By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Due to essential maintenance the subscribe/renew pages will be unavailable on Wednesday 26 October between 02:00- 08:00 BST/ 09:00 – 15:00 SGT/ 21:00- 03:00 EDT. Apologies for the inconvenience.
1Many studies have searched for traits that characterize successful invaders. Unfortunately, very few generalizations have emerged from this work. It seems that the traits of successful invaders are idiosyncratic and context-dependent. Unless we are to study each potential invader in each possible target community individually, we will need a new approach.
2We introduce a framework for predicting traits that are likely to confer success in a given ecosystem. Our approach considers the prevailing environmental conditions, the traits of resident species, and the traits of potentially invading species.
3Our approach can be applied to ecosystems where the environmental conditions and/or disturbance regime have recently changed, to predict the range of trait space occupied by (i) native species at risk of local extinction, (ii) native species that can persist under the present conditions, and (iii) successful invaders. Our approach can also be used to identify unoccupied viable trait space (i.e. vacant niches) that might be at risk of invasion.
4Synthesis. Understanding invasions resulting from rapid changes in environmental conditions and invasions resulting from the colonization of vacant niches would be a major step forward for invasion biology. The conceptual framework described here is not limited to plant invasions: the same approach can be used for any taxa (e.g. insects, fish, mammals and marine invertebrates) and could also be used to predict species responses to environmental change.
In 1965 and 1974, Herbert Baker listed a suite of characteristics that he thought the ideal weed might possess. These traits included high seed output under favourable conditions, rapid growth through the vegetative phase to flowering, and the capacity to disperse effectively over short and long distances (Baker 1965, 1974). Perhaps it was because this list of traits seemed so intuitively right that so many researchers have been attracted to the idea that there is a suite of traits that distinguish invasive plant species. The practical use of such research is also appealing: if we could identify traits that predict which species are likely to become highly successful, quarantine services could screen incoming plants and prevent costly and damaging invasions, and conservation organizations could target recently naturalized species before they become major problems.
In this paper, we describe a new approach to understanding and predicting the suite of traits that are likely to succeed in a given ecosystem. This approach considers the traits of resident species and the prevailing environmental conditions, as well as the traits of the invading species. We present the approach as a tool for understanding plant invasions, but the same conceptual framework should also work for other taxa (e.g. insects, fish, mammals and marine invertebrates).
Under what circumstances do invasive species usually establish?
The majority of plant invasions seem to occur in areas where environmental conditions such as disturbance regime or resource availability have recently changed (Davis et al. 2000; Facon et al. 2006). For instance, the sclerophyll shrublands of Sydney, Australia, are seldom colonized by invasive species except in areas where well-drained, nutrient-poor soils are modified by the additional water and nutrients associated with urban run-off (Lake & Leishman 2004). In New Zealand, few introduced species are able to establish in the shaded understorey characteristic of the forest systems that dominated the land prior to the arrival of humans (McGlone 1989; Jesson et al. 2000; Lloyd et al. 2006). However, areas where the frequency of disturbance has been increased (e.g. by fire, agriculture or roadside mowers) are dominated by introduced species such as Ulex europaeus, Cytisus scoparius and Hieracium pilosella.
From an evolutionary perspective, it is unsurprising that few introduced species are able to colonize a new environment unless there has been a substantial change in environmental conditions. The resident native species will have undergone many generations of selection for traits that enhance their success under the prevailing environmental conditions (including the prevailing disturbance regime). Most introduced species will have evolved under quite different conditions (different competitors, a different disturbance regime, and different soil fertility, rainfall and temperature). The chances that an introduced species will be able to successfully establish and compete with resident species are extremely small, unless: (i) the environmental conditions have recently changed, (ii) the invading species fills a niche that was not occupied by any of the resident species, and/or (iii) release from natural predators and pathogens allows the invading species to achieve a fitness equal to or greater than that of competing native species (Blossey & Nötzold 1995; Keane & Crawley 2002). In the following section, we introduce a model that we hope can be used to understand and predict invasions under the first two of these scenarios.
Foundations of the model: predicting the range of plant traits that will succeed under given environmental conditions
The mean and variance of plant trait distributions differs between groups of species growing under contrasting environmental conditions. For instance, species’ mean seed mass is log-normally distributed across the species coexisting in a community (Leishman et al. 2000). The mean of this log-normal distribution increases along a latitudinal gradient from temperate regions to the tropics (Lord et al. 1997), and as one moves from ecosystems with low net primary productivity to ecosystems with high primary productivity (Fig. 1). Similar shifts can be seen in other traits: for instance in species with evergreen leaves, specific leaf area tends to increase with increasing rainfall, and leaf life span decreases with increases in mean annual temperature (Wright et al. 2005).
Recent developments in informatics mean that it is now possible to build a predictive model for trait frequency distributions under different environmental conditions. To build such a model we need large plant trait data sets and detailed information on the environmental conditions at the sites from which the trait data were collected. Data bases that contain information on important ecological traits for thousands to tens of thousands of species are becoming common. For instance, we have seed mass data for around 13 000 species (Moles et al. 2005b), genome size data for around 5000 species (Beaulieu et al. 2007) and specific leaf area data for over 2000 species (Wright et al. 2004). These trait data sets can be combined with information on the locations in which species occur, using information from source papers, and from large syntheses of occurrence data (e.g. Salvias; http://www.salvias.net/pages/index.html). The resulting geo-referenced trait data can be combined with information on the environmental conditions at each site from both ground-based measurements (e.g. New et al. 1999) and from ever-more sophisticated climate modelling software (e.g. BIOME 4, Kaplan et al. 2003).
We can use a multivariate statistical approach to combine these large plant trait data bases with information on the environmental conditions under which the traits occur (including variables such as soil fertility, precipitation, temperature and disturbance frequency). Thus, we can use existing empirical data to parameterize a model that gives a predicted distribution of traits under given environmental conditions. For simplicity, we show the distribution of a single trait in Figs 1 and 2 (a model graph). In reality, the model would need to consider the distribution of multiple ecologically important traits simultaneously in multidimensional trait space. A species’ position in this multidimensional space will reflect aspects of its ecological strategy (Reich et al. 2003). The model could include interaction terms to account for correlations among variables.
Case 1: Predicting invasions in response to change in environmental conditions
It should be possible to predict the range of new strategies that will succeed after a change in environmental conditions by comparing the expected distribution of traits under the old conditions with the expected distribution of traits under the new conditions at a site. To calculate these expected distributions, one would run the model described above once for the original (undisturbed) conditions at a site, and again for the present conditions, and overlay the two results (Fig. 2). This reveals three zones of trait space:
Zone 1 (dark grey) is a region of trait space that used to be viable under the original conditions at the site that is no longer viable under the present conditions. This zone is expected to be occupied by native species that are now at risk of local extinction.
Zone 2 (grey) is a region of trait space that is viable under both the present conditions, and the original conditions at the site. This zone is expected to be occupied predominantly by native species. As these native species have undergone generations of selection for success under the prevailing conditions, they are likely to be relatively resistant to invasion.
Zone 3 (light grey) is a region of trait space that was not viable under the original conditions at the site, but which is viable under the present conditions. It represents newly opened viable trait space (unoccupied niches). In the past, this space would have been filled mostly as a consequence of evolutionary changes in resident species, and by migration of species from neighbouring communities, especially those with environmental conditions more similar to the present conditions at the site. In the present day, many ecosystems are exposed to a constant rain of introduced species’ propagules (Lockwood et al. 2005). Some of these propagules may be from species that evolved under environmental conditions similar to the present conditions at the site. If the present conditions are more similar to the conditions under which the introduced species evolved than to the conditions under which the native species evolved, then the introduced species will be better able to occupy the vacant niches than will the previous resident species.
We illustrate this approach with an example from New Zealand. About 3000 years ago, 85–90% of New Zealand was forested, and disturbances to forest structure were relatively rare (McGlone 1989). In contemporary environments, the temperature, precipitation and soil fertility are still relatively similar to what they were in pre-human times – but in many parts of New Zealand (especially roadsides and paddocks) the disturbance frequency has dramatically increased (McGlone 1989). The range of strategies (and thus traits) viable in pre-human forests will clearly be very different from the range of viable strategies in present-day roadsides or paddocks. Zone 1 (Fig. 2) in this case is populated by the long-lived, tall forest species. Zone 2 (a small region for highly disturbed areas like roadsides or paddocks, since the magnitude of the shift between the present and the past conditions is so great) is populated by the few opportunistic native species, such as those that used to colonize areas after major disturbances (e.g. landslides or volcanic eruptions). Zone 3 is populated by short-lived, fast-growing colonists. Interestingly, most of the introduced species that succeed under the new regime of high-frequency anthropogenic disturbance in New Zealand come from Europe and Asia (Williams et al. 2000), where they evolved under a regime of high-frequency anthropogenic disturbance.
The model can also help us to distinguish circumstances in which invasive species may be causing or hastening declines of native species from cases in which invasive species are simply colonizing sites that are no longer suitable for previously resident native species (the latter occurs when native species occupy Zone 1 in Fig. 2). In the first situation, removing invasive species will enhance the native species’ recovery. In the second situation, the recovery of native populations can only be achieved by restoring the disturbance regime and environmental conditions at the site to levels that the original residents can tolerate. In this case, focusing research and control efforts on the successful colonists rather than on the changes in environmental conditions that allows them to establish is akin to taking an aspirin to control a headache when the underlying problem is actually a brain tumour (although it may sometimes be necessary to remove the introduced species in order to restore the original conditions).
Case 2: Predicting invasions of previously vacant niches
Some of the most dramatic invasions happen when an introduced species is able to fill a vacant niche in an ecosystem (e.g. the introduction of terrestrial mammals to New Zealand). Our model can be used to predict these invasions.
One would begin by running the model once for the present-day conditions at the site. This will give a distribution of potentially viable trait values. Next, one would overlay the trait values for the present occupants of the site. At first, differences between the observed and expected data will mostly reveal additional variables that need to be included in the model. Once the model has been refined, differences between the observed and expected trait distributions will reveal trait combinations that are expected to succeed under the present conditions that are not being utilized by the present occupants of the site. These areas are vacant niches – areas of trait space that are likely to be vulnerable to invasion. Because it is evolutionarily simpler for species to modify existing structures than to evolve entirely new structures or physiological pathways, most vacant niches will not be gaps in the distribution of traits with continuous values (such as offspring size, plant height or wood density), but rather the possession of traits such as nitrogen fixation, C4 photosynthesis or the ability to produce a particular kind of defensive chemical (traits that are either present or absent).
We illustrate this approach with an example from Hawai’i. Hawai’i is a tropical archipelago with high rainfall, warm temperatures (at least at sea level) and substrates that are rich in many nutrients, but relatively poor in nitrogen. If we were to use our model to compile information on the traits of early successional species growing under similar conditions in other parts of the world, we would almost certainly find nitrogen-fixers among the viable strategies (see Aidar et al. 2003; Kreibich et al. 2006 for examples from Atlantic forests). However, if we were to overlay the trait distribution of native Hawaiian species before European colonization, we would find a striking absence of early successional vascular nitrogen-fixers (Vitousek et al. 1987). The fact that early successional nitrogen-fixing strategies could flourish in Hawai’i was unfortunately demonstrated by the introduced nitrogen-fixer Myrica faya (Vitousek et al. 1987).
Both the present model and limiting similarity theory predict that successful invaders should be functionally different from resident native species (Emery 2007). Our model extends limiting similarity theory by predicting which traits will differ between invading species and native species, as well as the likely direction and magnitude of these differences.
The framework we propose here includes information on the environmental conditions under which invasive species are establishing, the traits of the resident species and the traits of the invading species. We think this holistic approach to invasion ecology will yield useful generalizations, and allow us to move beyond the present situation where each invasive species and each ecosystem have to be studied individually. Our approach also provides a framework for predicting the traits of native species that are likely to be threatened by future environmental changes. Of course, we do not expect this approach to explain every single invasion. There will always be unique situations in the ecological world. However, we believe that this model will allow us to predict invasions that result from changes in environmental conditions, and invasions that result from occupation of previously vacant niches. This would be a major step forward for invasion biology.