffusion of ions and simple amino acids along a concentration gradient (Teste et al., 2015). Although we did not measure N or δ¹⁵N in the sediment, the rapid transformation of inorganic N to plant organic N is a major mechanism for short-term nitrogen retention (White and Howes, 1994a). Root decomposition could also contribute to N transfer; seasonal translocation of N from above to belowground biomass during senescence is potentially important to internal N cycling of other tidal marsh species, like Spartina alterniflora (Morris, 1980; Hopkinson and Schubauer, 1984). Physiological integration where N is transferred to ramets in clonal plants, such as Triglochin, was recently confirmed in marsh plants (Do, 2018).

While sharing its N, Triglochin can also limit the growth of its neighbors. Plants grown in pots +Triglochin had reduced shoot biomass; the dominants, S. pacifica and F. salina were particularly affected. Plants grown in +Triglochin pots had on average 44% less S. pacifica and 45% less F. salina shoot biomass than in –Triglochin pots by the end of the greenhouse experiment (Figure 4). These findings are consistent with earlier results (Table 3), Shoot N content in samples collected by Callaway et al. (2003) adjacent to our field plots ranged from 3.1 ± 0.2 to 5.1 ± 0.1% for Triglochin, and between 7.7 (J. carnosa) and 17.4% (F. salina) for the remaining species. Biomass did not differ between ±Triglochin in field plots (Figure 2), unlike in greenhouse pots with limited root space.

At least three mechanisms could account for the decreased growth of plants grown with Triglochin we observed here: direct competition, allelopathy, and indirect interactions that favor other species. In companion greenhouse and field experiments, we tested for competition between Triglochin and S. pacifica with varied N and water levels, using biomass to assess outcomes. Although there was evidence of facilitation in a short-term (6 mo.) experiment, the year-long interaction showed competition across the stress gradient (Morzaria-Luna and Zedler, 2014). Species-level traits explained responses of both Triglochin (with its high N uptake) and S. pacifica (with its large and perennial aboveground biomass). N supply was the only factor that affected outcomes; neither inundation nor time of interaction had any effect. By reducing the growth of its neighbors, Triglochin could facilitate canopy gaps, where the annual S. bigelovii could persist (Varty and Zedler, 2008). We attribute the observed reduction in shoot growth of plants in +Triglochin treatments to competition that we tracked using labeled N, as recommended by McKane et al. (2002). Although %N in plant tissue did not vary between ±Triglochin treatments, total N accumulation in F. salina and S. pacifica was higher in −Triglochin pots (0.6 ± 0.03 and 0.7 ± 0.1 g N m⁻²) than in +Triglochin (0.4 ± 0.05 and 0.3 ± 0.04 g N m⁻²) suggesting the additional growth of these plants was due to enhanced N availability.

The impact of N released from litter might be highest for S. bigelovii, the only annual species on the salt marsh plain (Zedler et al., 1999). N released by Triglochin during summer could be taken up by S. bigelovii seedlings, with senescing S. bigelovii releasing N in winter and spring, when Triglochin can take it up. Salicornia bigelovii might also take up higher amounts of N, since it was able to remove 23% of total N applied in a microcosm experiment (Brown et al., 1999). Lindig-Cisneros et al. (2003) found that where N was added to a restored S. foliosa marsh, S. bigelovii recruited rapidly, reaching biomass of up to 80 g m². Thus, Triglochin could be extremely important in facilitating both the recruitment and establishment of S. bigelovii seedlings. Further research could assess the impact of seasonal Triglochin litter decomposition on S. bigelovii seedlings.

Dominant species are often listed as restoration targets, and they are typically defined as those with extensive canopy cover or higher aboveground biomass (Zedler, 2001). In this study, however, we focused on a subordinate species that is hard to find after its leaves have senesced and washed away with the outgoing tide. Triglochin is often ignored; here, we confirm that it can dominate assemblages by accumulating N in its roots and shoots, even when it produces very little cover or biomass, and that it can share some of its N with neighbors and simultaneously limit their biomass. We had expected N-sharing or competition, but not both. The evidence, however, is compelling.

N-accumulation by Triglochin and subsequent effects on plant growth have immediate implications for the restoration of the plant community composition and its ecosystem functions. The salt marsh plain is dynamic, so species are not likely to remain where they are planted. A clonal species like Triglochin could concentrate N over a large area, translocating N between shoots and roots, creating long-lived nutrient-enriched patches and contributing significantly to spatial variability in available N (Baye, 2007; Do, 2018). The differential ability of plant species to compete for nutrients within nutrient-rich patches created by Triglochin could shift species abundances and increase diversity (Grime, 1997). As Triglochin shoots die back by mid-summer, they could decompose faster and release more nutrients than surrounding species because of their high N content (Quested et al., 2003). In greenhouse pots +Triglochin, 32% of the total N pool was contained in the litter (Figure 6).

Restoration is underway in several southern California salt marshes, to help compensate for major historical losses and in anticipation of future needs as sea level rises (SLR) and other anthropogenic impacts (Thorne et al., 2018). Even under the projected “low SLR scenario” for Tijuana Estuary, higher marsh vegetation would be submerged by the end of this century; and under the “high SLR scenario” all tidal marsh habitats would convert to mudflat by 2110 (Thorne et al., 2018). Thus, the upper margins of wetland salt marshes will need to extend into what is now upland. Soils that are low in N and other components (nutrients, mycorrhizae, seed banks) can be amended to support future salt marsh vegetation, but the most efficient plans will be based on knowledge of interactions, such as the ability of Triglochin to capture and share N with its native neighbors (Figures 3, 5).

We recommend planting Triglochin to increase plant diversity directly, adding seedlings and seeds to restoration plantings, and to restored and remnant marshes where Triglochin was more abundant historically (1989–2004 data show high variability at Tijuana Estuary (Zedler and West, 2008). We also recommend planting Triglochin to increase diversity indirectly, by reducing dominance where tidal action is being restored or created. We suggest collecting Triglochin seeds during peak months of fruiting (May–June) to build up a reserve for future use, e.g., longevity tests during storage beyond 3 years (Sullivan and Noe, 2001). At the Tidal Linkage, Triglochin recruitment was limited even though adult plants were nearby (Lindig-Cisneros and Zedler, 2002). Restoration plantings could omit biomass dominants (especially S. pacifica) to allow Triglochin to “creep” across the marsh plain with less competition. Additional considerations are needed for large-scale plantings that could be needed following the catastrophic loss of an entire salt marsh (e.g., through erosion, burial, oil or other toxic spills).

We continue to endorse Callaway’s (2005) call for “rigorously designed scientific experiments that identify cause-effect relationships for the development of salt marshes.” We suggest that including Triglochin in restoration can support the development and maintenance of desired ecosystem functions, especially in bare graded substrates and sandy soils, where N is likely to be limiting. We encourage long-term, strategic research and planning as a southern California “insurance policy” for salt marsh restoration.