Plant communities in harsh sites are less invaded: a summary of observations and proposed explanations

Within the invasion ecology literature, it is often noted that abiotically stressful environments are typically less invaded by non-native plants than nearby less-stressful environments. However, until now no one had collected and summarized examples of this pattern. This paper first compiles evidence that plant communities in many harsh habitats are less invaded, and then synthesizes possible explanations for this pattern. We discuss that harsh sites may be less invaded because, compared to moderate sites, they may receive lower propagule pressure, particularly from well-suited plants, and because their abiotic and biotic characteristics may make them inherently more resistant to invasion.


Introduction
Discerning the patterns and underlying causes of plant invasions is a central goal of invasion ecology. Many studies have attempted to identify characteristic traits of invasive plant species and invasible communities (e.g. Rejmá nek 1989;Lonsdale 1999;Alpert et al. 2000;Theoharides and Dukes 2007;Chytrý et al. 2008;Rejmá nek et al. 2013;Gioria and Osborne 2014). In addition to identifying which habitats tend to be highly invaded-often disturbed and high-resource sites-these and other authors often note that some habitats are characteristically less invaded than others-specifically, abiotically stressful, or 'harsh', sites. However, support for this assertion is often given in the form of individual examples, and there has not yet been a detailed compilation of harsh habitats that are reported to be less invaded. Furthermore, there has not yet been a synthesis of hypotheses for why harsh habitats may be less invaded than more moderate environments. In this paper, we provide a summary of evidence for the phenomenon of low invadedness of harsh habitats and discuss hypotheses for understanding this pattern. We first compile an annotated list of harsh habitats reported to have low levels of invasion by non-native plants, and then outline two major classes of hypotheses for why harsh habitats may be less invaded-propagule limitation mechanisms and invasion resistance mechanisms-and discuss their management implications.
We define harsh habitats as sites with the regular presence of one or more abiotic stressors, which can include naturally low levels of critical plant resources (e.g. nitrogen, light, oxygen or water), presence of toxins (e.g. heavy metals or salts) or temperature extremes (Table 1). Harsh habitats share the characteristic of abiotic stress, which limits the rate of resource acquisition, growth or reproduction in resident plants (Grime 1989). To persist in harsh habitats, plants must possess resource conservation or acquisition strategies to adapt to resource scarcity (e.g. increased resource use efficiency, nutrient resorption, modified roots; Funk 2013), or tolerance/avoidance strategies to deal with chemical toxicity (e.g. synthesis of detoxifying metabolites and proteins, ion compartmentalization or exclusion; Raskin et al. 1994;Parida and Das 2005) or freezing (e.g. production of 'antifreeze' proteins; Pearce 2001).
We consider a harsh site less invaded if the richness, cover and/or biomass of non-native invasive species is lower compared with similar, less harsh sites. When a stressor occurs in discrete patches (e.g. edaphic stressors or dense forest shade), less harsh sites can include the habitat matrix adjacent to patches of harsh habitat. When a stressor is caused by continuous climatic variation (e.g. low temperatures or aridity), less harsh sites are defined by their position along the stress gradient. Because native species richness or cover may also decline with greater abiotic stress, when available we include information on whether richness or cover of invasives in harsh sites is lower relative to natives. In cases where the proportion of invasive species out of the total species pool declines with increasing stress, this suggests that there is something unique about the phenomenon of invasion that causes this pattern (e.g. different traits of invasive vs. native species or differences in propagule pressure between harsh and moderate sites). Regardless of the corresponding trend in natives, differences in total invasive richness or abundance across a stress gradient provide useful information to managers trying to exclude invasive species.
Throughout this paper we use the term 'invasives' to mean non-native plant species capable of establishing, spreading and causing ecological and/or economic damage, and the term 'non-natives' to simply indicate species outside of their native range. Although we focus on invasive species, we sometimes include examples of non-natives that are not reported to be problematic because (i) evidence of impacts may not be well known or described and (ii) such species may eventually become invasive after a lag period (Simberloff 2009). We use the term 'non-native(s)' when there is a lack of evidence of impacts or when describing multiple non-indigenous species together, some of which may not be known invasives. We also note that hypotheses for low invadedness of harsh sites can also apply to patterns of invasion by native species (Alpert et al. 2000), and discuss certain cases where relevant.
After compiling a list of examples, we discuss two major classes of hypotheses that have been proposed to explain why harsh habitats are less invaded. (i) Propagule limitation mechanisms suggest that characteristics of harsh sites, such as isolation, small size and lower rates of human visitation and disturbance, may reduce their exposure to non-native propagules, which in turn limits successful invasions (Lockwood et al. 2005;Simberloff 2009). (ii) Invasion resistance mechanisms invoke the stressful conditions of harsh habitats as either direct or indirect causes of reduced invasion (Alpert et al. 2000;Shea and Chesson 2002). Specifically, the pool of potential invaders may either be physiologically intolerant of the stressful conditions, and/or the stressful conditions might increase the impacts of biotic resistance from resident native species. These two classes of hypotheses are not mutually exclusive, and in many cases both may be in effect. While we do not seek to test the relative importance of these hypotheses in particular harsh habitats or overall, we note that their implications for management of plant invasions in harsh habitats may differ.

Nitrogen-poor sites
Many natural ecosystems are relatively low in available soil nitrogen (N), and N is often a limiting factor for plant growth (Vitousek and Howarth 1991). Naturally low N levels may contribute to why some terrestrial ecosystems, such as calcareous and sandy grasslands (Kolb et al. 2002) and arid and semi-arid communities (McLendon and Redente 1991;Brooks 2003), have low cover of non-natives. Inversely, areas of higher fertility in relatively low-nutrient systems such as alpine communities have a higher mean number of non-native species, compared with lower fertility areas (McDougall et al. 2005). Higher N availability is linked with higher rates of invasion generally: the frequently observed pattern of disturbance favouring invasive species over natives has often been attributed to increased availability of N after disturbance (Hobbs et al. 1988;reviewed in Davis et al. 2000). Experimental studies have shown that increases in soil N tend to benefit invasive species more than native species in a variety of terrestrial systems, including both those severely and moderately poor in nitrogen (Wedin and Tilman 1996;Kolb et al. 2002;Abraham et al. 2009;Liancourt et al. 2009; see also review by James et al. 2011), implying reduced success of invasives relative to natives in N-poor sites. Experiments have shown similar patterns in aquatic systems: compared with native macrophytes (aquatic plants), non-native submersed macrophyte biomass (Chase and Knight 2006) and asexual propagule performance (Xie et al. 2010) were greater in high N treatments. Greater relative performance of non-natives over natives in N-enriched treatments has been found in wetland plants as well (Rickey and Anderson 2004;Holdredge et al. 2010). Moreover, anthropogenic N enrichment in the form of agricultural activities or roadside pollutants increase biomass and richness of N-loving invasive species over native species in low-N plant communities such as calcareous grasslands in the Netherlands (Willems 2001;Lee and Power 2013), sandy grasslands of Hungary (Tö rö k et al. 2014), coastal grasslands in California (Kolb et al. 2002) and arid (Brooks 2003) and semi-arid ecosystems in western North America (McLendon and Redente 1991). Similarly, Maron and Connors (1996) concluded that the presence of an N-fixing shrub species facilitates greater cover by non-native species in a California coastal prairie at the expense of the number and cover of native species.
Additional evidence that low nitrogen levels hinder invasion is that intentional N impoverishment carried out by restoration practitioners, especially in grassland systems, can reduce the impact of invasive plants, again not always in the context of systems that are considered particularly poor in nitrogen (Morgan 1994;Paschke et al. 2000;Prober et al. 2005;Steers et al. 2011;Tö rö k et al. 2014;reviewed by James et al. 2011). Reducing soil N can increase abundance of native species relative to invasive (or 'weedy') species, although success has been mixed (Wilson and Gerry 1995;Burke et al. 2013; also reviewed by Corbin et al. 2004;James et al. 2011). Techniques for N-impoverishment include biomass removal (Willems 2001;Perry et al. 2010), topsoil removal (Buisson et al. 2008) and carbon addition (Wilson and Gerry 1995;Reever Morghan and Seastedt 1999;Burke et al. 2013).

Phosphorus-poor environments
Phosphorus (P) limitation may be a barrier to invasion, though it is often confounded with co-limitation of other nutrients (Kueffer et al. 2008;Haubensak and D'Antonio 2011). Numerous studies have found a positive relationship between soil P levels and both richness and cover of non-native species, and a negative relationship between native species richness and increased P.  Rozema et al. (1997) For example, P-enrichment in naturally low-P Australian urban bushland was associated with higher non-native richness and lower native richness (Lake and Leishman 2004). Morgan (1998) found much higher soil P levels at the invaded edges of a Themeda trianda remnant Australian grassland than in the less invaded interior.
In observational studies in Australian Banksia woodland (Fisher et al. 2006), subtropical wetland (Boughton et al. 2011) and riparian zones in southeast China (Xiao-Yun et al. 2006), high-P environments were more invaded than low-P environments. Dimitrakopoulos et al. (2005) found that soil P was positively correlated with aboveground biomass of non-native species in a Mediterranean grassland. Experimental evidence has shown that increased P is associated with greater success of invasive species relative to natives in both terrestrial and aquatic systems. As with soil N, experimentally increasing soil P tends to lead to superior performance of invasives relative to natives. For example, in a greenhouse experiment Leishman and Thomson (2005) found that when nutrients were added to low P soils in Hawkesbury Sandstone communities in Australia, invasive species had higher survival and growth rates than natives. Leishman et al. (2004) found that increased soil P in stormwater runoff areas in Australia (defined as .150 mg kg 21 ) resulted in significantly increased invasive species cover proportional to natives. Cherwin et al. (2009) found that both absolute and relative cover of invasive grasses (Bromus spp.) in experimental plots in a Colorado grassland were positively associated with P addition. In serpentine soils, P fertilization increased the invasion rate and dominance of invasives within 2 years (Huenneke et al. 1990).
Phosphorus is often the most limiting nutrient in freshwater aquatic systems, and enrichment of waters with P and other nutrients is thought to increase invasibility (Engelhardt 2011). For example, invasive Hydrilla verticillata (Royle) (hydrilla) presence was positively correlated with total P in Florida lakes (Gu 2006), and hydrilla had a competitive advantage over native Vallisneria americana (American eelgrass) in higher P (and other nutrient) soils, while eelgrass was the stronger competitor in nutrient-limited soils (Van et al. 1999).

Saline, sodic or alkaline environments
Environments with highly saline, sodic and/or alkaline soils, such as tidal or inland salt marshes and alkali sinks, have been anecdotally observed to be less invaded (Baker 1986). Salinity, sodicity and alkalinity can negatively affect plant growth through ion toxicity, effects on osmotic potential and interference with plant nutrition. These sites are also often temporarily inundated, which can subject plants to anoxia.
Though wetlands are generally susceptible to invasion (Zedler and Kercher 2004), cover of non-natives decreased along a gradient of increasing salinity in Southern California salt marshes, while native cover increased with salinity (Uyeda et al. 2013). Other studies have found that particular wetland invaders are unable to establish in highly saline areas (e.g. Zedler et al. 1990;He et al. 2012; but see Daehler and Strong 1996). Human alteration of hydrological processes through shoreline development and water management practices (e.g. controlled flooding) can decrease salinity and facilitate invasion by non-native brackish species (Zedler et al. 1990;Meslé ard et al. 1993;Konisky and Burdick 2004;Silliman and Bertness 2004). Application of salt in these areas has been proposed as a method for controlling nonnatives without harming the native plant community (Kuhn and Zedler 1997;Uyeda et al. 2013). Conversely, increased salinity has been implicated in the success of invasive salt-tolerant tamarisk (Hultine et al. 2010).
Research on invasion of sodic or alkali soils is sparse, though it appears that these sites can act as edaphic refuges for native species (Dawson et al. 2007). For example, non-native Lolium multiflorum (ryegrass; now Festuca perennis) that dominated a non-sodic matrix was virtually absent in adjacent alkali sinks, whereas native Hemizonia pungens ssp. pungens was more abundant in the alkali sinks (Veblen and Young 2009). It is likely that amelioration of naturally sodic soils for the purpose of agriculture (for example, by addition of gypsum: see Qadir et al. 2001) may displace native species and facilitate invaders, though experimental evidence for this is currently lacking.

Serpentine sites
Serpentine soils are edaphically harsh, often characterized by low macronutrient (N, P, S) and micronutrient (Ca) content, high concentrations of toxic heavy metals (Cr, Ni), low Ca : Mg ratio and low soil moisture (Whittaker 1954;Chiarucci and Baker 2007). While western North American grasslands are highly invaded relative to other habitat types (e.g. forests), interspersed serpentine 'islands' have been historically less invaded relative to surrounding non-serpentine grasslands (Kruckeberg 1954;Walker 1954;Baker 1986). Many native species that were widespread prior to invasion by Mediterranean annuals in California's grasslands find refuge in serpentine soils (Kruckeberg 1984;Huenneke et al. 1990), where natives are generally more abundant than on nonserpentine soils, either in absolute terms or relative to non-natives (McNaughton 1968;Harrison 1999a). In northern California's Sedgewick Reserve, native species richness was higher than invasive species richness on spatially isolated rocky serpentine outcrops/hummocks, while the converse was true in the surrounding serpentine grassland. This was attributed to abiotically stressful conditions on shallow, lower-nutrient hummocks (Gram et al. 2004). Within serpentine sites in a northern California grassland, species diversity of invasives relative to natives increased with increasing P, Ca : Mg ratios, soil depth and water-holding capacity. Moreover, there was higher absolute and relative native species richness in serpentine than non-serpentine meadows (Harrison 1999b). Serpentine plant communities in New Caledonia are similarly depauperate in non-natives (Jaffre 1992). Experimental manipulations have shown that amelioration of serpentine stressors tends to favour invasive species over natives. In California, nutrient addition to serpentine soils generally increased biomass and fecundity of invasive grasses to a greater extent than native species (Huenneke et al. 1990;O'Dell and Claassen 2006;Going et al. 2009). Hobbs et al. (1988) found that invasive grasses almost entirely replaced native forbs in fertilized serpentine plots. These studies, along with the observations that anthropogenic N deposition is linked to greater invasion of serpentine sites (Turitzin 1982;Weiss 1999Weiss , 2006, suggest that N limitation may be the most important stressor controlling invasion in these systems.

Xeric sites
Water availability often limits plant growth and acquisition of other resources in terrestrial environments (Chapin et al. 1987). Across soil moisture gradients, xeric locations generally tend to have a lower proportion of invaders than mesic locations (Cronk and Fuller 1995;Rejmá nek et al. 2013). This pattern is evident in a wide variety of systems, including temperate forests (Rejmá nek et al. 2013), Hawaiian grasslands (Goergen and Daehler 2001), temperate grasslands (Larson et al. 2001), semi-arid savannahs (Harrington 1991) and deserts (Bashkin et al. 2003;Miller et al. 2006;Brooks 2009). When arid landscapes contain patches with greater soil moisture, these patches are often more highly invaded (e.g. Stohlgren et al. 2001). The importance of moisture in determining patterns of invasion is also evident from temporal fluctuations in water availability; for example, wet years tend to favour non-natives in arid and semi-arid systems (Burgess et al. 1991;Hobbs and Mooney 1991;Dyer and Rice 1999). In contrast, however, Lortie and Cushman (2007) showed that the proportion of exotic species increased along a natural coastal dune gradient of decreasing soil moisture (and nitrogen).
Experimental evidence suggests that soil moisture may influence the relative performance of native and invasive species. Daehler (2003) found natives to have superior physiological performance compared with invasives under water-limited conditions in 8 of 12 studies reviewed.
Similarly, water additions favoured non-natives over natives in a New Zealand grassland (White et al. 1997) and in shortgrass steppe (Milchunas and Lauenroth 1995).
However, while there appears to be a general trend of invaders responding positively to water addition, this is not always the case. Seabloom et al. (2003) found that year-round water additions did not significantly increase invasion of non-native annual grasses into native perennial stands in a California grassland, but instead enhanced native perennial grass reestablishment within stands of non-native annuals. Similarly, water additions had no significant impact on invasion or competitive interactions between several noxious invasives and monocultures of 10 native perennials in a water-limited system in western Montana (Maron and Marler 2008) or invasion of Holcus lanatus into perennial grass stands in California coastal grasslands (Thomsen and D'Antonio 2007).
There is also some question as to whether the pattern of lower richness of invasive species at drier sites holds with respect to cover of invasive species. Historically, deserts have been considered more resistant to invasion than more mesic biomes due to perceived physiological stress of chronic water limitation (Baker 1986;Loope et al. 1988). In a global analysis of plant species data from nature reserves, Lonsdale (1999) found that sites in deserts and savannas had proportionally fewer nonnative species than those in more mesic biomes. However, many arid and semi-arid rangelands are now highly invaded, often by a small number of species. For example, Bromus tectorum (cheatgrass) has become the dominant weed of the US Great Basin (Knapp 1996) and is able to invade even isolated and undisturbed locations (Belnap and Phillips 2001). Other examples from deserts include Brassica tournefortii (Asian mustard) and Pennisetum ciliare (buffelgrass) (Rejmá nek et al. 2013). Thus, if well-adapted non-natives can reach arid sites, they may become successful invaders.

Rocky outcrops and shallow soils
In rocky and/or shallow soils, patterns of invasion may be affected by a number of stressful conditions, including reduced access to water and nutrients, extremes in soil temperatures and mechanical resistance to soil disturbance (Houle and Phillips 1989). Soil depth influences plant community dynamics (Grime and Curtis 1976), and shallow soils often have lower invasive cover relative to natives than adjacent deep soils (Kolb et al. 2002;Talluto and Suding 2008;Emam 2015;but see MacDougall et al. 2006). Similarly, rocky outcrops often have a lower proportion of invasive species than the surrounding matrix, such as in non-native dominated coastal scrub grasslands of California (Steers 2011), piedmont of the southeastern USA (Wyatt 1997), South African inselbergs (Hopper 2002) and shallow-soil sites in New Zealand (Wiser and Buxton 2008). However, some shallow-soil sites in Australia (Hopper 2002) and tropical inselbergs in parts of Africa and Latin America (Porembski 2000) are highly invaded, at least in terms of absolute numbers of species. Invasion in these contexts may be due to increased human disturbance and/or higher fertility of interstitial soils compared with other shallow-soil rocky outcrops (Porembski 2000;Hopper 2002). The latter hypothesis suggests that nutrient limitation may drive invasion patterns in these sites.

Periodically inundated environments
Terrestrial systems that flood on a seasonal or periodic basis, such as vernal pools, ephemeral riparian areas and wetland margins, are characterized by another form of harshness: plants that inhabit these systems must tolerate or avoid periods of anoxia. Many plants that are invasive in upland habitats are intolerant of prolonged submersion and are therefore unable to establish permanently in adjacent inundation-prone areas. Vernal pools in the Central Valley of California, for example, are often regarded as 'islands' of native plant communities within a matrix of heavily invaded grassland (Holland and Jain 1988). The deepest (most heavily inundated) pools generally have the lowest richness of invasives and the lowest cover relative to natives (Gerhardt and Collinge 2003), likely due to reduced growth and reproduction of invasives when inundated (Gerhardt and Collinge 2007).
Similarly, although wetlands and riparian areas are generally highly invaded (Zedler and Kercher 2004), some periodically inundated riparian areas and wetland margins are less invaded than surrounding uplands. Evidence for reduced performance of non-natives with increasing inundation intensity occurs in both absolute terms (e.g. Stokes 2008;Tanentzap et al. 2014) and relative to natives (e.g. Mills et al. 2009). Wetlands that are substantially less invaded than surrounding uplands may experience particularly harsh inundation. Indeed, a study comparing the invasibility of riparian wetlands before and after implementation of flow control (amelioration of inundation intensity) found that when inundation became less intense, cover of non-native species increased, both in absolute terms and relative to natives (Catford et al. 2011). Further, in a vernal pool mesocosm experiment, deeper pools-which experienced more intense inundation-were more resistant to invasion (Collinge et al. 2011). Similarly, Tanentzap et al. (2014) found that ephemeral wetland plant communities that experienced more intense flooding contained a smaller proportion of non-native species compared with native species. Comparable dynamics exist in rice fields, and rice farmers have long employed periodic flooding to control (primarily non-native) weeds (Smith and Fox 1973;Williams et al. 1990).
Despite their harshness, periodically inundated areas may in some cases be prone to invasion. Although inundation may generally displace invasive species more readily than natives, it can also displace natives (e.g. Collinge et al. 2011) and increase availability of space, light and other resources, thereby ameliorating other forms of stress (e.g. dry, low-light or low-nutrient conditions) linked with resistance to invasion. In studies finding long-term dominance of invasives in inundated areas, the success of invasives may be due to a superior ability to colonize disturbed, post-inundation habitats (Barden 1987;Collinge et al. 2011) or a superior ability to compete under either inundated or non-inundated conditions (Taylor and Ganf 2005;Schooler et al. 2010). Some work suggests that despite being invasible, inundated areas may exhibit a pattern of pulsed 'resetting' of invasion each time an area floods (Flores et al. 2005;Stokes 2008;Mills et al. 2009), implying a dynamic equilibrium of partial invasion that may persist indefinitely.

Bogs
Bogs receive water mainly from precipitation, are characteristically low in nutrients and oxygen and are highly acidic. These stressors may limit invasibility to nonspecialized plants, either singly or in combination. While most other types of freshwater wetlands are highly invaded, bogs typically have few, or even zero, non-native invaders (Zedler and Kercher 2004;Rejmá nek et al. 2013). Lambdon et al. (2008) found that only roughly 10 % of all the non-native plants found in Europe occurred in mires, bogs and fens, making them the least invaded non-marine ecosystems across Europe. Similarly, in a comprehensive analysis of the flora of Catalonia, the Czech Republic and Great Britain, Chytrý et al. (2008) found that bogs were among the habitats with the lowest proportion of non-native plants: the percentage of neophytes (plants that arrived after 1500 AD) out of the total flora was zero for Catalonia and the Czech Republic, and only 0.2 % for Great Britain.
Observational evidence suggests that invasions that do occur in bogs can often be attributed to disturbances that ameliorate at least one stressor and/or increase the spread of non-native propagules. Loope et al. (1992) reported that in montane bogs in Haleakala National Park, Hawaii, disturbance by feral pigs allowed several invasive plants to displace native plants. Undisturbed bogs were less invaded, and native plants tended to recover in bogs where pigs were excluded. In a tamarack bog in Ohio, USA, hydrologic changes that resulted in an AoB PLANTS www.aobplants.oxfordjournals.org amelioration of a variety of stressors (e.g. increasing nutrient levels and pH) were implicated in the invasion of both native and non-native species (Miletti et al. 2005).
While invasions of non-native species into bogs are rare, more common is the encroachment of native vegetation, which is again generally attributed to anthropogenic reductions in stressors. Experimental manipulations often demonstrate the importance of multiple combined stressors in reducing the invasibility of bogs to encroaching natives. For example, in a mesocosm study, Limpens et al. (2003) concluded that N deposition improved performance of vascular plants, but that expansion into bogs was still limited by a high water table and low P. Similarly, in a fertilization experiment in a desiccated bog, Tomassen et al. (2004) concluded that while N deposition could enhance invasion of a grass-Molinia caerulea-invasion of other vascular species would occur only with concurrent increases in P.

Shaded terrestrial systems
Low-light terrestrial environments such as closed-canopy forests have long been considered resistant to invasion (Baker 1986;Rejmá nek 1989;Von Holle et al. 2003), and evidence suggests that many shaded ecosystems are indeed less invaded. For instance, forests with closed canopies in California and Europe have lower proportions of non-native species than nearby open habitats (Rejmá nek et al. 2013). Shaded forest understories often have lower total abundances of non-natives compared with forest edges (e.g. in old-growth forest in Indiana, USA; Brothers and Spingarn 1992), and certain invasive species may only be found in canopy gaps within dense forest (e.g. Ailanthus altissima (tree of heaven) in old-growth forest in New York, USA; Knapp and Canham 2000). Additionally, lower percentages of non-native weeds have been found in roadside habitats with closed canopies relative to roadsides with open canopies (Frankel 1977;Parendes and Jones 2000).
Low light is an abiotic stress that is usually directly related to a biotic cause: the shading species. Despite a potentially confounding role of competition, there is compelling evidence that low invadedness of highly shaded habitats may be caused, at least in part, by poorer performance of invasive species relative to natives in lowlight conditions, which has been demonstrated in many experiments that manipulate light alone. For example, in Hawaii, experimental reduction of light availability reduced growth and biomass of invasive grasses to a greater extent than native tree and shrub species (Funk and McDaniel 2010). In a review of studies comparing the performance of native to non-native plant species, Daehler (2003) found that native performance equalled or exceeded that of non-natives in low-light conditions in 7 of 10 studies.
The knowledge that invasive species often perform more poorly than natives under greater shade stress has been used as a restoration tool. For example, retaining plantation trees has been used in tropical and subtropical forests to suppress invasive grasses (Parrotta et al. 1997;Loh and Daehler 2008). The extent of shading likely determines the chances of invasion success; for instance, establishment of potentially invasive Miscanthus species is inhibited only at very low-light levels (West et al. 2014).
Despite the general pattern of lower relative performance of invasive species in highly shaded terrestrial environments, a number of studies have shown that certain invasive species outperform natives under all light conditions (Pattison et al. 1998;Zheng et al. 2009), and that increasing shade does not necessarily reduce invader abundance (Baars and Kelly 1996;Cabin et al. 2002). Thus, as in all stressful habitats, invasibility of low-light environments depends largely on the traits of the nonnative species (Martin et al. 2009;Funk 2013).

Shaded wetland, riparian and aquatic systems
As in terrestrial systems, low light availability appears to limit invasions of non-native plants in wetland and riparian zones-systems that are otherwise typically highly invaded (Alpert et al. 2000;Zedler and Kercher 2004). Canopy gaps and high-light areas created by flooding and sediment deposition are often implicated in the high invasibility of wetlands and riparian zones (Zedler and Kercher 2004;Schnitzler et al. 2007), suggesting that light limitation reduces invasibility. Field surveys corroborate this. For example, the presence of an invasive grass, Glyceria maxima (reed sweet-grass), along stream banks in Victoria, Australia, decreased with greater riparian overstory cover and was absent in areas of highest riparian shading (Loo et al. 2009). Similarly, the invasive grass Phragmites australis (common reed) was significantly less likely to be found in roadside drainage ditches in Quebec, Canada when dense woody cover was present (Albert et al. 2013). In a meta-analysis of non-native species diversity in European riparian forests, Schnitzler et al. (2007) implicated greater light availability as one reason that willow-poplar communities had a higher number and percentage of non-native species than other riparian community types.
Experimental work, though sparse, also points to a limiting effect of light on invasions in wetlands. For example, Maurer and Zedler (2002) found that rhizome establishment and biomass of the invasive Phalaris arundinacea (reed canarygrass) were reduced under greater canopy shading under both field and greenhouse conditions in Wisconsin, USA. As in terrestrial environments, shade may give natives a competitive advantage over invasive species in wetlands. In a greenhouse experiment, Chen et al. (2013) simulated the effects of different understory light levels on the growth of native and invasive mangroves, finding that while invasive mangroves responded negatively to high levels of shade, native mangroves did not.
Low light may be less important in reducing plant invasions in purely aquatic systems. While light limitation often influences the distribution of submersed macrophytes (Bornette and Puijalon 2011), evidence that light limitation hinders non-native submersed macrophytes more than natives or reduces invasibility is lacking or mixed (e.g. Ali et al. 2011). In a test of photosynthetic rates under different light intensities of one non-native and six native submersed macrophytes from Lake George, NY, USA, Madsen et al. (1991) classified the non-native Myriophyllum spicatum (Eurasian watermilfoil) as highlight adapted and all six natives as 'shade-tolerant'. However, shading of up to 94 % incident light did not hinder establishment success of invasive M. spicatum in artificial stream channels in California, USA (Zefferman 2014).

High elevations
High-elevation regions are characterized by low temperatures, a short growing/productive period (daily or seasonal), high UV exposure, low available soil nutrients, increased water stress and, in some locations, daily freeze-thaw cycles. These stressors may help explain the observation that high-elevation areas have relatively few non-native species (Alpert et al. 2000;Rejmá nek et al. 2013). Surveys spanning more than 14 locations over seven continents found a general pattern of decreasing numbers of non-native species from low-and midelevations to high elevations (Western and Juvik 1983;  Reviews of country-and continent-wide distributions of invasive and non-native species in the Czech Republic, China and North America have also found that mountainous or high-elevation areas have fewer non-native species than lower regions (Baker 1986;Feng and Zhu 2010;Pyšek et al. 2012).
It is less clear whether this pattern holds for proportional representation of non-natives within the species pool. Native species richness typically declines at high elevations as well (but see Becker et al. 2005, Alexander et al. 2011, and studies that report information on native species richness typically do not explicitly report proportional richness. However, non-native richness was found to decline with elevation both absolutely and relative to native plant richness in the Italian Alps (Marini et al. 2012(Marini et al. , 2013, the Swiss Alps (Becker et al. 2005), in woody species of the Kashmir Himalayas (Khuroo et al. 2011), in roadside weeds of Hawaiian Islands (Western and Juvik 1983;Jakobs et al. 2010), in roadside vegetation of Chile (Alexander et al. 2011) and in grasses of the northern Andes .2000 m (Giraldo-Cañ as 2010). Similarly, Palmer (2006) found that non-native species richness had a sharper decline with elevation than native species richness in North America. However, in the Canary Islands, studies suggest an absolute but not a relative decline in non-native plant richness at high elevations (Arévalo et al. 2005;Alexander et al. 2011).

High latitudes
Like high elevations, high-latitude regions (.508N or S) exert stresses associated with low temperatures that can prevent or slow the establishment, growth and lifecycle completion of plants. High latitudes also exhibit high seasonal variation in photoperiod, with long periods of low light energy.
Surveys along latitudinal gradients at the scale of continents or countries have shown that non-native species richness decreases above 408. This decline with latitude has been observed for naturalized species (introduced species that can sustain a population without requiring repeated reintroductions, sensu Richardson et al. 2000) in all biomes and vegetation community types in continental Europe (Sax 2001), Chile (Fuentes et al. 2013) and for all non-native plants in the contiguous USA (Stohlgren et al. 2005). In riparian forests in Europe, latitudes above .508N were host to the fewest non-natives Schnitzler et al. (2007). Because native species richness also decreases with increasing latitude, it remains unclear whether the proportion of non-native to native species typically declines with latitude in these cases. However, for North American flora, Palmer (2006) found that non-native plant richness is more negatively affected by latitude than native richness.
The Sub-Antarctic and Antarctic regions also show a general pattern of decreasing non-native plant richness with higher latitude, but not necessarily relative to native richness. In a survey of 25 Sub-Antarctic Islands, densities of non-native species per island (richness/area) decreased with latitude of the island, but proportion of non-native to total richness did not (Chown et al. 1998). The Antarctic continent only has two naturalized vascular plants and these are restricted to the milder maritime region (Frenot et al. 2005), but Antarctica has only two native vascular plants (Robinson et al. 2003 For the Arctic, it has been noted that naturalized nonnatives represent a low to null percentage of species in regional floras (Elven 2011), but research has not explicitly compared this with proportions of non-natives in adjacent lower-latitude areas. Hotspots of non-native richness and abundance in areas of high human activity do exist. For example, areas around settlements on the Arctic archipelago of Svalbard (74 -818N) have 28 -37 nonnative species present (native species: 165), but it is unclear how many of these are sustained only because of continuous reintroductions or conditions maintained by humans (Ware et al. 2012).
Experiments, though rare, suggest that the harsh conditions imposed by high elevations and latitudes could be responsible for low invasibility. For example, it has been shown with lab simulations of polar temperatures and field sowing experiments at polar latitudes (.608) that vascular plant species introduced to the Arctic or Antarctic (whether from temperate or sub-polar regions) often either do not germinate, or fail to reproduce and form a sustainable population (Smith 1996;Ware et al. 2012). Furthermore, winter snowpack depth has been shown to limit the establishment and population growth rates of two invasive shrubs at higher elevations in the Sierra Nevada mountains of California (Stevens and Latimer 2015).

Hypotheses
The preceding sections provide considerable evidence that abiotically stressful sites are less invaded than more moderate environments [summarized in Supporting Information]. Although exceptions occur, this pattern holds across multiple dimensions of plant stress. However, potential explanations of this phenomenon include more than abiotic stress alone, because successful invasion is ultimately influenced not only by the abiotic environment, but also by species traits and propagule pressure (Lonsdale 1999;Theoharides and Dukes 2007). Here we propose a framework for explaining the pattern of low invadedness of harsh sites (Fig. 1). We divide explanations for this pattern into two broad classes. Propagule limitation mechanisms assume that non-native species are less likely to be transported to, or disperse among, harsh sites. Invasion resistance mechanisms assume that potential invaders are not necessarily dispersallimited but instead limited by the abiotic stressors of the harsh site or by interactions with resident species. In reality, these classes of mechanisms occupy two ends of a continuum in which dispersal and site characteristics are both important, and hybrid mechanisms occur when propagule limitation and invasion resistance mechanisms operate in concert. These mechanisms are relevant to the invasibility of any ecosystem, but here we discuss them in the context of harsh sites specifically.
Two lines of empirical evidence can help distinguish propagule limitation mechanisms from invasion resistance mechanisms in harsh sites. First, the observation of increased invasion through time without an accompanying reduction in a stressor suggests that propagule limitation may have been at least partially responsible for initially low levels of invasion. Second, if experimental amelioration of a stressor results in greater success of invaders without an accompanying increase in propagule pressure, it suggests that the stressor is indeed a direct or indirect cause of invasion resistance. Many of the studies we reviewed imply invasion resistance by showing increased invader abundance or competitive performance following reduction of stressors. However, studies that simply correlate richness or cover of invaders with metrics of harshness are generally not able to distinguish the relative importance of propagule limitation vs invasion resistance. In the following sections, we discuss in greater detail the differences in the evidence for, and underlying causes of, propagule limitation and invasion resistance as drivers of low invadedness of harsh sites.

Propagule limitation mechanism
The inverse relationship between plant invasion and harshness could be driven by greater propagule limitation at harsh sites compared with more moderate sites. This could include low numbers of propagules arriving at one time, and/or low frequencies of introductions (Simberloff 2009). The propagule limitation mechanism allows for the possibility that any non-native species could invade harsh sites if introduced in sufficient numbers. For instance, low numbers of non-native propagules reaching harsh sites are often implicated in low levels of invasion, without the assumption that the propagules belong to stress-tolerant species (Lonsdale 1999;Alpert et al. 2000;Rejmá nek et al. 2013).
Harsh sites may experience particularly low propagule pressure of non-native species because people (i.e. potential vectors) may visit harsh habitats at lower rates (Alpert et al. 2000) due to cultural/aesthetic preferences or practical or institutional limitations. Harsh sites may be more likely than moderate sites to be designated as reserves, either because they are rare or because they are less useful for development (Scott et al. 2001), thus reducing human intrusion. Evidence exists for reduced human impacts at a range of harsh sites, including low nutrient and rocky soils (Scott et al. 2001;Lavergne et al. 2005), and arid, high-latitude and high-elevation regions (Scott et al. 2001;Sanderson et al. 2002).
Evidence for propagule limitation includes a positive correlation between human use and plant invasion among similarly harsh sites. For example, areas of higher human occupancy or use in high-latitude regions tend to have greater non-native species richness compared with areas of lower human presence at similar latitudes (Chown et al. 1998(Chown et al. , 2005Fuentes et al. 2013). Pickering et al. (2007) found a positive association between tourism infrastructure and presence of non-native species in highelevation alpine regions of an Australian national park. We do note that sites more visited by humans may also be more disturbed, thus confounding propagule limitation with invasion resistance mechanisms (see below).
Additional evidence to support the propagule limitation mechanism includes successful invasion of cosmopolitan invaders into harsh sites following their introduction (e.g. Alexander et al.

Invasion resistance mechanisms
Invasion resistance mechanisms refer to the ecosystem properties that reduce invasibility of a site to non-native species, and we divide these into two main types: abiotic resistance and biotic resistance (see also Alpert et al. 2000;Rejmá nek et al. 2013). Abiotic resistance plays a direct role in reducing the invasibility of a harsh ecosystem, while biotic resistance operates indirectly, as mediated by resident species.
Abiotic resistance. The abiotic resistance mechanism requires that the degree of physiological stress imposed by a harsh site prevents the establishment or population growth of potential invaders, even in the absence of negative interactions with native species. Potential invaders can be intolerant of many different stressors, including critical resource scarcity (e.g. low nutrients, water, light), environmental conditions that slow or halt metabolism (e.g. temperature extremes, anoxia) or the presence of toxins (e.g. heavy metals, salts) (Table 1; see also review by Alpert et al. 2000).
Signals of abiotic resistance involve failure by potential invaders to establish in harsh sites despite the presence of propagules and absence of competitors. Because of the ethical difficulties of experimentally introducing nonnative propagules into harsh habitats, however, experimental tests that could definitively exclude propagule limitation as an operative mechanism are rare (but see Ross et al. 2008;Stevens and Latimer 2015). Instead, experiments that show increased invasion of harsh sites when stressors are ameliorated are the most common and conclusive evidence for abiotic stress reducing invasibility, particularly when controlling for the level of competition or when conducted in the absence of competition from natives. For example, both field and laboratory soil transplant experiments in the Antarctic have shown that non-native non-vascular plants are already present in the soil propagule bank and will germinate and survive when temperature and moisture are raised (Kennedy 1996). Many studies outlined throughout the 'Harsh Habitats' sections above show a positive effect of increases in water, light and nutrients on the richness or cover of invasive species in sites where these resources are limiting.
Biotic resistance. Under the biotic resistance mechanism, negative interactions between potential invaders and resident native species reduce invasion risk (Elton 1958; Levine et al. 2004). Here we argue that these negative interactions may play a particularly important role in harsh sites. This is because invasive plants commonly have life history traits that are associated with highresource costs, such as rapid growth rates, high specific leaf area and low water-use efficiency (reviewed in Blumenthal 2006;Rejmá nek et al. 2013), which may make them especially vulnerable to biotic resistance in harsh sites.
Competition with native plant species (especially in the establishment phase) is one example of biotic resistance. Although competitive interactions are often considered less important in stressful ecosystems (Grime 1977), species competition can determine whether or not coexistence is possible even under harsh conditions (Chesson and Huntly 1997). In harsh sites, resource scarcity may limit the competitive advantage conferred to fast-growing potential invaders, and biotic resistance may become the proximate cause that prevents their invasion. Stresstolerant natives may outcompete stress-intolerant invaders if invaders need to devote a greater proportion of resources to metabolic maintenance (survival) and therefore have fewer resources available for growth and fecundity (Shea and Chesson 2002). Similarly, if stressadapted natives are more efficient at assimilating scarce resources, they may drive resources down to levels that are inhospitable for potential invaders.
It can be difficult to disentangle whether potential invaders of harsh sites are physiologically intolerant to abiotic conditions rather than competitively inferior to the native plant community, and in many cases both mechanisms may be in play. The most effective means of distinguishing abiotic and biotic resistance is through controlled experiments that simultaneously manipulate harsh conditions and native abundance. For example, through factorial experiments, Going et al. (2009) and Gerhardt and Collinge (2007) showed that abiotic stressors and competitors had additive negative effects on non-natives in serpentine grassland sites and vernal pools, respectively. It is also important to note that in certain cases, native biota may be sources of abiotic stress as well as competitors, thereby muddying the distinction between abiotic and biotic resistance mechanisms. Examples include low light in the understory because of shading by canopy vegetation, and reduced pH in bogs due to excretion of acids by mosses. In a meta-analysis of the effects of biotic resistance on invasive plant species, Levine et al. (2004) concluded that competitive exclusion rarely keeps invaders out entirely but often limits their abundance. Therefore, if competitive exclusion is the primary mechanism driving low invasion in a harsh site, we may be more likely to see evidence in terms of low cover rather than low richness of non-natives.
Biotic resistance to non-native plants may also come from resident herbivores and pathogens. Native plants adapted to harsh conditions often have enhanced chemical or structural defences due to the high resource cost of repairing damaged tissue, while high resource-adapted invaders are less likely to be resistant to disease and herbivory (Chapin 1991). Therefore, resident enemies may contribute to the pattern of lower invadedness in harsh sites by suppressing potential invaders, which could lead to competitive exclusion by better-adapted native plants, or to mortality even in the absence of competitors (Shea and Chesson 2002).
Furthermore, while 'enemy release' is often cited as a mechanism behind successful invasion of non-native plants (Blossey 2011), enemy release may be less important in sites that are naturally low in resources. Nonnative plants that are well adapted to high resource conditions are likely to benefit most from enemy release, because plants with resource-grabbing life history strategies generally allocate energetic resources towards high growth rates rather than defensive chemicals (Blumenthal 2006). Therefore, enemy release may increase the competitive abilities of fast-growing nonnatives in high resource sites, but this advantage may be less important in low resource conditions where nonnatives are unable to capitalize on available resources (Blumenthal 2006;Blumenthal et al. 2009). This potential difference in the effects of enemy release in high and low resource sites may contribute to why harsh sites are relatively less invaded than moderate sites. Experiments that manipulate herbivore or pathogen abundance (e.g. exclusion experiments) on plant communities across a stress gradient would be most useful in distinguishing the role of resident enemies and/or enemy release in contributing to the pattern of low invadedness of harsh sites, but correlative studies can yield insights as well (e.g. Blumenthal et al. 2009;Seabloom et al. 2009).
Hybrid mechanism: species -site mismatch Species -site mismatch is a hybrid mechanism that assumes invasion resistance mechanisms are in effect, but focuses on propagule pressure of species that are 'well-matched' to harsh sites. While there may be propagule pressure from non-native species at a harsh site, in a species -site mismatch scenario there is a low probability that these propagules are from species having the necessary resource conservation/acquisition or stress tolerance/avoidance traits required to thrive in a harsh habitat. As a result, harsh sites could be effectively propagule-limited, because they are receiving few, if any, propagules that are true potential invaders.
There are many explanations for why harsh sites may receive fewer propagules of well-matched species. Lower human visitation in harsh sites, as discussed above, could result in lower probabilities of transferring well-matched propagules between sites with similar abiotic stressors. Humans may also directly bias propagule pressure: intentionally introduced plants are often chosen for traits-such as fast growth and maturation rates-that make them easy to cultivate in moderate environments where humans are likely to live and work (Colautti et al. 2006;Martin et al. 2009;Rejmá nek et al. 2013). Similarly, species that thrive in human-dominated, disturbed landscapes (typically r-selected species) are more likely to be transported unintentionally by humans (Martin et al. 2009), while stress-tolerant, specialized plants that could invade harsh environments may be less likely to grow in human-dominated landscapes and therefore less likely to be transported to harsh sites.
Even when non-native stress-tolerant plants are successfully introduced to harsh habitats, their secondary dispersal to other isolated harsh sites may be limited by their dispersal traits. While many invasive plants have longdistance dispersal strategies [e.g. seeds that disperse by wind, water or animal movement (Schupp 2011)], it may be evolutionarily unfavourable for specialized stresstolerant species to disperse outside of their harsh environments when those habitats are isolated across the landscape (Cousens et al. 2008). Even without taking into account differences in dispersal traits, the fact that many harsh sites (e.g. bogs, serpentine soils or alpine sky-islands) are isolated and small in size means that these harsh habitat 'islands' may receive fewer naturally dispersing propagules from well-adapted non-native species from similarly harsh and previously invaded sites (sensu island biogeography theory, MacArthur and Wilson 1967).
Evidence of the species -site mismatch mechanism comes from an increasing number of studies where harsh environments have become invaded once appropriately matched species have arrived: Lepidium latifolium (perennial pepperweed) has become a notorious invader of wetlands and sodic soils (Renz and Blank 2004;Reynolds and Boyer 2010); deserts in the American west have recently been heavily invaded by arid-adapted Bromus tectorum (cheatgrass) (Mack 1981), Brassica tournefortii (Asian mustard) and Pennisetum ciliare (buffelgrass) (Rejmá nek et al. 2013); serpentine-tolerant ecotypes of Aegilops triuncialis (barbed goatgrass) have recently invaded many California serpentine sites (Lyons et al. 2010) and introductions of shade-tolerant non-natives have led to invasion of shaded forest understories in both tropical and temperate forests (Sanford et al. 2003;Martin and Marks 2006;Martin et al. 2009;Holmes et al. 2014).

Disturbance: evidence for invasion resistance mechanisms
The common observation that disturbance facilitates invasion (see review by Hobbs and Huenneke 1992) can be evidence for invasion resistance mechanisms when disturbance ameliorates harsh conditions. When resource limitation hinders potential invaders, disturbance can increase resource availability directly, alleviating abiotic resistance and/or reduce resource uptake by the native plant community, alleviating biotic resistance (Davis et al. 2000). Disturbances that facilitate invasions in harsh sites include soil disturbances by animals, such as feral pigs in Montane Hawaiian bogs (Loope et al. 1992), and gophers in serpentine grasslands (Koide et al. 1987;Hobbs and Mooney 1991). Physical disturbances like treefalls and logging in low-light forests (Rejmá nek 1989;Fine 2002;Burnham and Lee 2010) and flood scouring on periodically inundated stream banks (Barden 1987) can also increase recruitment and growth of invasives. When toxins are the dominant stressors in harsh habitats, disturbance may increase invasion in a system by removing toxins from the system. For example, shoreline development reduced salinity (and increased N) in New England salt marshes, leading to increased invasion by non-native Phragmites australis (Silliman and Bertness 2004). When a disturbance constitutes the dominant stressor, such as inundation, modification of existing disturbance regimes may increase invasion of the habitat (e.g. stabilization of river flow regimes in Australia favouring non-native species; Catford et al. 2011).

Caveats and counter-examples
While the majority of evidence we found supports the general pattern of low invadedness of harsh sites, there are an increasing number of studies documenting the AoB PLANTS www.aobplants.oxfordjournals.org ability of non-native species to invade harsh sites, albeit slowly in many cases (Williamson and Harrison 2002;Gerhardt and Collinge 2003;Zedler and Black 2004;Martin et al. 2009;Holmes et al. 2014). Furthermore, while we have provided many examples that support the hypothesis that native species are superior competitors in harsh sites, this pattern is not consistent (see Daehler 2003 for a review). In fact, in a meta-analysis, Van Kleunen et al. (2010) concluded that environmental stress did not have a significant effect on the relative performance of natives vs. invasives. We have provided many such counter-examples throughout this review, but these should not be considered a comprehensive list.
In certain cases, invasive plants may perform better than natives in harsh environments when the invaders themselves increase abiotic harshness in ways that favour conspecific recruitment. For example, Reinhart et al. (2006) found that increased shading from the invasive riparian tree Acer platanoides (Norway maple) suppressed native but not conspecific seedlings, and invasive species that increase soil salinity, such as Tamarix (saltcedar) species in the western USA (Glenn et al. 1998;Ladenburger et al. 2006) and Mesembryanthemum crystallinum (iceplant) in Mediterranean desert (El-Ghareeb 1991) can lower the relative establishment or growth rates of native species.
Furthermore, there are increasing published examples of populations of non-native species that may have evolved to become tolerant of greater stress. Examples include saline-/alkali-tolerant ecotypes of Festuca perennis (ryegrass; Dawson et al. 2007), cold-tolerant ecotypes of Echinochloa crus-galli (barnyard grass) in high latitudes (Roy et al. 2000), salt-tolerant ecotypes of B. tectorum (Scott et al. 2010) in deserts and heavy-metal-tolerant ecotypes of the grasses Anthoxanthum odoratum and Agrostis tenuis invading mine tailings (McNeilly 1968;Antonovics and Bradshaw 1970;Antonovics 2006). However, for the first three examples, it is not clear whether tolerance evolved after invasion, or if stress-tolerant genotypes arrived later. If the latter, these examples may more appropriately relate to the hypothesis of speciessite mismatch (see above), but in the context of well-suited genotypes rather than species.

Conclusions
Invasion of harsh environments by non-native plant species involves many, often interacting, factors. Overall, the descriptive and experimental evidence from a variety of ecosystems summarized in this paper show a pattern of lower plant invadedness in harsh sites. The extent to which this pattern will persist is unclear-in many systems it already appears to be breaking down, as non-native species continue to spread through human activities, expand their invasive ranges and evolve tolerance to local conditions. As some harsh sites are ameliorated by climate change, N deposition, eutrophication and other anthropogenic changes, they can be expected to become increasingly invaded.
Conservationists and restorationists have long taken encouragement from the fact that edaphically severe sites are often less invaded, thus becoming refuges for native endemic plant species. Understanding the mechanisms behind why harsh sites are less invaded can help managers determine appropriate actions to protect or restore these sites. To the extent that harshness of a site per se makes it resistant to invasion, restoration techniques that reinforce the role of the stressor may be effective. For example, soil impoverishment in eutrophied areas (Morgan 1994;Paschke et al. 2000;James et al. 2011;Tö rö k et al. 2014) and salt applications in salt marshes with anthropogenic freshwater inflows (Kuhn and Zedler 1997;Uyeda et al. 2013) have been used to shift plant communities away from invasive species to native species. However, if propagule limitation is the operative mechanism, management should be more directed towards identifying those species that may be able to successfully invade, or are already beginning to invade, and aggressively intervening to prevent their arrival or catch them in the early stages of invasion. This could include developing early detection and rapid response (EDRR) programmes, and increasing educational outreach and public awareness. Of course, addressing issues of propagule pressure and stress amelioration simultaneously are likely to be most effective at reducing invasions in any harsh site.

Contributions by the Authors
T.P.Y. conceived of this review paper. Each author wrote at least one section of the manuscript. E.Z. integrated these into a coherent manuscript, with help from J.T.S. All authors contributed edits.

Supporting Information
The following additional information is available in the online version of this article - Table S1. Summary of literature listed in this paper that provide observational evidence that harsh sites are less invaded and experimental evidence of invasion resistance. Full citations are listed in the 'Literature cited' of the main text. Tsuga canadensis dominated forest. Biological Invasions 12: