Non-Native Herbivores Promote Plant Invasions Away From Mountain Roads in the Andes

While the role of environmental filters, usually described by elevation as proxy, and anthropogenic disturbance as drivers of non-native plant diversity and abundance in mountains have been extensively studied, the impact of herbivores are less explored. Livestock grazing can facilitate the introduction of non-native species by seed dispersal and reduce biotic resistance due to consumption and trampling of native plants, even in the highest protected areas in the Andes. We here explored the effects of elevation, livestock and distance to the road on non-native and native plant distributions. Our results confirm the largely negative relationship of non-native plant richness and cover with elevation, with a peak in richness and cover at low to intermediate elevations. Similarly, we show a strong decline in non-native richness with increasing distance to the road, especially at low elevations, accompanied by a strong negative effect of roads on native species richness. Most importantly, however, we show that the presence of non-native herbivores greatly increases the cover of non-native species away from the roadside, identifying herbivore disturbance as a potential catalyst of non-native plant invasion into natural vegetation of high-Andean protected areas. Our results confirm the often-shown role of disturbance as driver of plant invasions in mountains, yet highlight the interactive effects of disturbance by roads and herbivory: roads funnel non-native species towards higher elevations, while non-native herbivores can promote non-native plant success away from the roadside and into the natural vegetation. Hence, regulating soil and non-native herbivory disturbance is important for minimizing plant invasions at high elevation in the Arid Andes.


Non-native plants in the Arid Andes
Fuentes-Lillo & Pauchard, 2019). Most of these studies highlighted that disturbance and environmental filtering due to increasing abiotic stress with elevation both play a critical role controlling non-native plant diversity in mountains (Alexander et al., 2011;Alexander et al., 2016).
Roads are a common disturbance factor in mountains and are known to facilitate the expansion of non-native plants to higher elevations (Alexander et al., 2011) and lead to biotic homogenization (Haider et al., 2020). Roads can affect various local abiotic and biotic factors including soil structure and chemistry (Mullerova et al., 2011), removal of native vegetation during its construction and maintenance (Gelbard & Belnap, 2003), and changes in microclimatic conditions . Roads can also facilitate non-native species introductions through the unintentional dispersal of seeds by vehicles (Ansong & Pickering, 2013a). As a result, mountain roads often host a higher diversity of non-native species than the surrounding landscape, yet with a progressive decline in non-native diversity with increasing elevation (Alexander et al., 2011). With their often disruptive effects on native species communities and positive effects on non-natives, mountain roads can truly funnel non-native species into mountain environments.
Nevertheless, invasion of non-native plant species away from mountain roads has been shown to be fairly limited up till now, among others due to the invasion resistance of undisturbed mountain vegetation , McDougall et al., 2018. As a result, the set of traits needed for non-native plant species to follow roads to high elevations is different from the one needed to invade undisturbed mountain vegetation (McDougall et al., 2018).
Domestic livestock is an important disturbance factor and can promote the introduction and establishment of non-native species acting as seed dispersal agents and through trampling and grazing disturbance (Loydi & Zalba, 2009;Ansong & Pickering, 2013b). These disturbances can reduce biotic resistance due to consumption and trampling, potentially facilitating the introduction and expansion of disturbance-resistant species, including many non-native species, even in the highest protected areas in the Andes Barros et al., 2020). As such, domestic livestock could potentially break the stalemate that limits plant invasion away from roadsides by reducing the resistance of the native vegetation and changing abiotic conditions in favor of the roadside invaders. Additionally, "alien alliances" have commonly been observed, where non-native herbivores promote non-native plants from their home range (Parker et al., 2006).

Non-native plants in the Arid Andes
Despite the recognized importance of environmental filtering on determining diversity of both natives and non-natives in mountains (McCain & Grytnes, 2010), plant richness patterns of natives vs non-natives can differ substantially. A recent global synthesis along elevation gradients highlighted that distribution patterns are not ubiquitously consistent, with differences across regions and between plots at roadsides and undisturbed further plots (Haider et al., 2018). For the Andes region, explicit comparisons of plant species richness along elevation gradients have shown that the most common pattern for non-natives and natives was a hump shaped curve, both in the roadside and the interior vegetation (Haider et al., 2018). However, in plots at the roadsides a monotonically decrease for native richness was detected in temperate and tropical forests in Argentina and Colombia (Sandoya et al., 2017;Haider et al., 2018). Roads' effect on plant richness patterns in the Arid Andes of Argentina have not yet been explored. In this region, although native plant communities are characterized by species adapted to extreme climate conditions at high elevations, non-native species are still present, potentially driven by human disturbances and livestock Barros et al., 2020).
Studying the combined effects of elevation, road disturbance and domestic livestock (and/or other non-native herbivores) on non-native plant distributions is thus critical to understand the interactive effects of these multiple disturbances, especially in regions where non-native herbivores are the main disturbance agents and a potential threat for plant invasions Barros et al., 2020;Liedtke et al., 2020). Understanding how both disturbance types separately and together affect native communities and promote non-native plant invasions can assist in the development of mitigation actions as well as preventing further plant invasions in higher elevation ecosystems. It can also contribute to disentangle the importance of theories about environmental filtering (elevation), biotic resistance (herbivores) and soil disturbances (i.e. roads) and their interactions (Catford et al., 2009).
We here aim to understand the effects of disturbance related to road and non-native herbivores on plant diversity and invasions in the Arid Andes of Argentina. Specifically, we ask: (1) How non-native and native plant richness are affected by elevation, road disturbance and non-native herbivores? and 2) How non-native cover is affected by non-native herbivores in conjunction with road disturbance along elevation? For the first question, we hypothesized that native and non-native plant richness differ in their distribution pattern along elevation. We predict that many non-native Non-native plants in the Arid Andes species would be more restricted in elevation than native species due to a lower adaptation to local environmental stress and the lowland provenance of most non-native species. We also expect that distance to the road and herbivores, as a surrogate of disturbance, would also affect positively nonnative richness by increased bare soil and propagule pressure. Instead, native richness would be negatively affected by these disturbance factors. For the second question, we expect that in conjunction with environmental filtering imposed by elevation, distance to the road and the abundance of non-native herbivores would also be important factors explaining non-native cover.
The three roads were selected due to their similarity in flora and bioclimatic conditions, and for the wide elevational gradient that they cover, ranging from 1755 to ~3919 m a.s.l. The gravel roads are open to traffic year-round, except at higher elevations (>2800 m a.s.l) during winter, where they can be temporarily closed due to snow cover.
Overall, the climate in the region is cold and dry, with strong variations in temperature and precipitation due to the broad elevational range and complex topography (Méndez, 2004). Low elevation areas are characterized by greater aridity and higher temperatures whereas higher elevation areas are cooler and wetter (Méndez, 2004). Precipitation mainly falls in autumn and winter, with precipitation totals ranging from 398 mm in Cordon del Plata to 120 mm in Villavicencio. On average, annual mean temperature ranges from 18°C in summer to 6° C in winter, with minimum temperatures during winter season below 0° C in these mountain roads (Cara et al., unpublished).
Soils in general are little developed and exhibit great spatial heterogeneity in depth and granulometric composition (Méndez, 2004;Méndez et al., 2006). Between 2400-3200 m a.s.l soils Non-native plants in the Arid Andes are seasonally frozen, with permafrost occurring at higher elevations (Roig et al., 2007). The region forms part of the Andean biogeographic region (Morrone, 2006) The provincial parks surrounding these mountain roads have limited past human use, and currently have very little human infrastructure except for a few mountain lodges. The main anthropogenic pressure results from regular touristic activities (sightseeing & mountain sports, use of mountain lodges), and livestock grazing, including horses and cows (Barros & Pickering, 2015).
Other non-native wild grazers include the European hare (Lepus europaeus). The only large native grazing mammal is the camelid Lama guanicoe (guanaco) . Small and medium native rodents are also common, including species of caviomorphs such as the genus Galea, and several genera of sigmodontinae rodents such as Phyllotis, Euneomys, Abrothrix, among others (Novillo & Ojeda, 2014) Field sampling design and data collection

Standardized vegetation surveys following the survey design of the international Mountain Invasion
Research Network (MIREN, Haider et al., 2021) were performed along the 3 mountain roads between spring 2016 and summer 2017. Each road was divided into 19 equally spaced elevation bands along the gradient, resulting in 20 sample sites (hereafter called 'transects') per road. Each transect had a "T" shape, consisting of three 50 x 2 m plots, one along the edge of the road (plot 1) and two others perpendicular to the road and leading away from the edge of the road (plot 2 and 3).

Non-native plants in the Arid Andes
We recorded the cover percentage of all angiosperms and gymnosperms in each plot in each transect, using visual estimation. Total vegetation cover was estimated by the sum of each species cover. Additionally, abundance of feces of ungulate and medium mammals and their identity and status (non-native or native) were noted. Non-native herbivores included horses and mules, cows, goats, and lagomorphs (European rabbits (Oryctolagus cuniculus) and European hare (Lepus europaeus)). Native mammals included rodents and the wild camelid Lama guanicoe.
Plant species were identified, verified and finally stored in the Ruiz Leal Herbarium MERL (IADIZA, Argentina). Plant species identification and classification as native or non-native was based on the Darwinion Institute Flora database (www.floraargentina.edu.ar) following criteria by Zuloaga et al. (2008). Data was collected, compiled and standardized following the protocol of the MIREN consortium network and submitted to the MIREN database (Haider et al., 2021).

Data Analyses
All the analyses were performed in R (R Core, 2018). We analyzed if native and non-native plant richness changed with elevation, abundance of non-native herbivores, and distance to the road (PLOT 1=near road, PLOT 2=middle, PLOT 3= interior) by fitting mixed generalized linear models (GLMMs) assuming a Poisson (log link). We used the same explanatory variables to analyze nonnative cover assuming a binomial (logit link) distribution. For GLMMs, we used glmer() function in the lme4 package (Bates et al., 2015). In all the models, road was used as a random factor due to our field design for data collection. Continuous predictor variables were centered and scaled in order to standardize predictions so that the units of the regression coefficients are the same. In order to explore predictive importance of fixed and random effects on plant richness and cover we applied r.squaredGLMM() function from MuMIn package (Barton, 2020) and calculated marginal and conditional R2, this is the variance explained by fixed effects and the variance explained by both fixed and random effects, respectively (Nakagawa & Schielzeth, 2013). To analyze the variance explained in GLMMs by unique or shared predictors we used partR2() function in partR2 package (Stoffel et al., 2021), this is calculated by monitoring the reduction in the fixed effect variance when a specified predictor is removed from the model, in relation to the total estimated variance.

Non-native plants in the Arid Andes
We analyzed plant richness separately for native and non-native species. In order to select the best relationship of richness and explanatory variables we compared models with or without quadratic elevation term, and models with and without the interaction of elevation and distance to the road, with and without the interaction between non-native herbivore abundance and distance to the road. We compared with Likelihood Ratio Tests the adjustment of models with and without interaction terms and to determine whether to keep or not interactive effects. Non-native plant cover was tested using as response variable the ratio of non-native cover: total plant cover to consider the expected low cover in harsh habitats in the high Andes, and to show the relative nonnative plant dominance in the total vegetation cover.

Results:
(1) General results A total of 360 plant species were recorded along the 3 surveyed mountain roads, of which 41 (11%) were non-native. Both native and non-native plant species richness decreased with elevation (Table   1, Table S1, Fig. 1). In 88% of the plots, at least one non-native plant species was detected.
Roadside plots hosted more than twice (2.33 times) the number of non-native species than the interior plots. The most common non-native species were from temperate European origin (Figs. S1, S2) and occupied an elevation range greater than 1000 m, with some of them distributed throughout the whole elevation gradient, including many common global invaders such as Taraxacum officinale and Cerastium arvense.
We recorded a total of 1639 animal droppings, of which 81% were from non-native herbivores, mainly horses or mules (41%), hares (20%) and cows (17%). Dung feces recorded from native herbivores were from guanacos (11 %) and micromammals (8%). At two of the roads, Manzano and Vallecitos, non-native herbivores (mainly cows and horses) were most abundant on the intermediate plots while along the third road they were more abundant (mainly European hare) in the roadside. The abundance of native and non-native herbivores were negatively correlated (Spearman correlation, S = 1139500, p = 0.02, rho= -0.17), hence only non-native herbivore counts were included in the following analyses.
(2) Native and non-native plant species richness

Non-native plants in the Arid Andes
Both native and non-native plant species richness decreased with elevation, but road distance had contrasting effects. Indeed, native species richness was on average 24% lower in the roadside than in interior plots, while non-native richness was 50% greater at roadsides. Non-native herbivores were not important in explaining native and non-native species richness patterns (Table S1). In GLMMs variance associated to the random factor (road identity) was more important for native richness (~20%) than for non-native richness (~4%, see Table S1a, b).
Importantly, while native species richness decreased monotonously (Fig 2a), non-native richness was best described by a hump-shaped pattern -with a peak in richness below the midelevation point at around 2000-2200 m a.s.l. (Fig. 2b). The interaction between distance to the road and elevation was important to describe native richness, but it was irrelevant for non-native richness. Increasing distance to the road resulted in a regular decline in non-native richness.
However, native richness was almost constant along the elevational gradient near the roadside, but decreased linearly in vegetation far away from the road. When analyzing variance partitioning for fixed effects in GLMMs, elevation explained most of non-native richness variance (71%) and distance to the road explained the rest of variance (7%). For native richness, the interaction of elevation and distance to the road explained 38% of the variance.

(3) Non-native plant cover
Non-native plant cover was explained by the interactions of elevation with road distance, and nonnative herbivores with road distance. The variance associated to the random factor (road identity) was very low (1%). Elevation negatively affected non-native cover, yet it decreased smoother at roadsides compared to interior plots (Table S2, Fig 3a). Overall, non-native cover was always greater in roadsides than in interior plots (~ 12 % higher, Fig. 3a). Non-native plant cover was positively affected by the abundance of non-native herbivores, but only at middle and interior plots.
These plots showed an increase in non-native cover with increasing herbivory from around 10 to 25% (Table S2, Fig. 3b). The variance was mainly explained by the interaction of elevation with road distance (16%), and a lower percentage (2%) was explained by the interaction of road distance with non-native herbivores.

Non-native plants in the Arid Andes
Our study confirms the findings from a growing body of literature showing the negative correlation of non-native plant species with distance to the road and elevation (Alexander et al., 2011;Seipel et al., 2012: Lembrechts et al., 2014Haider et al., 2018), yet goes further up by incorporating the role of non-native herbivores. Indeed, we found a positive effect of non-native herbivores on non-native cover in plots away from the road, highlighting their role in facilitating plant invasions into natural environments after their establishment in roadsides. These findings support previous research stating the importance of interacting effects, such as combined disturbance and climate, determining non-native plant invasions (Eschtruth & Battles, 2009;Geppert et al., 2021). Roads also positively affected non-native plant species richness, yet had a pervasive effect on native plants. These results emphasize the important role of recurrent disturbance from roads in creating novel communities dominated by non-native species in the extreme environment that characterize the high Andes.

Environmental filtering and road disturbance
Our results support previous research stating the importance of environmental filtering and disturbance in determining species richness patterns and non-native cover along elevation gradients (Alexander et al., 2011;Marini et al., 2012). The patterns found for non-native richness followed the most common patterns observed globally, with a curved hump-shape for non-natives and a linear decline with elevation for natives (Rahbek, 2005;McCain & Grytnes, 2010). This also supports previous research conducted in tropical and temperature parts of the Andes (Sandoya et al., 2017;Haider et al., 2018). The hump-shaped curve observed for non-natives can be attributed to the climatic harshness experienced at the two ends of the gradient in the Arid Andes, including higher aridity at low elevations and low temperatures at high elevations. Other studies have also observed that aridity can be a limiting factor for non-native plants success at lower elevations (Arevalo et al. 2005;Haider et al., 2010;Jakobs et al., 2010). At higher elevations, the combination of lower propagule pressure and lower temperatures may have limited non-native plant establishment, allowing only a subset of lowland species to occur at these elevations (Haider et al., 2010;Alexander et al., 2011;Marini et al., 2013;Averett et al., 2016;Haider et al., 2018). For native species, there was a lack of association between native richness and elevation in the roadside plots, probably explained by recurrent disturbance from roads. Non-native plants in the Arid Andes with more conservative strategies (Spellerberg, 1998;Tecco et al., 2010;Haider et al., 2018). Nonnatives recorded in our study area were mainly ruderal plants with broad tolerance to recurrent disturbance, also growing in other Andean regions and known as common global invaders (Fuentes-Lillo & Pauchard, 2019). Species were characterized by acquisitive strategies, mainly annuals or short-lived perennials herbs, with high relative growth rate, high seed reproduction and high Specific Leaf Area and low Leaf Dry Matter Content (Mazzolari et al., 2018). In contrast, native species, as observed in other mountain regions, tend to have more specific morphological and functional adaptations along the elevational gradient but with lower tolerance to recurrent human disturbance (Alexander et al., 2011). These are usually conservative species characterized by woody perennial forms, slow growth rate, reduced foliar area and high Leaf Dry Matter Content (Mazzolari et al., 2018). Hence the greater non-native richness at roadsides than interior plots, probably reflects the importance of their functional traits affecting their performance in roadside habitats. Other factors contributing to this pattern could be the higher propagule pressure on roads (Pauchard & Alaback, 2004), that can be associated to seed dispersal by vehicles and imported soil materials that increases seed banks of non-native species at roadside edges (Gelbard & Belnap, 2003; Pickering, 2013a)

The role of non-native herbivory on non-native plant invasions
In addition to the role of road disturbance, non-native herbivores can contribute to non-native plant success in mountains. We found that non-native herbivores, using dung abundance as a proxy, interacted with distance to the road, promoting plant invasions at further distances from the road edge. The impact of non-native herbivores on non-native plant invasion has been widely documented before (Maron & Vilà, 2001;Parker & Hay, 2005;Bobadilla et al., 2020;Geppert et al., 2021). Habitat modification due to trampling disturbance from domestic livestock, for example, can damage vegetation and expose soils favouring non-native plants more tolerant to disturbances Farrell & Fehmi, 2018); selective browsing and grazing on natives over nonnative species can indirectly promote their abundance (Mack, 1989;Parker et al., 2006). Other mechanisms that could be promoting non-native seed establishment beyond roads are seeds dispersal through the fur, cots and in the dung of animals, with many of the most frequent nonnatives recorded in our study (e.g. Taraxacum officinale, Verbascum thapsus, Rosa rubiginosa) known to be dispersed through livestock (Ansong & Pickering, 2013b;Dacar et al., 2019).

Non-native plants in the Arid Andes
Other factors related to herbivory that could be contributing to non-native invasion, is the relatively high abundance of non-native herbivores compared to natives. Non-native herbivores (mainly horses and cows) were the most dominant (81%) and the only native grazing large mammal, the guanaco, only accounted for 11% in our study area. Native and non-native herbivores were also negatively correlated, with decreasing abundance of natives at greater abundance of nonnatives. The lower abundance of large native herbivores, the guanaco, could be due to several reasons, including historical hunting pressure (Schroeder et al., 2013), differences in habitat selection (Ovejero et al. 2011;Esteban et al., 2012) and animal avoidance of areas with human pressure such as vehicle roads and settlements (Donadio & Buskirk, 2006;Acebes et al., 2012).
Previous research has shown that alteration of herbivore communities, including loss or reduction of native herbivores, can reduce biotic resistance to plant invasions because native herbivores tend to limit the abundance of non-native plants while non-native herbivores tend to promote their abundance (Parker et al., 2006). This is linked to the invasional meltdown hypothesis that associates biological invasions with facilitation processes between non-native species (Relva et al., 2010;Simberloff & Von Holle, 1999).
Unfortunately, the nature of our study does not allow us to identify the underlying mechanisms that explain the positive relationship between non-native herbivory and plant invasions. To better understand how herbivory, including native and non-native grazers, affect nonnative plant success in this system, further observational and experimental studies are required, including assessing feeding preferences of non-natives vs. native grazers e.g. (Relva et al., 2010;Morrison & Hay, 2011;Averill et al., 2016) as well as disentangling the direct effects (seed dispersal, trampling or consumption) and indirect effects (mutualistic interactions) of non-native herbivores on native and introduced plants (Villalobos & Zalba, 2010;Root et al., 2020).

Conclusion
Our study confirmed previous research assessing non-native and native plant distribution patterns along elevation gradients, observing similar patterns in the Arid Andes as those found globally, with largely declining richness of native and non-native plant species with increasing elevation. Our results also highlighted the importance of recurrent disturbance from roads in harsh environments, Non-native plants in the Arid Andes such as the Arid Andes, with strong negative impacts on native species richness in roadsides compared to areas further from the road, especially at lower elevations, yet significant increases in non-native species richness. Additionally, we found a positive effect of non-native herbivores on non-native plant success, highlighting their potential role as "alien allies" spreading non-native plants away from the roadside. The fact that non-native herbivory can promote invasions further from the roads and even at high elevations suggests that this disturbance potentially has a much broader area of impact than roadside disturbance itself. Consequently, implementing management guidelines aimed to minimize disturbance from roads and grazing of non-native herbivores is a key action for active restoration and to limit plant invasions in Andean mountains.

Figure 2
Predicted (lines) and observed (points) of (a) native and (b) non-native species richness along the elevation gradient and at different distances from vehicular roads (Plot 1=roadside, 2= middle and 3=interior) in the Arid Central Andes of Argentina. Because the explanatory variable had different range, elevation was standardized using the scale() function in R.

Figure 3
Non-native plant cover response predicted by binomial GLMMs using (a) elevation and (b) non-native herbivores as explanatory variables (color lines; Plot 1=roadside, 2= middle and 3=interior).

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. SuppMatAscheroBarrosetal.pdf