Potential for multiple lag phases during biotic invasions: reconstructing an invasion of the exotic tree Acer platanoides



    1. School of Forest Resources and Environmental Science, Michigan Technological University, 1400 Townsend Drive, Houghton, MI 49931, USA
    Search for more papers by this author

    1. School of Forest Resources and Environmental Science, Michigan Technological University, 1400 Townsend Drive, Houghton, MI 49931, USA
    Search for more papers by this author

Christopher R. Webster, School of Forest Resources and Environmental Science, Michigan Technological University, 1400 Townsend Drive, Houghton, MI 49931, USA (fax 906/487 2915; e-mail cwebster@mtu.edu).


  • 1Perennial woody invaders often form persistent patches that significantly alter the structure and composition of native plant communities. Given their long generation times compared with ruderal invaders, these species may experience prolonged establishment phases between successful introduction and spread. Gap dynamics of shade-tolerant invaders could lead to multiple lag phases during the invasion process.
  • 2In order to investigate the potential for long or multiple lag phases, we reconstructed the invasion of Acer platanoides (a shade-tolerant, invasive, exotic tree) on an 1130-ha temperate forested island in Lake Huron, USA. We measured and mapped the spatial location of every A. platanoides≥ 0·5 m in height that had successfully established within a 728-ha forested park on the island. A simple age–diameter relationship, developed from a randomly selected subsample of the population, was used to assign an establishment date to each individual.
  • 3Following a 34-year establishment phase, the area occupied by ≥ 1 A. platanoides ha−1 increased linearly at a rate of 5·6 ha year−1 for 35 years, after which range expansion slowed. Population growth lagged behind range expansion, with rapid population growth associated with infill between parents. During the expansion phase, numerous satellite populations established but contributed little to population growth and spatial expansion because of the long time required for them to become reproductive. These satellite populations will most probably accelerate population growth and spread once they reach reproductive age.
  • 4Roads and trails provided important corridors for propagule movement away from developed areas over the course of the invasion. They also appeared to facilitate longer distance dispersals than would be expected given the biology of the species.
  • 5Synthesis and application. Our results suggest that shade-tolerant invaders with long generation times may undergo long establishment phases as well as periodic lags during the expansion phase. These lags may provide windows of opportunity for control but could easily be misinterpreted as signs that the population has reached an equilibrium density or the geographical extent of its spread. Additionally, roads and trails may provide important corridors for movement of propagules via non-standard means of dispersal.


Introduced woody plants pose significant challenges to the conservation of both forest and grassland ecosystems world-wide (Myers 1983; Braithwaite, Lonsdale & Estbergs 1989; Vitousek & Walker 1989; Wyckoff & Webb 1996; Hutchinson & Vankat 1997; Grice, Radford & Abbot 2000; Lepßet al. 2002; Barton et al. 2004). Woody invaders often form persistent patches that significantly alter the structure and composition of native plant communities, and may ultimately disrupt ecosystem function. For example, the conversion of sedgeland communities in Australia to shrublands by the invasive shrub Mimosa pigra L. threatens several resident and migratory bird species that rely on these sedge-dominated communities (Braithwaite, Lonsdale & Estbergs 1989). Similarly, in eastern North America shade-tolerant woody invaders that form monotypic stands (e.g. Acer spp., Berberis thunbergii DC, Lonicera spp. and Rhamnus spp.) have been shown to reduce tree species diversity by inhibiting the recruitment of native species (Woods 1993; Hutchinson & Vankat 1997; Martin 1999; Silander & Klepeis 1999; Frappier, Eckert & Lee 2003). Nevertheless, most woody invaders in North America and elsewhere were intentional introductions and many are still widely planted as ornamental or commercial species (Nowak & Rowntree 1990; Silander & Klepeis 1999; Rouget et al. 2002).

In order to better understand which factors govern the invasion process following the successful establishment of an introduced species, numerous mathematical models have been developed to characterize the spread of invasive organisms (Higgins & Richardson 1996; With 2002). These models range from reaction–diffusion (Skellam 1951) and stratified–diffusion (Shigesada, Kawasaki & Takeda 1995) models to complex integrodifference equation (Neubert & Caswell 2000) and neutral landscape (Gardner et al. 1987; With 2002) models. The invasion process can also be described in terms of a simple logistic growth curve, representing the three phases of the invasion process: introduction, colonization and naturalization (Radosevich, Stubbs & Ghersa 2003). These phases roughly parallel the phases of stratified diffusion: establishment, expansion and saturation (Shigesada, Kawasaki & Takeda 1995). The expansion phase in stratified diffusion is theorized to reflect one of three distinct forms: (i) a constant linear expansion; (ii) a biphasic expansion with an initial slow linear expansion followed by a more rapid linear expansion; or (iii) an exponential increase (types 1–3 stratified diffusion, respectively; Shigesada, Kawasaki & Takeda 1995). The saturation phase is reached when the population reaches a quasi-threshold density where population growth stabilizes and the geographical extent of the invasion remains approximately constant. This threshold is reached when niche occupancy and available resources limit the rate of spread (Radosevich, Stubbs & Ghersa 2003) or the population has reached a geographical limit to expansion (Shigesada, Kawasaki & Takeda 1995).

The expansion phase, which is typically of great interest to practitioners and policy makers, is modulated by the life-history characteristics of the invader as well as native plant community and landscape structure. For example, forest fragmentation may facilitate the spread of ruderal exotic invaders, but the isolation of forest remnants may slow the invasion of shade-tolerant perennial invaders (Janzen 1983; Brothers & Spingarn 1992). Likewise, roads and water courses may provide corridors for invasion (Parendes & Jones 2000). Buckley et al. (2003) found that logging roads serve as the primary conduits of dispersal for invasive plants into the interior of managed forest stands. The distance that these vectors allow propagules to spread can be more important than the frequency with which they carry propagules (Neubert & Caswell 2000). Consequently, a species’ capacity for short- vs. long-distance dispersal is an important determinant of the shape of the invasion curve during the expansion phase. For example, species that establish primarily within the range of the parent population expand the occupied area from the periphery at a relatively constant rate resulting in a type 1 shape (Shigesada, Kawasaki & Takeda 1995). Types 2 and 3 shapes result from mixtures of long- and short-distance dispersal and are strongly influenced by where long- and short-distance dispersers originate along the invasion front and how they are dispersed (Shigesada, Kawasaki & Takeda 1995). Consequently, by examining the spread of invasive organisms across heterogeneous landscapes, we may be able to elucidate the comparative importance of long- and short-distance dispersal events, and identify landscape features that act as either vectors, which accelerate and/or facilitate invasion, or barriers to invasion (With 2002). This knowledge could be used to design more effective control strategies that can be applied during the expansion phase, which is all too often when the negative impacts of invaders are first recognized.

Because of their longevity and size, the movement, growth and reproduction of introduced woody plants can be relatively slow, and impacts may take a relatively long time to materialize compared with herbaceous and animal invaders with short generation times (Frappier et al. 2003). For example, Norway maple Acer platanoides L. was introduced into eastern North America from Europe during the mid-1700s as an ornamental shade tree and then widely planted during the latter half of the 20th century to replace Ulmus americana L. lost to Dutch elm disease. In spite of its long history in North America, A. platanoides has only recently been recognized as a serious threat to native forest plant communities (Martin 1999; Webb et al. 2000). In North America, this species is moderately long lived (100–150 years; Nowak & Rowntree 1990), and usually does not produce viable seed until 25–30 years of age (Gordon & Rowe 1982). Acer platanoides is shade-tolerant (Kloeppel & Abrams 1995) and can be an aggressive colonizer of forest understoreys (Webb & Kaunzinger 1993) that can persist as advance regeneration for extended periods of time (Webster, Nelson & Wangen 2005). Long time lags between introduction and range expansion may mask the invasive potential of many common ornamentals (Frappier et al. 2003). Conversely, changes in extrinsic factors, such as land use and disturbance regime, may facilitate rapid range expansion by previously non-invasive species.

In order to identify landscape features that influence the invasion of A. platanoides and to clarify the potential for long or multiple time lags during invasions of introduced trees, we reconstructed the spread of A. platanoides on a temperate forested island. The island setting allowed us to reconstruct the invasion process across a relatively large finite landscape while maintaining a high degree of resolution (the location of every A. platanoides≥ 0·5 m tall was mapped). Detailed tree-ring analysis enabled us to backdate the invasion to the first successful establishment within the island's forest and follow the process forward to the present. Additionally, the presence of a minimum of 3 km of lake in any direction helped to isolate the island from outside influences, such as additional unaccounted for seed sources. There are, however, limitations to this approach because the island setting provides a limited range of landscape features (the island is mostly forested). Nevertheless, this setting provided an unprecedented opportunity to improve our understanding of the invasion process as it relates to shade-tolerant exotic trees.

Materials and methods

study site

This study was conducted on Mackinac Island, Michigan, USA (45°51′N 84°37′W), which lies in the north-western reaches of Lake Huron. The island is approximately 1130 ha in extent, of which 728 ha have been protected as a state park. Originally designated as the second national park after Yellowstone in 1875, it was returned to the state in 1895 and became the first Michigan state park. The island is approximately 80% forested, reflecting a variety of temperate forest types from upland coniferous stands comprised predominately of northern white cedar Thuja occidentalis L. to hardwood-dominated stands of sugar maple Acer saccharum Marsh. and American beech Fagus grandifolia Ehrh. The soil substrate on the island consists primarily of brecciated limestone, resulting in a predominance of calcareous soils (Milstein 1987).

field techniques

The spatial location of every A. platanoides (≥ 0·5 m in height) within Mackinac Island State Park was mapped with the aid of a high-precision global positioning system (GPS). Individual trees were located by running parallel transects between road segments on the island. Distance between transects was adjusted based on visibility constraints but at no time exceeded 15 m. Individual tree locations were recorded with a Trimble GeoExplorer CE (Trimble Navigation, Sunnyvale, CA). Trees in groups were mapped by establishing a single GPS anchor point and then recoding the distance and azimuth from the anchor point to each individual in the group. Long GPS observation times were used to increase the number of satellites used to establish a position and facilitate position averaging. To improve accuracy, GPS points were downloaded and post-processed using Pathfinder Office v2·90 (Trimble Navigation) and two local continuously operating reference stations (CORS). After correction, anchor points were cross-referenced with the distance and azimuth information collected for each tree in order to calculate individual x, y coordinates.

Diameter at breast height (1·37 m, d.b.h.) was recorded for A. platanoides trees ≥ 5 cm d.b.h.; stem diameter was measured 10 cm above the ground for smaller trees. Only the largest stem in a sprout clump was measured. The population was then divided into five 10-cm diameter classes from which trees were randomly selected for crown measurement and age determination. Increment cores were collected from a random sample of approximately 10 trees from each diameter class (10–19·9 cm, 20–29·9 cm, 30–39·9 cm, 40–49·9 cm, and ≥ 51 cm). Two perpendicular cores were collected from each tree at 30 cm above ground level. Twenty saplings and poles < 10 cm d.b.h. were also randomly selected and felled. Stem disks were collected from each at ground level and at 30 cm. For crown measurements, a new random sample of 10 trees from each diameter class was drawn. On each tree selected for crown measurement, we measured the diameter of the crown along its widest axis and a perpendicular axis. Crown edges were located using a clinometer.

invasion reconstruction

We reconstructed the invasion of A. platanoides into the forest reserve (state park) based on tree age and the spatial location of each tree during our survey of the island in 2002. Tree ages for all individuals in the population were estimated based on the relationship between total age and stem diameter (detailed below). This approach allowed us to back-date the invasion to the first establishment and track its progress forwards until 1995, when the youngest trees in our sample established. However, a limitation of this approach was that no information on individual tree mortality during the invasion was available. Consequently, our reconstruction depicted the spread of successful establishments or survivors. For simplicity and ease of interpretation, the invasion was reconstructed in 5-year time steps.

Increment cores and stem disks collected from A. platanoides were air dried and sanded for radial increment analysis. Phloroglucinol dye in solution with ethyl alcohol and hydrochloric acid was used to help distinguish questionable rings. Annual rings were measured under a binocular microscope with a stage micrometer. Skeleton plots were graphed for each tree ring series and visually cross-dated against growing season precipitation data (Gaylord, MI, weather station; National Oceanic & Atmospheric Administration) and a mean chronology based on a set of cores with clearly identifiable rings (Yamaguchi 1991). A regression of tree age as a function of diameter was then developed [total age = exp(2·53 + 0·322 × ln(diameter)), F1,76 = 146·4, P < 0·001, R2 = 0·66] and used to estimate age from diameter for the remaining trees in the population (for additional detail on contemporary population age structure see Webster, Nelson & Wangen 2005).

Once ages had been assigned to all individuals, stem diameters were back-dated through time and used to estimate total crown area based on the relationship between stem diameter and mean crown radius observed in our sample: mean crown radius = (0·458 + 0·454 × ln(diameter) + 0·272 × D-type)2, where D-type equals 1 if the stem diameter measurement was taken at 1·37 m and 0 if measured at 10 cm (F2,49 = 416·4, P < 0·001, adj. R2 = 0·95). Two-dimensional circular estimates of crown area were derived by buffering each tree location in ArcMap (Release 8.3, 2002; ESRI, Redlands, CA) using the predicted mean crown radius. Standard diagnostics were performed for each regression and transformations were used when necessary to homogenize error variance (Neter et al. 1996).

Spatial And Statistical Analysis

Several analytical techniques were used to examine the spread of A. platanoides into the island's forest. First, density interpolations were performed for each time step. Density interpolation was performed using the density function of the spatial analyst extension in ArcView. The radius of a circular 1-ha plot (56·419 m) was used as the search radius. Larger radii lead to coarser density surfaces and estimates of local abundance. Based on this interpolation, we were able to develop a density surface for each time period and calculate the maximum amount of area registering a density of ≥ 1 A. platanoides ha−1. In order to allow direct comparison with the range radius expansions theorized by Shigesada, Kawasaki & Takeda (1995), we converted the area of ≥ 1 A. platanoides ha−1 during each time period to a range radius value. The range radius value was calculated as the square root of the area occupied divided by the square root of π (Andow 1993). This conversion reduces geographical range expansion to the radius of a single expanding circular entity, and allowed us to compare our results to range expansion patterns reported for other invasives based on a common standardized methodology.

In order to quantify the area directly influenced by A. platanoides, crown maps were created for each time step based on the location of each extant tree and its estimated total crown area. Individual crown areas were summed for each period (total crown area). An estimate of the actual area of influence was calculated by dissolving the barriers between overlapping crowns and then summing the area of the new polygons. This approach allowed us to examine how the area directly influenced by the population changed over time, and was relevant because the most profound impacts on native plant communities typically occur directly under the dense crowns of A. platanoides (Martin 1999; Webb et al. 2000). Linear regression techniques were used to examine trends in population growth, expansion and influence over time. Residual and normal probability plots were used to assess regression assumptions.

The spatial distribution of A. platanoides in relation to roads and development was also investigated at each time step. The distance from each extant A. platanoides stem to the nearest road and developed area was calculated at each time step and then separately for only new recruits (stems that were not present in the previous time step) using the Nearest Features Extension in ArcView (Jenness 2004). Spatial data layers for roads/trails and land-use classifications were obtained from the Michigan Geographic Data Library (2003) and Michigan Department of Natural Resources (1993, 2001), respectively. A combination of multiple-linear regression and one-way analysis of variance (anova) was used to examine the relationship between long- and short-distance dispersals and proximity to roads over the course of the invasion. As the exact locations of the original seed trees were not always known (numerous trees have probably been removed from the developed areas because of physical damage or interference with new construction), long-distance dispersals were defined as new recruits that were in the 90th percentile of distance from development during a given year. This analysis allowed us to evaluate the relative importance of roads as corridors for invasion during various phases of the invasion process.


a. platanoidespopulation growth and spread

Following establishment within the forest reserve on Mackinac Island in approximately 1938, the population of A. platanoides increased slowly in abundance until approximately 1965 (Fig. 1a). After 1965, the population grew exponentially until 1990, when population growth decelerated (Fig. 1a). During the peak growth phase, the population was adding 254 individuals year−1. Population growth was greatest on the coniferous forest type, which hosted 74% of the population in 1995 and added 0·44 trees ha−1 year−1 between 1980 and 1990. By comparison, population growth rates on the mixed conifer–hardwood and hardwood forest types were 0·37 and 0·07 trees ha−1 year−1, respectively, during the same time period.

Figure 1.

Trends in A. platanoides population expansion (a) and spread (b, c). Population spread is expressed as the area occupied by a minimum density of 1 A. platanoides ha−1 (b) and as a range radius (c), which is the square root of the area occupied by a minimum density divided by the square root of π. This conversion reduces spatial spread to a radial expansion rate of an expanding circular entity.

The total area occupied by at least 1 A. platanoides ha−1 (area of occupation) increased slowly following the initial invasion of the forest reserve (Figs 1b and 2). Prior to approximately 1955, less than 4 ha of parkland had a minimum density of 1 A. platanoides ha−1. Around 1955, the area of occupation began to increase linearly at a rate of 5·6 ha year−1 (1955–90, area of occupation = −0·42 + 5·56 (year), F1,7 = 3197·6, P < 0·001, R2 = 0·998). The increase in area of occupation preceded the explosion in population growth by nearly a decade. This initial expansion period lasted approximately 35 years (Fig. 1b). After 1990, spatial expansion of the population slowed to 1·9 ha year−1.

Figure 2.

Spatial location and density of A. platanoides over the course of the invasion.

Converting area of occupation to a range radius (radius of a singular expanding circular entity) at each time step also resulted in a three-phase increase similar to the increase in area of occupation (Fig. 1c). The range radius increased slowly until 1955 and then expanded rapidly for approximately 5 years (0·05 km year−1) before decelerating to a more modest linear expansion rate of 0·02 km year−1 for approximately 30 years (1960–90, range radius = 0·324 + 0·0217 (year), F1,6 = 414·8, P < 0·001, R2 = 0·988). Between 1990 and 1995, expansion slowed to 0·005 km year−1. This pattern of expansion was most similar to type 1 stratified diffusion (Fig. 1c).

The pattern of invasion during the establishment phase indicated that for more than a decade after A. platanoides began invading the island's forest most individuals were found in a few isolated patches (Fig. 2). The rapid increase in area of occupation that began in 1955 appeared to have resulted from the establishment of numerous new colonies (Fig. 2). By 1985, older colonies (25–30 years) were beginning to expand, but much of the continued increase in area of occupation appeared to be associated with the establishment of more new colonies (Fig. 2). While fewer new colonies were observed between 1985 and 1995, older existing populations began coalescing and increasing in density (Fig. 2). This infill resulted in rapid population growth between the late 1970s and early 1990s (Fig. 1a). As population growth and range expansion slowed during the 1990s, even fewer new colonies were observed, suggesting either a decrease in propagule spread or the saturation of easily invaded sites.

Given that a few trees per hectare may have only a modest effect on the native plant community, depending on their size, we calculated the crown area of each A. platanoides based on its estimated diameter during each time period. Following 23 years of population growth, less than 0·12 ha of the island's forest were shaded by A. platanoides (Fig. 3). However, as early colonists matured and the population began to grow, the area directly influenced by the crowns of A. platanoides increased exponentially (1955–95, area of influence (ha) = 0·0872 – 0·0791 (year) + 0·00739 (year2), F2,8 = 2976, P < 0·001, adj. R2 = 0·999). By 1995, 14 ha of A. platanoides crowns shaded 8·7 ha (Figs 3 and 4) or 1·3% of the forest area on the island. This rapid increase in area of influence illustrated how invasions, even of long-lived species, often take us by surprise (Fig. 4).

Figure 3.

Relationship between A. platanoides crown cover (area of influence) and time since invasion.

Figure 4.

Acer platanoides invasion and crown expansion along a heavily invaded belt transect (400 × 1500 m). Transect location is indicated by the open rectangle on the inset of the island, and is just north of the interface between the forest reserve and residential development. Roads and trails are represented by solid lines.

influence of landscape features on invasion dynamics

Acer platanoides was strongly associated with the interface of developed areas and native forest (Fig. 2). Over the course of the invasion, roads appeared to have been an important vector for the dispersal and successful establishment of A. platanoides away from developed areas (Fig. 4). Results of a multiple-regression of dispersal distance from developed areas as a function of distance from road and invasion year indicated that long-distance dispersal distances (top 10% of distances from developed areas for new recruits at each time step) increased significantly over time (P < 0·001) while maintaining a strong proximal correlation with roads (P < 0·001; Fig. 5). In 1995, long-distance dispersals along roads were nearly 150 m further from developed areas than successful long-distance dispersals in the forest matrix (50 m from a road edge; Fig. 5). Over time, the furthest 10% of new recruits from developed areas were slightly closer to roads than the closest 10% of new recruits to developed areas (20·6 ± 0·78 m vs. 23·0 ± 1·4 m). However, the difference was not significant (F1,895 = 2·24, P= 0·135), indicating that the successful long-distance establishments during any given year were no further from a road than new recruits establishing near developed areas.

Figure 5.

Relationship between successful long-distance dispersal events and distance from the nearest road or trail over the course of the invasion. Dotted lines represent solutions to a multiple regression of distance from development (DistDevelopment, m) for long-distance dispersals (top 10% of dispersal distances from developed areas for new recruits during each time period) as a function of distance from road (DistRoad, m) and invasion year (yr) (ln (DistDevelopment) = 0·065 – 0·0679 ln(DistRoad) + 1·49 ln(InvasionYear), F2,447 = 201·9, P < 0·001, adjusted R2 = 0·47).


a. platanoidespopulation growth and spread

In 1965, approximately 25 years after the first successful establishment of A. platanoides in the forest reserve on Mackinac Island and approximately 40 years since the first known introduction (Webster, Nelson & Wangen 2005), only scattered clumps of A. platanoides were present along the periphery of developed areas on the southern end of the island. At that time, it would have been difficult to find an A. platanoides in the forest, much less perceive the rapid population growth and range expansion that would ensue over the course of the next three decades. This prolonged lag phase highlights the dilemma posed by long-lived exotic woody invaders (Frappier et al. 2003). In the case of A. platanoides, one of the ‘most popular’ and widely planted urban street trees in eastern North America is belatedly coming into focus as a serious invader of native forests (Nowak & Rowntree 1990; Martin 1999; Webb et al. 2000). Similarly, well-documented lag phases between wide-scale planting and invasive spread have been reported for commercially important exotic Pinus spp. in the southern hemisphere (Richardson, Williams & Hobbs 1994; Rouget et al. 2004).

The population of A. platanoides on the island entered the expansion phase around 1955, as new individuals and colonies expanded the area occupied. The linear increase in area occupied and range radius of the population during the expansion phase are suggestive of type 1 stratified diffusion (Shigesada, Kawasaki & Takeda 1995). This pattern of range expansion is typically generated when offspring primarily establish within the neighbourhood of the parent population via short-distance dispersal events (Shigesada, Kawasaki & Takeda 1995). During the first phase of expansion on the island, most new colonies and recruits were proximate to developed areas that hosted source populations. However, by 1975 numerous satellite populations resulting from long-distance dispersal events were evident. Given the time required for these satellite populations to reach reproductive age (25–30 years for open grown trees; Gordon & Rowe 1982), they still contributed relatively little to range expansion and population growth. The apparent decline in the rate of population growth and spread of the contemporary population, even though large areas of forest remain uninvaded, suggests that most of the easily invaded sites near parent populations are moderately well saturated. This slow down could represent an additional lag phase in population growth and range expansion. Nevertheless, as satellite populations come of age, their expansion may become the primary driver of population growth and spread (Moody & Mack 1988; Neubert & Caswell 2000). This demographic shift could greatly accelerate the invasion and might best be described by a range expansion pattern similar to type 2 stratified diffusion. Type 2 stratified diffusion is characterized by a biphasic expansion phase with initially slow growth followed by a higher constant rate, which could reflect an increase in the importance of long-distance dispersals (Shigesada, Kawasaki & Takeda 1995). Nevertheless, multiple lag phases and/or shifts in expansion type are not features of contemporary stratified diffusion theory, but may represent logical theoretical extensions. At a metapopulation scale, the expansion of an invasive tree, such as A. platanoides, may again slow as older satellite populations converge with the advancing front and new non-reproductively active satellite populations are established by chance long-distance dispersal events. In other words, exotic species with long generation times may expand their ranges in relatively discrete pulses with multiple lag phases. Eventually, once a critical density of parent populations are established, range expansion would become continuous until all suitable sites have been occupied.

Given that A. platanoides can aggressively colonize (Wyckoff & Webb 1996; Martin 1999; Webb et al. 2000) and persist in shaded understoreys (Kloeppel & Abrams 1995; Sanford, Harrington & Fownes 2003), it can build up a large pool of advance regeneration in relatively undisturbed native forests. These understorey saplings are typically non-reproductive and may persist for upwards of 30 years (Bertin et al. 2005; Webster, Nelson & Wangen 2005). Consequently, the reproductively active portion of the population on Mackinac Island, during any given time period, was probably less than would be predicted based on tree age alone. However, saplings growing in canopy gaps or under high shade may be able to recruit into the overstorey nearly a decade before native shade-tolerant species (Webster, Nelson & Wangen 2005). Highly variable understorey residence times prior to canopy recruitment and subsequent seed production make it difficult to accurately model the invasion dynamics of shade-tolerant woody perennials in forested systems. Following successful introduction and establishment, these species may appear harmless for decades before being released by natural or anthropogenic disturbance (e.g. overstorey tree-fall gaps or logging). Nevertheless, canopy turnover rates in most forests would still result in the gradual conversion of native forests to exotic species through natural gap dynamics. Large-scale anthropogenic disturbances or catastrophic natural disturbances that remove most of the native canopy during a single event, however, could result in a rapid shift towards exotic dominance in forests with well-developed seedling layers of A. platanoides or other perennial woody invaders tolerant of shade (Hutchinson & Vankat 1997; Webb et al. 2000; Collier, Vankat & Hughes 2002; Yates, Levia & Williams 2004). Other exotic tree species that are intolerant of shade, such as Ailanthus altissima (Miller) Swingle, may be able to gain a foothold in large tree fall gaps and then invade rapidly following more intense disturbance (Knapp & Canham 2000; Call & Nilsen 2003). Consequently, it may be difficult to demonstrate the risk posed by exotic, woody perennials, especially if they are commercially important species, given that widespread range expansion and deleterious impacts may lag decades behind widespread introduction.

The potential for multiple lag phases during the expansion phase of population spread may require somewhat more complex invasion models, which incorporate population age structure, the temporal and spatial availability of habitat, and changes in dispersal mode as the population expands (Higgins, Richardson & Cowling 2001). For example, the theory of stratified diffusion could be expanded to include a punctuated or stepped expansion phase that is sensitive to both positive and negative feedbacks. In the case of shade-tolerant exotic trees, lags during the expansion phase may be strongly influenced by the long time needed to reach reproductive maturity, which is also influenced by the frequency of disturbances that facilitate canopy recruitment. For example, the availability of canopy gaps proximate to parent populations or probable long-distance dispersal sites may be an important determinant of lag behaviour because it influences the time needed for individuals to reach reproductive maturity (as is the case with suppressed understorey saplings awaiting the formation of a canopy gap). Similarly, spread may slow as the invasion radiates away from dispersal corridors and more readily invaded habitats into a less easily invaded matrix. All of these phenomena are scale dependent and may become obscured at increasingly large temporal and spatial scales. Additionally, at the metapopulation scale non-synchronous introductions could conceivably ‘average out’ lags displayed by individual populations. Nevertheless, lag behaviour during the expansion phase may have important implications for control and risk assessment efforts at local and regional scales.

vectors and barriers to establishment

While several studies have demonstrated that A. platanoides does not require disturbed edge environments for establishment or survival (Webb et al. 2000; Sanford, Harrington & Fownes 2003), road corridors appear to facilitate the dispersal of this species at least in forested landscapes (Anderson 1999). Most studies of invasive spread along roads have concluded that roads act as disturbances that facilitate the establishment of exotic species (Greenberg, Crownover & Gordon 1997; Parendes & Jones 2000; Silveri, Dunwiddie & Michaels 2001; Call & Nilsen 2003; Rentch et al. 2005), but the role of roads as corridors for the flow of propagules has been less clear (cf. Harrison, Hohn & Ratay 2002; Buckley et al. 2003). Our results suggest that roads may be associated with long-distance dispersal events that move A. platanoides well beyond the interface between developed areas and native forests. During each time period examined, the furthest dispersals from developed areas were along road and trail corridors. This association is probably the result of both standard and non-standard dispersal mechanisms. In developed areas, A. platanoides is commonly planted as a street tree and rains a tremendous amount of seed onto the road surface and passing vehicles. Consequently, in addition to samaras floating down road corridors that act as wind tunnels or simply blowing along the road surface, A. platanoides may hitchhike on passing vehicles. Transportation on the island is restricted to foot traffic, bikes, horses and horse-drawn wagons and carriages, all of which could effectively transport seeds varying distances. Horses may be especially influential in the transportation of seed as mixtures of seed and manure may also adhere to hooves and cart wheels. However, the consolidation of manure and a discontinued practice of transporting manure into the forest for composting (J. Dykehouse, Mackinac Island State Park, personal communication) are probably responsible for at least the early association between roads and A. platanoides observed in our reconstruction. Contrary to expectations, roads still appear to be one of the primary vectors for the dispersal of A. platanoides into the native forests on Mackinac Island. This result highlights the importance of human-mediated jump dispersal in the spread of invading organisms (Suarez, Holway & Case 2001).

Another possible interaction between horses and A. platanoides is that A. platanoides may benefit from a fertilization effect associated with the deposition of urine and manure along roads and trails. Nutrient additions in conjunction with disturbance may interact synergistically to promote invasions by exotic species (Kercher & Zedler 2004; Perry, Galatowitsch & Rosen 2004; Rickey & Anderson 2004). However, given the high nutrient use efficiency of A. platanoides compared with shade-tolerant native species (A. saccharum; Kloeppel & Abrams 1995) it is unclear how these factors may influence successful establishment of this species. Additionally, a strong affinity between A. platanoides and trail edges has been documented in areas without horse traffic (Anderson 1999). Nevertheless, high nutrient and resource levels may allow A. platanoides to compete against more aggressive native species or persist in otherwise hostile environments, and warrants further study.


Acer platanoides easily invaded forests adjacent to developed areas with ornamental plantings and appears capable of successfully spreading through contiguous forests by aggressively competing with native species for growing space (Webb et al. 2000). Our results suggest that rare long-distance dispersals and gap dynamics may ultimately govern invasion of this species into relatively undisturbed expanses of forests. Given the long establishment phase and the possibility of multiple lag phases during the expansion phase, one of the greatest risks associated with shade-tolerant exotics may be underestimating their potential as successful invaders.


We would like to thank the Mackinac Island Community Foundation, Mackinac Island State Park, and the city of Mackinac Island for their support in this endeavour. Emily Duerr, Jennifer Griggs, Roy Johnson and Kathryn Nelson provided invaluable assistance in data collection and processing. Yvonne Buckley, Linda Nagel and two anonymous referees provided helpful comments on an earlier draft of this manuscript. Financial support was provided by the School of Forest Resources and Environmental Science, Michigan Technological University, and the McIntire-Stennis Cooperative Forestry Program.