An Allee effect at the front of a plant invasion: Spartina in a Pacific estuary


  • Heather G. Davis,

    Corresponding author
    1. Population Biology Graduate Group, 2320 Storer Hall, University of California, Davis, California 95616, USA,
    2. Bodega Marine Laboratory, Box 247, Bodega Bay, CA 94923–0247, USA
    Search for more papers by this author
  • Caz M. Taylor,

    1. Ecology Graduate Group, 2132 Wickson Hall, University of California, Davis, California 95616, USA,
    Search for more papers by this author
  • Janie C. Civille,

    1. Ecology Graduate Group, 2320 Storer Hall, University of California, Davis, California 95616, USA,
    Search for more papers by this author
  • Donald R. Strong

    1. Section of Evolution and Ecology, 2320 Storer Hall, University of California, Davis, California 95616, USA, and
    2. Bodega Marine Laboratory, Box 247, Bodega Bay, CA 94923–0247, USA
    Search for more papers by this author

Heather G. Davis (tel. +1 530 754 6456; fax +1 530 752 1449; e-mail


  • 1Spartina alterniflora sets very little viable seed at the leading edges of an invasion in Willapa Bay, Washington, USA, where it was introduced c. 100 years ago. This largely outbreeding, rhizomatous grass recruits into previously unoccupied areas at low density, so young plants initially grow isolated from one another but eventually coalesce to form continuous meadows.
  • 2Isolated recruits set approximately one-tenth the seed of meadow plants at five sites, spread over the 230 km2 of Willapa Bay mudflats, and this seed germinated at only one-third the rate observed in meadow plants.
  • 3The consistent patterns suggested that the low seed set in the isolated plants was largely due to the demographic effects of density. Differences between sites in the incidence and amount of seed set and germination rate indicated, however, that there was some environmental influence.
  • 4These data imply that plants in newly invaded, low-density areas produce little viable seed until rhizomatous growth brings them into close contact. This Allee effect could substantially reduce the rate of invasion.


The demographic effects of density on individual reproductive success in exotic species can have profound environmental and economic consequences. Increasing density can raise the per capita rate of population growth when rare and thereby the rate of spatial spread (Hastings 1996). The Allee effect, defined as a positive relationship between any component of fitness and either numbers or density of conspecifics (Allee 1931; Stephens et al. 1999), could be a common, though largely uninvestigated, regulator of population growth in invasive species. Theory suggests, at the front of an invasion, where the diffusive effects of dispersal lead to low density, an Allee effect could slow or change the pattern of range expansion (Dennis 1989; Lewis & Kareiva 1993; Veit & Lewis 1996; Courchamp et al. 1999).

An Allee effect in plants could be caused by demographic, genetic or environmental mechanisms (Courchamp et al. 1999; Stephens et al. 1999). For instance, reduced seed set, and thus recruitment, can result when pollen is limited due to low densities of pollen donors (Silander 1978; Antonovics & Levin 1980). This can be compounded by full or partial self-incompatibility when self-pollen is rejected (e.g. Menges 1991). Even wind-pollinated species can be limited by lack of pollen (Knapp et al. 2001; Koenig & Ashley 2003). But empirical evidence of an Allee effect caused by pollen limitation is confined to species with animal pollinators (e.g. Lamont et al. 1993; Groom 1998). Inbreeding depression and loss of heterozygosity can also reduce individual fitness when plants are forced to self-pollinate (Lande & Schemske 1985; Schemske & Lande 1985; Charlesworth & Charlesworth 1987; de Jong et al. 1993).

The perennial Spartina alterniflora Loisel. (smooth cordgrass), a native to the Atlantic and Gulf coasts of North America, grows on intertidal mudflats in extensive monocultures that often reach 1.5 m in height. Where it has invaded Pacific estuaries north of San Francisco, it is the sole emergent vascular plant. The inflorescence (Fig. 1) has loosely overlapping branching spikes and appressed florets (Hickman 1993). Each floret (also referred to as a spikelet or flower) contains an ovule that can form a single seed. Seeds do not separate from the floral structure, although they do detach from the inflorescence, and are dispersed, both locally and over long distances, in wrack mats on winter and spring tides, leaving no seedbank (Woodhouse 1979). Although S. alterniflora is wind pollinated and native populations are almost completely self-incompatible (Somers & Grant 1981), the population at our study site shows greater self-compatibility (H. G. Davis, unpublished data). Nevertheless, the invasive population in San Francisco Bay has substantial genetic load and suffers from inbreeding depression when self-pollinated (Daehler 1999).

Figure 1.

S. alterniflora inflorescence in male (upper) and female phase (lower). From top to bottom right: floret in male phase, female phase and with a ripe seed. Artwork by Jun Bando.

This species was accidentally introduced to Willapa Bay, Washington, USA, c. 100 years ago and by 1997 it had colonized c. 60 of the 230 km2 of previously bare intertidal mudflats. Each recruit germinates from a single seed and grows rhizomatously into a circular plant, comprised of a single genet, and with time, these merge to form dense continuous meadows. Due to the low density of recruits and lack of other emergent plants on the open mud, the progress of invasions, which is reminiscent of the growth of bacterial colonies on a Petri plate, is readily observed in aerial photographs (Fig. 2).

Figure 2.

Oblique aerial photograph (courtesy of Fritzi Grevstad) of leading edge of S. alterniflora invasion at Willapa Bay.

This study explores the effects of the isolation of recruits and of colonization history upon fecundity in the invading population, by comparing plants at the leading edge of the invasion, with those that have merged to produce continuous meadows. First we asked whether each plant could set any seed at all. If it did, then we investigated the effect of isolation on seed production and germination. Furthermore, we explored whether the site in Willapa Bay from which a plant originated affected the incidence and amount of seed set, and its germination rate.


We measured the effects of recruitment isolation of S. alterniflora plants on seed set during 2000 and 2001 and on germination rate in 2000 at Willapa Bay, Washington, USA. The Bay was divided into five sites: Long Island (LI), Peninsula (PN), South Bay (SO), Palix River (PX) and Shoalwaters (SH) (Table 1). We selected isolated plants, genets that had grown from a single seed into roughly circular clumps (Fig. 2), that were at least 1 m from each other and were often separated by tens of metres, and plants from dense meadows that had coalesced at different times (Table 1). The isolated and meadow plants occupied approximately the same range of tidal elevation at three of the sites (LI, PN, PX), but meadow plants were slightly higher at SO and slightly lower at SH.

Table 1.  History and location of sites
SiteEarliest evidence of colonization*Earliest evidence of coalescenceLatitude/ longtitude
  • NA = not applicable.

  • *

    Data from historical aerial photographs of Washington State Department of Natural Resources and US Army Corps of Engineers.

  • Earliest dates that the sample sites were covered completely by S. alterniflora with no open mud visible in Washington State Department of Natural Resources historical aerial photographs.

  • Thomas Sheffer photo (1942, California Academy of Sciences Herbarium) shows S. alterniflora coalesced into a group c. 40 m diameter. Known growth rates would place the colonization date decades earlier.

  • §

    First date indicates earliest plants visible in photos. More plants colonized the area c. 1995. These plants had not coalesced by the time of this study (2000).

Long Island (LI)
Peninsula (PN)
South Bay (SO)
Shoalwaters (SH)
Palix River (PR)

Germination, seedling growth and plant coalescence into meadows occurred between 1992 and 1999 (J. C. Civille, unpublished data). Thousands of seedlings characterized by attachment to the floret, filamentous root morphology and the absence of large rhizome attachments to other culms, were sampled during this period.

In October 2000, we collected inflorescences from 10 isolated plants and 10 meadow plants at each site. In meadows we chose individuals at least 20 m apart, walking perpendicularly to the tide gradient. We sampled within 1–2 m2 areas to reduce the possibility of collecting from more than one genet. Where possible, we chose five inflorescences per plant that had not shattered and that were yellowing and snapped off easily (indicating that seeds were provisioned). We repeated sampling for four meadow and four isolated plants at three of the sites (LI, SO and PX) in autumn 2001.

To assess the role of increasing distance from the meadow, we laid out three parallel transects in 2000 approximately 50 m, 100 m and 150 m from the SO meadow, and collected inflorescences from 10 randomly selected plants per transect. Spartina alterniflora has recently colonized Gray's Harbor, c. 30 km north of Willapa Bay (Sayce et al. 1997). We obtained up to five inflorescences from each of eight isolated plants at this site in 2000. No conspecific was visible; all were isolated by more than 250 m although most were separated by kilometres.

seed set and germination rate

Seed set was measured first by stripping and counting all florets from each inflorescence. Then, 50 florets (or fewer with small inflorescences) were chosen at random and screened for the presence of seeds. Seed set is expressed as the proportion of florets with seeds.

Florets collected in 2000 were placed in separate small zip-lock bags perforated with needle holes. Bags containing florets and seed were submerged in a single opaque plastic bin containing 1.75 L of autoclaved seawater and 5.75 L of distilled water to yield a salinity of c. 8%. The bin was covered and placed in a refrigerator (4 °C) to simulate winter temperatures in the field. Every week the bags were drained for 2 hours and then resubmerged to oxygenate the water and reduce growth of microbes. After 2 months, by which time many of the seeds had begun to germinate, the bags were placed in a single layer in distilled water on a laboratory bench at room temperature. Number of germinations per inflorescence was scored 2 weeks later by counting the number of radicles. Germination rate for each inflorescence was calculated as the number of germinations per expected number of seeds (Eseeds) with Eseeds = (pro-portion of sampled florets with seeds) (total number of florets).

data analyses

We used a logistic analysis with a binomial distribution to determine the likelihood of observing any seed set. Amongst those plants that did set seed, we asked whether the factors that are associated with zero seed set are also associated with lower numbers of seeds. To do this, we used an unbalanced anova with the factors ‘isolation’ and ‘site’, the interaction, and the random variable ‘plant’, that is nested within the interaction. Germination rate was similarly analysed for all plants that had set seed. Differences between sites were explored using planned pairwise contrasts for the logistic analysis and post hoc Tukey pairwise comparisons for the anovas. Seed set data were log transformed and 0.001 was added to the germination rate and then log transformed. As we are using percentages as outcomes in seed set and germination rate, inflorescences with more florets have more information than those with fewer. To compensate for this, the seed set anova and logistic analysis weights the proportion with number of florets screened and the germination rate anova weights data points proportional to the total number of florets of each inflorescence. This weighting scheme does not change the degrees of freedom. We used SAS version 8 for the analyses (SAS Institute Inc., Cary, NC, USA).



Spartina alterniflora has spread to new areas of Willapa Bay by seed that floats on the tide, with hundreds of seedlings observed each year from 1992 to 1999 (J. C. Civille, unpublished data) but with virtually no recruitment by rhizome fragments. Seedlings that survived the first winter were almost all spaced at intervals of 1 m to a few tens of metres.

seed set

In 2000, both the incidence of seed set and the quantity of seed produced were much greater in meadow plants than in recently established, isolated plants. The mean incidence differed among sites though the pattern remained consistent; overall 92% of meadow plants produced seed vs. 37% of isolated plants (Fig. 3, Table 2). The overall seed set was 0.2 (± 0.01, n = 50 plants) for meadow plants vs. 0.02 (± 0.004, n = 49 plants) in isolated plants.

Figure 3.

Number of plants producing seed for each of the five sites. Twenty plants per site were observed, except for SH, where 10 meadow and nine isolated plants were observed. Sites (isolated and meadow) with different letters are statistically different at P < 0.05.

Table 2.  Relationship between incidence of seed set (present or absent) and isolation and site in S. alterniflora in Willapa Bay, Washington, in 2000. Type III analysis of effects of logistic model
  1. All sites except PX and SO are statistically different from each other at P < 0.05 according to planned contrasts.

Isolation15603.72< 0.0001
Site43329.03< 0.0001

Considering only those plants that had at least one seed, the proportion of florets containing a seed was 0.3 (± 0.02, n = 46) for meadow plants compared with 0.08 (± 0.01, n = 18) for recruits. In every site, meadow plants produced more seed than isolated plants, although some sites differed in mean seed set (Fig. 4, Table 3). The overall effect of site (Table 3) appears to be due solely to the very low production at LI (0.11 ± 0.02, n = 8) as there are no differences amongst the other sites (Fig. 4). The LSmean for site LI was not estimable, so direct comparisons were not possible.

Figure 4.

Seed set (a) and germination rate (b) (means + 1 SE) of S. alterniflora in 2000 for isolated and meadow plants with at least some seed for each of the five sites. Untransformed data were used to generate graph. Sites with the same letter are not statistically different at P < 0.05. ‘NE’ denotes no least-squares means were estimable. ‘0’ indicates no seed was set, ‘Ø’ indicates there were no seeds to test germination rate.

Table 3.  Relationship of seed set (seeds/50 or fewer florets) and germination rate (germinations/expected seeds) for the effects of isolation, site and plant for unbalanced nested anovas
AnalysisSourced.f.Type III MSFP
  1. All observations with seed set equal to zero were removed from seed set and germination data sets. Seed set and germination rate data were log transformed.

Seed setIsolation  1 1571.7325.80< 0.0001
Site  4  304.80 5.02  0.0011
Isolation × site  3  218.02 3.53  0.0186
Plant (isolation × site) 55   80.74 2.80< 0.0001
Error170   28.79  
Germination rateIsolation  1 120 68614.54  0.0002
Site  4  41 412 4.79  0.0012
Isolation × site  3  10 173 1.21  0.3087
Plant (isolation × site) 55  12 886 2.25< 0.0001

The 2000 transects at SO isolates, showed that the mean seed set decreased with distance from the meadow, from 0.04 ± 0.01 near the meadow, to 0.02 ± 0.01, and 0.01 ± 0.004. Five of the eight very isolated plants sampled at Gray's Harbor set seed, with mean production 0.02 ± 0.01.

The seed set in 2001 was extremely low, but isolated plants again produced fewer seeds (0.004 ± 0.001, n = 12) than meadow plants (0.009 ± 0.005, n = 12).


Germination rates paralleled the pattern of seed set (Fig. 4), with seeds from meadow plants more likely to germinate than seeds from isolates (0.34 ± 0.02, n = 46 vs. 0.13 ± 0.03, n = 18). While sites did differ in germination rate, this appears to be due solely to low germination at the oldest site, LI (0.01 ± 0.05, n = 8). As the LSmean was not estimable, comparisons with other sites were not possible. At SO, we found that the mean germination rate, as well as seed set, decreased with distance from the meadow. For plants that set at least one seed, germination rate decreased from 0.02 (± 0.01 n = 6) at the closest transect to 0.01 (± 0.01, n = 2) at the other two. No seed from Gray's Harbor germinated.


This study has documented the occurrence of an Allee effect during a large-scale invasion. The relatively isolated plants at the leading edge produce only c. one-tenth of the seed of well-established plants in high-density meadows; likewise, the germination was also reduced (c. 40% of the meadow plants). These differences were observed at all sites and were significant except for germination at PN.

Differences between sites indicate that some component of environmental variation probably affects seed set and germination rate. The ability to set any seed at all, showed only two of the sites to be the same (Fig. 3), whereas relative seed set and germination rate in plants that set at least some seed differed only between the longest site (LI) and the remainder (Figs 3 and 4). Despite these environmental differences, the Allee effect is consistent over the 43-km length of Willapa Bay, with isolated plants contributing much less to recruitment than plants in meadows.

The difference between seed set in meadow and isolated plants might be caused by pollen limitation. We have three observations that suggest that isolation decreases seed set. First, isolated plants were closer to the meadow at PN than at other sites, and more of these set seed, and those that did set more seed than isolated plants at other sites (Figs 3 and 4). Secondly, mean seed set decreased with distance of transects from the meadow at SO. Thirdly, the very isolated plants from Gray's Harbor had extremely low seed set.

Pollen limitation appears to reduce seed production for S. alterniflora in its native range (Bertness & Shumway 1992), possibly via floral predation, rather than plant density. However, S. alterniflora in Willapa Bay was not exposed to any specialist, or generalist, insect herbivores (Daehler & Strong 1997) until a recent biological control release (Grevstad et al. 2003). The only ovary/seed predator we observed on S. alterniflora was the fungus Claviceps purpurea (ergot), which prevents seed formation within colonized flowers, and this was very rare except at low densities in well-established meadows close to land. We saw no evidence of vertebrate grazing.

We are currently observing pollen traps and estimating stigma loads in addition to applying pollen addition and exclusion treatments to S. alterniflora in the field. We will use these data to derive a pollen dispersal density function to test how wind speed and direction, density and spatial arrangement of neighbours and floral elevation in the tidal gradient contribute to variation in seed production.

Pollen limitation over very short distances is a largely uninvestigated, although possibly common, feature of wind-pollinated plants (Koenig & Ashley 2003). Only a fraction of the ambient pollen may be viable as it is short-lived, particularly in grasses, and pollen quality differs between individual donors (see Kearns & Inouye 1993). Aquatic wind-pollinated species could be acutely vulnerable to pollen loss as pollen settles on water. Abiotic factors such as wind direction have a profound influence on the pollen deposition of grasses (e.g. Giddings et al. 1997), with a very limited upwind isolation restricting pollen availability (as occurs in grass crops, see Griffiths 1950).

Inbreeding depression could cause the lower germination rate of isolated plants in Willapa Bay and zero germination from Gray's Harbor. Even isolated plants had many inflorescences, and we infer that most of their seed set was due to geitonogamous self-pollination. In glasshouse trials, S. alterniflora plants from Willapa Bay were more self-compatible than were those from the native range (H. G. Davis, unpublished data), suggesting that the invasive populations may have a greater potential for selfing.

Seed of S. alterniflora disperses great distances by floating on water-borne wrack in winter within Willapa Bay. Some goes out to sea (Sayce et al. 1997) and the dominant current carries wrack northwards. Over the past decade seedlings have appeared hundreds of kilometres north of Willapa Bay, as far as the mouth of the Copalis River in Washington State (Sayce et al. 1997). At Willapa Bay, patches can grow as large as 8.5 m in diameter within a decade, reducing isolation, as do new recruits colonizing the open mud. Isolation and the open mud that separates recruits decrease as the plants grow together. Thus, spread of S. alterniflora is an example of stratified diffusion (Shigesada et al. 1995) with short-distance dispersal by vegetative growth and long-distance seed dispersal.

The result of incorporating an Allee effect into the process of stratified diffusion has not been studied, although in theory it could reduce the initial or overall invasion speed (Lewis & Kareiva 1993; Kot et al. 1996; Veit & Lewis 1996) and can halt invasion (Keitt et al. 2001). We are using the data from this field study to build analytical and simulation models describing the spread of S. alterniflora in Willapa Bay (Taylor et al., in press). We will then explore if the Allee effect could be primarily responsible for the long establishment phase of this invasion and evaluate different control or management strategies.


We are grateful to D. Ayres, W. Brown, T. Brownlee, B. Dumbauld, A. Fisher, T. Haring, L. Holcomb, J. Lambrinos, K. Sayce, A. Sears, C. Sloop, M. Watnik and M. Wecker for their facilitation of this research, and S. Barrett and two anonymous reviewers for comments on the manuscript. This work was funded by the U.S. Environmental Protection Agency, Office of Research and Development, National Center for Environmental Research, Science to Achieve Results (STAR) Program, Grant # U91580201-0 to HGD and the National Science Foundation Biocomplexity Grant # DEB0083583 (P. I. A. Hastings).