Environmental determinants of Phragmites australis expansion in a New Jersey salt marsh: an experimental approach

OIKOS 89: 59–69. Copenhagen 2000

Environmental determinants of Phragmites australis expansion in a New Jersey salt marsh: an experimental approach David Bart and Jean Marie Hartman

Bart, D. and Hartman, J. M. 2000. Environmental determinants of Phragmites australis expansion in a New Jersey salt marsh: an experimental approach. – Oikos 89: 59–69. Interdependence among disturbance events, ecosystem properties, and biological invasions often make causal relationships difficult to discern. For example, Phragmites australis invasion in mid-Atlantic salt marshes is often associated with disturbances that create well-drained features as well as with low sulfide concentrations, but explanations of these associations have been elusive. We tested experimentally: 1) that disturbances increasing wetland drainage facilitate Phragmites invasion by altering sulfide concentrations and salinity; 2) that translocation allows plants to spread beyond drainage areas; and 3) that plants can then lower edaphic stress through pressure ventilation of the rhizosphere and promote further expansion. At the invasion front, treatments of 1) severing rhizomes to halt translocation and 2) combined severing with clipping dead culms to limit ventilation of the rhizosphere killed most culms, but did not affect pore water chemistry. In already invaded areas, severing and clipping reduced culm height and panicle production, severing alone and in combination with clipping also raised sulfide and ammonium concentrations in the root zone. There were no treatment effects on plant performance or pore water chemistry along mosquito ditches, where sulfide concentrations were negligible. Small-scale hydrological alterations such as ditches appear to provide suitable sites for the establishment of Phragmites because soils are well-drained and are low in free sulfides. Subsequent expansion into more hostile areas occurs through translocation, with well-drained areas acting as sources for essential substances. Once established, the plant increases rhizosphere oxygenation and lowers sulfide concentrations. D. Bart and J. M. Hartman, Graduate Program in Ecology and E6olution, Rutgers, The State Uni6. of New Jersey, 113 Blake Hall, 93 Lipman Dri6e, New Brunswick, NJ 08901 -8524, USA ([email protected]).

Ecological literature on the relationship between invasive species, ecosystem properties, and disturbance tends to focus on three themes. The first elucidates the characteristics that make certain ecosystems more invasible than others. Such studies suggest that more fertile systems (Bridgewater and Backshall 1981, Vitousek 1990), and those that are more prone to natural breaks in plant cover are more invasible, while systems that have stressful edaphic conditions often resist invasion (Baker 1986). The second common theme suggests that anthropogenic disturbance often increases the likelihood of invasions. This assertion is not new, and since Elton’s (1958) classic work has been a major focus in

the study on invasives (Vitousek 1990). These studies have shown that anthropogenic disturbance provides mechanisms by which the invaders are dispersed (Mack 1986, Bart 1997), and can alter ecosystem characteristics such as trophic level interactions, resource availability (Orians 1986), and edaphic conditions (Ewel 1986) in ways that facilitate the establishment and spread of invasives. Finally, a relatively new focus has been on the effects the invaders have on ecosystem properties and processes. Examples document the effects of invasions of nitrogen fixers on nitrogen poor soils, and the effects of invaders on natural disturbance frequency and intensity (Vitousek 1986, 1990).

Accepted 6 September 1999 Copyright © OIKOS 2000 ISSN 0030-1299 Printed in Ireland – all rights reserved OIKOS 89:1 (2000)

59

Since the above-mentioned studies suggest that disturbances, ecosystem properties, and invasive species can be interdependent, interpreting the relevance of ecosystem-level data and disturbances in invasion ecology can be challenging. Because ecosystem-altering disturbances can also be associated with invasions, ecosystem properties associated with an invasion could be either a function of the disturbance that facilitated the invasion or the result of the invasion itself (Vitousek 1986). Similarly, a negative association between an invader and edaphic stress could also raise the question of whether the absence of edaphic stress is a measure of the invasibility of a site, the effects of the invader, or both (Chambers 1997). These conceptual challenges associated with interpreting ecosystem-level data in invasion ecology make explaining invasions of particular species in a given system problematic. An explanation of a single invasion involves the formation of a narrative history (Mack 1986), where the events and mechanisms that lead to the establishment and subsequent spread of an invader are documented and supported with evidence (Bart 1997). Historical accounts can provide associations between disturbance events and an invasion (Elton 1958, Mack 1986), but it is impossible to reject spurious relationships unless some mechanism can be demonstrated to link the disturbance to the advance of the invasive (Bart 1997). Since one mechanism that could connect disturbances to invasions is an alteration in ecosystem properties, it can be crucial to understand causal connections between the disturbance, the ecosystem property, and the advance of the invasive. The invasion of Phragmites australis (Cav) Trin. ex Steudel (hereafter referred to as Phragmites) into salt and brackish water marshes of the mid-Atlantic coast of the United States illustrates both the challenges and importance of interpreting this relationship. Phragmites is a native grass in the United States, where it usually inhabits wet meadows, riversides, and fresh water marshes (Marks et al. 1994). Sediment core analysis indicates that it was also a historical constituent of some salt marshes (Niering and Warren 1977), although other historical studies suggest that some marshes lacked the plant (Bart 1997). When present in salt marshes, Phragmites formed monospecific stands in the back marsh areas where salinity was low. As far back as the turn of the century, some salt marshes in New England and the mid-Atlantic coast of the United States experienced rapid and dramatic Phragmites expansion. While it is unclear whether the invasion began from new stands or from sudden expansion of old ones (Bart 1997), vigorous rhizomatous growth led to expanding ‘‘fronts’’. Behind these fronts, the tall culms of the plant would outshade most salt marsh plant species, at times creating monotypic stands of the grass. As it invaded, Phragmites altered the ability of marshes to support wildlife (Roman et al. 60

1984), decreased agricultural value of salt hay farms (Bart 1997), and increased the likelihood of large fires after the plant senesces (Hellings and Gallagher 1992). Phragmites invasion into salt marshes is interesting because the plant is spreading into an environment of reduced soils that exhibit high concentrations of salt and sulfides, all of which can be harmful to plants that are not adapted to them (Chalmers 1982, Pearson and Havill 1988). Explaining the invasion has usually been construed as a function of overcoming the adverse conditions of salt marshes (Roman et al. 1984; see Windham 1995, Bart 1997, Chambers 1997). Chambers (1997) found little correlation between areas that have been invaded and pore water salinity, but did find an association between invaded areas and low sulfide concentrations. Nevertheless, he was unsure whether this was a consequence of the invasion, the reason why the area was invaded, or both, because the higher productivity associated with Phragmites invasion can increase soil oxygenation (cf. Howes et al. 1986) and therefore inhibit sulfate reduction to sulfides. Furthermore, Phragmites can pressure ventilate its rhizosphere especially when outflow of gasses is increased though dead culms (Yamasaki 1987, Armstrong et al. 1992), potentially reducing sulfide stress further. Another complexity is that the invasion is often associated with alterations such as mosquito ditches (ditches built specifically to drain standing water from the marsh to eliminate the breeding habitat for mosquitoes), roadbeds, landfill edges, and dikes that both de-vegetate areas and create better drainage. Historical studies indicated that the invasion begins along these sites, from which Phragmites later spreads rhizomatously into less altered areas of the marshes (Bart 1997). These well-drained areas could exhibit lower sulfides, but the lower sulfides may not be as important as community-level interactions or propagule dispersal through maintenance of water control structures in explaining the pattern of invasion. It is also unclear whether these well-drained disturbed areas affect post-establishment spread of Phragmites into more poorly drained areas. As the plant spreads rhizomatously from the ditches, it may depend on translocation to offset nutrient uptake problems and to tolerate edaphic stress. The importance of translocation of nitrogen (Evans 1988, 1991, Alpert 1991), water (Salzman and Parker 1985, Alpert 1990, Evans 1991), photoassimilate (Ashmun et al. 1982, Alpert and Mooney 1986, Shumway 1995) and osmoregulatory ions (cf. Matoh et al. 1988) in allowing new ramets to survive edaphic stress and low nutrient conditions has already been documented in other plants. This is only possible if ramets in stressful conditions are connected to areas that could serve as a source for what they need to survive. Ditches and other well-drained areas could serve as such a source area, making them crucial for the spread of the plant after its establishment. OIKOS 89:1 (2000)

This paper is the result of a study designed to elucidate the mechanisms of Phragmites spread in a New Jersey salt marsh. Using field experimentation to eliminate translocation and limiting rhizosphere oxidation, we hypothesize that Phragmites would prosper along well-drained mosquito ditches regardless of treatment because sulfides and salt are low. The plant then spreads rhizomatously into areas exhibiting more stressful conditions. We further hypothesize that the plant initially supports its spread into highly stressful areas through translocation of substances such as oxygen, nitrogen, and osmoregulatory ions to offset edaphic stress, drawing upon the established stands along the ditches as a source area. We also hypothesize that once established, the higher productivity and pressurized ventilation of the rhizosphere should alleviate edaphic stress from sulfides and lead to better plant performance and increased nutrient utilization from the soil. This, in turn would support advance into new areas.

Methods Site description The research was conducted at the Riverbend Marsh of the Hackensack Meadowlands, Secaucus, New Jersey, USA. The site was selected because examination of aerial photographs indicated a pattern of invasion from mosquito ditches to the interior of a high marsh. This site also has several relatively healthy patches of high marsh vegetation (Distichlis spicata and Spartina patens mixed with some Salicornia sp. and others) that are being actively invaded by Phragmites. There is also a lack of confounding anthropogenic activities such as salt hay farming and burning that would result in differential success of Phragmites along mosquito ditches and in the high marsh (see Bart 1997).

were defined as dominated ( \ 75% cover) by high marsh vegetation with sparse Phragmites. Furthermore, the invasion front patches had to exhibit growth of culms in areas that were at least 0.25 m from any standing culms from previous years. The invasion front patches were selected and demarcated first, based on the clarity of the invaded versus uninvaded areas. A transect was measured from the middle of the patch to the ditch, and the heavily invaded patch was placed halfway between the invasion front and the ditch. All patches were 4 m ×1m. Once demarcated, all patches were split into four plots, each measuring 1 m×0.5 m, and randomly assigned a treatment: control, severing and clipping treatment, one as severing alone, and one as clipping alone. During late April all plots assigned for clipping, or severing and clipping, were clipped of all standing dead culms to a point just below the marsh surface. Any culms that could not be clipped below the surface were sealed with silicone sealant to impede airflow. From May through June, all plots that were assigned for severing or severing and clipping had their rhizomes and roots severed to a depth of 80 cm using a post driver and a long handled ice pick. 80 cm was chosen, as it would allow for deeper severing than the lowest recorded rhizomes (60 cm). The entire perimeter of each plot was severed to ensure that no rhizomatous connections existed. In late June, all plots had three tension lysimeters placed at 10, 20 and 30 cm in depth to collect water from the Phragmites root zone. The lysimeters were placed in the ground either with a hammer or by first penetrating the ground with a 4-cm diameter galvanized pipe driven by a post driver. In November 1998, the relative elevation of all plots was determined using a surveyor’s transit and rod. The rod was placed to the marsh surface and all aboveground biomass was pushed away to avoid inaccurate readings. All readings are relative to an arbitrary point in the center of a Spartina patens patch.

Plot selection and preparation In early April 1998, five excavations were conducted in Phragmites invaded areas to establish the depth of the roots and rhizomes of the site. Roots in the interior patches (n =3) extended down to 60 cm in depth, while roots at the ditches (n=2) varied from 40 to 55 cm in depth. Selection of suitable sites was made in April 1998. Four patches were selected in each of the following three zones: ditch side, heavily invaded areas, and invasion front. Ditch side plots were within 0.5 m of an actively draining mosquito ditch that had 100% cover of Phragmites. Heavily invaded areas were more than 5 m from a ditch were almost entirely covered by Phragmites, with the previous years’ standing dead culms reaching a height of at least 1 m. Invasion front patches OIKOS 89:1 (2000)

Sample collection Pore water Pore water was collected during low tide near quarter moon for the first two sampling periods. Lack of water in the upper depths of the marsh led us to change the sampling dates to half moon. Samples were taken once every other week. On a collection date, accumulated water was driven out of the lysimeters using a pressurized nitrogen tank, with a slight overpressure to keep the sample anoxic. A handheld vacuum pump was then employed to extract fresh pore water. The lysimeters were allowed to draw water for about 30 min before sampling. Samples were taken from one of the stopcocks using a 60-ml syringe. 61

The water drawn was then filtered using a syringe filter fitted with a Watman GF/F (0.7 mm) filter to remove particulates and suspended organics. Plant height In October 1998, five randomly selected culms from each plot were measured for height from the marsh surface to the base of the inflorescence (when applicable). In plots with fewer than five surviving culms, all the culms were measured. Biomass In November 1998, all biomass was removed from a 0.25-m2 subplot located toward the center of the each plot. All biomass was weighed, dried in a drying oven at 65°C, and weighed again after 5 d.

Sample preparation and analysis Sulfides In the field, 6.25 ml of filtered pore water was added to 0.5 ml of Cline’s reagent (Cline 1969) in a 60-ml HDPE sealable bottle for sulfide analysis. The bottle was inverted once to assure proper mixing. Standards for colorometric analysis were prepared from stock standards by combining 0.024 g of sodium sulfide (Na2S×9H2O) with 100 ml of deoxygenated de-ionized water (water was de-ionized using an ‘‘EPure’’ system). Standards were prepared at concentrations of 0, 10, 20, 30, and 40 mM using the same deoxygenated distilled water. All samples were analyzed within 24 h (when possible) using a Perkin-Elmer Lambda 12 spectrophotometer set at 670 nm. Ammonium Between 4 and 20 ml of pore water for ammonium analysis were placed in 20-ml HDPE sealable bottles and frozen within 10 h of sampling. Within 4 h of analysis the samples were thawed. The samples were fixed with an indophenol reagent mixture and were kept in the dark for 4 to 24 h to allow the color to develop.

The samples were analyzed with a Perkin-Elmer Lambda 12 spectrophotometer at 640 nm. The standards for the colorometric analysis were 0, 0.125. 0.5, 5, 10, 20, and 60 mM of ammonium chloride. The analysis otherwise followed standard methods (Pilling 1991). Salinity After expelling the filtered water for sulfides, several drops of water were applied to a hand held refractometer (91 ppt) directly in the field to measure salinity.

Data analysis All porewater data was subjected to repeated measures analysis (SAS Proc GLM repeated measures statement) to determine the effects of time on the data as segregated by all permutations. If no time effect was discerned, data were pooled and subjected to ANOVA (SAS Proc GLM), otherwise each run was treated independently. The biomass, culm height, and panicle count data were analyzed through ANOVA (SAS Proc GLM).

Results Elevation In Riverbend Marsh all invasion front plots (mean= 0.008 m above base point) were slightly higher than the invaded (mean= −0.062 m) and ditch side plots (mean= − 0.094 m) (F2 = 7, p=0.0024). There were no differences in elevation between treatments within each zone (F11 = 1.02, p =0.4530).

Above-ground vegetation characteristics Control plot culm height was significantly greater in the ditch than the heavily invaded areas away from the ditches, the heavily invaded zone was significantly greater than the front (Table 1), and treatment effect

Table 1. Control plot ANOVA and LSD means separation test results for plant performance indicators by zone of invasion. Different lower-case letters indicate significant differences (p =0.05) from LSD test. Zone

Front Invaded Ditch

Culm height (m) Mean9 SE

n

Above ground biomass (g dry weight/m) Mean 9 SE

0.91 9 0.08 a 1.5290.04 b 2.219 0.13 c

20 20 15

9.5 9 5.90 a 460.20 9109.67 b 643.39 9 175.10 c

F= 41.16, pB0.0001, df=2

62

F =19.05, p = 0.0015, df =2

Panicles n

Mean9SE

n

4 4 3

29 1.41 a 18.5 96.59 ab 23.798.41 b

4 4 3

F =3.87, p =0.0669, df=2

OIKOS 89:1 (2000)

Table 2. ANOVA and LSD means separation results for culm height, biomass, and panicle production by treatment. Different lower-case letters indicate significantly different groupings (LSD test, pB0.05). Zone

Treatment

Culm height (m) Mean9 SE

Front

Control Clipped Severed Severed/Clipped

0.91 9 0.08 0.679 0.07 0.51 9 0.06 0.62 9 0.06

a ab b‡ b‡

n

n

Mean 9SE

n

16 12 7 11

9.5 95.90a 4.57594.575a 5.925 94.86a 6.959 4.63a

4 4 4 4

2 91.41a 0.5 90.5a 0a 0a

4 4 4 4

F = 5.35, p = 0.0033, df =3 Invaded

Control Clipped Severed Severed/Clipped

1.52 9 0.04 1.30 9 0.04 1.17 9 0.07 0.76 9 0.04

a b b c

F= 0.17, p =0.91, df =3 20 20 20 20

F = 41.16, pB0.001, df = 3 Ditch

Control Cut Severed Severed/Clipped

2.21 9 0.13a 2.24 9 0.13a 2.42 9 0.1a 2.37 9 0.07a

Panicles/m2

Above ground biomass (g dry weight/m2) Mean9 SE

460.20 9 109.67a 340.09954.91a 378.73933.34a 219.73 952.09a

F =2.65, p =0.1392, df =2 4 4 4 4

F= 2.12, p =0.1507, df = 3 15 15 15 15

F = 1.13, p = 0.3442, df = 3

643.39 9 175.10a 698.56 983.52a 1336.59 9170.47b 691.99 9211.12a F= 4.07, p =0.049, df = 3

18.5 96.59 5.5 91.89 2.5 91.04 2.25 90.85

a b b b

4 4 4 4

F= 4.26, p =0.0317, df =3 3 3 3 3

23.7 98.41a 23 92.89a 26.33 96.74a 28.33 93.76a

3 3 3 3

F= 0.13, p =0.9379, df =3

‡ indicates high mortality of individual culms in at least three of the replicates.

varied with the zone. Culm height in the ditch showed no treatment effects, while the height of invaded area controls was greater than plots that were clipped and severed, which in turn were greater than the severed/ clipped plots. In the invasion front severing and combining severing and clipping killed most culms within two weeks. Those that survived were significantly shorter than the control plots (Table 2). Above-ground biomass was significantly greater in the severed than the other treatments along the ditches. In the heavily invaded areas, biomass decreased in the severed and clipped plots, but the differences were not significant. No treatment effects were found along the invasion front (Table 2). The timing of panicle development did not differ in the control plots of each zone. The number of panicles in the control plots was greater in the ditches than the interior, which in turn were greater than the front, but these differences were not significant (Table 1). In the invasion front, there were no significant differences between treatments in panicle count, although this could be an artifact of the absence of panicles at one replicate’s control plot. The invaded area’s control plots had greater numbers of panicles than the clipped, severed, and severed/clipped plots. Furthermore, panicle development in the severed and severed clipped plots was delayed by more than a month. The ditch side plots showed no treatment effect in the number of panicles (Table 2). OIKOS 89:1 (2000)

Salinity Repeated measures analysis revealed a time effect (Wilk’s Lambda, F4 = 55.69, pB 0.0001), so each sample run was treated independently. On the 3 – 4 September and the 17 – 18 September runs there was significantly higher pore water salinity in the ditch sides’ and the heavily invaded areas’ control plots than those of the invasion front (Table 3). Although the invaded areas’ control plots were always slightly higher than any treatment, the differences were not significant until 15 – 16 October. At no time was there a treatment effect on salinity at the invasion front, while in the 3 –4 September and the 17 – 18 September runs the ditches exhibited significantly higher salinity in the controls and severed plots than in the clipped or severed/clipped plots (Table 4). These last relationships are suspect due to the amount of missing data resulting from dry lysimeters.

Sulfides There were no time effects for any combination of measures from 17 July to 18 September (Wilk’s Lambda, F2 = 0.4015, p= 0.63), so data were pooled for the following analysis. As shown in Table 5, the controls of the ditch side plots exhibited lower sulfide concentrations than those of either the interior or the invasion front regardless of 63

64

OIKOS 89:1 (2000)

9 11 8

12.5 93.08 a 14.549 2.66 a 23.8 91.2 a F=0.23, p= 0.7997, df = 2

n

Mean 9 SE 17–20 July

F= 1.04, p = 0.3939, df = 2

16.5 9 1.61 a 17.37 9 1.25 a 14.5 9 0.5 a

Mean 9 SE 30 July–3 August 10 8 2

n

F= 8.32, p = 0.0108, df =2

20.18 9 0.92 a 26.1 9 1.33 b 23 9 0.58 ab

Mean9SE 3–4 September

Salinity (ppt)

12 10 3

n

F =3.91, p= 0.0361, df =2

21.490.89 a 25.589 1.16 a 24.49 0.4 a

Mean9 SE 17–18 September 10 12 5

n

F =2.65, p=0.0965, df =2

23.2791.05 a 26.2791.27 a 26.279 0.25 a

Mean9SE 15–16 October

11 11 3

n

Treatment

Control Clipped Severed Severed/Clipped

Control Clipped Severed Severed/Clipped

Control Clipped Severed Severed/Clipped

Zone

Front

Invaded

Ditch

F= 0.05, p =0.9564, df = 3

14.5 9 0.5 a 15.25 9 1.89 a 16.5 90.87 a 15 90 a

23.8 91.2 a 14 92.26 a 129 1.64 a 12.78 9 1.71 a F= 0.26, p= 0.853, df =3

F= 0.11, p =0.9530, df =3

F= 0.97, p= 0.9530, df =3 8 8 6 9

17.37 9 1.25 a 15.86 91.6 a 15.62 9 1.82 a 14.71 9 1.6 a

a a a a

11 8 11 9

a a a a

14.54 9 2.66 7.75 9 2.62 12.692.14 119 3.36

16.5 91.61 16.4 91.35 16.11 91.58 15.7 91.36 F= 0.37, p = 0.7772, df = 3

9 8 11 12

12.5 93.08 a 12.89 9 2.21 a 9.5 92.5 a 10 92.45 a

Mean9 SE 30 July–3 August

F= 0.59, p= 0.6235, df = 3

n

Mean 9SE 17–20 July

Salinity (ppt)

2 4 4 1

8 7 8 7

10 8 9 10

n

a a a a

F= 3.48, p = 0.0758, df =3

23 9 0.58 ab 23 90.84 ab 25.5 90.72 a 22 9 1 b

F= 1.6, p =0.2095, df = 3

26.1 91.33 a 23.6 9 1.29 a 21.27 9 1.09 a 22 91 a

F= 0.15, p =0.9293, df = 3

20.18 90.92 21.3 9 0.92 22.36 9 1.82 21.67 91.39

Mean9SE 3–4 September

3 5 6 2

10 10 11 11

12 11 11 12

n

a a a a

a a a a

F= 4.08, p= 0.0677, df = 3

24.49 0.4 ab 23.891.2 ab 25.49 0.6 a 22.89 1.39 b

F =1.35, p = 0.2731, df = 3

25.589 1.16 22.29 1.43 23.369 0.78 23.29 0.94

F= 0.3, p= 0.8242, df = 3

21.490.89 21.290.86 20.899 1.08 22.2791.24

Mean9SE 17–18 September

5 5 5 5

12 10 11 10

10 12 9 11

n

a a a a

a ab b b

F= 0.04, p =0.9615, df =3

26.2790.25 a 2691 a 26.2590.25 a 27 90 a

F= 4.08, p =0.0141, df = 3

26.2791.27 24.990.91 23.649 0.69 23.0990.84

F= 1.00, p =0.4045, df = 3

23.2791.05 22.2790.87 23 90.85 2491.00

Mean9SE 15–16 October

3 2 4 1

11 10 11 11

11 11 11 11

n

Table 4. ANOVA and LSD means separation test results for depth averaged porewater salinity by treatment. Different lower-case letters indicate significantly different (p=0.05) groupings from LSD test.

Front Invaded Ditch

Zone

Table 3. ANOVA and LSD means separation test results for control plots’ depth average porewater salinity by zone. Different lower-case letters indicate significantly different (p= 0.05) groupings from LSD test.

Table 5. Control plots by zone ANOVA and LSD means separation test results for porewater sulfides. Lower-case letters indicate significantly different (p= 0.05) groupings from LSD test. Zone

Sulfides (mM) Mean 9SE

Front Invaded Ditch

10 cm

n

20 cm

n

30 cm

n

1130.66995.55 a 125.289 101.81 b 9.6790.62 b

9 9 2

1936.82 9216.71 a 1020.352 9 294.24 b 5.94 9 1.59 c

9 10 6

1712.57 9256.02 a 1081.22 9 170.50 a 7.45 9 2.57 b

10 10 5

F= 31.71, pB0.0001, df = 2

F= 13.34, p =0.0002, df = 2

the depth. The interior plots exhibited significantly decreased sulfide concentrations over those of the front at all depths. As shown in Table 6, the invasion front exhibited no treatment effects on the sulfide concentrations at any depth. At 10 cm the control plot invaded area sulfide concentrations were significantly lower than either the severed or severed/clipped plots. There were no differences at 20 cm, while at 30 cm the controls were significantly less than the cut plots (Table 6). In the ditch side plots, the 10 cm severed and clipped plots exhibited lower sulfides than the severed/clipped plots, at 20 cm the cut plots had greater sulfide concentrations than at all other sites, and at 30 cm there were no differences. The differences found at 10 cm may be the result of the extremely small sample sizes, so the

F= 12.55, p = 0.0002, df =2

analysis is suspect. No depth or treatment in the ditches showed sulfide concentrations above 40 mM, 10 to 100 times lower than the invaded or invasion front areas.

Ammonium Repeated measures analysis showed no time effects from 17 July to 16 October, (Wilk’s Lambda, F3 = 0.7101, p=0.59) so data were pooled. At 10, 20 and 30 cm, ammonium concentrations for the control plots were significantly greater at the invasion front than in the invaded plots or ditch (Table 7). There were no differences in ammonium concentration at 10, 20, or 30 cm for any treatment at the invasion front (Table 8). At 10 cm, the heavily invaded plots’ severed clipped and clipped plots had significantly higher ammonium con-

Table 6. Treatment by zone ANOVA and LSD means separation test results for porewater sulfides. Different lower-case letters indicate different groupings elucidated by the LSD test, significant at p = 0.05. Zone

Treatment

Sulfides (mM) Mean9 SE

Front

Control Clipped Severed Severed/Clipped

10 cm

n

20 cm

1130.69 9 95.55 a 606.829 171.13 a 1053.079 200.58 a 1222.69 9 263.19 a

9 8 10 8

1936.82 9 216.71 1745.36 9170.88 2042.26 9 201.37 1082.75 9 202.61

F= 1.91, p= 0.1483, df = 3 Invaded

Control Clipped Severed Severed/Clipped

125.289 101.81 a 275.209 72.69 ab 702.399 226.8 bc 952.72 9 257.54 c

Control Clipped Severed Severed/Clipped

9.679 0.62 6.38 9 3.34 5.179 0.95 12.979 1.56

ab a a b

F= 4.83, p= 0.048, df = 3

OIKOS 89:1 (2000)

30 cm

9 10 9 12

1712.57 9256.02 1779.80 9 345.14 1818.63 9 250.28 2062.68 9 116.56

F= 0.59, p = 0.625, df =3 9 9 8 8

F = 4.77, p= 0.0078, df = 3 Ditch

a a a a

n

1020.35 9 294.24 794.56 9 150.86 1034.49 9 191.14 1098.54 9 266.32

a a a a

10 10 10 11

a ab ab b

10 11 12 12

F =0.39, p =0.763, df =3 a a a a

10 6 11 9

F= 0.21, p = 0.8904, df = 3 2 2 3 3

n

5.949 1.59 a 17.6493.5 b 5.191 90.32 a 7.786 9 2.19 a F= 6.3, p =0.0032, df = 3

1081.21 9 170.46 1695.77 9 178.44 1261.42 9206.66 1606.40 9170.44

F= 2.37, p = 0.0846, df =3 6 7 7 5

7.45 92.56 11.5194.00 9.42 9 2.64 9.27 92.58

a a a a

5 7 7 6

F =0.27, p =0.8495, df = 3

65

Table 7. Control plots by zone ANOVA and LSD means separation test results for porewater ammonium. Lower-case letters indicate significantly different (p=0.05) grouping from LSD test. Zone

Ammonium (mM) Mean9SE

Front Invaded Ditch

10 cm

n

20 cm

n

30 cm

n

262.35931.24 a 19.1195.28 b 4.8490.00 b

10 12 1

305.42 9 45.56 a 29.18 93.54 b 40.31 910.27 b

10 14 6

294.02 956.77 a 76.21 936.59 b 17.96 97.27 b

12 12 3

F= 34.90, pB0.0001, df = 2

F =33.59, pB0.0001, df =2

centrations than the controls. At 20 cm the severed clipped and severed plots had significantly higher ammonium concentrations than the control. At 30 cm there were no significant differences between treatments in ammonium concentrations. The ditch plots’ ammonium concentrations were not significantly different for any treatment at any depth.

Discussion Mosquito ditches and Phragmites invasion Bart (1997) clearly showed that the Phragmites invasion is associated with ditches and other well-drained features, but could not establish whether rhizome dispersal

F =7.15, p =0.0037, df = 2

or less stressful conditions connected these features with the invasions’ start. This study suggests that mosquito ditches and similar well-drained disturbances lower sulfides in the immediately adjacent marsh sediment, while salinity stays the same. The sulfides are lower probably because these well-drained areas are too oxygenated to support sulfate reduction to sulfides (cf. Agosta 1985, Harvey et al. 1987). It is unclear why the salinity is high at the mosquito ditches in comparison to those recorded along tide creeks by Agosta (1985). Because culm height, biomass, and panicle production were high along the ditches regardless of treatment, the sulfide findings suggest that the mosquito ditches alter edaphic conditions in a way that benefits the growth of Phragmites. Lower sulfide concentrations caused by the better drainage create less toxic condi-

Table 8. ANOVA and LSD means separation results for treatment by zone porewater ammonium. Lower-case letters indicate significantly different (p= 0.05) groupings from LSD test. Zone

Treatment

Ammonium (mM) Mean9 SE 10 cm

Front

Control Clipped Severed Severed/Clipped

262.35 9 31.24 350.649 58.13 381.459 72.71 302.75 9 53.94

a a a a

n

20 cm

10 9 11 12

305.42 945.56 390.81 9 50.46 397.77 9 69.12 470.9 9 64.29

F= 1.71, p= 0.1684, df = 3 Invaded

Control Clipped Severed Severed/Clipped

19.119 5.28 a 140.109 51.02 b 81.56 9 26.32 ab 119.75 9 37.70 b

Control Clipped Severed Severed/Clipped

4.849 0.00 a 16.369 5.96 a 28.569 11.97 a 11.309 0.00 a F= 0.72, p= 0.5794, df = 3

66

30 cm

10 13 11 13

294.02 956.77 302.26 9 29.12 358.53 9 54.67 337.91 936.56

F= 1.27, p =0.2969, df = 3 12 8 10 8

F= 3.04, p= 0.04, df = 3 Ditch

a a a a

n

29.189 3.54 a 86.35 9 50.52 ab 189.64 945.01 b 169.34 956.13 b

40.31910.27 a 55.19 24.39 a 69.47 949.90 a 138.10 9125.08 a F=0.55, p =0.6521, df = 3

a a a a

12 14 12 14

F =0.45, p =0.7174, df = 3 14 6 12 12

76.21936.59 102.20 9 34.37 85.48 9 21.07 160.14 9 44.31

a a a a

12 11 14 13

F= 3.67, p =0.3164, df = 3

F =3.67, p =0.02, df =3 1 4 3 1

n

6 7 6 4

17.969 7.27 a 96.79 9 49.93 a 16.74 9 4.40 a 95.55 948.40 a

3 9 7 5

F =1.07, p =0.3828, df = 3

OIKOS 89:1 (2000)

tions for Phragmites (Fu¨rtig et al. 1996) that allow for increased growth. Thus, while historical studies have shown that there is propagule dispersal to these welldrained areas through ditch and dike construction (Bart 1997), this study demonstrates that the sites to which the propagules are dispersed present better conditions for growth than unaltered high marshes.

Mechanisms for spread into areas away from the ditches The invasion front and heavily invaded areas exhibited higher sulfide and higher ammonium concentrations than the ditches, regardless of treatment. Severing rhizomatous connections along the invasion front resulted in death of many of the culms, with no associated change in edaphic conditions. One hypothesized reason for the death of the culms from severing is that the culms are infected by pathogens (Windham pers. comm.). This is unlikely for two reasons. First, severing produced no negative biological effects at the ditches. Second, there have been no documented cases of pathogen invasion confounding an experiment so consistently as to mimic differential responses along a stress gradient (Pennings pers. comm.). The alternative reason for the death of culms at the invasion front after severing rhizomatous connection is that the plant’s survival in newly emerging areas is dependent on translocation between ramets. Clonal integration has already been proven to be important in producing even-sized stands of Phragmites (Hara et al. 1993), suggesting that translocation occurs in this plant. It has also been documented in other species that highly stressful conditions can be overcome through clonal integration (Shumway 1995). Since sulfide concentrations of the level measured at the invasion front can be toxic and inhibit ammonium uptake in Phragmites (Fu¨rtig et al. 1996, Chambers et al. 1998), it is possible that the translocation of nitrogen, carbohydrates, and oxygen might be essential for growth at the invasion front. If translocation is to be accepted as the mechanism that allows Phragmites to invade areas away from the ditches, it is essential to find an area that could be a source for the translocated substances. Initially, the mosquito ditch sides could be such a source area. The better drainage and less stressful conditions would allow for ample uptake of nutrients and storage of photoassimilate, and taller culms would allow for increased pressurized flow of oxygen to the new ramets. Also, the pattern of invasion (ditch side invasion followed by rhizomatous spread) makes ditches the only plausible source area in this study. More research is needed to demonstrate what substances are translocated to help Phragmites survive at the invasion front. OIKOS 89:1 (2000)

Phragmites invasion, sulfides, and pore water ammonium Once established in a poorly drained area, it seems that rhizomatous connections, and to a lesser extent standing dead culms, decrease sulfide and ammonium concentrations at higher depths, and that removing such structures decreases plant performance. There are two mechanisms by which Phragmites might decrease pore water sulfide concentrations. The first is that the increased productivity of the plant leads to better soil aeration through increased flow of oxygen to the rhizosphere (Howes et al. 1986) or through ground water drawdown resulting from higher evapo-transpiration rates. The second is that pressurized ventilation of the rhizosphere does the same. The greatest increases in sulfides were found when the ability of the plant to ventilate its rhizosphere through old dead culms coupled with severing rhizomes. Therefore, the data suggest that pressure ventilation is partially responsible for the decrease in sulfides compared to uninvaded areas. Nevertheless, the greatest increases in sulfide concentrations were associated with severing rhizomes. There are three possibilities that might explain this increase. First, Phragmites can transport oxygen for some distance through rhizomes (Weisner and Strand 1996). It is possible that much of the rhizosphere oxygenation for a given area comes from long-distance transport. The second possibility is that inhibiting translocation decreases productivity, which would prevent productivity induced soil aeration or groundwater drawdown. Finally, it is possible that leakage of organic compounds facilitated sulfate reduction. This last possibility is unlikely given that the lysimeters were clustered in the center of the plots and therefore the unsevered plots’ lysimeters were often just as close to a severed edge as the lysimeters in the severed plots. At present we can only suggest that soil aeration through heavy Phragmites invasion is likely responsible for the treatment effects on porewater sulfide concentrations in heavily invaded areas, and that the plant is therefore changing this edaphic condition. Severing rhizomes and clipping dead culms caused an increase not only in sulfides, but in ammonium as well. There are two ways to explain this increase. First, it is possible that the more stressful conditions (higher sulfides) both suppressed growth and inhibited ammonium uptake by decreasing ATP production. This has been documented in other plants (Pearson and Havill 1988, Koch and Mendelssohn 1989, Koch et al. 1990). Phragmites has shown diminished ATP production (Fu¨rtig et al. 1996) and ammonium uptake (Chambers et al. 1998) at sulfide concentrations found in the severed/clipped plots in the heavily invaded areas. A second possibility is that higher ammonium concentrations may result from decreased denitrification and more reduced conditions caused by the decreased pro67

ductivity of the plant when severed and clipped of standing dead culms (cf. Chambers 1997). This is unlikely because recent work has demonstrated that in Phragmites coupled nitrification/denitrification processes are secondary to uptake in the loss of soil nitrogen, especially after peak biomass is reached (Starink et al. 1999). The fact that severing rhizomes and clipping dead culms decreased stem height, biomass, and panicle production as well as increased sulfide concentrations to levels proven to be toxic to Phragmites suggests that pre-establishment sulfide concentrations were toxic to Phragmites. The data suggest that the effects the plant has on sulfide concentrations also lead to better growth of the plant, forming a feedback loop similar to Howes et al. (1986). Suggestions that this feedback loop between Phragmites and aeration might be important in understanding the plant’s success in high sulfide systems are not new (Fu¨rtig et al. 1996, Chambers 1997), although to our knowledge this has never been tested. We have demonstrated clearly that Phragmites is lowering sulfide concentrations in the upper surface of the marsh below toxic levels and that this lowering of sulfides is associated with increased growth of Phragmites, lending support to claims that the loop exists and that it is relevant to the plants’ performance in harder-to-invade areas.

Disturbance in invasion ecology: gradients or mechanisms? Associations between disturbed areas and biological invasions are among the oldest generalizations to appear in invasion ecology (Elton 1958). While some studies focus on classes of disturbances such as trophic level disruptions or hydrological alterations (Ewel 1986, Orians 1986), the focus of other studies is to associate intensity and frequency of disturbance to invasions with little attention to how particular disturbances relate to particular invasions (Mooney et al. 1986). We feel that a focus on disturbance gradients precludes a priori the importance of rare or localized disturbance events in determining the course of biological invasions and in shaping ecosystem properties. The focus of this study is on a particular type of disturbance that increased drainage and diminished a particular edaphic stress. This type of disturbance only has to happen once and initially affects small areas of a marsh, but leaves long-lasting changes in edaphic conditions. In contrast, repeated and intense disturbances, such as salt hay harvest, cattle grazing, and burning have not proven efficacious in triggering or facilitating the invasion (Bart 1997). This historical data led us to focus on drainage as central in spurring the invasion. Orians (1986) suggests that narrowing studies of disturbance to how particular types relate to invasions of 68

particular species would allow ecologists to understand why some species are capable of invading while others are not. We suggest that focusing on disturbance events associated with the invasion rather than gradients of disturbance is the necessary first step in causal explanation of associations between ecosystem properties and a particular invasion. Once a particular type of disturbance is historically linked to an invasion, we can suggest mechanisms by which this disturbance could account for the invasion as well as change ecosystem properties.

Ecosystem properties and invasion ecology: a call for experimental approaches The mechanistic linkage between mosquito ditches, ecosystem properties, and the invasion of Phragmites is very complex. While many studies have attempted to understand the relationship between ecosystem properties and plant invasions through simple correlation (Roman et al. 1984, Windham 1995, Chambers 1997, Templer et al. 1998), the resulting confusion between cause and effect makes the utility of such studies limited when the goal is to assert that the invasion has an effect on ecosystem properties (cf. Vitousek 1986) or in understanding site invasibility (cf. Chambers 1997). We suggest that while establishing associations between invasions, ecosystem properties, and disturbances is an important first step, manipulative experimental design is needed to establish why the association exists. These experiments can be simple manipulations of species distribution for determining an invasion’s effects on ecosystem properties (Vitousek 1990). Other approaches (such as the one presented here) involve manipulating mechanisms that are suspected to affect the invasion and/or the invasive’s ability to change ecosystem properties. Acknowledgements – We would like to thank the members of the Hartman lab for field and laboratory assistance, with special thanks to Leslie Shank, Nancy Hanna, Mary Yurlina, Chris Liliola and Marc Knowlton. We would like to thank Sybil Seitzinger, Renee Styles, and Claire Reimers for laboratory instructions. We would also like to thank Randy Chambers, Joan Ehrenfeld, Peter Morin, Kelly Smith, Andrew Vayda, and Bradley Walters for reviews and helpful discussions. Financial support for this project was made possible through the Meadowlands Environmental Research Institute (MERI) and the Hackensack Meadowlands Development Commission (HMDC).

References Agosta, K. 1985. The effects of tidally induced change in the creek-bank water table on pore water chemistry. – Estuarine Coastal Shelf Sci. 21: 398 – 400. Alpert, P. 1990. Water sharing among ramets in a desert population of Distichlis spicata (Poaceae). – Am. J. Bot. 77: 1648 – 1651. OIKOS 89:1 (2000)

Alpert, P. 1991. Nitrogen sharing among ramets increases clonal growth in Fragaria chiloensis. – Ecology 72: 69 – 80. Alpert, P. and Mooney, H. A. 1986. Resource sharing among ramets in the clonal herb Fragaria chiloensis. – Oecologia 70: 227 – 233. Armstrong, J., Armstrong, W. and Beckett, P. 1992. Phragmites australis: Venturi- and humidity-induced pressure flows enhance rhizome aeration and rhizosphere oxidation. – New Phytol. 120: 197–207. Ashmun, J. W., Thomas, R. J. and Pitelka, L. F. 1982. Translocation of photoassimilates between sister ramets in two rhizomatous forest herbs. – Ann. Bot. 49: 403 – 415. Baker, H. G. 1986. Patterns of plant invasion in North America. – In: Mooney, H. A. and Drake, J. (eds), Ecology of biological invasions of North America and Hawaii. Springer-Verlag, New York, pp. 44–57. Bart, D. J. 1997. The use of local knowledge in understanding ecological change: a study of salt hay farmers’ knowledge of Phragmites australis invasion. – MA, Rutgers the State Univ. of New Jersey, New Brunswick, NJ. Bridgewater, P. B and Backshall, D. J. 1981. Dynamics of some Western Australian ligneous formations with special reference to the invasion of exotic species. – Vegetatio 46: 141 – 148. Chalmers, A. G. 1982. Soil dynamics and productivity of Spartina alterniflora. – In Kennedy, V. S. (ed.), Estuarine comparisons. Academic Press, New York, pp. 231 – 246. Chambers, R. M. 1997. Porewater chemistry associated with Phragmites and Spartina in a Connecticut tidal marsh. – Wetlands 17: 360–367. Chambers, R. M., Mozdzer, T. J. and Ambrose, J. C. 1998. Effects of salinity and sulfide on the distribution of Phragmites australis and Spartina alterniflora in a tidal marsh. – Aquat. Bot. 62: 161–169. Cline, J.D. 1969. Spectrophotometric determination of hydrogen sulfide in natural waters. – Limnol. Oceanogr. 14: 454 – 458. Elton, C.S. 1958. The ecology of invasions by animals and plants. – Methuen, London. Evans, J. P. 1988. Nitrogen translocation in a clonal dune perennial, Hydrocotyle bonariensis. – Oecologia 77: 64 – 68. Evans, J. P. 1991. The effects of resource integration on fitness related traits in a clonal dune perennial, Hydrocotyle bonariensis. – Oecologia 86: 268–275. Ewel, J. J. 1986. Invasibility: lessons from south Florida. – In: Mooney, H. A. and Drake, J. (eds), Ecology of biological invasions of North America and Hawaii. Springer-Verlag, New York, pp. 214–230. Fu¨rtig, K., Ru¨egsegger, A., Brunold, C. and Bra¨ndle, R. 1996. Sulphide utilization and injuries in hypoxic roots and rhizomes in common reed (Phragmites australis). – Folia Geobot. Phytotax. 31: 143–151. Hara, T., Van Der Toorn, J. and Mook, J. H. 1993. Growth dynamics and size structure of shoots of Phragmites australis, a clonal plant. – J. Ecol. 81: 47–60. Harvey, J. W., Germann, P. W. and Odum, W. E. 1987. Geomorphological control of suburface hydrology in the creekbank zone of tidal marshes. – Estuarine Coastal Shelf Sci. 25: 677–691. Hellings, S. E. and Gallagher, J. L. 1992. The effects of salinity and flooding on Phragmites australis. – J. Appl. Ecol. 29: 41 – 49. Howes, B. L., Dacey, J. W. H. and Groehringer, D. D. 1986. Factors controlling the growth form of Spartina alterniflora: feedbacks between above-ground production, sediment oxidation, nitrogen, and salinity. – J. Ecol. 74: 881 – 898. Koch, M. S. and Mendelssohn, I. A. 1989. Sulfide as a soil phytotoxin: differential responses in two marsh species. – J. Ecol. 77: 565–578.

OIKOS 89:1 (2000)

Koch, M. S., Mendelssohn, I. A. and McKee, K. L. 1990. Mechanism for the hydrogen sulfide-induced growth limitation in wetland macrophytes. – Limnol. Oceanogr. 35: 399 – 408. Mack, R. N. 1986. Alien plant invasion into the intermountain west: a case history. – In: Mooney, H. A. and Drake, J. (eds), Ecology of biological invasions of North America and Hawaii. Springer-Verlag, New York, pp. 191 – 213. Marks, M., Lapin, B. and Randall, J. 1994. Phragmites australis (P. communis): threats, management, and monitoring. – Nat. Areas J. 14: 285 – 294. Matoh, T., Matsushita, N. and Takakashi, E. 1988. Salt tolerance of the reed plant Phragmites communis. – Physiol. Plant. 72: 8 – 14. Mooney, H. A., Hamburg, S. P. and Drake, J. A. 1986. The invasions of plants and animals into California. – In: Mooney, H. A. and Drake, J. (eds), Ecology of biological invasions of North America and Hawaii. Springer-Verlag, New York, pp. 250 – 272. Niering, W. A. and Warren, R. S. 1977. Our dynamic marshes: vegetation changes as revealed by peat analysis. – Connecticut College Arboretum Bulletin 22. Orians, G. H. 1986. Site characteristics favoring invasions. – In: Mooney, H. A. and Drake, J. (eds), Ecology of biological invasions of North America and Hawaii. Springer-Verlag, New York, pp. 133 – 148. Pearson, J. and Havill, D.C. 1988. The effects of hypoxia and sulfide on culture-grown wetland and non-wetland plants. – J. Exp. Bot. 39: 431 – 439. Pilling, I. 1991. Analysis of ammonium in fresh and sea water. – Academy of Natural Sciences Division of Environmental Research procedure P-20-01. Roman, C. T., Niering, W. A. and Warren, R. S. 1984. Salt marsh vegetation change in response to tidal restriction. – Environ. Manage. 8: 141 – 150. Salzman, A. G. and Parker, M. A. 1985. Neighbors ameliorate local salinity stress for a rhizomatous plant in a heterogenous environment. – Oecologia 65: 273 – 277. Shumway, S. W. 1995. Physiological integration among clonal ramets during invasion of disturbance patches in a New England salt marsh. – Ann. Bot. 76: 225 – 233. Starink, M., van Rijswijk, P., Middelburg, J. and Cappenberg, T. 1999. Nitrification-denitrification in the rhizosphere of emergent macrophytes: fiction or facts? – Paper presented at EUREED, International conference on Phragmitesdominated wetlands, their function and sustainable use. April 18 – 23, 1999, Trˇebonˇ, Czech Republic. Templer, P., Findlay, S. and Wigand, C. 1998. Sediment chemistry associated with native and non-native emergent macrophytes of a Hudson River marsh ecosystem. – Wetlands 18: 70 – 78. Vitousek, P. M. 1986. Biological invasions and ecosystem properties: can species make a difference? – In: Mooney, H. A. and Drake, J. A. (eds), Ecology of biological invasions of North America and Hawaii. Springer-Verlag, New York, pp.163 – 176. Vitousek, P. M. 1990. Biological invasions and ecosystem processes: towards an integration of population biology and ecosystem studies. – Oikos 57: 7 – 13. Weisner, S. E. B and Strand, J. A. 1996. Rhizome architecture in Phragmites australis in relation to water depth: implications for within-plant oxygen transport distances. – Folia Geobot. Phytotax. 31: 91 – 97. Windham, L. 1995. Effects of the Phragmites australis invasion on aboveground biomass and soil properties in brackish tidal marsh of the Mullica River, NJ. – MS, Rutgers, The State University of New Jersey, New Brunswick, NJ. Yamasaki, S. 1987. Oxygen demand and supply in Zizania latifolia and Phragmites australis. – Aquat. Bot. 29: 205– 215.

69