Volume 90, Issue 2 p. 291-302
Free Access

Effects of disturbance on germination and seedling establishment in a coastal prairie grassland: a test of the competitive release hypothesis

Heli M. Jutila

Corresponding Author

Heli M. Jutila

US Geological Survey, National Wetlands Research Center, 700 Cajundome Boulevard, Lafayette, LA 70506, USA

Present address and correspondence: Heli M. Jutila, Äijälänkatu 4B 12, 13210 Hämeenlinna, Finland (tel. 358 3–625 1492, e-mail [email protected]).Search for more papers by this author
James B. Grace

James B. Grace

US Geological Survey, National Wetlands Research Center, 700 Cajundome Boulevard, Lafayette, LA 70506, USA

Search for more papers by this author
First published: 28 June 2008
Citations: 121


1 We evaluated the responses of native grassland sods to a variety of types of disturbance in order to assess hypotheses about the competitive effects of established vegetation on seed germination and seedling establishment. In particular, we consider whether germination is more responsive to the magnitude and duration of vegetation removal (competitive release) or to individual disturbance types (specific effects).

2 Field-collected sods of coastal tallgrass prairie were subjected to no manipulation, cutting with clippings left, cutting with clippings removed (hayed), burning, and complete destruction of established vegetation under greenhouse conditions. The emergence and fate of seedlings, as well as light penetration through the canopy, were followed for a period of 4.5 months.

3 Total seedling emergence increased from cut to control, hayed, burned and plants-removed treatments. Several periods of increased seedling emergence suggested responses to both light penetration and seasonal change.

4 Species richness was lowest in cut sods and highest in sods that had plants removed or were burned. Rarefaction analysis showed that these differences were largely those expected from differences in seedling number, except for the cut treatment, which produced fewer species per seedling than other treatments.

5 Indicator species analysis and ordination methods revealed that seedling community composition overlapped strongly across all treatments, although the area of ordination space did increase with increasing numbers of seedlings.

6 Overall, most of the effects of disturbance could be explained by cumulative light penetration to the soil surface, an indicator of total competitive release, although a few specific effects could be found (particularly for the cutting treatment). Thus, these results generally support the competitive release hypothesis.


It is well known that disturbance plays a primary role in determining both the composition and diversity of grassland communities (Chaneton & Facelli 1991; Collins 1992; Turkington et al. 1993). Both the system being affected and the disturbance itself can influence the nature of the induced changes (Pickett & White 1985; Petraitis et al. 1989) via selective damage or mortality (Hulme 1994), altered competitive interactions (Belsky 1992; Clarke et al. 1996), effects on abiotic conditions (Bazely & Jefferies 1986; Jutila 1997; Ford & Grace 1998), and effects mediated through altered trophic interactions (Coffin et al. 1998; Wootton 1998).

Although disturbance is known to have specific and complex effects, one element, an abrupt reduction in the intensity of competition, is assumed to be general and such competitive release is a primary element of the intermediate disturbance hypothesis (Grime 1973; Connell 1978), as well as other models of community dynamics such as the dynamic equilibrium model (Huston 1979, 1994). Plants and litter have been shown to inhibit germination for many, though not all, plant species (Foster 1964; Thompson et al. 1977; Silvertown 1980) and the suppression of germination by established vegetation has been described as a cryptic form of competition (Grace 1999); this is thus an important interaction when considering the effects of disturbance.

Communities are strongly influenced by periodic disturbances, of which one effect is to remove plant biomass, and fire, grazing, trampling, burrowing and digging, as well as human activities such as mowing, haying or tilling, are all common agents of disturbance in grasslands (Hulbert 1988; Jutila b. Erkkilä 1999; Maret & Wilson 2000). Examination of the results of experimental studies (Grace 1999) indicates that the effects of disturbance on species diversity in grasslands are generally consistent with the intermediate disturbance hypothesis, though with some exceptions, and that disturbance history is particularly critical in such responses (Milchunas et al. 1988; Hobbs & Mooney 1995; Collins et al. 1998).

Both grazing and fire have unique effects on species composition and diversity in many cases (Belsky 1992; Collins 1992; Noy-Meir 1995; Hartnett et al. 1996; Jutila 1997; Collins et al. 1998; Grace & Jutila 1999). Indeed fire, which is one of the key forms of natural disturbance for many ecosystems, including grasslands, may have unique effects that cannot be reproduced by other means. In some ecosystems, particularly arid shrublands, some forms of disturbance, most notably fire and its impact through heat, charred wood and smoke, have highly specific effects on germination (e.g. Thanos & Rundel 1995; Keeley & Bond 1997; Keeley & Fotheringham 1998; Van Staden et al. 2000). Similar studies in grasslands are comparatively few (Morgan 1998; Maret & Wilson 2000) and it is currently not known whether such specific effects occur in these ecosystems.

For the purposes of nature conservation, it is important to know the degree to which species in a system depend on specific forms of disturbance or whether various types of disturbance have equivalent effects. Many native grasslands are subject to active management and practices are based on a variety of criteria and constraints. Mowing, haying and grazing are commonly used periodically (Bakker 1989) to reduce competitive effects. While some authors have argued that prescribed burning should be the preferred form of management, others have suggested that a variety of forms of disturbance can have equivalent effects (e.g. Collins et al. 1998). The potentially adverse effects of disturbance, particularly when intense and/or frequent, must also be given careful consideration.

We evaluated the responses of sods from a native grassland to a variety of types of disturbance in order to compare hypotheses about the competitive effects of established vegetation on seed germination and seedling establishment. We consider the degree to which germination responses are simply a result of the amount and duration of vegetation removal (the competitive release hypothesis), rather than being specifically related to individual disturbance types (the specific effects hypothesis). Coastal prairie sods were used to make detailed observations of germination and seedling establishment in dense prairie vegetation. High-quality conservation sites are restricted to a few locations that are remote from our facility and the experiment was therefore conducted under controlled, glasshouse conditions, although this somewhat restricts its direct relevance to field community dynamics, which are influenced by periodic drought, herbivory and other extrinsic factors. This study therefore sacrifices a degree of realism in order to make a detailed examination of the responses to various forms of competitive release.



Sods representing microcosms of a native grassland were collected from a coastal prairie at the University of Houston Coastal Centre (Galveston County, Texas, USA) in May 2000. This prairie, which is dominated by Schizachyrium scoparium (Michx.) Nash, has been managed by annual spring mowing for the past 30 years and has not been burned during that time period. We selected one of three potential sampling sites and established five linear transects spanning an area of 20 × 100 m2. Approximately 150 possible collection points were identified and sods were collected from 53 of these using a table of random numbers (to provide a minimum of 50 for the experiment, allowing for problems). Sods, 28 cm in diameter and approximately 17.5 cm deep, were collected in the centre of a 1-m square, and sods were placed in pots whose sides were trimmed so that approximately 2 cm remained above the soil surface, and transported to the NWRC.


Ten sods were randomly assigned to each of five treatments: (i) no manipulation, ‘control’; (ii) cutting with clippings left behind, ‘cut’; (iii) cutting with clippings removed, ‘hayed’; (iv) burned; and (v) complete destruction of established vegetation using systemic herbicides, ‘plants removed’. The extra samples were added to the burned group (total 12 sods) and the control group (total 11). Sods were arranged randomly within a temperature-controlled glasshouse and were repositioned biweekly to ensure randomization of spatial location effects and thus to use a complete random design (CRD).

Sods were watered approximately once a day, and since water stress was minimal the results may not apply to water-stressed periods in field. Greenhouse temperatures generally tracked outdoor variations except for being about 5 °C cooler in the summer and about 5–10 °C warmer in November and December, and sods were exposed to ambient day length. Six control pots filled with heat-sterilized prairie soil were located among the sods and monitored to estimate seed dispersal into (and within) the glasshouse. Although 22 seedlings were detected in the control pots, most were of species not found in the study sods (Cyperus iria 3, Dichanthelium oligosanthes 1, Echinocloa grus-gallii 2, Fimbristylis miliaceae 1, Kyllinga brevifolia 1, Panicum dichotomiflorum 3, Tridens strictus 1, plus a further 11 dicots that died before they could be identified). Flowering and fruit production were regularly monitored and ripening seeds were removed from sods so that any seedlings that germinated represent the soil seed bank at the time of collection.


Approximately 2.5 months were allowed for recovery from transplantation and adjustment to regular watering before initiation of treatment. An inventory of all species present in the sods was conducted during this time and a complete evaluation of vegetation and seedlings in all sods (census 0) was made immediately prior to the experiment (25–27 July 2000, see Table 1).

Table 1. Census timetable showing the timing of treatments and census dates in 2001
Control Cut Hayed Burned Plants removed
Number of sods 11 10 10 12 10
Treatment date 28 July 1 August 31 July 28 July 11 August
Census 0 date 25–27 July 25–27 July 25–27 July 25–27 July 26–27July
Days after treatment (mean, standard error equals zero unless shown)
Census 0 −1.9 ± 0.50 –5.8 ± 0.48 −4.8 ± 0.48 –2 ± 0.5 −15.7 ± 0.42
Census 1 11 6 7 10 11
Census 2 17 12 13 16 17
Census 3 25 21 21 24 24
Census 4 31 27 28 31 31
Census 5 38 34 35 38
Census 6 45 41 42 45 45
Census 7 59 55 56 59 56
Census 8 70 66 67 70 73
Census 9 88 84 84 87
Census 10 115.6 ± 0.50 113.0 115.1 ± 0.18 118.7 ± 0.61 116.4 ± 0.48
Census 11 144 ± 1.09 134.5 ± 0.50 136.7 ±  0.42 143.5 ± 0.50 137.4 ± 0.48

For the cut treatment the vegetation was cut at the soil surface and the clippings were then cut into 3-cm-long pieces and placed on the soil surface, whereas for the hayed treatment they were removed, sorted by species, dried and weighed.

For the burned treatment sods were combusted outdoors and then returned to the glasshouse. On a hot and sunny afternoon (28 July 2000) they were buried in the ground flush with the surface in the centre of a 4.7-m2, 30-cm-tall open chamber (essentially a ring) that was free of rooted vegetation but that contained 1 kg m−2 of dried, seed-free Triticum aestivum L. straw (comparable with prairie standing biomass). Fire temperatures were estimated using temperature-sensitive tablets wrapped in aluminium foil and placed at the soil surface within the sods and suspended 20 cm above the sods (tablets for 93, 150, 204, 260, 315, 370 and 400 °C). The 12 sods were burned, three at a time. Temperatures averaged 319.8 ± 33.5 °C at 20 cm above the ground and 267.5 ± 62.2 °C at ground level, well within the range of natural fire temperatures for temperate grasslands (Wright & Bailey 1982).

For the plants-removed treatment, established vegetation was killed without greatly disturbing the soil, by spraying twice with glysophate herbicide at 11-day intervals and removing all remaining dead and dying plant material after a further 3 days (the starting date for this treatment). Differences in starting dates (Table 1) are accounted for by expressing the results in terms of days or observation periods since treatment initiation. Each treatment was similarly adjusted to ensure equal duration (Table 1) and they were harvested between 134 and 144 days after application.


Sods were monitored qualitatively on a daily basis and formally re-censused 11 times, with intervals increasing from weekly to monthly later in the experiment (Table 1). Seedlings were marked and counted and identified to species or genus as soon as possible. Colour-coded pins were used to determine both mortality and new germinants at each sample period. Overall vegetation height and aerial coverage, plant cover and height by species, vegetative growth and flowering, and amount of litter, were also measured. The shading effects of biomass were estimated non-destructively by measuring photosynthetically active solar radiation (i) above the vegetation, (ii) below the canopy at 5 cm above the soil surface but always above the litter layer, and (iii) at the soil surface (below the litter layer). A Decagon Ceptometer equipped with a 1-cm-wide wand sensor, programmed to record light readings over its 20-cm length, was used. A 1 cm by 1 cm door was created in the side of each pot and was sealed shut, except when inserting the wand laterally along the soil surface. Percentage of full sunlight passing through the canopy and that reaching the soil surface was calculated. The sum of light penetration to the soil surface was obtained for each pot by summing all measured absolute values of a pot together.

At the end of the experiment (4.5 months) seedlings, including roots, were harvested individually and established vegetation (either remaining or re-growing from the original established plants) was harvested to the ground surface. Shoot length was determined for all the seedling genets and adult ramets. Material was separated by species, dried and weighed. The dry weight of litter was also determined. For certain analyses, species were subdivided by life-form and life-history groups. The number of species of seedlings per sod is referred to as its species richness and the number of species of seedlings per treatment as total species richness.


Statistical analyses were performed using the Statistical Analysis System (1989–96) (SAS Institute Inc., Carey, North Carolina, USA). Prior to analysis, all data were examined for homogeneity of variance and normality of residuals and transformations (typically logarithmic) were performed as needed. For most variables the overall response was assessed, e.g. for total seedling emergence during the study, and temporal dynamics were analysed using repeated-measures procedures. For many of the analyses, including all the repeated-measures analyses, the SAS MIXED procedure was used to mitigate the problem that heteroscedascity could not be eliminated through data transformations. Effects of individual treatments were assessed using a posteriori least square contrasts with Bonferroni adjustments as needed (P-values presented are the adjusted values).

Immediate effects were first evaluated by comparing values immediately before and after treatment applications (censuses 0 and 1) and temporal dynamics that took place following treatment application were then assessed (censuses 1–11). Only seedlings that emerged following the initiation of the treatments were included in the analyses. In a few analyses, GLM procedures were used to evaluate treatment differences at specific time-points (e.g. censuses 3, 8 and 11).

In order to augment the analysis of treatment effects on species richness, the number of species per seedling was evaluated using rarefaction procedures and the ECOSIM program (Gotelli & Entsminger 2001) to determine whether any differences observed might simply be due to differences in the number of seedlings that emerged in different treatments: 95% confidence intervals were generated using re-sampling procedures.

Characterizations of species-specific responses to treatments were performed using the PC-ORD software system (McCune & Mefford 1995). In order to determine whether individual species responded differently to individual treatments, indicator species analysis was performed using the method of Dufrêne & Legendre (1997). Vegetative dissimilarities among seedling communities in individual sods were assessed using non-metric multidimensional scaling (NMS) (see Grace et al. 2000).



Errors were detected in the light measurements at census 11 and were therefore excluded. Data are expressed as percentage of full sun penetrating either through the upright canopy (i.e. using measurements taken immediately above the litter layer) or to the soil surface (i.e. measurements were taken beneath the litter layer). At census 0, light penetrations did not differ significantly between treatments.

When relative light penetration to both below and above litter were compared before and immediately after treatment (censuses 0 and 1), influence of time, treatment and their interaction was evident (all P < 0.0001). In the control treatment, light values showed no significant change between censuses 0 and 1. Light penetration through the canopy (Fig. 1a) showed a significantly smaller change in the control (P < 0.0001) than in the other treatments, which were not separable. A similar initial response was seen in light penetration to the soil surface (Fig. 1b), although control and cut treatments were not significantly different. In the remaining treatments where litter was much reduced, light values below litter were much higher.

Details are in the caption following the image

The change in relative light intensity for different treatments during the experiment: (a) below canopy but above litter, and (b) below litter. The vertical bars indicate standard errors of mean.

Differences among treatments in light parameters were maintained over time (censuses 1–10; Fig. 1, Time, Treatment, and Time × Treatment, all P < 0.0001). Light penetration through the canopy was consistently lower in control and consistently higher in plants-removed sods than in all other treatments (Fig. 1a; P < 0.0001). At the soil surface (Fig. 1b), the light dynamics of control and cut treatments were not substantially different from each other (P = 0.5592), but levels were lower than in hayed and burned treatments, which were again indistinguishable from one another (P = 0.6474). The light values in the plants-removed treatment remained at a significantly higher level than all others (P < 0.0001).


Altogether 3453 seedlings were observed in the 53 sods (equivalent to an average of 1058.07 ± 848.25 (1 SE) seedlings m−2). The highest number in a single sod was 452 (7340.62 seedlings per m2), recorded in the plants-removed treatment. The majority of the seedlings were monocots (62.55%) and most of the identified species were perennials (97.14%). A significant number of the seedlings that emerged in control and cut treatments were located towards the periphery of the sods, perhaps representing an edge effect. However, because all sods were subjected to the same manipulations, and the effect was neither highly uniform nor precisely defined, we made no attempt to separate it.

Based on repeated measures analysis of variance, the total number of seedlings that emerged per sod during the course of the experiment differed significantly among treatments (P < 0.0001 for between-subject effects). The cut treatment, which involved leaving the vegetative clippings in the sod, produced the fewest seedlings (final average = 7.90 1.70 vs. 28.09 18.64 in controls). Hayed (54.80 15.86), burned (81.92 13.41) and plants-removed treatments (151.20 38.51) produced increasingly more seedlings (all significantly higher than control). Monocots and dicots produced similar patterns to those for total seedlings except dicot numbers were lower and more variable, and only the cut treatment differed significantly. The number of perennial seedlings also differed significantly among treatments in the same way as the total (cut < control < hayed = burned < plants-removed, Student-Newman-Keuls test). For annual seedlings, which constituted only a very small fraction of the total, no indication of significant treatment effects could be detected.

Cumulative seedling emergence values show how differences among treatments varied over time (Fig. 2). anovas performed for individual times revealed that significant differences among treatments had developed by the second census (cut treatment had significantly fewer and burned and plants-removed treatments more seedlings than the rest).

Details are in the caption following the image

The change in the cumulative numbers ( ± SE) of seedlings for different treatments during the experiment.

A more detailed understanding of the temporal dynamics of seedling emergence was obtained by examining the emergence of new germinants throughout the study (Fig. 3). To allow for differences in time between censuses, data are expressed as number of seedlings emerged per day. Time, Treatment, and Time × Treatment terms in the repeated measures anova were all highly significant (P < 0.0001). Patterns were similar to those found for overall emergence, except that rates for cut and control were no longer different (P = 0.7767). Once again, plant removal resulted in significantly higher numbers than all other treatments (P < 0.0001) but burning was not distinguishable from haying. Time had a significant effect for all treatments. Across all treatments the numbers of new seedlings increased during the final third of the experiment (Fig. 3), and in control and cut treatments, few seedlings emerged prior to this period. In other treatments the number of new germinants per day showed more temporal variability. Thus, contrary to expectations, there was not one single germination peak after treatment, but several periods when germination was elevated. This occurred in spite of decreasing light penetration (in most treatments) due to vegetation re-growth, indicating that much germination was delayed after maximal competitive release.

Details are in the caption following the image

New germinants per day in different treatments over time.

In order to consider the degree to which treatment effects on light penetration might explain differences among sods in seedling emergence, the sum of light penetration to the soil surface was regressed against total seedling emergence. While one (plants removed) sod deviated substantially, there was a fairly strong relationship between light and seedling emergence (Fig. 4). Further analyses based on the residuals from the regression line were conducted to find out if the treatments (added to regression model as dummy variables) would have additional explanatory power after removing the light effects. This was not the case and thus specific treatments explain no more variation than did light values.

Details are in the caption following the image

Regression between the number of seedlings emerged during the experiment and the cumulative value of light penetration to the soil surface for different treatments. The equation for the line is y = 0.0223x – 16.763, R2 = 0.4019 and p < 0.01.

While treatments showed significant effect on percentage mortality of seedlings, values were variable and effects weak. Average values were highest in the control sods (19.30 ± 4.36) but differed significantly only from the cut treatment (7.95 ± 3.22) (hayed = 7.91 ± 1.58; burned = 12.41 ± 2.32; plants removed = 9.39 ± 2.07).


Plant removal produced the greatest amount of total seedling biomass, much greater than the amount produced in any other treatment (Table 2). This was due to both greater numbers of seedlings per sod and a significantly higher average biomass per individual seedling for monocots, which comprised the majority of total seedling biomass. When both monocots and dicots were studied, the average biomass per individual seedling did not differ between treatments (Table 2) and biomass differences between control, cut, hayed and burned treatments were never significant. Ten species of seedlings contributed more than 1 g to total biomass: nine of these were graminoids and Paspalum plicatulum, Schizachyrium scoparium and Setaria glauca made the greatest contributions.

Table 2. Seedling biomass and species richness per sod at the end of the experiment. Values with different letters are significantly different at P > 0.05 (Student-Newman-Keuls test)
Variable Control (n = 11) Mean ± SE Cut (n = 10) Mean ± SE Hayed (n = 10) Mean ± SE Burned (n = 12) Mean ± SE Plants removed (n = 10) Mean ± SE
Averaged sum of dicot seedling biomass per sod in g  0.11 ± 0.09a 0.02 ± 0.01a 0.07 ± 0.05a  0.03 ± 0.01a 0.59 ± 0.32a
Averaged sum of monocot seedling biomass per sod in g  0.07 ± 0.03b  1.73 ± 1.72b  0.67 ± 0.60b  0.12 ± 0.02b   9.49 ± 0.82b
Averaged sum of seedling biomass per sod in g 0.17 ± 0.09b  1.75 ± 1.72b  0.74 ± 0.66b  0.16 ± 0.03b  10.08 ± 0.79a
Average biomass of a dicot per sod in mg 48.07 ± 45.91a 7.20 ± 6.65a  5.07 ± 4.49a 1.16 ± 0.49a  6.00 ± 2.44a
Average biomass of a monocot per sod in mg 3.11 ± 1.66b  47.11 ± 46.42b  24.13 ± 22.40b 2.76 ± 1.23b 190.16 ± 59.87a
Average biomass of a seedling per sod in mg 13.96 ± 11.40a  48.13 ± 46.32a  18.00 ± 16.91a 1.46 ± 0.26a    77.99 ± 13.26a
Number of species per sod 7.64 ± 1.29c  3.10 ± 0.59d  10.00 ± 1.37bc  12.58 ± 1.46b 17.90 ± 1.48a

The summed seedling heights per sod were significantly higher in the plants-removed treatment than in other treatments, reflecting the same phenomena as for biomass, and involving both number and size, but in this case also for mono- and dicots separately. Not only were seedlings taller in the plants-removed treatment, but they were also more advanced, with flowering and seed-bearing plants found only in this treatment. Even then, few individuals flowered (1.4 ± 1.28 individuals sod−1 and 0.93% of the seedlings), but significantly more than in the other treatments. The five species that flowered were Ludwigia palustris (one individual), Oxalis dillenii (eight), Paspalum plicatulum (two), Rudbeckia hirta (two) and Setaria glauca (one). S. glauca is an annual, R. hirta can be either annual, biennial or perennial and the others are perennials.


On average, 10.28 species of seedlings were identified per sod. While not all morphologically distinct taxa could be identified to species, the majority were and these were predominantly dicots (58.90%) and perennials (93.87% of the total species richness). Species richness was highest in the plants-removed treatment (Table 2). Because of difficulties in identifying seedlings to species early in the experiment, it was not possible to track with confidence the dynamics of species richness over time and treatment effects were assessed from final species richness.

A total of 76 species of seedlings were observed. The cut treatment produced the fewest species (13), with moderate values for control, hayed and burned treatments (32, 35 and 44 species, respectively), and plant removal the most (52). Analysis of variance results indicated that there was a significant difference between treatments in species richness (P < 0.0001), with both lowest (cut treatment) and highest values (plants removed) differing significantly from all others.

The monocot and perennial species richness showed the same pattern as did all species combined. For dicots the highest species richness was found in the plants-removed treatment and although species richness values were in the same order, control and cut treatments did not differ significantly from one another. For annuals, however, while the plants-removed treatment produced significantly more species, the other treatments did not differ.

Approximately 56% of the variance among sods in species richness could be explained by variations in the cumulative value of light penetration to the soil surface (P < 0.0001; Fig. 5). Further analyses (as for seedling number) showed that the cut treatment explained an additional 5.3% of the species richness data (P < 0.0072); in a multivariate regression the effect of cumulative light penetration to the soil surface and cut treatment combined to explain 60.8% of the species richness data (P < 0.0001).

Details are in the caption following the image

Regression between the total species richness by treatments at the end of experiment and cumulative light penetration to the soil surface. The equation for the line is y = 0.0022x + 2.3315, R2 = 0.5635 and P < 0.0001.

Because differences among treatments in species richness may be related to differences in the numbers of seedlings, rarefaction was used to standardize the numbers of species to the numbers of seedlings found. The cut treatment produced fewer species than expected for a given number of seedlings compared with all other treatments. The hayed treatment did not differ significantly from controls, but its species richness was lower than either the burned or plants-removed treatments; no other differences were detected.


Indicator species analysis showed that 16 taxa (12 of them identified to species) showed a significant effect of treatment on the number of germinating seedlings. Three taxa of specific interest also had significantly different numbers of seedlings between treatments according to Kruskal–Wallis tests. Four of the significant indicator taxa (Ambrosia psilostachya, Eragrostis ssp., Ludwigia palustris and Scleria ciliata) were found only in the plants-removed treatment and a further eight (Andropogon virginicus, Dichanthelium oligosanthes, D. sphaerocarpon, Evolvulus sericeus, Oxalis dillenii, Paspalum plicatulum, Rhynchospora ssp. and Setaria glauca) occurred there in higher numbers than in the other treatments. The plants-removed treatment also had higher numbers than the control for seedlings of three other taxa (including Sabatia campestris).

None of the species were indicators solely for the fire treatment, although three species were most abundant there. Paspalum plicatulum, Rhynchospora ssp., Sabatia campestris and two unidentified taxa were significantly more abundant in burned sods than in controls, but they were also increased in the plants-removed treatment. None of the species were indicators only for haying, although three species (Rhynchospora ssp. and Schizachyrium scoparium) did show a significant increase compared with the control. Only five species (Paspalum plicatulum and four unnamed taxa) had a significantly different number of seedlings in the burned than in the hayed treatment (greater in each case). According to the results of the indicator species analysis, one grass species was a negative indicator solely for the cut treatment, from which it was absent, while being found in moderate numbers in other treatments. The abundance of three other species (Aster ssp., Dichanthelium spherocarpon and an unnamed taxon) was significantly decreased (to zero) by cutting and leaving the litter when compared with the control treatment. Biomasses of individual species generally paralleled the species richness data and are not therefore presented.

Higher-order ordinations did not reveal additional differences among treatment groups (even though they decreased the stress value indicating better goodness of fit) and, for simplicity, results of the two-dimensional NMS-ordination are therefore presented. There was substantial overlap among treatments in species composition (Fig. 6), but drawing envelopes around all the points associated with a single treatment, members of the cut treatment occupied the least two-dimensional ordination space and plant removal the most space, albeit with all groups overlapping, primarily in a common area. Total ordination space occupied by a treatment group was estimated using the ARC-View software program and subsequently plotted against the corresponding total number of seedlings and total species richness. Although data points are limited, significant relationships were apparent (P = 0.0049, R2 = 0.95 for total numbers of seedlings and P = 0.0340, R2 = 0.82 for total species richness per treatment).

Details are in the caption following the image

Ordination results for seedling communities. Species with less than five seedlings were excluded and thus only the 40 most common species were used for the ordination.



The results of this study indicate a negative influence of litter on germination, which was partly due to effects other than decreased light. Cutting and leaving the litter resulted in the smallest number of germinating seedlings and species of all treatments. Species richness was lower in cut than control sods, even after adjusting for the production of fewer seedlings (using rarefaction analysis), suggesting that the extra litter inhibited many species from germinating. Interestingly, mortality of seedlings that did emerge was significantly lower in the cut treatment than in controls, further suggesting that the effect of the extra litter was to reduce germination. This effect was not simply due to reduced light, as levels at the soil surface did not differ from controls.

Plant litter is known to be an important influence on germination and establishment (e.g. Facelli & Pickett 1991; Foster & Gross 1998): it can be a physical barrier (Grime 1979) and can also alter microclimate at the soil surface, such as dampening diurnal temperature oscillations, an important cue for germination (Thompson et al. 1977), and lowering average temperature, which could result in slower germination. The decomposition of the litter left in the cut treatment may have contributed to the increased germination at the end of the experiment, although seasonal changes, such as cooler nights and shorter days, may have been involved towards the end (see also Olff et al. 1994).

The practice of haying, which involves both cutting and removing vegetation, often leads to increased plant establishment and diversity (Bakker 1989; Kull & Zobel 1991; Ekstam & Forshed 1992; Meyer & Schmid 1999). In this study, both the number of seedlings and species richness increased compared with controls, although for the latter the effect was neither large (c. 25%) nor statistically significant. Mortality, total seedling biomass and average seedling biomass were not affected by haying.

A large number of studies have reported effects of burning on species composition and species richness of mature vegetation (see references in Wright & Bailey 1982; Whelan 1995; Bond & van Wilgen 1996), or examined the effects of fire on germination and plant establishment (e.g. Thanos & Rundel 1995; Tyler 1995; Benwell 1998). Most have documented the germination responses of individual species rather than entire plant communities and few have considered temperate grasslands (references in Collins & Wallace 1990; Morgan 1998; Maret & Wilson 2000). In this study, fire increased seedling emergence by more than twofold, with the effect appearing rapidly. Burned sods also produced a higher species richness compared with controls, so that the overall effect on germination was positive despite fire potentially killing some seeds in the soil. The unique effects of fire (heat, smoke, etc.) were estimated by comparing hayed and burned treatments, both of which involved a removal of above-ground vegetation. Although the cumulative number of seedlings was higher in burned sods than in hayed ones throughout the experiment, final numbers did not differ significantly, possibly because unique short-term fire effects became less important than long-term release from light competition (similar for both treatments). For species richness such unique fire effects were not significant. The greater total species richness in the burned treatment compared with the control appeared to be attributable to greater emergence. This result, along with the overlap in community composition (Fig. 6) and the presence of few significant indicator species, suggests that in this study, burning allowed free space for germination like haying or plant removal, but had few unique effects on either species richness or the particular species that established (see also Collins et al. 1998). This is in agreement with the hypothesis that most plant responses to fire are not actually unique adaptations to fire per se, but to overall responses to disturbances in general (Bond & van Wilgen 1996).

Plant removal often increases plant establishment and growth (Gurevitch & Unnasch 1989; Goldberg & Barton 1992) and here had a major impact on most parameters compared with controls. Competition from other plants had a significant influence on germination as well as on establishment and growth of seedlings, with removal also allowing fastest growth and greater phenological development (it was the only treatment in which species flowered).

It is important to note that removal of plants may also have deleterious effects on germination and seedling establishment, suggesting a nurse effect of established vegetation (Ryser 1993; Callaway 1997; Wied & Galen 1998; Levine 1999; Olff et al. 1999; Jutila 2001). There was no indication for such a facilitation effect here, as more than twice as many species were found in the plants-removed treatment than in controls. However, facilitation is most often observed in environments where drought stress is a major reason for mortality of seedlings (Ryser 1993; Wied & Galen 1998; Jutila 2001) and such effects were excluded here.


The traditional view that sexual reproduction plays a minor role in prairie dynamics is counterintuitive given the vast number of species and individuals found in many types of grassland. Our results indicate that seedling germination can indeed be fairly high when the conditions are optimal. The average number of seedlings m−2 (1058.07 ± 848.25, 1 SE), higher than that reported by Schott & Hamburg (1997), indicates the importance of soil seed bank and the need for seed bank and seedling studies in grasslands.

Models that assume that recruitment is largely regulated by shading (e.g. Grace 2001) predict rapid responses to disturbance: treatments such as haying and burning, that increase light penetration, were expected to trigger germination followed by a decline as the canopy re-established. However, Figs 2 and 3 show little relation to the light values in Fig. 1(b), but a tendency for germination to increase in the last 2 months. We interpret this effect, seen clearly in control sods, as a seasonal change, perhaps brought on by cooler weather and/or declining day length, as light penetration in control sods did not change substantially over time. The delayed impact of competitive release on germination (reported also after fire by Benwell 1998) may have important implications for grassland management.


To a substantial degree, the effects of cutting, haying, burning and plant removal in this study can be ascribed to general effects of disturbance on competitive release. The cumulative light penetration to the soil surface (used as a measure of the degree of competitive release) explained 40.5% of the variation in number of seedlings and 56.3% of the variation in species richness, probably because of the fundamental importance of light to germination and seedling establishment (Weiner et al. 1997; Grace 1999, 2001). Only cutting had a specific treatment effect and it was manifested only for species richness, explaining an additional 5.3% of the variance. While a few unique effects of individual treatments were found on specific species, the importance of such effects was considerably less than we had anticipated. Much attention has been given to studies that have reported specific mechanisms whereby fire can influence germination of seeds in Mediterranean habitats (e.g. Thanos & Rundel 1995; van Staden et al. 2000). However, mowing and grazing may be as important as fire in promoting diversity in prairie grasslands (Collins et al. 1998). The roles of haying and grazing have been earlier recognized for semi-natural grasslands as well (e.g. Bakker 1989; Ekstam & Forshed 1992). The effects of burning (at least the summer burning used here) differed little from those of haying, suggesting that fire acted primarily as a form of competitive release.

It is important to point out that, although light penetration correlated well with treatment effects, below-ground competition may also be important. The hayed treatment and plants-removal treatment only differed initially in the below-ground competition, which was completely excluded in plant removal but stayed in hayed (above-ground competition was initially excluded in both of them). However, rapid vegetative re-sprouting and the quicker decline in light levels in the hayed treatment prevented separation of the effects of above- vs. below-ground competition. Current knowledge suggests that below-ground competition would affect establishment and growth of seedlings rather than germination and it is likely that the more rapid growth and more advanced phenology of seedlings in the plants-removed treatment was partly due to the lack of below-ground competition. Other factors, in addition to light penetration to the soil, are certainly needed to explain in full the variation in numbers of seedlings and in species richness.

Overall, competitive suppression of germination by established vegetation appears to play a major role in regulating recruitment in this coastal tallgrass prairie. Increasing competitive release allowed a greater number of seedlings to be produced, resulting in greater species richness and more variable community composition among replicate sods. This finding supports the view that the regulation of species richness by community biomass results not only from effects of competitive exclusion, but also from suppression of germination (Grace 1999).


We thank Beth Vairin, Roy Turkington and two anonymous reviewers for their comments on the manuscript. We also thank Larry Allain, Brian Fontenot, Karri Jutila, Sharon King and Tracey McDonnell for their help in both the field and glasshouse. This research was supported by the Finnish Academy and the US Geological Survey. Co-operation by the University of Houston and especially Professor Glenn Aumin for access to the field site is greatly appreciated.