Mechanisms of exclusion of native coastal marsh plants by an invasive grass

Authors


*Present address and correspondence: Todd E. Minchinton, Institute for Conservation Biology and School of Biological Sciences, University of Wollongong, NSW 2522, Australia (tel. + 61 (2) 4221 5188; fax + 61 (2) 4221 4135; e-mail tminch@uow.edu.au).

Summary

  • 1Determining the mechanisms by which invasive species exclude natives is critical for conserving and restoring native populations in impacted habitats. In recent decades the grass Phragmites australis has been aggressively invading coastal marshes of North America, with monocultures often replacing diverse assemblages of plants.
  • 2Our objective was to quantify how P. australis modifies the abiotic (soil and light conditions) and biotic (litter and shoots) environment and to determine the mechanisms by which it excludes two common forbs, the annual chenopod Atriplex patula var. hastata and the perennial aster Solidago sempervirens, from the highest tidal elevations of a brackish marsh in southern New England, USA.
  • 3In a 3-year field experiment we added seeds of both forb species to stands of P. australis, where we manipulated shoots and litter in an orthogonal design, and to uninvaded marsh areas dominated by the rush Juncus gerardi, where we manipulated the shoots of the marsh vegetation. In general, seedling establishment and the number of plants surviving until the end of the growing season were substantially greater in areas not invaded by P. australis, and both shoots and litter limited the abundance of forbs within stands.
  • 4Forbs surviving within stands of P. australis grew larger and produced more seeds than those in uninvaded areas, indicating that changes to the soil resulting from invasion do not preclude the survival of established forbs. This was confirmed by a glasshouse study where the performance of forbs in soil collected from within stands of P. australis was better than in soil from areas dominated by J. gerardi.
  • 5Similar to many invasive grasses in terrestrial communities, P. australis excludes native forbs through competition, modifying the biotic environment of the marsh at both the ground (litter) and above-ground (shoots) levels. Our results suggest that successful invaders, such as P. australis, are likely to be the ones that can engineer habitats in multiple ways and limit populations of native species across several critical stages of their life history.

Introduction

The increasing spread of invasive species is dramatically changing ecological communities across coastal estuarine and marine landscapes (Carlton 1989; Ruiz et al. 2000; Bertness et al. 2002; Grosholz 2002). Investigations quantifying the patterns of spread and impacts of species invasions have revealed that invaders may alter the abiotic and biotic environment, ecological interactions among species and the functioning of ecosystems, frequently resulting in the exclusion of native species (D’Antonio & Vitousek 1992; Mack et al. 2000; Minchinton & Bertness 2003). Less well-understood are the mechanisms by which invaders exclude native populations and the stage of life history at which these mechanisms limit the demographic processes of the native species (Byers et al. 2002). Experimental studies to determine the nature of ecological interactions between invasive and native species are necessary for conserving and restoring native species in impacted habitats (Parker et al. 1999; Byers & Goldwasser 2001).

Some of the best insights into the manner by which invasive species exclude natives come from studies in terrestrial grassland communities (e.g. Kolb et al. 2002; Corbin & D’Antonio 2004; Milton 2004). Changes to environmental conditions often give invasive grasses in particular an advantage that leads to the exclusion of natives through competition for key limiting above- and below-ground resources such as nutrients, space and light (Daehler 2003; Levine et al. 2003; Vilà & Weiner 2004). Grasses have the potential to be similarly invasive in coastal estuaries (e.g. Kuhn & Zedler 1997; Minchinton 2002a,b; Minchinton & Bertness 2003; Davis et al. 2004), but relatively little is known about how they exclude native plants in the intertidal landscape or whether the mechanisms of exclusion are similar to those in terrestrial communities.

Over the past century the grass Phragmites australis (often called common reed and hereafter referred to as Phragmites) has been aggressively invading coastal marshes of North America, displacing the diverse assemblage of native plants and often forming vast monocultures (see reviews in Marks et al. 1994; Tiner 1997; Chambers et al. 1999; Meyerson et al. 2000; but see Ostendorp 1989 for concern over the decline of Phragmites in Europe). In southern New England, USA, Phragmites has been a minor component of the assemblage of plants along the terrestrial border of freshwater and brackish marshes for several thousands of years (Niering et al. 1977; Clark 1986; Orson et al. 1987). It is now spreading into coastal marshes from which it has historically been absent, as well as to more seaward locations within brackish and salt marshes where abiotic conditions are supposedly physiologically stressful (Amsberry et al. 2000). This has been attributed to the recent cryptic invasion by a non-native genotype of Phragmites that apparently has a much broader tolerance of environmental conditions (Saltonstall 2002, 2003). Concurrently, anthropogenic modification of coastal marshes, particularly the clearing of vegetation along the terrestrial-marsh ecotone and increased nutrient load, appears to be accelerating the spread of Phragmites (Bertness et al. 2002; Minchinton & Bertness 2003; Silliman & Bertness 2004).

Correlative surveys in southern New England have revealed that marshes dominated by Phragmites have fewer species of plants than those without Phragmites (Keller 2000; Meyerson et al. 2000; Silliman & Bertness 2004). There is little experimental evidence, however, substantiating a causal relationship between the invasion of Phragmites and the decline in the abundance and diversity of native marsh plants (but see Burdick & Konisky 2003; Minchinton & Bertness 2003). Much of the plant diversity in coastal brackish and salt marshes of southern New England is due to a suite of halophytic forbs that typically co-occur with Phragmites along the terrestrial border of the marsh (Tiner 1987; Brewer et al. 1997; Rand 2000). Phragmites is typically the largest plant in these marshes and therefore it is simple to assume that it excludes the smaller forbs through shading. Occasionally, however, adult forbs are found within undisturbed stands of Phragmites, indicating that some species may grow under the canopy of Phragmites, and suggesting that the mechanisms of exclusion may be more complicated. Indeed, as Phragmites spreads, it not only creates a canopy over the other marsh plants, but it also engineers the habitat by increasing the accumulation of plant litter on the substratum and altering the physico-chemical conditions of the soil (e.g. Windham & Lathrop 1999; Meyerson et al. 2000). Through these abiotic and biotic modifications to the marsh, Phragmites may exclude native forbs at different stages of their life history by limiting demographic processes (i.e. dispersal and supply of seeds, establishment of seedlings, survival and reproductive output of adults) necessary for sustaining local populations.

Here we present the results of investigations designed to reveal the mechanisms by which Phragmites excludes native halophytic forbs in a coastal brackish marsh in Massachusetts, USA. We chose two species of common and abundant forbs with different life histories, the annual chenopod Atriplex patula var. hastata (hereafter referred to as Atriplex) and the perennial aster Solidago sempervirens (hereafter referred to as Solidago). First, we did a quantitative field survey to test the hypothesis that the abundance of seeds, seedlings and adults of Atriplex and Solidago is negatively related to the abundance of Phragmites. We then tested hypotheses about how changes to abiotic and biotic conditions due to the presence of Phragmites might limit the abundance of these forbs at different stages of their life history. A field experiment over three growing seasons tested whether shoots and litter of Phragmites, or their combination, limit the establishment, survival and reproductive output of the forbs. We supplied seeds of both forb species to areas in the marsh not invaded by Phragmites and also within manipulated stands of Phragmites. A glasshouse study tested the hypothesis that the composition of the soil, per se, limits establishment and survival, by placing seeds of both forb species onto blocks of soil collected from within stands of Phragmites and from uninvaded areas in the marsh.

Materials and methods

location and plants studied

The study was carried out between May 1998 and September 2000 in a brackish tidal marsh at the Adolph Rotundo Wildlife Preserve along the Palmer River in Rehoboth, Massachusetts, USA (41°47′ N, 71°16′ W) (Amsberry et al. 2000). The marsh is in a developed area, crossed by two highways, and bordered by forest, farms and residential and commercial buildings. The marsh is intersected by tidal creeks and has been ditched at regular intervals from land to sea. The plants in the marsh are typical of brackish and salt marshes of southern New England, with characteristic bands of perennial turf-forming grasses and rushes along a gradient of elevation (e.g. Niering & Warren 1980; Bertness & Ellison 1987; see Tiner 1987 for taxonomic authorities). The banks of tidal creeks and the low marsh are dominated by the grass Spartina alterniflora. The high marsh is occupied by a dense matrix of grasses and rushes, including the grass Spartina patens and the rush Juncus gerardi along its lower and upper borders, respectively, with the grass Distichlis spicata interspersed throughout (hereafter these three species are collectively referred to as marsh turf). Within this matrix, particularly at the highest elevations of the marsh, is a relatively diverse assemblage of halophytic forbs (Tiner 1987; Rand 2000). The highest elevations of the marsh, which comprise the terrestrial border of the high marsh and the levees of tidal creeks, are also dominated by Phragmites, and the shrub Iva frutescens is occasionally present.

Phragmites is a clonal grass with annual shoots and perennial rhizomes, which spreads vegetatively to form monospecific stands. Two study sites were chosen where large, solitary stands of Phragmites along the levee of tidal creeks were spreading into the high marsh dominated by marsh turf (primarily J. gerardi, but S. patens and D. spicata were occasionally present) and a suite of forbs. Stands extended tens of metres along each tidal creek (site 1, 150 m; site 2, 100 m) and tens of metres into the marsh (site 1, 36 m; site 2, 24 m). Observations indicate that both stands are at least a decade old (see Minchinton & Bertness 2003) and comprised of the non-native strain of Phragmites (see Saltonstall 2002). The two species of forbs selected, Atriplex and Solidago, are common and abundant in areas of the high marsh dominated by J. gerardi (Tiner 1987; Rand 2000). Atriplex is an annual chenopod with seeds that are typically dispersed by water and Solidago is a perennial aster with wind-dispersed seeds that can also float in water.

patterns of abundance of forbs, marsh turf and phragmites

To quantify the observed patterns of decreasing abundance of marsh turf and forbs with increasing abundance of Phragmites, surveys were done at each site along a transect from the high marsh, dominated by J. gerardi and forbs, to the levee of the tidal creek dominated by Phragmites. The transect was divided into the high marsh adjacent to the edge of the Phragmites stand and dominated by J. gerardi (zone 1, hereafter referred to as the Juncus zone), the edge of the Phragmites stand adjacent to the Juncus zone (zone 2) and the levee of the tidal creek dominated by Phragmites (zone 5, hereafter referred to as the Phragmites zone): the area remaining between zone 2 and zone 5 was divided into two equal zones (3 and 4). Therefore, zone 1 had the greatest density of Juncus and no Phragmites, zones 2–4 had progressively less Juncus and more Phragmites, and zone 5 had no Juncus and the greatest density of Phragmites (see Fig. 1a–d). Each zone extended 50 m alongshore and was 6 m wide (except that zones 3 and 4 at site 1 were 12 m wide because of the greater width of the stand at that site), and the central 3-m band within each zone was sampled.

Figure 1.

Mean (± SE) (a) density of Phragmites shoots, (b) biomass of Phragmites litter, (c) density of shoots of marsh turf, (d) biomass of shoots of marsh turf, (e) density of adult Atriplex (none found in quadrats at site 1), (f) density of adult Solidago, (g) density of Atriplex seeds, and (h) density of Solidago seeds from the Juncus zone (zone 1) to the Phragmites zone (zone 5).

In June 1998, the abundance of Phragmites litter (dead shoots and leaves on the substratum) and marsh turf was estimated separately by collecting the above-ground plant material from each of four randomly located quadrats (15 cm × 15 cm) in each zone. The shoots of each species of marsh turf were counted, and then these samples and those for Phragmites litter were dried to a constant mass at 50 °C and weighed. The numbers of recently recruited seedlings of Atriplex and Solidago were also counted in these quadrats. At the end of the first growing season (September 1998) when most plants are reproductive, the density of Phragmites shoots (live and dead), adult Atriplex and adult Solidago was measured in five randomly located quadrats (1 m × 1 m) in each zone. The number of species of plants was also counted in each zone.

patterns of seed supply of forbs

To quantify patterns of seed dispersal of forbs in relation to the abundance of Phragmites, 10 seed traps were randomly placed in each of the five zones at each site (see Rand 2000). The traps consisted of two, circular plastic plates (surface area about 314 cm2 per plate), placed back-to-back and held together with cable ties. The outer surfaces of the plates were covered with a thin layer of Tanglefoot™ insect trap coating. Traps were attached to wire stakes and positioned 10 cm above and perpendicular to the marsh substratum so that both seeds dispersed by wind at low tide and those dispersed by water at high tide were caught. Traps were placed in the marsh in August 1998 before Atriplex and Solidago had gone to seed and collected 3 months later in November 1998 after seeds had dispersed. Seeds were identified and counted in the laboratory under a dissecting microscope. For each species, a two-factor analysis of variance (anova) was used to determine differences in seed supply (transformed to natural logarithms) among zones (five zones) and between sites.

influence of soil, shoots and litter of phragmites

We determined whether Phragmites limits the abundance of forbs after the arrival of seeds by examining effects on the establishment, survival and growth of Atriplex and Solidago and the reproductive output of Atriplex.

The influence of soil

Ten blocks of soil (10 cm × 10 cm × 10 cm) were extracted from the Phragmites and Juncus zones at each site, giving a total of 40 blocks of soil. Blocks were placed individually in plastic pots in the glasshouse at Brown University, and above-ground vegetation was clipped to the level of the soil. Fifty Atriplex or 250 Solidago seeds collected from the marsh in the previous year were added to each of 20 pots, giving five replicate pots per zone per site for each species. Different numbers of seeds were used because germination success was expected to be different for each species. The pots were randomly arranged in an array and their positions were re-randomized at least once per week. Plants were maintained at ambient light and temperature and watered daily with fresh water. By removing the physico-chemical (e.g. salinity, flooding, light) and biological (e.g. above-ground vegetation) differences that would be present between zones in the field, variation specifically due to below-ground differences in the composition of the soil (including root matter) could be studied. Note that, because we did not manipulate Phragmites, differences in the soil between the Phragmites and Juncus zones may be due to factors other than the presence of Phragmites (although Phragmites is known to modify its soil environment; see Windham & Lathrop 1999; Meyerson et al. 2000); nevertheless, comparing the performance of forbs on these soils tests whether soil properties can, by themselves, exclude forbs from the Phragmites zone.

The study was carried out from August to December 1998, and the establishment and mortality of plants was calculated as the percentage of seeds that established and the percentage of established plants that died, respectively. After 3 months, most of the Atriplex plants had flowered and set seed, and at this time plants were recorded as producing seeds or not, harvested, dried to a constant mass at 50 °C and weighed. An additional month was allowed for Solidago seedlings to grow larger and, after 4 months, Solidago were similarly harvested and weighed. For each species, a two-factor anova was used to determine the effect of zone (Juncus or Phragmites zone) and site on the establishment, survival, density and biomass of Atriplex and Solidago grown in the glasshouse and on the percentage of Atriplex plants with seeds.

The influence of shoots and litter

The presence or absence of Phragmites shoots and Phragmites litter was manipulated in an orthogonal design in the Phragmites zone. Similarly, the presence or absence of shoots of the species comprising the marsh turf (primarily Juncus) was manipulated in the Juncus zone, where Atriplex and Solidago are normally found. Litter was not present in the Juncus zone and therefore the influence of this factor was not examined in the Juncus zone, resulting in differences in experimental design between zones. Ten plots were randomly located in each of the Phragmites (2 m × 3 m) and Juncus (0.5 m × 1.5 m) zones at each site and separated by at least 3 m. Half of these plots were randomly selected in each zone and the shoots of Phragmites or marsh turf were clipped as close to the substratum as possible (1–2 cm) using hedge clippers, and the other five plots were left as uncut controls. Shoots that regrew were clipped at regular intervals during the growing season.

Cages into which Atriplex and Solidago seeds would be added were placed into the middle of each plot and spaced 50 cm apart. Two replicate cages (both without litter) were added to each plot in the Juncus zone and four arranged in a square (two with litter and two without litter) to each plot in the Phragmites zone, giving a total of 40 cages in the Phragmites zone and 20 cages in the Juncus zone at each site. Phragmites litter naturally covers much of the substratum in the Phragmites zone, but in varying amounts, so the litter was removed from all cages and a standardized amount based on the means of survey estimates (17.7 g dry biomass/225 cm2 or 24.6 g per cage) was added to two randomly selected cages in each plot in the Phragmites zone.

Cages were cylindrical (26 cm tall, 314 cm2 area) and made of galvanized hardware cloth (1.3 cm mesh size) lined with row crop cover cloth, which allowed water and light to enter the cages, but did not allow seeds in or out. Inevitably, the availability of light within the cages was diminished by about 28% by the cloth (authors’ unpublished data), but the cloth was only in place for 6 weeks as seedlings established, all replicates had cloth, and it is impossible to manipulate these seeds without such cages. Each cage was divided in half with a wall of hardware cloth lined with row crop cover cloth so that Atriplex seeds could be added to one side and Solidago seeds to the other. The tops of cages were covered with lids made of white polyester organza fabric attached with elastic bands. This prevented seeds from escaping and allowed seedlings to be counted without removing the cage.

Using seeds collected from the marsh in the previous year, 100 Atriplex and 500 Solidago seeds were added to their respective halves of each cage in June 1998, and then the establishment of seedlings was monitored about every 2 weeks. Six weeks after adding the seeds, establishment had ceased and therefore the dividing walls and row cover cloth lining the cages were removed. The wire skeleton of the cages remained in place to exclude litter until plants were harvested at the end of the growing season. After Atriplex had set seed at the end of the growing season (September 1998), plants of both species were harvested. Atriplex were counted, the percentage with seeds was noted, the number of seeds per plant was counted, and then the plants were dried to a constant mass at 50 °C and weighed. Solidago were counted, harvested, dried and weighed as described for Atriplex, but no seeds were available for counting because Solidago is a perennial and did not become reproductive in its first year.

Throughout the growing season in 1998, Atriplex seeds that had been added at the beginning of the experiment could be seen on the soil within the cages and, in spring 1999, a new cohort of seedlings of both Atriplex and Solidago emerged within the exact area of the cages. It was clear that these seedlings came from seeds that had been added in the previous year (and not from naturally dispersed seeds or seeds of experimental plants) because these species of forbs exhibit extremely localized dispersal and seedlings are rare except near the parental plant (see Rand 2000), seedlings of these species were never found during surveys in this study, and experimental plants had been harvested before their seeds had dispersed. Therefore, to determine whether similar results would be obtained for this second cohort, manipulations of shoots and litter were implemented again and continued until the end of the growing season of 1999. The only difference from the previous year was that cages used to exclude litter were not lined with row cover cloth because we were not concerned about escaping seeds. Establishment and survival were monitored and, at the end of the growing season (September 1999), plants were harvested and data collected as in the previous year. We again checked for seeds and new seedlings in spring and at the end of the growing season of 2000, but none was found.

Ultimately, relatively few seedlings established or survived in some treatments (see Results). Therefore, to increase the power of statistical tests, data from both cohorts were pooled across years, which was appropriate because differences among treatments were qualitatively similar in 1998 and 1999. For each replicate, data were combined across years by summing the number of plants that established (or survived) in each of the two years. Establishment of plants was calculated as the percentage of the number of seeds added in June 1998 that established, and survival as the percentage of established plants that survived to be harvested at the end of the growing season.

Because of the necessary differences in the experimental design between zones (due to the absence of litter in the Juncus zone), two separate analyses were done to test for the effect of shoots and litter of Phragmites on the establishment and survival of forbs. In the first, a four-factor, nested anova was used to determine the effect of litter (present or removed), shoots (present or removed), plots (nested within the factors for the effects of shoots and sites), and sites in the Phragmites zone. In the second, litter was excluded as a factor, and a three-factor anova was used to determine the effect of shoots (present or removed; for shoots of Phragmites in the Phragmites zone and for shoots of marsh turf in the Juncus zone), zone (Juncus or Phragmites), and site. For this second analysis, the average values for the two cages in each plot were used as replicates.

Environmental conditions

Relevant abiotic factors that might limit the performance of forbs were measured in the presence and absence of shoots within each zone at each site, including edaphic conditions (salinity, redox potential, moisture and grain size), sedimentation rate, availability of light and relative tidal height. These measurements were taken to help explain the results by determining how the abiotic environment varied spatially across the marsh and with experimental manipulations of the vegetation. Soil porewater salinity was measured in each plot during neap tides on several occasions during the growing seasons of 1998 and 1999 and, due to the similarity of the relative differences among treatments across dates, only those from August 1998 are presented. Salinity was determined by extracting pore water from the top 2 cm of a small core of soil and measuring it using a hand-held NaCl refractometer with a precision of ±1 g kg−1. Soil redox potential, an indicator of soil oxygen availability, and soil moisture were measured in each plot about 1 week after a spring tide in October 1998. Soil oxygen availability was estimated by removing a core (1 cm diameter × 5 cm long) of soil, inserting a platinum redox electrode (filled with 4 mol L−1 KCL saturated with Ag/AgCl reference solution) into the hole, and measuring the redox potential using a portable meter (Orion pH ISE meter, model 230 A). Soil moisture was estimated by removing two cores (1.5 cm diameter × 6 cm long) of soil from each plot, wrapping them in aluminium foil to retain the water, and then weighing them before and after drying to a constant mass at 50 °C. Soil moisture was calculated as the percentage difference between the wet and dry weights. The same cores were used to compare soil grain sizes and these were pooled across plots (about 33 and 51 g dry biomass per sample in the Juncus and Phragmites zones, respectively) and wet sieved to determine percentage composition of the following grain size classes: macro-organic (> 2 mm), very coarse sand (1–2 mm), coarse sand (0.5–1 mm), medium sand (0.25–0.5 mm), fine sand (0.125–0.25 mm), very fine sand (0.063–0.125 mm), and silt and clay (< 0.063 mm).

Sedimentation rates were measured in each plot from September to December 1998. In each plot, we placed two sediment traps, consisting of 50 mL centrifuge tubes, into the ground so that their tops were level with the soil, thus recording sedimentation that would be experienced by seedlings. Tube openings were covered with nylon screen (about 1 mm mesh size) to prevent macro-invertebrates and large debris from entering. After 90 days we withdrew the sediment traps, and the sediment was dried to a constant mass at 50 °C and weighed. The availability of light was quantified between 11.00 and 14.00 on a cloudless day in August 1998 using a light meter (LI-COR solar monitor, model 1776). Two instantaneous measurements (µE m−2 s−1) were taken under and immediately above the litter in each plot, and these were expressed relative to measurements in the open marsh, which were unobstructed by marsh vegetation or litter. Light measurements were only done at site 2, but given the similarities in the densities of shoots and abundance of litter between sites, results should be similar at site 1. We determined the relative differences in elevation among zones and between sites by measuring water depth during spring high tides at 10 random locations within each zone at each site several times during the study. A three-factor anova was used to determine the effect of shoots (present or removed, for shoots of Phragmites in the Phragmites zone and for shoots of marsh turf in the Juncus zone), zone (Juncus or Phragmites zone) and site on soil salinity, soil redox potential, soil moisture and sedimentation rate. Grain size distributions and light availability were not analysed statistically.

statistical analyses

For all statistical analyses, site and plot were considered random factors and zone, shoots and litter were considered fixed factors. Where appropriate, data were transformed to their natural logarithms or to the arcsine of their square roots before analysis. Student-Newman-Keuls (SNK) multiple comparisons tests were used to locate significant differences among treatment means when there were significant interactions in the anova (at P < 0.05 for both SNK tests and anova), and the results of these tests are simply described in the text.

Results

patterns of abundance of forbs, marsh turf and phragmites

There was a gradual and substantial increase in the density of Phragmites shoots from the Juncus zone in the high marsh to the Phragmites zone at the edge of the tidal creek (Fig. 1a). As shoot density increased, so did the biomass of Phragmites litter on the substratum, with a particularly pronounced increase in the Phragmites zone (Fig. 1b). In contrast, the density and biomass of shoots of marsh turf (> 75%Juncus by dry biomass, but S. patens and D. spicata were also present) decreased gradually from the Juncus to the Phragmites zone (Fig. 1c,d). The density of adult Atriplex (which was present but not recorded in quadrats at site 1) and adult Solidago declined more abruptly, with few individuals extending more than 6 m into the stand of Phragmites (Fig. 1e,f). Moreover, no seedlings of Atriplex or Solidago were found in any of the quadrats sampled. The number of species of plants declined from the Juncus to the Phragmites zone at both sites, with none of the 10 species recorded in the Juncus zone (Atriplex, D. spicata, I. frutescens, Juncus, Potentilla anserina, Scirpus spp., Solidago, S. patens, Typha angustifolia, Triglochin maritimum) present in the Phragmites zone.

patterns of seed supply of forbs

An order of magnitude more Solidago seeds were trapped than Atriplex seeds, but seed supply for both species roughly paralleled the abundance of adults and declined precipitously from the Juncus zone to the Phragmites zone (Fig. 1g,h). For both species, differences in seed supply between sites varied among zones (Atriplex, F4,90 = 8.2, P < 0.001; Solidago, F4,90 = 15.8, P < 0.001; results for interaction between zone and site in anova). In the Juncus zone, significantly fewer Atriplex seeds were caught at site 1, where adults were uncommon, than at site 2, and no Atriplex seeds were trapped beyond zone 2, which extended only 6 m into the stand of Phragmites (Fig. 1g, SNK tests). Densities of Solidago seeds in zones 1 and 2 were significantly greater at site 1 than at site 2, reflecting adult abundances, whereas densities of seeds of Solidago in other zones were significantly greater at site 2 than at site 1 (Fig. 1h, SNK tests). The spike in the density of Solidago seeds in the Phragmites zone at site 2 (Fig. 1h) probably occurred through the supply of seeds from plants located only a few metres away on the other side of the tidal creek. This result is important because it attests to the efficacy of the traps, with large numbers of seeds caught even in areas with great densities of Phragmites shoots.

influence of soil, shoots and litter of phragmites

Environmental conditions

There was on average less than a 1 cm difference in elevation between the Juncus zone at site 1, the Phragmites zone at site 1, and the Phragmites zone at site 2. The Juncus zone at site 2 was, however, 6 cm lower than these other three zones. Differences in soil porewater salinity among treatments were relatively small (Fig. 2). In the presence of shoots, salinity was similar between zones, whereas salinity in the Juncus zone was significantly greater than in the Phragmites zone where shoots had been removed (Fig. 2, Table 1, SNK results for interaction between zone and shoots). Salinity in the Juncus zone was significantly greater where shoots had been removed than where they were present, whereas the reverse was true in the Phragmites zone (Fig. 2, Table 1, SNK results for interaction between zone and shoots). Soil redox potential was significantly greater in the Juncus zone than in the Phragmites zone at site 1, but the reverse was true at site 2, where the low tidal elevation in the Juncus zone corresponded with a low redox potential (Fig. 2, Table 1, SNK results for interaction between site and zone). Although soil moisture was significantly greater in the Juncus zone than in the Phragmites zone and at site 1 than at site 2, these differences among treatments were relatively small (Fig. 2, Table 1). At both sites, sediments accumulated at a significantly faster rate in the Phragmites zone than in the Juncus zone, and this difference was more pronounced at site 1 (Fig. 2, Table 1, SNK results for interaction between site and zone). There was no effect of removing shoots on soil redox potential, soil moisture or sediment accumulation (Fig. 2, Table 1). The primary difference in the composition of the soil was the macro-organic component (roots and rhizomes), whose mean across sites was about four times greater in the Juncus zone (18.6% of soil mass) than in the Phragmites zone (4.5% of soil mass) (Table 2). The availability of light in plots without shoots and litter in the Phragmites zone or without shoots of marsh turf in the Juncus zone was 91% (91.1 ± 2.7% light; mean ± SE, n = 5) and 98% (97.9 ± 1.3% light; mean ± SE, n = 5) of that measured under full sunlight in the open marsh, respectively. In contrast, the presence of shoots reduced light availability to only 2% of full sunlight in both the Phragmites (1.6 ± 0.6% light; mean ± SE, n = 5) and Juncus zones (1.7 ± 0.8% light; mean ± SE, n = 5). Similarly, litter diminished light in the Phragmites zone to less than 3% of that available in the open marsh, regardless of the presence (0.4 ± 0.3% light; mean ± SE, n = 5) or absence (2.5 ± 0.7% light; mean ± SE, n = 5) of shoots.

Figure 2.

Mean (± SE) soil porewater salinity (g salt kg−1 seawater), soil redox potential, percentage soil moisture and sedimentation rate in the presence or absence of shoots in the Juncus (J) and Phragmites (P) zones at each of two sites.

Table 1.  Analyses of soil salinity, soil redox potential, percentage soil moisture and biomass of sediment accumulated in the presence or absence of shoots in the Juncus and Phragmites zones at each of two sites (see data in Fig. 2). Results are estimates of mean squares (MS) and probability levels (P) of analysis of variance. Data were not transformed
Sourced.f.SalinityRedoxMoistureSediment
MSPMSPMSPMSP
  • *

    To increase the power of the test, the estimate of MS used in the denominator of the F-ratio is a pooled estimate from the MS of the S × Z × Sh interaction and the residual, and then the effect of Z × Sh was tested with 1 and 33 d.f. (see Winer et al. 1991 for pooling procedures).

Site: S 1  9.0  0.17318 727< 0.00144.90.0486 5570.053
Zone: Z 1 67.6 0.4104 520  0.8171509.20.04668 1410.194
Shoot: Sh 1  4.2  0.2752 481  0.05639.80.3861 0240.376
S × Z 1 38.0  0.00751 883< 0.0018.00.3926 7670.049
S × Sh 1  0.9  0.66320  0.90519.20.1894600.598
Z × Sh 1126.0< 0.001*491  0.70232.40.3042 1240.035
S × Z × Sh 1  1.6  0.5611 925  0.2438.60.37460.950
Residual32  4.6 1 363 10.6 1 624 
Table 2.  Percentage frequency distribution of grain sizes of the soil in the Juncus and Phragmites zones at each of two sites
Particle typeGrain size (mm)Site 1Site 2
JuncusPhragmitesJuncusPhragmites
Macro-organic     > 221.4 5.315.7 3.7
Very coarse sand        1–2 7.8 3.3 2.7 0.7
Coarse sand    0.5–1 6.7 8.9 7.510.3
Medium sand   0.25–0.5 8.4 6.1 7.212.7
Fine sand  0.125–0.2511.310.412.018.1
Very fine sand  0.063–0.12511.412.2 9.610.6
Silt and clay< 0.06333.053.845.343.9

The influence of soil

In the glasshouse, seedlings of Atriplex and Solidago established and grew successfully in both Juncus and Phragmites soil. There was no significant influence of soil type on percentage establishment for either species (Fig. 3, Table 3). Mortality for both species was minimal, except for Solidago seedlings in Juncus soil from site 1, which had significantly greater mortality than seedlings in the other treatments (Fig. 3, Table 3, SNK results for interaction between site and zone). The density of plants at the end of the experiment did not vary significantly among soil types for either species (Fig. 3, Table 3). In contrast, the total biomass of plants per pot, which integrates both the density and individual biomass of plants, was significantly greater in Phragmites than in Juncus soil for Solidago, with a similar non-significant pattern for Atriplex (Fig. 3, Table 3). The percentage of Atriplex plants grown in Phragmites soil that produced seeds was more than double and significantly greater than that for plants grown in Juncus soil (Fig. 3, Table 3).

Figure 3.

Mean (± SE) percentage establishment, percentage mortality, density, biomass, and percentage of plants with seeds (Atriplex only; ‘*’ indicates not measured), for Atriplex and Solidago grown in the glasshouse in soil from either the Juncus or Phragmites zone at each site.

Table 3.  Analyses of percentage establishment, percentage mortality, density, biomass, and percentage of plants with seeds (Atriplex only), for Atriplex and Solidago grown in the glasshouse in soil from the Juncus or Phragmites zones at two sites (see data in Fig. 3). Results are estimates of mean squares (MS) and probability levels (P) of analysis of variance. Density and biomass were transformed to their natural logarithms and proportions to the arcsine of their square roots
Source (d.f.)Site: S (d.f. = 1)Zone: Z (d.f. = 1)S × Z (d.f. = 1)Residual (d.f. = 16)
Species and variableMSPMSPMSPMS
Atriplex
 Establishment0.0270.3010.0250.3530.0090.5350.023
 Mortality0.0250.4640.1840.1620.0120.6040.044
 Density0.3470.4320.6250.2800.1380.6180.534
 Biomass2.3860.0280.8420.4030.4550.3080.410
 Percentage with seeds0.0130.7491.4520.0350.0040.8540.125
Solidago
 Establishment0.00010.9060.00420.4520.00310.4280.0047
 Mortality0.4220.0290.5870.4480.4220.0290.074
 Density0.5720.3660.0840.7060.3400.4830.660
 Biomass0.5200.5782.9230.0160.0020.9731.615

The influence of shoots and litter

On average, establishment of seedlings was greater for Atriplex than for Solidago, at site 1 than at site 2, in the Juncus zone than in the Phragmites zone, and where shoots or litter had been removed than where they were present (Fig. 4). In the Phragmites zone, litter dramatically and significantly reduced the establishment of Atriplex and Solidago, and this result was consistent in the presence or absence of shoots and at both sites (Fig. 4, Table 4, SNK results for interaction between shoots and litter). In contrast, at both sites the effect of removing Phragmites shoots on the establishment of Atriplex and Solidago was dependent on the presence or absence of litter (Fig. 4, Table 4). In the presence of litter, there was no effect of removing Phragmites shoots on the establishment of forbs, whereas in the absence of litter, the removal of shoots significantly increased the establishment of both species (Fig. 4, Table 4, SNK results for interaction between shoots and litter). For both species, establishment in the Juncus zone was not significantly affected by removing shoots (Fig. 4, Table 5, SNK results for interaction between zone and shoots for Atriplex). For both Atriplex and Solidago, establishment at both sites was significantly greater in the Juncus zone than in areas without litter in the Phragmites zone (Fig. 4, Table 5, SNK results for interaction between zone and shoots for Atriplex and site and zone for Solidago).

Figure 4.

Mean (± SE) percentage establishment and percentage survival of Atriplex and Solidago in the presence or absence of shoots and litter (L) in the Phragmites zone and the presence or absence of shoots in the Juncus zone at each of two sites.

Table 4.  Analyses of percentage establishment and percentage survival of Atriplex and Solidago in the presence or absence of shoots and litter in each of five plots in the Phragmites zone at each of two sites. Results are estimates of mean squares (MS) and probability levels (P) of analysis of variance. Data were not transformed
Sourced.f.Establishment (%)Survival (%)
AtriplexSolidagoAtriplexSolidago
MSPMSPMSPMSP
  • *

    To increase the power of the test, the estimate of MS used in the denominator of the F-ratio is a pooled estimate from the MS of S × Sh × L, P(S × Sh) × L, and the residual, and then the effect of Sh × L was tested with 1 and 47 d.f. (see Winer et al. 1991 for pooling procedures).

Site: S 1 1740.143 2.240.186338.3 0.32275.90.153
Shoots: Sh 1 2110.35119.400.149 41.4 0.414 1.50.510
S × Sh 1  800.311 1.100.347 23.9 0.789 1.50.833
Plot(S × Sh): P(S × Sh)16  730.016 1.180.012323.9 0.06433.70.775
Litter: L 130750.10923.980.072468.5 0.53322.70.519
P(S × Sh) × L16  180.888 0.460.534338.5 0.05240.10.646
S × L 1  920.038 0.310.420577.6 0.21025.60.436
Sh × L 1 3120.004* 3.120.040  0.8< 0.001 0.00.445
S × Sh × L 1   50.606 0.010.871  0.0 1.000 0.01.000
Residual   32  0.49 174.7 48.4 
d.f. in residual   40    40    34  31 
Table 5.  Analyses of percentage establishment and percentage survival of Atriplex and Solidago in the presence or absence of shoots (and in the absence of litter in the Phragmites zone) in the Juncus and Phragmites zones at each of two sites. Results are estimates of mean squares (MS) and probability levels (P) of analysis of variance. Data were not transformed
Sourced.f.Establishment (%)Survival (%)
AtriplexSolidagoAtriplexSolidago
MSPMSPMSPMSP
  • *

    To increase the power of the test, the estimate of MS used in the denominator of the F-ratio is a pooled estimate from the MS of the S × Z × Sh interaction and the residual, and then the effect of Z × Sh was tested with 1 and 33 d.f. (see Winer et al. 1991 for pooling procedures).

Site: S16160.010 59.80.0182.60.8631630.022
Zone: Z134970.176243.50.32916.40.8492220.493
Shoots: Sh19310.069 22.60.05713.00.8892340.405
S × Z12810.074 78.70.008278.60.0832130.010
S × Sh1110.717  0.20.892419.80.0361280.041
Z × Sh13480.048*  3.30.452219.70.1421810.493
S × Z × Sh1900.303  2.50.61911.30.7211730.019
Residual 82   9.7 86.9 28 
d.f. in residual 32     32  31  31 

The overall survival of forbs was low, with only 155 plants (1.29% of seeds) surviving in the Juncus zone and 25 (0.042% of seeds) in the Phragmites zone over both years (Fig. 4, Table 6). In the Phragmites zone at both sites, the presence of shoots significantly reduced the survival of Atriplex in the absence of litter, but there were no effects of shoots on Atriplex in the presence of litter (Fig. 4, Table 4, SNK results for interaction between shoots and litter). Similarly, in the absence of litter, removing shoots increased the survival of Atriplex in both the Phragmites and Juncus zones at both sites, but this effect was only significant at site 1 (Fig. 4, Table 5, SNK results for interaction between site and shoots). Given the small number of Solidago surviving in the Phragmites zone, it is not surprising that there were no statistical differences in survival among any of the treatments (Fig. 4, Table 4). Survival of Solidago in the Juncus zone at site 1 was significantly greater where shoots had been removed than where they were present but, as for Atriplex, this effect was not significant at site 2 where few plants survived (Fig. 4, Table 5, SNK results for interaction between site, zone and shoots).

Table 6.  Total number of Atriplex and Solidago (and total number with seeds) surviving until the end of the growing season in the presence (+L) or absence (–L) of litter and presence (+Sh) or absence (–Sh) of shoots in the Juncus and Phragmites zones at each of two sites. Mean (± SD) dry biomass (mg) per plant, total dry biomass for all plants (mg), mean (± SD) number of seeds per plant for plants producing seeds, and total number of seeds produced by all plants for Atriplex and Solidago surviving until the end of the growing season in both years at both sites
Forb and zoneTreatmentPlantsBiomass (mg)Seeds
LitterShootsSite 1Site 2TotalTotal with seedsMeanSDTotalMeanSDTotal
Atriplex
 Phragmites+L+Sh 02  2 220718504 14217857356
+L–Sh 54  9 23642843 2776511
–L+Sh 10  1 1460 46042 42
–L–Sh 82 10 5683815 62568 38389210204461
 Juncus +Sh 10  1 024 24   
–Sh442 461628201 27175112
Solidago
 Phragmites+L+Sh 00  0       
+L–Sh 01  1 1 1   
–L+Sh 01  1 40 40   
–L–Sh 01  1 79 79   
 Juncus +Sh 40  4 3211   
–Sh986104 18171 856   

There was substantial variation in the size and reproductive output of individual plants (Table 6). Similar to the glasshouse study, however, Atriplex were significantly larger and produced significantly more seeds in the Phragmites zone than in the Juncus zone (Table 6, independent t-test using plants producing seeds for all treatments in the Phragmites zone and, to balance the design, for 10 randomly selected plants from the Juncus zone: biomass, t = 4.8, P < 0.001, n = 10; seeds, t = 3.0, P = 0.008, n = 10; all data were ln-transformed). The largest plants for either species occurred where both litter and shoots had been removed in the Phragmites zone. For Atriplex, the largest individual had 5980 leaves, a dry biomass of 50.4 g, and produced 2541 seeds.

Discussion

Abundances of the two study forbs and marsh turf generally decreased as the abundance of shoots and litter of Phragmites increased within the marsh, suggesting that Phragmites is displacing forbs and marsh vegetation. Investigations demonstrated that Phragmites limits the abundance of Atriplex and Solidago through competitive interactions that affect fundamental demographic processes at multiple life-history stages of the forbs. The supply of seeds and establishment of seedlings were dramatically reduced within stands of Phragmites compared with areas of the high marsh dominated by Juncus, but the post-recruitment environment within stands of Phragmites imposes the most severe constraints on these forbs. The influence of litter and shoots of Phragmites, rather than changes to the soil associated with the invasion of Phragmites, appear to be the dominant means by which Phragmites competitively excludes forbs from this coastal brackish marsh. Both species of forbs followed the same patterns of abundance and responded in the same way to manipulations, suggesting that the results here may be generally applicable to the suite of forbs in coastal marshes threatened by this invader.

dispersal limitation

The patterns of dispersal and seed supply paralleled those of the adult forbs, with a dramatic reduction in the number of seeds within stands of Phragmites only metres away from source populations of Atriplex and Solidago. The decline in the density of trapped seeds from the Juncus to the Phragmites zone might reflect the increasing distance to source populations of forbs or an enhanced physical barrier to dispersal as Phragmites shoots became more dense from the edge to the back of the stands. Extremely localized dispersal of seeds is a more likely explanation because another study examining the dispersal of the same species of forbs in a salt marsh unobstructed by Phragmites also found that the majority of seeds dispersed very short distances (Rand 2000). Moreover, we observed at one site that seeds from nearby Solidago only metres away on the other side of the tidal creek could penetrate the densest part of the stand. Regardless of the exact explanation, these patterns indicate that, as Phragmites stands increase in area and exclude source populations of adult forbs, dispersal and seed supply of forbs into these stands will become increasingly limited. The maintenance of local populations of forbs through a seed bank is unlikely. We observed the potential for a limited seed bank as forbs emerged in the year following seed addition, but after 2 years seeds were exhausted (see also Rand 2000). Moreover, apart from where we added seeds, seedlings were not found within stands of Phragmites. Dispersal or seed limitation is thus an important process precluding the recruitment of forbs within stands of Phragmites.

establishment limitation

Supplying seeds dramatically increased the density of seedlings of Atriplex and Solidago within stands of Phragmites, substantiating the hypothesis that dispersal can limit forb abundance. Forbs seeded in Phragmites and Juncus soil under benign conditions of freshwater and abundant sunlight in the glasshouse established with similar success, demonstrating that the soil per se (i.e. finer sediments, reduced macro-organic component) within stands of Phragmites does not directly limit recruitment of Atriplex or Solidago. In contrast to the glassouse results, when we removed the above-ground biotic influences of litter and shoots in the field, Atriplex and Solidago established more successfully in the Juncus than in the Phragmites zone. This suggests that the physico-chemical conditions of the soil associated with tidal influences within stands of Phragmites limit the establishment of forbs compared with Juncus-dominated areas of the marsh not invaded by Phragmites. Rand (2000) found that these same species of forbs establish more successfully at higher than at lower tidal elevations of the marsh, where there is greater soil oxygen availability (see also Windham & Lathrop 1999). Redox potentials and establishment success of both species were greater in the Juncus zone at site 1 than in the Phragmites zone at site 1 and the Juncus zone at site 2, which is consistent with the results of Rand (2000). In contrast, the Juncus zone at site 2 was at a lower tidal elevation and had lower redox potentials than the other zones, but seedlings of both forbs still established more successfully in this zone than in both the Phragmites zones. Oxygen availability alone therefore cannot account for all the differences in establishment success of forbs between the Juncus and Phragmites zones, but it is probably one of several important edaphic conditions. Sedimentation rate was substantially greater in the Phragmites than in the Juncus zone, which is also a possible explanation for the reduced establishment of forbs within stands of Phragmites.

Within stands of Phragmites, shoots and, particularly, litter dramatically reduced the establishment of both forb species, with their combination having the greatest negative effect. Litter reduced the transmittance of light, but only to the same degree as the presence of shoots in the absence of litter, indicating that its influence on establishment is not only through shading. The negative impact of litter on the establishment of plants is multifaceted (see Facelli & Pickett 1991; Minchinton 2002a), but it is likely that litter suspended by the tides simply rips out or smothers fragile, recently established seedlings.

Floating plant litter, or wrack, is common in these coastal marshes (primarily the grass S. alterniflora, but also Phragmites), and forbs commonly find refuge in bare patches that are generated where wrack stranded in the open marsh smothers native marsh turf, but is subsequently removed by the tides (Bertness & Ellison 1987; Ellison 1987; Brewer et al. 1997; Minchinton 2002a). The residency time of litter within stands of Phragmites appears to be much longer than that which typically produces bare patches in the open marsh, and few plants establish in the marsh where wrack remains in place (e.g. Minchinton 2002a). Within stands, the shoots of Phragmites, which die annually but may remain upright for several years, trap the litter and, consequently, much of it remains there until it decomposes. Therefore, although seasonal wrack deposition in the open marsh may have a positive effect on forbs by providing habitat refuges in bare patches, the continual presence of litter within stands of Phragmites has a negative impact on the colonization of forbs similar to the chronic stranding of wrack in the open marsh (Minchinton 2002a). The supply and retention of a continual cover of litter on the substratum is likely to be a primary mechanism by which Phragmites excludes forbs.

In the absence of litter, the presence of shoots within stands of Phragmites had a negative influence on the establishment of forbs. This effect of shoots on establishment was not as evident where shoots had been removed in the Juncus zone, even though there was a similar reduction in the availability of light (see also Rand 2000). Therefore, environmental conditions during establishment, such as sedimentation and the presence of litter and shoots, place another constraint on the recruitment of forbs after the arrival of seeds within Phragmites stands.

survival limitation

Although the dispersal of seeds and establishment of seedlings limited the abundance of forbs growing within stands of Phragmites, further severe constraints on survival occurred once seedlings had become established, and few seedlings survived to the end of the growing season. In the absence of litter, mortality of established forbs between zones and sites was linked to soil oxygen availability, with the waterlogged Juncus zone at site 2 showing the greatest mortality of seedlings and the lowest redox potentials. Similarly, Rand (2000) found increasing mortality of these forbs with decreasing tidal elevation and redox levels. The Phragmites and Juncus zones at site 1 were at similar tidal elevations, yet the Phragmites zone had lower redox potentials, indicating that Phragmites may reduce the survival of forbs through limitations on soil oxygen availability. More studies measuring soil redox potentials in different marsh vegetation types at equivalent tidal levels are needed, however, to validate this conclusion, which is based on a single comparison between two zones.

Although litter impeded the establishment of seedlings in stands of Phragmites, once forbs became established they appeared to survive equally well in areas with or without litter. Because so few individuals survived, however, these results should be viewed tentatively. One clear effect on survival was that seedlings of both species survived better where shoots of Phragmites (in the absence of litter) or marsh turf had been removed (see also Rand 2000). Similar to the establishment of seedlings, shoots of Phragmites and Juncus probably reduce survivorship by shading the forbs. Interestingly, Atriplex within stands of Phragmites was able to survive and reproduce in areas with litter and shoots, highlighting the importance of microhabitats as temporary refuges within stands of Phragmites.

Despite fewer forbs surviving within stands of Phragmites compared with the Juncus zone, individuals of both species grew substantially larger and, for Atriplex, more than twice as many individuals became reproductive, and these produced on average two orders of magnitude more seeds. For both species, individuals grew largest and produced the most seeds in the absence of litter and competing shoots. These results are identical to those in the glasshouse, providing further evidence that Phragmites soil might be better for the growth of forbs than Juncus soil. Either Phragmites soil contains more nutrient resources or the lower biomass of roots and rhizomes reduces below-ground competition for limiting resources in the immediate environment.

invasion of phragmites and the exclusion of coastal marsh plants

The field experiments of Minchinton & Bertness (2003) have documented that Phragmites can out-compete the dominant grasses and rushes comprising the marsh turf in only a couple of years (see also Burdick & Konisky 2003). This study has demonstrated that Phragmites is also a superior competitor to two species of forbs and that the continued invasion of Phragmites is likely to result in the exclusion of local populations from coastal brackish and salt marshes of southern New England, USA, with the possibility of local extinctions of vulnerable species. As for invasive plant species in terrestrial communities (Levine et al. 2003; Vilà & Weiner 2004), exclusion of coastal marsh forbs by Phragmites appears to occur through competitive processes, particularly the impacts of shoots and their eventual breakdown into litter. The performance of invasive plants in terrestrial habitats is typically increased under disturbed conditions, enhancing their competitive advantage over natives (Daehler 2003). With increased production of Phragmites under elevated nutrient loads (see Minchinton & Bertness 2003) and the continued urbanization of estuaries, Phragmites is expected to become a stronger competitor, displacing resident marsh plants at an accelerated rate.

Although it may be simple to assume that large and dominant invasive species, such as Phragmites, competitively exclude smaller plants by overgrowing them, results here demonstrate that this is only one mechanism. Production and retention of Phragmites litter and, perhaps, increased sedimentation may limit the abundance of forbs even in the absence of shoots. Importantly, Phragmites is a particularly successful invader because it engineers the marsh habitat, modifying both the abiotic and biotic environments in diverse ways and reducing forb populations by limiting demographic processes across multiple stages of their life history. Understanding the key mechanisms by which invasive species exclude natives and determining the life-history stage most sensitive to invasion will allow managers to target their conservation and restoration efforts in the control of invasive species.

Acknowledgements

The authors gratefully acknowledge the formidable field efforts of Erin Siska and the advice of Tatyana Rand. We also thank Fred Jackson, Ben Pister and Liz Selig for help in the field, laboratory and glasshouse. Research was supported by funds from the Andrew Mellon Foundation, the Natural Sciences and Engineering Research Council of Canada, and the Rhode Island Sea Grant College Program.

Ancillary