Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Environmental gradients shift the direction of the relationship between native and alien plant species richness

Environmental gradients shift the direction of the relationship between native and alien plant... Introduction Understanding the factors that underpin the relationship between native and alien plant species richness is of central importance in invasion biology because it provides a means to predict the vulnerability of ecological communities to invasion (Levine & D'Antonio, ; Lonsdale, ; Richardson & Pyšek, ) and the likelihood of impacts on biodiversity, for example biotic homogenization (Lambdon et al ., ; Winter et al ., ). There is an emerging consensus that the relationship between native and alien plant richness is scale dependent, often being negative when assessed across small spatial grains and extents but positive as the scales of observation increase (Fridley et al ., , ; Hulme, ). The explanation for this changing relationship, termed the ‘invasion paradox’ (Fridley et al ., ), is framed in the context of a resident native community being invaded by alien species. For small spatial grains (e.g. plots < 100 m 2 ) and small extents (e.g. < 10 km 2 ), where the environment can be regarded as relatively homogenous and biotic interactions are likely to influence species co‐occurrence, sites with more resident native species are better able to resist invasion by aliens through competitive exclusion, leading to a negative relationship between species richness (Levine & D'Antonio, ; Herben et al ., ). For larger spatial grains (e.g. plots ≥ 100 m 2 ) and larger extents (e.g. ≥ 10 km 2 ), encompassing greater spatial heterogeneity, variation among plots in native species richness primarily reflects the variation in underlying environmental conditions, including changes in resource availability, levels of disturbance or proximity to propagule sources (Stohlgren et al ., ; Fridley et al ., ; Hulme, ). Alien species should respond to these large‐scale gradients in a similar manner to native species such that sites where conditions favour high (or low) native richness should likewise favour high (or low) alien richness, leading to a positive relationship between the two. The evidence to date supports the expectation that native–alien richness relationships are positive at large plot sizes, which is usually interpreted as the result of both native and alien plant species responding to similar gradients in resource availability and habitat heterogeneity at a broad scale (Stohlgren et al ., ). However, studies that use small plots, while more suited to identifying patterns associated with biotic interactions between native and alien plant species, typically show more variable outcomes with both positive and negative relationships common (Stohlgren et al ., , ). This variability has been interpreted as a statistical problem associated with very small plots (1–10 m 2 ) that fail to adequately sample the plant community resulting in high variance in native and alien plant richness because of high spatial turnover in species composition (Stohlgren et al ., ). Nevertheless, variability in the native–alien richness relationship might also arise for ecological reasons. A wealth of studies have highlighted that native and alien species can differ in their distribution, particularly in relation to anthropogenic impacts that can alter the representation of species through changes in the regional species pool via increased propagule pressure of aliens (McKinney, ; Arévalo et al ., ; Lockwood et al ., ; Aikio et al ., ), alterations in the disturbance regime through fire and grazing (Hobbs & Huenneke, ; D'Antonio, ; Keeley et al ., ), changes in soil nutrient status as a consequence of atmospheric or agricultural fertilization (Dukes & Mooney, ), other forms of land management (e.g. highly managed or semi‐natural pastures; Boughton et al ., ) or land‐use history (e.g. ongoing reforestation; Parker et al ., ). Considerable effort has been invested in disentangling the role of environmental factors from anthropogenic factors in determining the richness of native and alien plants (Gilbert & Lechowicz, ; Carboni et al ., ; Bartomeus et al ., ), given that human pressure is generally correlated with better climate (Hanspach et al ., ; Marini et al ., , ). While native and alien species richness can be positively associated with some anthropogenic gradients (e.g. human population density; Marini et al ., ), they can be negatively or not associated with others (e.g. land management intensity; Boughton et al ., ). Consequently, positive and/or negative relationships between native and alien plant richness may be found within the same landscape depending on the character, magnitude and variation in the dominant environmental or anthropogenic gradients. In this study, we explore the relationship between native and alien plant species richness in over 1200 vegetation plots (36 m 2 ) systematically distributed across a heterogeneous landscape ( c . 1000 km 2 ) on Banks Peninsula, New Zealand. The Banks Peninsula has a variable topography (0–920 m a.s.l.) associated with strong gradients in climate, land‐use history and management, and distribution of human population. We used these data to address the following questions: What is the overall relationship between native and alien plant species richness and how strongly is it shaped by variation in anthropogenic and environmental gradients? Do similar native and alien relationships hold in plant communities that have either experienced relatively high or low human impact and are respectively dominated by either alien or native species? What is the relative contribution of environmental and anthropogenic gradients to the relationship between native and alien plant species richness? Methods Study area Banks Peninsula ( c . 1000 km 2 ) in the south‐east coast of the South Island, New Zealand (43°33′–43°54′S, 172°37′–173°7′E), comprises the eroded remnants of two large shield volcanoes, creating a highly varied topography that ranges in altitude from 0 to 920 m above sea level (see Fig. S1 in Supporting information). Soils are derived from basaltic volcanic rock and loess (Sewell et al ., ) and are typically well drained and of moderate to high fertility (Speight, ; Williams, ; Sewell et al ., ; Wilson, ). Annual rainfall ranges from 600 mm at the driest low‐elevation sites to 2000 mm at higher elevations, and mean daily temperature ranges from 8 to 13°C. Banks Peninsula was almost entirely forested prior to human colonization of New Zealand, but following European settlement in the mid‐1800s forest was extensively cleared and converted to grassland for farming, such that by 1920, < 1% of the original forest cover remained (Wilson, , ). In less intensively farmed areas, typically at higher elevation and in less‐accessible locations, forest clearance and burning has led to forest replacement by grassland dominated by native tussocks, particularly native Festuca spp., Poa spp. and Chionochloa spp. These tussock grasslands are typically associated with an intertussock sward comprising a mixture of native and alien grasses and herbs. On more accessible and productive, typically low‐ to mid‐elevation sites, more intensive burning, oversowing with pasture species, fertilizer addition and livestock grazing led to the removal of native tussocks and their replacement by grasslands dominated almost entirely by improved alien pasture species. During the last several decades, some areas of less intensively managed farmland have been abandoned and left to regenerate back to native shrubland and forest. The present landscape thus comprises a mosaic of about 10% original or regenerated native forest, 5% native shrubland (< 6 m tall) and 85% grassland ranging from less‐modified areas of semi‐native tussock grassland to highly modified pastures dominated by alien grasses such as Lolium perenne , Dactylis glomerata and Cynosurus cristatus (Wilson, , ). This gradient in vegetation, from native forest to shrubland to semi‐native tussock grassland to alien‐dominated pasture, covaries with elevation but principally reflects a gradient in anthropogenic impacts, moving from sites less modified by burning, grazing, fertilizer addition and oversowing of improved alien pasture species to sites that have been heavily modified by these processes. Human population density is low in the region, with approximately 7000 people living on the Peninsula and largely concentrated in three major settlements: Akaroa, Diamond Harbour and Little River (Statistics New Zealand, ). Floristic data We used data from a comprehensive floristic survey conducted between 1983 and 1988 that comprised a total of 1260 plots systematically located at each intersection of a regular 1000 × 1000 yard grid ( c . 920 × 920 m) drawn over the entire Banks Peninsula (Wilson, ). A 6 × 6 m plot, a common scale for vegetation sampling (Mueller‐Dombois & Ellenberg, ), was located at the intersection of each grid, within which the species composition of all vascular and non‐vascular plant species was recorded and each species was ranked according to its relative abundance in the plot. The plot aspect and slope were also recorded. The 36‐m 2 plot size is consistent with recommendations for adequately sampling grassland vegetation in New Zealand (Hurst & Allen, ) and Europe (Chytrý & Otýpková, ; Otypková & Chytry, ). Furthermore, given that there is no agreed optimum plot size at which to assess the relationship between native and alien species richness, a plot size of 36 m 2 appears a reasonable compromise, being large enough to adequately sample the community while not being so large as to mask patterns arising from biotic interactions. In our analyses, we considered only vascular plant species and excluded 33 plots without any vascular species, leaving a total of 1227 plots. Plant species were classified as native or alien to New Zealand following the standard definitions (New Zealand Plant Names database – available at http://nzflora.landcareresearch.co.nz ; Parsons et al ., ; Mahon, ), and we calculated the total number of native and alien species per plot, which we used as our response variables. We classified each plot as either ‘alien dominated’ or ‘native dominated’ based on whether the species ranked as the most abundant in each plot was an alien or a native species, respectively. We further classified each species as a tree (woody species ≥ 6 m tall), shrub (woody species < 6 m tall including lianes), fern, herb or grass using the trait categories in Ecological Traits of New Zealand Flora (available at http://ecotraits.landcareresearch.co.nz/ ) and calculated the relative proportions of these groups in each plot. Species nomenclature follows Flora of New Zealand (Moore & Edgar, ; Healy & Edgar, ; Allan, ; Webb et al ., ; Edgar & Connor, ). Explanatory variables For each plot location, we estimated the values of climate, environmental and human‐related variables from spatially explicit data layers in a GIS database (Table S1) that might explain variation in species richness. We initially identified nine climate variables known to influence the growth and distribution of plants but problems of colinearity (e.g. r > 0.5) resulted in selecting only three: annual precipitation, growing degree days and potential solar radiation as key climate variables that captured the major precipitation and temperature gradients. Climate variables were obtained from the National Institute of Water and Atmospheric Research (NIWA) 500 × 500 m resolution climate maps (Tait, ; Tait & Zheng, ). Potential solar radiation (a measure of the amount of radiation per unit area reaching the earth's surface as a proportion of the amount received at the equator) was calculated from latitude, aspect and slope of each plot (Kaufmann & Weatherred, ). We selected four other environmental variables: elevation, distance to the nearest river and stream, and mean soil pH. Elevation data were obtained from a Digital Elevation Model (DEM) downscaled to a resolution of 10 m but was not used in some analyses because of its high colinearity with climate variables (e.g. r ≥ 0.5). Riparian areas are known to be vulnerable to invasion by alien species, especially when subjected to human‐induced disturbances (Aguiar et al ., ; Parks et al ., ; Maskell et al ., ). We therefore included the distance of each plot to the nearest permanent river (a large natural permanent flowing water body) or stream (a perennial or intermittent tributary of a permanent river), as defined in the New Zealand River Environment Classification (REC; Ministry for the Environment, ). An estimate of mean soil pH at each plot (0.2–0.6 m depth) was obtained from polygon layers derived from stereo photograph interpretation, field verification and measurement as part of the 1:63.000/1:50.000 scale New Zealand Land Resource Inventory survey (NZLRI database; Landcare Research, ) integrated with the National Soils Database (NSD; Wilde, ). As human habitation and roads may be important sources and conduits of alien plant dispersal (Timmins & Williams, ; Hobbs, ; Sullivan et al ., ; McKinney, ), we included four human‐related variables: distance to nearest built‐up area, local population density, and distance to the nearest paved or unpaved road. We calculated the distance (m) of each plot centre to the nearest built‐up area (or buildings) and estimated the human population density proximate to each plot using the 1991 New Zealand Census book (Statistics New Zealand, ). We georeferenced and photointerpreted nine orthorectified aerial photographs from the early 1990s (2.5 × 2.5 m spatial resolution) of Banks Peninsula derived from Land Information New Zealand (LINZ – available at http://www.linz.govt.nz/ ) to identify built‐up areas. Those areas were classified as areas with at least three houses or other buildings in an area of at least 1.012 ha (U.S. Geological Survey Land Cover Institute definition). To ensure accuracy of the locations of settlements from aerial photographs (such as built‐up areas or buildings), particularly those close to vegetation plots, we undertook field verification. To determine whether land‐use history and management of the Peninsula shape the relationship between native and alien species richness, we also included the relative proportion of native trees in each plot which reflects a gradient in anthropogenic impact from less‐modified areas of native forest and shrubland with a high native woody component to more heavily modified grasslands. To quantify how additional biotic factors shape the relationship between native and alien plant species richness, we also included separately native richness and alien richness as explanatory variables. Statistical analysis All spatial data were stored and extracted using ArcGIS 9.3 (Esri, ), and all statistical analyses were performed in R (2.13.0; R Development Core Team, ). We first quantified the relationship between native and alien richness across all plots using Spearman's rank correlation, because this measure is less sensitive and more robust than Pearson's correlation to outliers. Once we had verified that any outliers were not sampling errors, we then examined the relationship between native and alien richness separately for plots where the dominant plant species was either an alien or a native. We also assessed the correlations between native and alien species richness and the proportions of trees, shrubs, herbs and grasses because variation in the representation of these life forms reflect a gradient in intensity of past land use. We then fitted a multiple regression model to identify factors that could explain the variation in native and alien species richness. To account for spatial autocorrelation, we fitted the regression models with a spatial autocorrelation structure using generalized least squares (GLS; Legendre, ; Dormann, ). We assessed the potential influence of spatial autocorrelation on parameter estimates by modelling different spatial correlation structures (Pinheiro & Bates, ) and using the Akaike Information Criterion (AIC; Akaike, ; Burnham & Anderson, ; Johnson & Omland, ) to identify the best model (Pinheiro et al ., ). We assessed the degree to which our models accounted for unexplained spatial variation by plotting a semi‐variogram of the normalized residuals. We also examined whether explanatory variables showed a nonlinear relationship to the response by testing for the importance of quadratic terms. Only soil pH showed a strong nonlinear relationship with species richness, so we included this variable along with its quadratic term in the multiple regression model. Given the large elevation range on Banks Peninsula and the covariance of anthropogenic impacts with elevation, we also examined how the native–alien species richness relationship varied across this gradient by examining the correlation between native and alien species richness separately for plots in five elevational bands (0–100, 101–200, 201–300, 301–400 and > 400 m a.s.l.) chosen to ensure that each band had an approximately equal number of plots. Stratifying plots by elevational bands ensures key climate variables (such as temperature and precipitation) remain within a narrow range and allows us to examine the relationship between native and alien richness having controlled for this variation (Hanspach et al ., ; Marini et al ., , ). Results Relationships between native and alien species richness Although slightly more native (368) than alien (311) vascular plant species were recorded in the 1227 plots on Banks Peninsula, on average over twice as many alien (16.4 ± 0.19) as native species (7.9 ± 0.23) were found per plot (Wilcoxon rank sum test: W = 120, P < 0.001). Across all plots, native and alien plant species richness were significantly negatively correlated (Spearman's rank correlation, ρ = −0.126, d.f. = 1227, P < 0.001; Fig. a). However, fitting a cubic smoothing spline to the data suggested that the relationship was nonlinear. For plots with fewer than about 10 native species, the relationship between native and alien richness appeared positive, while for plots with more than 10 native species, the relationship was strongly negative. Over 60% ( n = 739) of plots were classified as alien dominated, with the majority being grassland plots dominated by introduced pasture species such as Lolium perenne (dominant in 189 plots), Cynosurus cristatus (89 plots) or Dactylis glomerata (83 plots). Alien‐dominated plots typically had low native species richness and comprised the majority of plots with fewer than 10 native species. The 488 plots dominated by a native species included modified tussock grassland dominated by Poa cita (94 plots) and Rytidosperma clavatum (52 plots), along with native forest and shrubland communities dominated, for example, by Kunzea ericoides (53 plots). Native‐dominated plots included the majority of plots with more than 10 native species. The separation of plots into those dominated by alien or native species largely accounted for the nonlinear relationship between native and alien species richness seen across all plots (Fig. a). For alien‐dominated plots, with low native but high alien species richness per plot (3.9 ± 0.15 and 17.5 ± 0.19, respectively), there was a significant positive relationship between native and alien richness (ρ = 0.26, P < 0.001; Fig. b). In contrast, the species richness relationship was stronger and significantly negative in native‐dominated plots (ρ = −0.34, P < 0.001; Fig. c) with similar mean values of alien and native richness per plot (14.7 ± 0.36 and 14.1 ± 0.4, respectively). Relationship between native and alien plant species richness across the B anks P eninsula, N ew Z ealand in: (a) all 1227 plots (ρ = −0.126, P < 0.001); (b) alien‐dominated plots (739 plots, ρ = 0.26, P < 0.001); (c) native‐dominated plots (488 plots, ρ = −0.34, P < 0.001). Grey points are individual plots; black points show the mean value of alien species richness for each value of native species richness. Solid lines show a cubic smoothing spline fitted to the full data set. Across all plots, the dominant species life forms shift along the gradient of increasing native species richness. Plots with low native species richness have a higher proportion of alien grass and herbaceous species (ρ = −0.59 and ρ = −0.27, P < 0.001 respectively), while plots with high native species richness contain a higher proportion of native trees and shrubs (ρ = 0.59 and ρ = 0.7, P < 0.001, respectively, Table ). These patterns remain when native‐ and alien‐dominated plots are examined separately. Native‐dominated plots with low native species richness have a higher proportion of grass and herbaceous species (ρ = −0.69 and ρ = −0.39, P < 0.001, respectively), while those with high native species richness contain a higher proportion of tree and shrub species (ρ = 0.62 and ρ = 0.69, P < 0.001, respectively). For alien‐dominated plots, the same gradient is apparent where plots with high native species richness have a higher proportion of native tree and shrub species (ρ = 0.29 and ρ = 0.51, P < 0.001, respectively), while plots with high alien species richness have more grass and herbaceous species (ρ = 0.32 and ρ = 0.37, P < 0.001, respectively). Parameter estimates from multiple regression models [generalized least squares ( GLS ) with spatial correlation structures] predicting native and alien species richness within: (i) all plots (d.f. = 1227), (ii) alien‐dominated plots (d.f. = 739), and (iii) native‐dominated plots (d.f. = 488) with climate, environmental and human‐related explanatory variables Variables All plots Alien‐dominated plots Native‐dominated plots Alien richness Native richness Alien richness Native richness Alien richness Native richness Total N 311 368 282 217 194 345 Mean N 16.4 ± 0.19 7.9 ± 0.23 17.5 ± 0.19 3.9 ± 0.15 14.7 ± 0.36 14.1 ± 0.4 Growing degree days 3.39 *** − 2.99 *** 2.01 − 2.61 ** 3.11 *** − 5.26 *** Solar radiation 0.87 *** − 1.44 *** 0.04 − 0.86 ** 1.28 *** − 2.12 *** Distance to buildings −0.06 0.08 − 0.25 ** 0.21 * 0.11 −0.16 Distance to unpaved roads 0.09 0.15 ** −0.11 0.17 ** 0.15 * 0.05 Distance to streams 0.14 ** −0.06 0.21 * 0.21 * 0.17 * − 0.22 ** Proportion tree/plot − 0.33 *** 0.59 *** − 0.32 *** 0.29 *** − 0.61 *** 0.62 *** Soil pH 1.96 *** 1.62 *** 2.05 *** 1.02 * 1.25 1.67 * Soil pH 2 − 1.28 *** − 1.05 *** − 1.34 *** − 0.65 * −0.83 − 1.08 * Alien richness NA − 0.58 *** NA 1.41 *** NA − 0.49 *** Native richness − 0.18 *** NA 0.75 *** NA − 0.72 *** NA R 2 0.11 0.23 0.14 0.21 0.19 0.25 Spatial correlation structures Spherical Exponential Exponential Spherical Exponential Gaussian In all cases, the explanatory and response variables were transformed [log10( x + 1)] to ensure normality and to deal with zero values. Explanatory variables were then standardized to zero mean and standard deviation one so that parameter estimates were comparable. All explanatory variables (see Methods), were tested but only those variables that were statistically significant in at least one model are shown. Significant variables in a given model are shown in bold (*** P < 0.0001, ** P < 0.001, * P < 0.05). Also shown are the coefficients of determination ( R 2 ), the total number of alien and native species (total N ), and the mean number of native and alien species per plot (Mean N ) (±SE). Determinants of native and alien richness Low but significant spatial autocorrelation was consistently found in the residuals of our GLS, and thus, we report results based on these spatial models. Plots with high native richness (containing a higher proportion of tree and shrub species) were more likely to occur at cooler sites (typically at higher elevation) with low solar radiation (steeper, south‐facing slopes), intermediate in soil pH, in areas with lower alien richness that were further away from unpaved roads (Table ). In contrast, plots with higher alien species richness (dominated by grass and herbaceous species) occurred on warmer sites (typically at lower elevation) with high solar radiation (drier north‐facing slopes) that had low native species richness and intermediate soil pH (Table ). Hence, at a broad scale, plots with high native and alien species richness were spatially separated and tended to occupy different parts of the landscape. However, these edaphic factors only accounted for 21% and 9% of the variation in native and alien species richness, respectively. Alien richness in the native richness model and vice versa explained a small but significant amount of additional variation (increasing the variation accounted for to 23% and 11%, respectively; Table ), indicating that unmeasured factors linked with biotic suitability further shaped species richness patterns. These relationships were also evident within alien‐ and native‐dominated plots (Table ). Within each of these groups, alien richness tended to be higher on warmer (lower elevation), drier north‐facing slopes while native richness tended to be higher on cooler (higher elevation) sites on south‐facing slopes, with both alien and native richness higher at intermediate soil pH and distant from streams. Alien richness was low, and native richness was high, when there was a greater number of tree species per plot. The major difference was that, having controlled for other factors in the model, native and alien richness were positively associated in alien‐dominated plots (increasing the total variation accounted for from 14% to 21% respectively), but negatively associated in native‐dominated plots (19% to 25% of total variation accounted for). In addition, alien richness increased and native richness declined significantly with proximity to buildings in alien‐dominated plots and with distance to streams in native‐dominated plots. Alien species richness was generally higher than native species per plot across the elevation gradient (Table ). However, native species richness increased with elevation for all plots and for plots dominated by either native or alien species, such that the only occasion mean native richness was greater than alien richness was at the highest elevations (> 400 m a.s.l.) within native‐dominated plots. In contrast, alien species richness was less influenced by elevation and appears to have a unimodal relationship with a slight peak at mid elevations (Table ). Across all plots, native and alien richness were significantly and positively associated up to 200 m a.s.l., but this relationship became increasingly negative at higher elevations, becoming significantly so above 400 m a.s.l. (Fig. ). Splitting the analysis into alien‐ and native‐dominated plots separately revealed that this trend reflects the positive relationship between native and alien richness in alien‐dominated plots below 300 m a.s.l. and the negative relationship in native‐dominated plots above 300 m a.s.l. Within each elevational band and across the entire elevation gradient, there was a consistent positive relationship between native and alien species richness in alien‐dominated plots, and a consistent negative relationship in native‐dominated plots (Fig. ). Change in S pearman's rank correlation coefficients (black dots with 95% confidence intervals) of native versus alien species richness within five elevational bands (0–100, 101–200, 201–300, 301–400 and > 400 m a.s.l.). Solid line shows native–alien relationship within all plots ( n = 1227). Dotted line shows native–alien relationship within alien‐dominated plots ( n = 739). Broken line shows native–alien relationship within native‐dominated plots ( n = 488). The horizontal dotted line shows value of ρ = 0. The superscript refers to the statistical significance of correlations (*** P < 0.001, ** P < 0.01, * P < 0.05). Total number of alien and native species and mean number of species per plot within: all plots, alien‐dominated and native‐dominated plots in each separate elevational band. Total number of plots in each elevational band are shown Elevational bands (m) 0–100 101–200 201–300 301–400 > 400 Alien Native Alien Native Alien Native Alien Native Alien Native All plots Tot. species 292 220 256 237 218 191 168 156 265 261 Mean species/plot 16.2 4.7 17.7 7.4 17.2 7.3 16.7 8.5 14.5 12.1 Tot. plots 296 260 223 172 276 Alien‐dominated plots Tot. species 219 143 154 131 146 114 97 81 123 108 Mean species/plot 17.0 2.5 18.1 4.1 17.9 3.5 17.5 4.1 16.9 6.3 Tot. plots 219 154 146 97 123 Native‐dominated plots Tot. species 68 73 100 105 67 76 69 75 135 151 Mean species/plot 13.1 11.0 16.9 12.3 15.8 14.4 15.7 14.2 12.6 16.7 Tot. plots 77 106 77 75 153 Discussion Previous interpretation of the sign and magnitude of the relationship between native and alien species richness has largely centred on the ‘invasion paradox’ that addresses how the shape and strength of the native–alien richness relationship can change with spatial grain and extent (Levine & D'Antonio, ; Shea & Chesson, ; Fridley et al ., ; Herben et al ., ). We show, however, that at a constant grain and extent, the relationship between native and alien richness differs between plant communities subject to relatively high or low human impact that are respectively dominated by either alien or native species. Such variation in the native–alien richness relationship at small plot sizes has previously been attributed to statistical problems associated with high turnover of species leading to high variation in species richness among plots and thus inconsistent relationships (Stohlgren et al ., , ; Fridley et al ., ). Our data do show high among‐plot variation in both native and alien richness (Fig. ), but we nevertheless find highly significant relationships with the sign of that relationship shifting from positive to negative in going from alien‐ to native‐dominated plots. This shift could not be fully explained by changes in any of the environmental, climatic or human variables that we measured, although the explanatory variables did a consistently better job explaining native than alien richness. This might be expected if aliens were reasonably ubiquitous as a result of human impacts. Indeed, unlike other studies that typically find a marked decline in alien richness with increasing elevation (Alexander et al ., ; Marini et al ., , ), we observed relatively little change (Table ). Grasslands across the entire elevation range, for example, tended to contain a similar suite of common alien species (e.g. Lolium perenne , Dactylis glomerata and Anthoxanthum odoratum ). Within native‐dominated plots, the gradient of increasing native species richness coincided with a shift from plots at warmer, lower elevation on northerly aspects to plots at cooler and higher elevation sites on south‐facing aspects, with a correspondingly greater proportion of trees. This gradient most likely reflects a legacy of past land‐use, with less‐modified or regenerating areas of native forest and shrubland occurring in less‐accessible and less‐productive higher‐elevation south‐facing sites, while mid‐elevation warmer sites tend to support more modified native‐dominated grassland. Why then is alien species richness negatively correlated with native species richness along this gradient? The traditional interpretation would be that high native richness drives the sign of the relationship and confers resistance to invasion by alien species (biotic resistance). However, while alien species richness declines along this gradient, plots with high native richness still have, on average, a high proportion of alien species (about one‐half to one‐third of species), suggesting that these sites are readily invaded. Instead, it may not be high native richness per se that confers resistance to invasion, but the fact that higher native richness coincides with a shift from grassland to remnant or regenerating native forest and shrubland. The understory of plots dominated by native woody vegetation may be less susceptible to invasion by alien grass or herbaceous species better adapted to more open environments, which comprise the bulk of the alien flora. The lower number of alien species and the higher number of native species in these native‐dominated communities may thus reflect a shift in vegetation structure, from grassland to shrubland/forest, rather than being a function of biotic resistance linked with the number of species. Forested plots with high alien richness may also be in areas regenerating after agricultural abandonment, highlighting the potential for historical factors such as land‐use change to influence current native–alien plant relationships (Parker et al ., ). For alien‐dominated plots, we see a positive association between native and alien species richness, which is commonly attributed to both native and alien species responding in a similar manner to underlying environmental gradients associated with plant performance (Gilbert & Lechowicz, ; Richardson et al ., ). However, in our study, only one variable (soil pH) appeared to influence native and alien richness similarly (Table ). None of the remaining environmental variables we measured could fully explain the covariance between native and alien species richness. Thus, is there any evidence that native and alien species richness covary along either anthropogenic or climate gradients? Stratifying by elevation helps disentangle the potentially confounding effects of covariance among anthropogenic and environmental variables (Marini et al ., ). The relationship between native and alien species richness was consistently positive or negative for alien‐dominated and native‐dominated plots, respectively, even when the variation in climate was constrained within fixed elevational bands. This suggests that anthropogenic effects shape these relationships more strongly than climate. Nevertheless, the strength of the positive and negative relationships changed with elevation, suggesting that the magnitude of anthropogenic effects also vary with elevation. In contrast to the findings of Boughton et al . ( ) who found management intensity resulted in negative relationships between native and alien species, we interpreted our positive relationship to be a function of the intensity of management. Sites with low native and alien species richness are dominated by alien pasture grasses that are intensively managed through grazing, ploughing and fertilizer application to favour just a few highly productive fast growing alien species (for example Lolium perenne and Trifolium spp. swards). Less intensive management may allow pastures to be invaded by other alien and native species, leading to a positive relationship between the two, although aliens dominate in these more intensively managed systems. Thus, the positive relationship is driven by patterns in the persistence of native species along a gradient of management intensity that influences alien species richness to a much lesser extent. With increasing elevation, climate variables might be expected to exert a greater influence on native and alien plant distributions and to affect these in a similar manner (Stohlgren et al ., ; Marini et al ., ; Pauchard et al ., ). However, across the large elevation gradient, while alien species richness showed a unimodal relationship that could be attributed to higher elevations becoming increasingly inclement, native richness progressively increased with elevation. While we might have expected a similar unimodal relationship for native species (Marini et al ., ), the linear relationship undoubtedly reflects the fact that much of the native diversity has been removed by forest clearance at low to mid elevations (Wilson, ). The outcome is that at low elevation, where most plots are heavily modified and dominated by alien species, native and alien richness are positively correlated, while at higher elevations, there is stronger spatial segregation and hence a negative correlation, with less‐modified remnants of native forest vegetation tending to have more native and fewer alien species. Conclusion Much of the discussion to date regarding the drivers of native and alien species richness suffers from the fact that the grain and extent of studies are rarely independent and the grain size covaries with the spatial extent examined (Hulme, ). This prevents adequate assessment of the local and regional drivers on patterns of species richness. Our study is one of the few that examines patterns of species richness at a relatively fine grain (36 m 2 ) over a large spatial extent ( c . 1000 km 2 ) (c.f. Stohlgren et al ., ; Chen et al ., ). Our results do not provide strong evidence of biotic resistance associated with higher native species richness limiting alien plant invasions, although this might only be expected to be found at even finer grain sizes (Levine & D'Antonio, ; Herben et al ., ). In contrast, our study confirms an increasing and recent body of evidence (Parker et al ., ; Boughton et al ., ) that indicates contemporary and historical anthropogenic impacts strongly shape both negative and positive relationships between native and alien species richness, especially where such impacts covary with climate gradients. The impact of local management effects (e.g. land clearance, grazing) may be less discernible at larger grain sizes (> 1 km 2 ) where other broad‐scale environmental factors are likely to shape patterns in species richness. Acknowledgements This research was funded by the New Zealand Tertiary Education Commission via a PhD scholarship grant and Bio‐Protection Research Centre Writing Scholarship. The authors are very grateful to Hugh Wilson for allowing use of the floristic data from his Banks Peninsula survey; Andrew Tait; Susan Wiser; Brad Case, Takayoshi Ikeda and the Lincoln University Spatial Ecology Group for GIS and statistical support. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Diversity and Distributions Wiley

Environmental gradients shift the direction of the relationship between native and alien plant species richness

Download PDF
11 pages

Loading next page...
 
/lp/wiley/environmental-gradients-shift-the-direction-of-the-relationship-yVlqry9KYR

References (81)

Publisher
Wiley
Copyright
Copyright © 2013 Blackwell Publishing Ltd
ISSN
1366-9516
eISSN
1472-4642
DOI
10.1111/j.1472-4642.2012.00939.x
Publisher site
See Article on Publisher Site

Abstract

Introduction Understanding the factors that underpin the relationship between native and alien plant species richness is of central importance in invasion biology because it provides a means to predict the vulnerability of ecological communities to invasion (Levine & D'Antonio, ; Lonsdale, ; Richardson & Pyšek, ) and the likelihood of impacts on biodiversity, for example biotic homogenization (Lambdon et al ., ; Winter et al ., ). There is an emerging consensus that the relationship between native and alien plant richness is scale dependent, often being negative when assessed across small spatial grains and extents but positive as the scales of observation increase (Fridley et al ., , ; Hulme, ). The explanation for this changing relationship, termed the ‘invasion paradox’ (Fridley et al ., ), is framed in the context of a resident native community being invaded by alien species. For small spatial grains (e.g. plots < 100 m 2 ) and small extents (e.g. < 10 km 2 ), where the environment can be regarded as relatively homogenous and biotic interactions are likely to influence species co‐occurrence, sites with more resident native species are better able to resist invasion by aliens through competitive exclusion, leading to a negative relationship between species richness (Levine & D'Antonio, ; Herben et al ., ). For larger spatial grains (e.g. plots ≥ 100 m 2 ) and larger extents (e.g. ≥ 10 km 2 ), encompassing greater spatial heterogeneity, variation among plots in native species richness primarily reflects the variation in underlying environmental conditions, including changes in resource availability, levels of disturbance or proximity to propagule sources (Stohlgren et al ., ; Fridley et al ., ; Hulme, ). Alien species should respond to these large‐scale gradients in a similar manner to native species such that sites where conditions favour high (or low) native richness should likewise favour high (or low) alien richness, leading to a positive relationship between the two. The evidence to date supports the expectation that native–alien richness relationships are positive at large plot sizes, which is usually interpreted as the result of both native and alien plant species responding to similar gradients in resource availability and habitat heterogeneity at a broad scale (Stohlgren et al ., ). However, studies that use small plots, while more suited to identifying patterns associated with biotic interactions between native and alien plant species, typically show more variable outcomes with both positive and negative relationships common (Stohlgren et al ., , ). This variability has been interpreted as a statistical problem associated with very small plots (1–10 m 2 ) that fail to adequately sample the plant community resulting in high variance in native and alien plant richness because of high spatial turnover in species composition (Stohlgren et al ., ). Nevertheless, variability in the native–alien richness relationship might also arise for ecological reasons. A wealth of studies have highlighted that native and alien species can differ in their distribution, particularly in relation to anthropogenic impacts that can alter the representation of species through changes in the regional species pool via increased propagule pressure of aliens (McKinney, ; Arévalo et al ., ; Lockwood et al ., ; Aikio et al ., ), alterations in the disturbance regime through fire and grazing (Hobbs & Huenneke, ; D'Antonio, ; Keeley et al ., ), changes in soil nutrient status as a consequence of atmospheric or agricultural fertilization (Dukes & Mooney, ), other forms of land management (e.g. highly managed or semi‐natural pastures; Boughton et al ., ) or land‐use history (e.g. ongoing reforestation; Parker et al ., ). Considerable effort has been invested in disentangling the role of environmental factors from anthropogenic factors in determining the richness of native and alien plants (Gilbert & Lechowicz, ; Carboni et al ., ; Bartomeus et al ., ), given that human pressure is generally correlated with better climate (Hanspach et al ., ; Marini et al ., , ). While native and alien species richness can be positively associated with some anthropogenic gradients (e.g. human population density; Marini et al ., ), they can be negatively or not associated with others (e.g. land management intensity; Boughton et al ., ). Consequently, positive and/or negative relationships between native and alien plant richness may be found within the same landscape depending on the character, magnitude and variation in the dominant environmental or anthropogenic gradients. In this study, we explore the relationship between native and alien plant species richness in over 1200 vegetation plots (36 m 2 ) systematically distributed across a heterogeneous landscape ( c . 1000 km 2 ) on Banks Peninsula, New Zealand. The Banks Peninsula has a variable topography (0–920 m a.s.l.) associated with strong gradients in climate, land‐use history and management, and distribution of human population. We used these data to address the following questions: What is the overall relationship between native and alien plant species richness and how strongly is it shaped by variation in anthropogenic and environmental gradients? Do similar native and alien relationships hold in plant communities that have either experienced relatively high or low human impact and are respectively dominated by either alien or native species? What is the relative contribution of environmental and anthropogenic gradients to the relationship between native and alien plant species richness? Methods Study area Banks Peninsula ( c . 1000 km 2 ) in the south‐east coast of the South Island, New Zealand (43°33′–43°54′S, 172°37′–173°7′E), comprises the eroded remnants of two large shield volcanoes, creating a highly varied topography that ranges in altitude from 0 to 920 m above sea level (see Fig. S1 in Supporting information). Soils are derived from basaltic volcanic rock and loess (Sewell et al ., ) and are typically well drained and of moderate to high fertility (Speight, ; Williams, ; Sewell et al ., ; Wilson, ). Annual rainfall ranges from 600 mm at the driest low‐elevation sites to 2000 mm at higher elevations, and mean daily temperature ranges from 8 to 13°C. Banks Peninsula was almost entirely forested prior to human colonization of New Zealand, but following European settlement in the mid‐1800s forest was extensively cleared and converted to grassland for farming, such that by 1920, < 1% of the original forest cover remained (Wilson, , ). In less intensively farmed areas, typically at higher elevation and in less‐accessible locations, forest clearance and burning has led to forest replacement by grassland dominated by native tussocks, particularly native Festuca spp., Poa spp. and Chionochloa spp. These tussock grasslands are typically associated with an intertussock sward comprising a mixture of native and alien grasses and herbs. On more accessible and productive, typically low‐ to mid‐elevation sites, more intensive burning, oversowing with pasture species, fertilizer addition and livestock grazing led to the removal of native tussocks and their replacement by grasslands dominated almost entirely by improved alien pasture species. During the last several decades, some areas of less intensively managed farmland have been abandoned and left to regenerate back to native shrubland and forest. The present landscape thus comprises a mosaic of about 10% original or regenerated native forest, 5% native shrubland (< 6 m tall) and 85% grassland ranging from less‐modified areas of semi‐native tussock grassland to highly modified pastures dominated by alien grasses such as Lolium perenne , Dactylis glomerata and Cynosurus cristatus (Wilson, , ). This gradient in vegetation, from native forest to shrubland to semi‐native tussock grassland to alien‐dominated pasture, covaries with elevation but principally reflects a gradient in anthropogenic impacts, moving from sites less modified by burning, grazing, fertilizer addition and oversowing of improved alien pasture species to sites that have been heavily modified by these processes. Human population density is low in the region, with approximately 7000 people living on the Peninsula and largely concentrated in three major settlements: Akaroa, Diamond Harbour and Little River (Statistics New Zealand, ). Floristic data We used data from a comprehensive floristic survey conducted between 1983 and 1988 that comprised a total of 1260 plots systematically located at each intersection of a regular 1000 × 1000 yard grid ( c . 920 × 920 m) drawn over the entire Banks Peninsula (Wilson, ). A 6 × 6 m plot, a common scale for vegetation sampling (Mueller‐Dombois & Ellenberg, ), was located at the intersection of each grid, within which the species composition of all vascular and non‐vascular plant species was recorded and each species was ranked according to its relative abundance in the plot. The plot aspect and slope were also recorded. The 36‐m 2 plot size is consistent with recommendations for adequately sampling grassland vegetation in New Zealand (Hurst & Allen, ) and Europe (Chytrý & Otýpková, ; Otypková & Chytry, ). Furthermore, given that there is no agreed optimum plot size at which to assess the relationship between native and alien species richness, a plot size of 36 m 2 appears a reasonable compromise, being large enough to adequately sample the community while not being so large as to mask patterns arising from biotic interactions. In our analyses, we considered only vascular plant species and excluded 33 plots without any vascular species, leaving a total of 1227 plots. Plant species were classified as native or alien to New Zealand following the standard definitions (New Zealand Plant Names database – available at http://nzflora.landcareresearch.co.nz ; Parsons et al ., ; Mahon, ), and we calculated the total number of native and alien species per plot, which we used as our response variables. We classified each plot as either ‘alien dominated’ or ‘native dominated’ based on whether the species ranked as the most abundant in each plot was an alien or a native species, respectively. We further classified each species as a tree (woody species ≥ 6 m tall), shrub (woody species < 6 m tall including lianes), fern, herb or grass using the trait categories in Ecological Traits of New Zealand Flora (available at http://ecotraits.landcareresearch.co.nz/ ) and calculated the relative proportions of these groups in each plot. Species nomenclature follows Flora of New Zealand (Moore & Edgar, ; Healy & Edgar, ; Allan, ; Webb et al ., ; Edgar & Connor, ). Explanatory variables For each plot location, we estimated the values of climate, environmental and human‐related variables from spatially explicit data layers in a GIS database (Table S1) that might explain variation in species richness. We initially identified nine climate variables known to influence the growth and distribution of plants but problems of colinearity (e.g. r > 0.5) resulted in selecting only three: annual precipitation, growing degree days and potential solar radiation as key climate variables that captured the major precipitation and temperature gradients. Climate variables were obtained from the National Institute of Water and Atmospheric Research (NIWA) 500 × 500 m resolution climate maps (Tait, ; Tait & Zheng, ). Potential solar radiation (a measure of the amount of radiation per unit area reaching the earth's surface as a proportion of the amount received at the equator) was calculated from latitude, aspect and slope of each plot (Kaufmann & Weatherred, ). We selected four other environmental variables: elevation, distance to the nearest river and stream, and mean soil pH. Elevation data were obtained from a Digital Elevation Model (DEM) downscaled to a resolution of 10 m but was not used in some analyses because of its high colinearity with climate variables (e.g. r ≥ 0.5). Riparian areas are known to be vulnerable to invasion by alien species, especially when subjected to human‐induced disturbances (Aguiar et al ., ; Parks et al ., ; Maskell et al ., ). We therefore included the distance of each plot to the nearest permanent river (a large natural permanent flowing water body) or stream (a perennial or intermittent tributary of a permanent river), as defined in the New Zealand River Environment Classification (REC; Ministry for the Environment, ). An estimate of mean soil pH at each plot (0.2–0.6 m depth) was obtained from polygon layers derived from stereo photograph interpretation, field verification and measurement as part of the 1:63.000/1:50.000 scale New Zealand Land Resource Inventory survey (NZLRI database; Landcare Research, ) integrated with the National Soils Database (NSD; Wilde, ). As human habitation and roads may be important sources and conduits of alien plant dispersal (Timmins & Williams, ; Hobbs, ; Sullivan et al ., ; McKinney, ), we included four human‐related variables: distance to nearest built‐up area, local population density, and distance to the nearest paved or unpaved road. We calculated the distance (m) of each plot centre to the nearest built‐up area (or buildings) and estimated the human population density proximate to each plot using the 1991 New Zealand Census book (Statistics New Zealand, ). We georeferenced and photointerpreted nine orthorectified aerial photographs from the early 1990s (2.5 × 2.5 m spatial resolution) of Banks Peninsula derived from Land Information New Zealand (LINZ – available at http://www.linz.govt.nz/ ) to identify built‐up areas. Those areas were classified as areas with at least three houses or other buildings in an area of at least 1.012 ha (U.S. Geological Survey Land Cover Institute definition). To ensure accuracy of the locations of settlements from aerial photographs (such as built‐up areas or buildings), particularly those close to vegetation plots, we undertook field verification. To determine whether land‐use history and management of the Peninsula shape the relationship between native and alien species richness, we also included the relative proportion of native trees in each plot which reflects a gradient in anthropogenic impact from less‐modified areas of native forest and shrubland with a high native woody component to more heavily modified grasslands. To quantify how additional biotic factors shape the relationship between native and alien plant species richness, we also included separately native richness and alien richness as explanatory variables. Statistical analysis All spatial data were stored and extracted using ArcGIS 9.3 (Esri, ), and all statistical analyses were performed in R (2.13.0; R Development Core Team, ). We first quantified the relationship between native and alien richness across all plots using Spearman's rank correlation, because this measure is less sensitive and more robust than Pearson's correlation to outliers. Once we had verified that any outliers were not sampling errors, we then examined the relationship between native and alien richness separately for plots where the dominant plant species was either an alien or a native. We also assessed the correlations between native and alien species richness and the proportions of trees, shrubs, herbs and grasses because variation in the representation of these life forms reflect a gradient in intensity of past land use. We then fitted a multiple regression model to identify factors that could explain the variation in native and alien species richness. To account for spatial autocorrelation, we fitted the regression models with a spatial autocorrelation structure using generalized least squares (GLS; Legendre, ; Dormann, ). We assessed the potential influence of spatial autocorrelation on parameter estimates by modelling different spatial correlation structures (Pinheiro & Bates, ) and using the Akaike Information Criterion (AIC; Akaike, ; Burnham & Anderson, ; Johnson & Omland, ) to identify the best model (Pinheiro et al ., ). We assessed the degree to which our models accounted for unexplained spatial variation by plotting a semi‐variogram of the normalized residuals. We also examined whether explanatory variables showed a nonlinear relationship to the response by testing for the importance of quadratic terms. Only soil pH showed a strong nonlinear relationship with species richness, so we included this variable along with its quadratic term in the multiple regression model. Given the large elevation range on Banks Peninsula and the covariance of anthropogenic impacts with elevation, we also examined how the native–alien species richness relationship varied across this gradient by examining the correlation between native and alien species richness separately for plots in five elevational bands (0–100, 101–200, 201–300, 301–400 and > 400 m a.s.l.) chosen to ensure that each band had an approximately equal number of plots. Stratifying plots by elevational bands ensures key climate variables (such as temperature and precipitation) remain within a narrow range and allows us to examine the relationship between native and alien richness having controlled for this variation (Hanspach et al ., ; Marini et al ., , ). Results Relationships between native and alien species richness Although slightly more native (368) than alien (311) vascular plant species were recorded in the 1227 plots on Banks Peninsula, on average over twice as many alien (16.4 ± 0.19) as native species (7.9 ± 0.23) were found per plot (Wilcoxon rank sum test: W = 120, P < 0.001). Across all plots, native and alien plant species richness were significantly negatively correlated (Spearman's rank correlation, ρ = −0.126, d.f. = 1227, P < 0.001; Fig. a). However, fitting a cubic smoothing spline to the data suggested that the relationship was nonlinear. For plots with fewer than about 10 native species, the relationship between native and alien richness appeared positive, while for plots with more than 10 native species, the relationship was strongly negative. Over 60% ( n = 739) of plots were classified as alien dominated, with the majority being grassland plots dominated by introduced pasture species such as Lolium perenne (dominant in 189 plots), Cynosurus cristatus (89 plots) or Dactylis glomerata (83 plots). Alien‐dominated plots typically had low native species richness and comprised the majority of plots with fewer than 10 native species. The 488 plots dominated by a native species included modified tussock grassland dominated by Poa cita (94 plots) and Rytidosperma clavatum (52 plots), along with native forest and shrubland communities dominated, for example, by Kunzea ericoides (53 plots). Native‐dominated plots included the majority of plots with more than 10 native species. The separation of plots into those dominated by alien or native species largely accounted for the nonlinear relationship between native and alien species richness seen across all plots (Fig. a). For alien‐dominated plots, with low native but high alien species richness per plot (3.9 ± 0.15 and 17.5 ± 0.19, respectively), there was a significant positive relationship between native and alien richness (ρ = 0.26, P < 0.001; Fig. b). In contrast, the species richness relationship was stronger and significantly negative in native‐dominated plots (ρ = −0.34, P < 0.001; Fig. c) with similar mean values of alien and native richness per plot (14.7 ± 0.36 and 14.1 ± 0.4, respectively). Relationship between native and alien plant species richness across the B anks P eninsula, N ew Z ealand in: (a) all 1227 plots (ρ = −0.126, P < 0.001); (b) alien‐dominated plots (739 plots, ρ = 0.26, P < 0.001); (c) native‐dominated plots (488 plots, ρ = −0.34, P < 0.001). Grey points are individual plots; black points show the mean value of alien species richness for each value of native species richness. Solid lines show a cubic smoothing spline fitted to the full data set. Across all plots, the dominant species life forms shift along the gradient of increasing native species richness. Plots with low native species richness have a higher proportion of alien grass and herbaceous species (ρ = −0.59 and ρ = −0.27, P < 0.001 respectively), while plots with high native species richness contain a higher proportion of native trees and shrubs (ρ = 0.59 and ρ = 0.7, P < 0.001, respectively, Table ). These patterns remain when native‐ and alien‐dominated plots are examined separately. Native‐dominated plots with low native species richness have a higher proportion of grass and herbaceous species (ρ = −0.69 and ρ = −0.39, P < 0.001, respectively), while those with high native species richness contain a higher proportion of tree and shrub species (ρ = 0.62 and ρ = 0.69, P < 0.001, respectively). For alien‐dominated plots, the same gradient is apparent where plots with high native species richness have a higher proportion of native tree and shrub species (ρ = 0.29 and ρ = 0.51, P < 0.001, respectively), while plots with high alien species richness have more grass and herbaceous species (ρ = 0.32 and ρ = 0.37, P < 0.001, respectively). Parameter estimates from multiple regression models [generalized least squares ( GLS ) with spatial correlation structures] predicting native and alien species richness within: (i) all plots (d.f. = 1227), (ii) alien‐dominated plots (d.f. = 739), and (iii) native‐dominated plots (d.f. = 488) with climate, environmental and human‐related explanatory variables Variables All plots Alien‐dominated plots Native‐dominated plots Alien richness Native richness Alien richness Native richness Alien richness Native richness Total N 311 368 282 217 194 345 Mean N 16.4 ± 0.19 7.9 ± 0.23 17.5 ± 0.19 3.9 ± 0.15 14.7 ± 0.36 14.1 ± 0.4 Growing degree days 3.39 *** − 2.99 *** 2.01 − 2.61 ** 3.11 *** − 5.26 *** Solar radiation 0.87 *** − 1.44 *** 0.04 − 0.86 ** 1.28 *** − 2.12 *** Distance to buildings −0.06 0.08 − 0.25 ** 0.21 * 0.11 −0.16 Distance to unpaved roads 0.09 0.15 ** −0.11 0.17 ** 0.15 * 0.05 Distance to streams 0.14 ** −0.06 0.21 * 0.21 * 0.17 * − 0.22 ** Proportion tree/plot − 0.33 *** 0.59 *** − 0.32 *** 0.29 *** − 0.61 *** 0.62 *** Soil pH 1.96 *** 1.62 *** 2.05 *** 1.02 * 1.25 1.67 * Soil pH 2 − 1.28 *** − 1.05 *** − 1.34 *** − 0.65 * −0.83 − 1.08 * Alien richness NA − 0.58 *** NA 1.41 *** NA − 0.49 *** Native richness − 0.18 *** NA 0.75 *** NA − 0.72 *** NA R 2 0.11 0.23 0.14 0.21 0.19 0.25 Spatial correlation structures Spherical Exponential Exponential Spherical Exponential Gaussian In all cases, the explanatory and response variables were transformed [log10( x + 1)] to ensure normality and to deal with zero values. Explanatory variables were then standardized to zero mean and standard deviation one so that parameter estimates were comparable. All explanatory variables (see Methods), were tested but only those variables that were statistically significant in at least one model are shown. Significant variables in a given model are shown in bold (*** P < 0.0001, ** P < 0.001, * P < 0.05). Also shown are the coefficients of determination ( R 2 ), the total number of alien and native species (total N ), and the mean number of native and alien species per plot (Mean N ) (±SE). Determinants of native and alien richness Low but significant spatial autocorrelation was consistently found in the residuals of our GLS, and thus, we report results based on these spatial models. Plots with high native richness (containing a higher proportion of tree and shrub species) were more likely to occur at cooler sites (typically at higher elevation) with low solar radiation (steeper, south‐facing slopes), intermediate in soil pH, in areas with lower alien richness that were further away from unpaved roads (Table ). In contrast, plots with higher alien species richness (dominated by grass and herbaceous species) occurred on warmer sites (typically at lower elevation) with high solar radiation (drier north‐facing slopes) that had low native species richness and intermediate soil pH (Table ). Hence, at a broad scale, plots with high native and alien species richness were spatially separated and tended to occupy different parts of the landscape. However, these edaphic factors only accounted for 21% and 9% of the variation in native and alien species richness, respectively. Alien richness in the native richness model and vice versa explained a small but significant amount of additional variation (increasing the variation accounted for to 23% and 11%, respectively; Table ), indicating that unmeasured factors linked with biotic suitability further shaped species richness patterns. These relationships were also evident within alien‐ and native‐dominated plots (Table ). Within each of these groups, alien richness tended to be higher on warmer (lower elevation), drier north‐facing slopes while native richness tended to be higher on cooler (higher elevation) sites on south‐facing slopes, with both alien and native richness higher at intermediate soil pH and distant from streams. Alien richness was low, and native richness was high, when there was a greater number of tree species per plot. The major difference was that, having controlled for other factors in the model, native and alien richness were positively associated in alien‐dominated plots (increasing the total variation accounted for from 14% to 21% respectively), but negatively associated in native‐dominated plots (19% to 25% of total variation accounted for). In addition, alien richness increased and native richness declined significantly with proximity to buildings in alien‐dominated plots and with distance to streams in native‐dominated plots. Alien species richness was generally higher than native species per plot across the elevation gradient (Table ). However, native species richness increased with elevation for all plots and for plots dominated by either native or alien species, such that the only occasion mean native richness was greater than alien richness was at the highest elevations (> 400 m a.s.l.) within native‐dominated plots. In contrast, alien species richness was less influenced by elevation and appears to have a unimodal relationship with a slight peak at mid elevations (Table ). Across all plots, native and alien richness were significantly and positively associated up to 200 m a.s.l., but this relationship became increasingly negative at higher elevations, becoming significantly so above 400 m a.s.l. (Fig. ). Splitting the analysis into alien‐ and native‐dominated plots separately revealed that this trend reflects the positive relationship between native and alien richness in alien‐dominated plots below 300 m a.s.l. and the negative relationship in native‐dominated plots above 300 m a.s.l. Within each elevational band and across the entire elevation gradient, there was a consistent positive relationship between native and alien species richness in alien‐dominated plots, and a consistent negative relationship in native‐dominated plots (Fig. ). Change in S pearman's rank correlation coefficients (black dots with 95% confidence intervals) of native versus alien species richness within five elevational bands (0–100, 101–200, 201–300, 301–400 and > 400 m a.s.l.). Solid line shows native–alien relationship within all plots ( n = 1227). Dotted line shows native–alien relationship within alien‐dominated plots ( n = 739). Broken line shows native–alien relationship within native‐dominated plots ( n = 488). The horizontal dotted line shows value of ρ = 0. The superscript refers to the statistical significance of correlations (*** P < 0.001, ** P < 0.01, * P < 0.05). Total number of alien and native species and mean number of species per plot within: all plots, alien‐dominated and native‐dominated plots in each separate elevational band. Total number of plots in each elevational band are shown Elevational bands (m) 0–100 101–200 201–300 301–400 > 400 Alien Native Alien Native Alien Native Alien Native Alien Native All plots Tot. species 292 220 256 237 218 191 168 156 265 261 Mean species/plot 16.2 4.7 17.7 7.4 17.2 7.3 16.7 8.5 14.5 12.1 Tot. plots 296 260 223 172 276 Alien‐dominated plots Tot. species 219 143 154 131 146 114 97 81 123 108 Mean species/plot 17.0 2.5 18.1 4.1 17.9 3.5 17.5 4.1 16.9 6.3 Tot. plots 219 154 146 97 123 Native‐dominated plots Tot. species 68 73 100 105 67 76 69 75 135 151 Mean species/plot 13.1 11.0 16.9 12.3 15.8 14.4 15.7 14.2 12.6 16.7 Tot. plots 77 106 77 75 153 Discussion Previous interpretation of the sign and magnitude of the relationship between native and alien species richness has largely centred on the ‘invasion paradox’ that addresses how the shape and strength of the native–alien richness relationship can change with spatial grain and extent (Levine & D'Antonio, ; Shea & Chesson, ; Fridley et al ., ; Herben et al ., ). We show, however, that at a constant grain and extent, the relationship between native and alien richness differs between plant communities subject to relatively high or low human impact that are respectively dominated by either alien or native species. Such variation in the native–alien richness relationship at small plot sizes has previously been attributed to statistical problems associated with high turnover of species leading to high variation in species richness among plots and thus inconsistent relationships (Stohlgren et al ., , ; Fridley et al ., ). Our data do show high among‐plot variation in both native and alien richness (Fig. ), but we nevertheless find highly significant relationships with the sign of that relationship shifting from positive to negative in going from alien‐ to native‐dominated plots. This shift could not be fully explained by changes in any of the environmental, climatic or human variables that we measured, although the explanatory variables did a consistently better job explaining native than alien richness. This might be expected if aliens were reasonably ubiquitous as a result of human impacts. Indeed, unlike other studies that typically find a marked decline in alien richness with increasing elevation (Alexander et al ., ; Marini et al ., , ), we observed relatively little change (Table ). Grasslands across the entire elevation range, for example, tended to contain a similar suite of common alien species (e.g. Lolium perenne , Dactylis glomerata and Anthoxanthum odoratum ). Within native‐dominated plots, the gradient of increasing native species richness coincided with a shift from plots at warmer, lower elevation on northerly aspects to plots at cooler and higher elevation sites on south‐facing aspects, with a correspondingly greater proportion of trees. This gradient most likely reflects a legacy of past land‐use, with less‐modified or regenerating areas of native forest and shrubland occurring in less‐accessible and less‐productive higher‐elevation south‐facing sites, while mid‐elevation warmer sites tend to support more modified native‐dominated grassland. Why then is alien species richness negatively correlated with native species richness along this gradient? The traditional interpretation would be that high native richness drives the sign of the relationship and confers resistance to invasion by alien species (biotic resistance). However, while alien species richness declines along this gradient, plots with high native richness still have, on average, a high proportion of alien species (about one‐half to one‐third of species), suggesting that these sites are readily invaded. Instead, it may not be high native richness per se that confers resistance to invasion, but the fact that higher native richness coincides with a shift from grassland to remnant or regenerating native forest and shrubland. The understory of plots dominated by native woody vegetation may be less susceptible to invasion by alien grass or herbaceous species better adapted to more open environments, which comprise the bulk of the alien flora. The lower number of alien species and the higher number of native species in these native‐dominated communities may thus reflect a shift in vegetation structure, from grassland to shrubland/forest, rather than being a function of biotic resistance linked with the number of species. Forested plots with high alien richness may also be in areas regenerating after agricultural abandonment, highlighting the potential for historical factors such as land‐use change to influence current native–alien plant relationships (Parker et al ., ). For alien‐dominated plots, we see a positive association between native and alien species richness, which is commonly attributed to both native and alien species responding in a similar manner to underlying environmental gradients associated with plant performance (Gilbert & Lechowicz, ; Richardson et al ., ). However, in our study, only one variable (soil pH) appeared to influence native and alien richness similarly (Table ). None of the remaining environmental variables we measured could fully explain the covariance between native and alien species richness. Thus, is there any evidence that native and alien species richness covary along either anthropogenic or climate gradients? Stratifying by elevation helps disentangle the potentially confounding effects of covariance among anthropogenic and environmental variables (Marini et al ., ). The relationship between native and alien species richness was consistently positive or negative for alien‐dominated and native‐dominated plots, respectively, even when the variation in climate was constrained within fixed elevational bands. This suggests that anthropogenic effects shape these relationships more strongly than climate. Nevertheless, the strength of the positive and negative relationships changed with elevation, suggesting that the magnitude of anthropogenic effects also vary with elevation. In contrast to the findings of Boughton et al . ( ) who found management intensity resulted in negative relationships between native and alien species, we interpreted our positive relationship to be a function of the intensity of management. Sites with low native and alien species richness are dominated by alien pasture grasses that are intensively managed through grazing, ploughing and fertilizer application to favour just a few highly productive fast growing alien species (for example Lolium perenne and Trifolium spp. swards). Less intensive management may allow pastures to be invaded by other alien and native species, leading to a positive relationship between the two, although aliens dominate in these more intensively managed systems. Thus, the positive relationship is driven by patterns in the persistence of native species along a gradient of management intensity that influences alien species richness to a much lesser extent. With increasing elevation, climate variables might be expected to exert a greater influence on native and alien plant distributions and to affect these in a similar manner (Stohlgren et al ., ; Marini et al ., ; Pauchard et al ., ). However, across the large elevation gradient, while alien species richness showed a unimodal relationship that could be attributed to higher elevations becoming increasingly inclement, native richness progressively increased with elevation. While we might have expected a similar unimodal relationship for native species (Marini et al ., ), the linear relationship undoubtedly reflects the fact that much of the native diversity has been removed by forest clearance at low to mid elevations (Wilson, ). The outcome is that at low elevation, where most plots are heavily modified and dominated by alien species, native and alien richness are positively correlated, while at higher elevations, there is stronger spatial segregation and hence a negative correlation, with less‐modified remnants of native forest vegetation tending to have more native and fewer alien species. Conclusion Much of the discussion to date regarding the drivers of native and alien species richness suffers from the fact that the grain and extent of studies are rarely independent and the grain size covaries with the spatial extent examined (Hulme, ). This prevents adequate assessment of the local and regional drivers on patterns of species richness. Our study is one of the few that examines patterns of species richness at a relatively fine grain (36 m 2 ) over a large spatial extent ( c . 1000 km 2 ) (c.f. Stohlgren et al ., ; Chen et al ., ). Our results do not provide strong evidence of biotic resistance associated with higher native species richness limiting alien plant invasions, although this might only be expected to be found at even finer grain sizes (Levine & D'Antonio, ; Herben et al ., ). In contrast, our study confirms an increasing and recent body of evidence (Parker et al ., ; Boughton et al ., ) that indicates contemporary and historical anthropogenic impacts strongly shape both negative and positive relationships between native and alien species richness, especially where such impacts covary with climate gradients. The impact of local management effects (e.g. land clearance, grazing) may be less discernible at larger grain sizes (> 1 km 2 ) where other broad‐scale environmental factors are likely to shape patterns in species richness. Acknowledgements This research was funded by the New Zealand Tertiary Education Commission via a PhD scholarship grant and Bio‐Protection Research Centre Writing Scholarship. The authors are very grateful to Hugh Wilson for allowing use of the floristic data from his Banks Peninsula survey; Andrew Tait; Susan Wiser; Brad Case, Takayoshi Ikeda and the Lincoln University Spatial Ecology Group for GIS and statistical support.

Journal

Diversity and Distributions – Wiley

Published: Jan 1, 2013

There are no references for this article.