Oecologia (2001) 128:594–602 DOI 10.1007/s004420100709
C. Noemí Mazia · Enrique J. Chaneton Claudio M. Ghersa · Rolando J. C. León
Limits to tree species invasion in pampean grassland and forest plant communities Received: 22 August 2000 / Accepted: 22 March 2001 / Published online: 19 May 2001 © Springer-Verlag 2001
Abstract Factors limiting tree invasion in the Inland Pampas of Argentina were studied by monitoring the establishment of four alien tree species in remnant grassland and cultivated forest stands. We tested whether disturbances facilitated tree seedling recruitment and survival once seeds of invaders were made available by hand sowing. Seed addition to grassland failed to produce seedlings of two study species, Ligustrum lucidum and Ulmus pumila, but did result in abundant recruitment of Gleditsia triacanthos and Prosopis caldenia. While emergence was sparse in intact grassland, seedling densities were significantly increased by canopy and soil disturbances. Longer-term surveys showed that only Gleditsia became successfully established in disturbed grassland. These results support the hypothesis that interference from herbaceous vegetation may play a significant role in slowing down tree invasion, whereas disturbances create microsites that can be exploited by invasive woody plants. Seed sowing in a Ligustrum forest promoted the emergence of all four study species in understorey and treefall gap conditions. Litter removal had species-specific effects on emergence and early seedling growth, but had little impact on survivorship. Seedlings emerging under the closed forest canopy died within a few months. In the treefall gap, recruits of Gleditsia and Prosopis survived the first year, but did not survive in the longer term after natural gap closure. The forest community thus appeared less susceptible to colonization by alien trees than the grassland. We conclude that tree invasion in this system is strongly limited by the availability of recruitment microsites and biotic interactions, as well as by dispersal from existing propagule sources. E.J. Chaneton (✉) · C.M. Ghersa · R.J.C. León IFEVA – Departamento de Ecología, Facultad de Agronomía, Universidad de Buenos Aires, Av. San Martín 4453, 1417 Buenos Aires, Argentina e-mail:
[email protected] Fax: +54-11-45148730 C.N. Mazia Departamento de Producción Vegetal, Facultad de Agronomía, Universidad de Buenos Aires, Av. San Martín 4453, 1417 Buenos Aires, Argentina
Keywords Alien plants · Competition · Disturbance · Recruitment · Seed addition
Introduction Plant invasions are a key component of global environmental changes (Vitousek et al. 1997). Invasions from alien plant species are often triggered by changes in land-use patterns (Heywood 1989; Vitousek et al. 1997), and there is growing evidence that they represent a major threat to native communities (e.g. Hobbs and Mooney 1986; Archer et al. 1988; D’Antonio and Vitousek 1992). Studying invasions is important both from a conservation perspective (Hobbs and Huenneke 1992), and because it can help us to understand basic processes such as recruitment limitation and community assembly (Crawley 1987; Levine and D’Antonio 1999). Plant invasions are influenced by processes operating beyond the limits of the focal community and by factors controlling establishment within the community (Johnstone 1986; Tilman 1997; Levine 2000). Human activities have eliminated most natural barriers to dispersal, increasing the propagule pressure from alien plants worldwide (Heywood 1989; Williamson 1996). Therefore, accounting for intra-community interactions becomes critically important to predict invasion dynamics (Crawley 1987; D’Antonio 1993; Davis et al. 2000). In this paper, we examine the effects of disturbance on grassland invasibility by alien trees in the Inland Pampas of Argentina. We also address disturbance-invasion relations in cultivated forest habitat, because of the role that forestry stands play as foci of dispersal of introduced trees in agricultural landscapes (Richardson 1998). Disturbances generate opportunities for plant invasions via their disruptive effect on species interactions and resource availability (Hobbs 1989; Réjmánek 1989; Hobbs and Huenneke 1992; D’Antonio 1993; Burke and Grime 1996; Davis et al. 2000). By altering physical and biotic barriers to recruitment, disturbances increase the
595
chances of invasion in otherwise “closed” plant communities (Johnstone 1986; Crawley 1989). In herbaceous systems, disturbances impinge on various processes that affect colonization by woody species, including aboveand below-ground competition (De Steven 1991a; Hill et al. 1995; Van Auken and Bush 1997), inhibition of germination by plant litter (Facelli and Pickett 1991), herbivory (De Steven 1991b; Gill and Marks 1991), and facilitation from established plants (De Steven 1991b; Berkowitz et al. 1995). Likewise, natural tree recruitment in forest communities can be associated with disturbance at various scales, through variation in microsite characteristics (Molofsky and Augspurger 1992), patchhabitat conditions (e.g. understorey vs. gap; Gray and Spies 1997), or predation (Schupp 1988). While disturbances may increase community invasibility, plant lifehistory traits such as seed size, growth rate and shade tolerance determine the ability of different species to exploit newly created gaps (Molofsky and Augspurger 1992; Bazzaz 1996). Seedling recruitment may represent a major filter to the spread of an invasive species (Johnstone 1986), a process often limited by the interaction of seed and microsite availability (Eriksson and Ehrlén 1992). Unfortunately, most experiments on recruitment limitation only document the early seedling phase (Turnbull et al. 2000), which tends to obscure the fate of establishing cohorts (cf. Tilman 1997). Mortality factors acting beyond the seedling stage may be especially important during invasions by long-lived plants, as post-recruitment processes determine the success of isolated colonization events (e.g. Archer et al. 1988). So far, few studies have experimentally evaluated the role of disturbance-limited recruitment in grassland and forest communities challenged by the same suite of plant invaders (see Turnbull et al. 2000). During the last two centuries, native grasslands of the Inland Pampas have been reduced to small, semi-natural fragments through conversion to farmland and domestic grazing (Soriano 1992). Landscape management increased the extent of forest areas, creating novel propagule sources and corridors for the spread of alien tree species (Ghersa and León 1999). In the Inland Pampas, tree invasions affect disturbed grasslands, fallow fields, and marginal habitats (León and Anderson 1983; Facelli and León 1986; Dascanio et al. 1994). Introduced tree species also invade, or regenerate within, forestry plantations (Parodi 1942; Ellenberg 1962). Despite the potential impact of woody plant invasions on remnant grassland fragments, factors controlling tree establishment in the subhumid pampas are still poorly understood. It has been proposed that tree invasion in this system is limited by competition from resident grasses (Parodi 1942; Facelli and León 1986). Other studies suggested that dispersal may be limiting tree colonization in otherwise suitable habitats (León and Anderson 1983; Montaldo 1993). However, these ideas have not been tested in experiments allowing for propagule availability and identity of potential invaders. Moreover, no study has yet ex-
amined microsite effects on recruitment in forest stands that may act as “stepping stones” for the spread of introduced tree species. This study investigated factors limiting tree invasion in the Inland Pampas by monitoring the establishment of four alien tree species in semi-natural grassland and cultivated forest stands. We hypothesized that tree invasion rates in grassland remnants are limited by low seed supply and interference from herbaceous vegetation. These factors interact to decrease the probability of an invader finding adequate establishment microsites (Crawley 1989), and thus act in narrowing the “window” of invasion (Johnstone 1986; Myster 1993). Here, we focus on the role of disturbance in removing local barriers to plant invasion once propagules of alien species become available. The effects of disturbance on tree seedling emergence, growth, and survival were evaluated by adding seeds of various species to intact and disturbed vegetation patches. Our experiments were designed to create different levels of microsite limitation to recruitment (Crawley 1990; Eriksson and Ehrlén 1992), by simulating conditions associated with disturbances common in grassland and forest habitats, such as grazing, cultivation, animal burrows and windfalls. We also assessed the survival of establishing trees for an extended, 5-year period, to determine the longer-term fate of early recruits.
Materials and methods Study sites and species The study was conducted at Estancia San Claudio in Carlos Casares, 400 km west of Buenos Aires, Argentina (36°S, 61°5′W). The region is extensively devoted to agriculture and cattle husbandry. The mean annual rainfall is 911.5 mm (1978–1993); mean temperatures vary from 7.2°C in July to 23.8°C in January. The dominant soil is a well-drained sandy loam. Mesic grasslands, including a diverse array of caespitose perennial grasses, constituted the original vegetation type in this area (Parodi 1942; Soriano 1992); at present, herbaceous communities are confined to semi-natural remnants, old fields and pastures at various stages of succession. Forests do not occur naturally in this region (Parodi 1942; Cabrera 1976). Woodlots and tree lines have been extensively planted on roadsides, crop fields, recreational areas, and around farmhouses (Ghersa and León 1999). The four species used in this study were Gleditsia triacanthos L., Ulmus pumila L., Ligustrum lucidum Ait., and Prosopis caldenia Burk. (hereafter called by their genera). These were selected to encompass different origins, life-history traits, and natural history in the study area (Table 1). Gleditsia and Ulmus are known to colonize old fields and grasslands (Facelli and León 1986; Ghersa and León 1999). Gleditsia also invades forestry stands and riparian corridors. Ligustrum has become widely naturalized in cultivated forests (Ellenberg 1962), and can invade abandoned fields and disturbed woodland (Montaldo 1993; Dascanio et al. 1994). Prosopis caldenia is native to semi-arid woodlands of the “Espinal” phytogeographic province (Cabrera 1976) and is a frequent invader of xeric grasslands on the western part of the Inland Pampas (León and Anderson 1983; Peláez et al. 1992). There is no account of its natural occurrence in the study area and it is, therefore, “alien” to mesic pampean grasslands. In early 1994, we selected three study plots: one (12×16 m) was located in a semi-natural grassland site, while the other two (each 8×11 m) were marked in a forestry stand. The grassland and forest sites were 5 km apart; the distance from the grassland to the
596 Table 1 Tree species used to study the invasibility of semi-natural grassland and forest communities in the Inland Pampas, Argentina Species
Family
Habitat/origin
Seed mass (mg)
Dispersal mode
Natural populations Grassland
Forest
Gleditsia triacanthos
Fabacea
Mesic/North America
200
Cattle
Seedlings common around several adults
Seedlings rare; scattered adults
Ligustrum lucidum
Oleacea
Mesic/China
65
Birds
Seedlings absent; two juveniles present
Seedlings common; canopy dominant
Ulmus pumila
Ulmacea
Mesic/Siberia
20
Wind
Absent, but planted in nearby woods
Absent, but planted in nearby woods
Prosopis caldeni
Fabacea
Xeric/South America
30
Cattle
Absent from study area
Absent from study area
nearest woodlot was 1.5 km. Study plots were fenced to 1.2 m height and 0.2 m depth using 7-cm wire mesh, to exclude livestock and burrowing mammals (armadillos). The grassland plot was established in a 4-ha enclosure that had remained protected from grazing since 1979; the site had not been cultivated for the prior 50 years. Vegetation in this enclosure resembled that of midsuccessional old fields (Omacini et al. 1995), being dominated by Lolium multiflorum, Bromus unioloides, Sorghum halepensis, Festuca arundinacea, and Carduus acanthoides. The total aboveground grassland biomass is 1,700 g m–2, including an 8-cm-deep litter layer. The forest stand was a 50-year-old, 10-ha plantation of Ligustrum lucidum and Robinia pseudoacacia. Forest plots were chosen to represent two contrasting habitat patches, referred to as “understorey” and “gap”. The former represented conditions under closed forest canopy, while the latter was a large, natural treefall gap. These plots were located within 100 m of each other and were 50 m away from the forest edge. Experimental design and measurements Seeds of the four study species were planted in mid-October (grassland) and early November (forest) 1994. In laboratory tests, seeds showed high germination rates (80–88%), except for Ulmus that germinated poorly under laboratory conditions (20%, at 20–30°C for 14 days). Our treatments were intended to represent disturbances relevant to each community and therefore differed between grassland and forest sites. The grassland experiment included three treatments: plant canopy and soil left intact (control); all aboveground vegetation removed to 3 cm height (clipped); and canopy removed, roots severely disturbed with a spade, and soil raked to 10 cm depth (disturbed). The control and disturbed treatments represented two extremes with regard to possible interference from resident vegetation on tree establishment. Topsoil disturbance also created a microrelief similar to that associated with cultivation. In clipped patches, seedlings faced no aboveground competition but could still be affected by the presence of litter or belowground competition. Clipping was conducted regularly until March 1996; care was taken not to damage emerging tree seedlings. The 12 species-by-treatment combinations were randomly assigned to sixty, 0.75×1-m quadrats arranged in five blocks. Adjacent quadrats were 50 cm apart. Fifty seeds of the designated species were planted in rows within each quadrat, except for Ulmus that was sown at 100 seeds quadrat–1 to allow for its low germination potential. Thus we added in total 750 seeds per species (Ulmus= 1,500 seeds) to the grassland. Seeds were always put in contact with the soil surface. In the forest experiment, we established five replicate blocks in the understorey and gap plots. Each block was divided into four, 0.75×1-m quadrats to which the study species were randomly assigned. These quadrats were split into two 50×75-cm halves receiving one of two litter treatments: (1) intact litter, or (2) removal
of litter, including leaves and woody debris. Quadrats were separated by 50-cm-wide walkways. In total, each forest plot comprised eight species-by-treatment combinations arranged in a splitplot design (Steel and Torrie 1980). This spatial layout was dictated by the size of the natural gap. Species were sown at 50 seeds quadrat–1 (25 on each half) for a total of 500 seeds per species, with Ulmus being added at double that density. In intact quadrats, seeds were planted beneath the litter layer. The forest litter was 3 cm thick and had a dry mass (mean±SD) of 1,206±548 g m–2 in the understorey and 1,640±565 g m–2 in the gap. Treatments simulated natural heterogeneity in forest litter distribution and scraping by armadillos. The understorey vegetation was too sparse to warrant a clipping treatment. We recorded the number of tree seedlings at 20-day intervals within the first 5 months of sowing. Newly emerged seedlings were tagged to assess their survivorship. For each species, percentage seedling emergence was calculated until January 1995 (early summer), when seedlings stopped emerging from sown quadrats. To evaluate the dispersal limitation of recruitment, we monitored natural background emergence of each study species using the quadrats in which seeds of that species had not been planted. This was most important in the forest where Ligustrum and Gleditsia contributed to the local seed rain (Table 1). Emergence rates for these two species were adjusted by subtracting the mean seedling number recorded in paired control quadrats. Still, natural seedling dynamics did not affect the general outcome of the experiments (see Results). Survival was recorded at monthly intervals during the first post-emergence summer (January–March 1995) and over the following growing season (September 1995–March 1996). We assessed seedling mortality for three periods: the first summer from January to March 1995, autumn– spring 1995 (April–December), and the second summer from January to March 1996. These were conditional mortality rates as they were calculated from the number of seedlings present at the end of the preceding period. Seedling growth was recorded during the 1995–1996 growing season by measuring stem height for five randomly chosen plants per quadrat. To determine the longer-term survival of planted cohorts, all plots were re-sampled in March 2000, ≈5 years after emergence. At that point, grassland patches had totally recovered from our disturbances, while the forest gap plot was visually no longer distinguishable from the surrounding vegetation. Hence, patterns of seedling mortality on disturbed grassland and forest patches incorporated the influence of vegetation regrowth after March 1996. Statistical analysis Given that disturbance treatments were replicated within habitats, not across several similar sites, data were analysed separately for grassland and forest plots to avoid pseudoreplication implicit in any joint statistical analysis of unreplicated habitat types. Therefore differences between grassland and forest, as well as between
597 forest understorey and gap, should be interpreted with caution. For the grassland experiment, percentage seedling emergence was analysed using a mixed ANOVA model with blocks (n=5). Blocks were treated as a random effect as they were drawn from a larger population of possible locations within the grassland (Bennington and Thayne 1994). Tree species and disturbance treatment were included as fixed main effects. For each forest plot, data were analysed using a split-plot, mixed ANOVA model with blocks (n=5) as a random effect, and tree species (main plot) and litter treatment (sub-plot) as fixed effects. We calculated SEs for splitplot designs according to Steel and Torrie (1980) and tested for differences between means using t-tests. Percent emergence data were arcsine square-root transformed prior to analyses. Species showing very low emergence rates were excluded from statistical tests. Patterns in the number of seedlings surviving a given study period were analysed for each separate habitat using ANOVA models as described above. Seedling counts were square-root (x+0.5) transformed before analysis. Where appropriate, survival rates were compared using χ2-tests. Differences in cumulative height growth of seedlings in March 1996 were evaluated using mixed-model ANOVAs on log (x+1) data. Environmental conditions Measurements were taken in December 1994 to describe the shortterm impact of disturbance on microhabitat conditions for the seedling emergence phase. In grassland, we measured soil gravimetric moisture (%) in two quadrats per treatment in each block by taking three, 10-cm-deep soil cores. We recorded photosynthetic photon flux density (PPFD) at ground level using a LI-COR quantum sensor (Lambda Instruments, Neb.). Canopy and topsoil disturbances decreased soil water content by 4–7% relative to intact grassland (two-way ANOVA: P=0.003). Disturbance also increased light penetration to the soil surface (P
