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

Learn More →

Disentangling the relative effects of environmental versus human factors on the abundance of native and alien plant species in Mediterranean sandy shores

Disentangling the relative effects of environmental versus human factors on the abundance of... Introduction Many factors govern plant community assembly and invasibility ( Richardson & Pyšek, 2006 ). Understanding which specific features affect native plant assemblages is ever more important in light of the increasing global loss of biodiversity. In general, environmental factors that have been found to influence patterns of species richness at a regional scale include climate, landscape heterogeneity, spatial patterns, geomorphologic processes and level of protection ( van Rensburg , 2002 ; Hawkins , 2003 ; Davies , 2005 ; Moser , 2005 ; Thuiller , 2006a ). Moreover, human influence is nowadays one further important agent of change for species richness and diversity patterns ( Maestre, 2004 ; Gaston, 2005 ). Human population density and proxies thereof have often been shown to correlate positively with species richness in response to similar factors, such as productivity ( Balmford , 2001 ; Araújo, 2003 ; Chown , 2003 ), a congruence that suggests a marked conflict between conservation and development. However, depending on spatial scale and overall level of urbanization, intense human disturbance can finally lead to an inversion of this tendency with troubling effects on native biodiversity (e.g. Koh , 2006 ). At the same time, the abundance of alien species is also strongly affected by features of the recipient environment influencing community invasibility ( Rejmánek , 2005 ; Thuiller , 2010 ). Pinpointing abiotic and socio‐economic factors associated with greater abundances of neophytes and with higher levels of invasion can thus help elucidate the threat of biotic exchange ( Dark, 2004 ; Richardson , 2005 ). Human‐related factors tend to be particularly important ( O’Shea & Kirkpatrick, 2000 ; McKinney, 2002 ; Pyšek , 2002 ; Liu , 2005 ; Thuiller , 2006b ). This is consistent with theories that relate urbanization and other human activities to higher levels of propagule pressure (a measure of the number of individuals released into a region to which they are not native; Lockwood , 2005 ; Wilson , 2007 ; Britton‐Simmons & Abbott, 2008 ) which increase chances of alien species establishment ( Alston & Richardson, 2006 ). Studies considering both introduced and native species have also generally found a greater importance of factors associated with human disturbance for alien species ( Deutschewitz , 2003 ; Arévalo , 2005 ). Regional‐scale observational studies consistently find a positive native–alien species richness correlation. Such emergent patterns have been partially explained in terms of ecological processes including species interactions and comparable responses to the same or covarying external factors ( Stohlgren , 1999, 2003 ; Sax, 2002 ; Fridley , 2004 ; Davies , 2005 ; Richardson , 2005 ). However, how environmental and human determinants differentially influence alien species distribution patterns in contrast to native ones has not yet been convincingly clarified, particularly in very stressful ecosystems. Extrinsic factors that favour high numbers of native species may also directly increase niche opportunity for invaders or different specific determinants may facilitate the establishment of alien species ( Davis , 2000 ; Thuiller , 2010 ). We use Mediterranean coastal sand dunes as target ecosystems to tease apart environmental determinants from human‐related parameters underlying patterns of richness for native and alien species. Sandy shores are a typical example of a severely constrained ecotone in which both kinds of determinants are particularly intense. Extreme physical stress and disturbance factors (e.g. sand burial, wind or salinity) act at very small spatial scales shaping community zonation along the dune profile in a gradient from pioneer communities on the beach to shrublands and macchias on fixed dunes ( Wilson & Sykes, 1999 ; Acosta , 2006, 2008 ). The total local species pool is moreover filtered by severe environmental conditions acting at the broad scale including summer droughts and coastline movements ( Forey , 2008 ). At the same time, human activities in littoral areas, including suburban development and touristic exploitation, have been historically widespread ( Brown & McLachlan, 2002 ; Defeo , 2009 ). The aim of this work is therefore to explore how the interplay of environmental and human factors determines native–alien patterns of plant richness at a regional scale but focusing specifically on coastal dune habitats. Considering the level of transformation of Mediterranean coasts, we hypothesize a strong influence of human activity on both the native and alien flora. In particular, for alien species we may expect to find a strong role of human‐mediated introduction and propagule pressure as the main drivers of invasions, whereas environmental factors, generally more important determinants of patterns of native species richness, should play a secondary role. To address this objective, we used an explicit treatment of multicolinearity together with a multimodal inference generalized additive framework accounting for spatial structures in the data. Randomization procedures were then used to estimate the absolute contribution of each selected variable to pattern of richness. Methods Study area The study was conducted along 250 km of the Italian Tyrrhenian coast within the limits of the region Lazio (42°23′ N, 11°39′ E to 41°11′ N, 13°20′ E). More than 90% of this coastline is composed of sandy beaches, although some rocky promontories and river outlets can be found. In this region, the coastal dune system has a relatively simple structure, consisting generally of embryo dunes followed by a main mobile dune ridge and a fixed dune ridge. This area is characterized by a Mediterranean climate, with high summer temperatures, dry summers and most of its annual rainfall distributed in autumn–winter ( Carranza , 2008 ). Plant richness data Plant richness data was obtained from a survey of the vascular flora of central Italian coastal dunes in 3′ by 5′ grid‐cells (about 35 km 2 ) which followed the European Cartographic Project protocol ( Ehrendorfer & Hamann, 1965 ; Acosta , 2008 ). Since the focus was on coastal dune vegetation, sampling of the grid was limited to the geologic class of holocenic dunes, as outlined from a geologic map ( Accordi & Carbone, 1988 ) and from orthophotographs. Fifty‐six grid cells fall within the limits of the region Lazio; however 10 were removed because they either did not contain holocenic dunes or were too strongly urbanized. Within each grid cell, all vascular plant species (natives and aliens) were recorded wherever they occurred on recent dunes ( Acosta , 2008 ). Only neophytes (introduced after the 15th century) were classified as aliens in this study. Considering that the more non‐native species are introduced in a region, the higher is the probability that some of them will establish and become invasive, patterns revealed by analysing the overall number of introduced species are relevant for gaining a broad‐scale view of invasion success. In fact, several authors have explicitly relied on overall numbers of exotic species recorded in floras or in field sampling (irrespective of their estimated naturalization degree) as an indication of the threat of biological invasions ( Pyšek , 2002 ; Sobrino , 2002 ; Deutschewitz , 2003 ). We thus included in our classification of alien species both naturalized species (those with establishing, sustainable populations) and ‘casuals’ (those relying on repeated introduction by humans ( Pyšek , 2004 ) (see Table S1). After verifying dependence on sampled holocenic dune area (A), we standardized richness (S) based on the logarithmic species–area relationship derived from the Arrhenius power function S = cA z (1921 – where c and z are constants varying among taxa and types of ecosystems), which best fitted our data ( Lepš & Štursa, 1989 ; Lomolino, 2000 ). We therefore used residuals of LogS vs. LogA regressions for further analyses, since these represent variability unexplained by sampled dune surfaces. External correlates To analyse the main determinants of richness of both natives and neophytes in sandy shores, we calculated candidate variables within buffers of sampled areas. In a GIS environment, we created separately in each grid cell 1.5‐km buffer areas surrounding sampled holocenic dunes, which were mapped along the coast as outlined previously. We chose 1.5‐km buffers to consider only effects of local neighbouring factors. Firstly, we classified the main possible factors in environmental/physical drivers (climate and coastline movements) and in human drivers (land cover, urbanization and tourism). For each driver, we selected a first set of variables of interest for a total of 25 predictors (see Table S2). For climatic variables (average annual temperature, minimum January temperature, maximum July temperature and annual precipitation), we used mean climate grids derived from the statistical downscaling to 100‐m resolution (as described in Zimmermann , 2007 ) of 1‐km Worldclim monthly maps ( Hijmans , 2005 ). The movement of the coastline between 1977 and 1998 (percentage eroding, stable or advancing) was estimated from recent updates of the atlas of Italian beaches ( CNR, 1999 ; Aucelli , 2006 ). Percentages of urban, agricultural and natural cover were calculated from the CORINE 2000 land cover map, which was obtained from the Italian Institute for Environmental Research and Protection (ISPRA – available at http://www.pcn.minambiente.it/ ) as was the percentage of protected areas. In addition, we considered several variables related both to propagule pressure for alien introduction and to disturbance for native species. Population density and a series of indicators of urban development (total density of buildings, density of buildings built before 1945 and density of buildings built after 1971; density of employees of construction activities) and of tourism pressure (density of employees of hotels, bars, restaurants, campsites, travel and tourism agencies and bathing establishments) were obtained from the 2001 census by the Italian Central Institute of Statistics (ISTAT – available at http://www.istat.it ) at the level of census subunits, which represent the finest spatial resolution of Italian demographic surveys. Density of roads was calculated from a planar graph of the road network of the region. To reduce colinearity, we only kept for model fitting the most meaningful and least interrelated variables identified by calculating a Pearson’s correlation matrix. Variables related to population density and tourism pressure were strongly interrelated ( r > 0.5) so we used only population density (per ha) as a proxy of overall level of human pressure and urban development as a whole. Only the density of buildings built after 1971 (per 10 m 2 ) indicating the degree of recent urban development was independent and was treated separately. This variable in the coastal context is a good proxy of coastal urban sprawl involving the proliferation of garden areas with many planted exotic species. Natural and agricultural land cover (%) were negatively correlated, so we only included natural cover ( r = −0.71). Protected areas (%) were not correlated with natural areas. To represent the progress of the coastline, we included the percentage of advancing coast (%– inversely related to retreating coastline; r = −0.4). We included annual precipitation (mm – inversely related to minimum winter temperature; r = −0.86) to represent climatic influences. We thus used a total of six variables (population density, natural cover, protected areas, recent development, precipitation and advancing coast) for model fitting. Model fitting To explain the spatial pattern of native and alien richness as a function of environmental and human variables, we fitted generalized additive models (GAM) with two degrees of smoothing in a multimodal‐inference framework, making inference from more than one single ‘optimal’ model ( Link & Barker, 2006 ; Dormann , 2008 ). Model weighting ( w i ) was based on small‐sample bias adjustment of Akaike Information Criteria (AICc). We estimated the weight of evidence with which each predictor explains the response variable ( w pi ) as the sum of the AICs weights ( w i ) over all models in which the selected predictor appeared. The predictor with the highest w pi (the closest to 1) gets the highest weight of evidence to explain the response variable (the highest relative importance). Since spurious correlations between a predictor and a response variable may occur because of random or unexplained noises, a permutation procedure was then carried out to test the absolute weight of evidence of each presupposed important variable. The absolute weight of evidence in favour of a given variable was calculated by subtracting the median value of 1000 randomized w pi from the original w pi . Only determinants with positive values of absolute weights were considered important drivers ( Thuiller , 2007 ). We evaluated the overall explanatory power of our analysis by regressing observed richness values against fitted values and extracting the R ‐square of the relationship. To derive fitted values from our models, we averaged the predictions from each submodel weighted by the model w i . Spatial autocorrelation Given our sampling design and the fact that invasions often start from few introduction sites, we tested for spatial autocorrelation (SA) using a Moran’s I test (calculated in relation to the first nearest neighbour of each observation). In case of significant autocorrelation in the residuals of native and alien richness, we added to the model a spatial autocovariate term (SAC) which expresses the potential influence of the response variable at neighbouring sampling sites on local response values ( Dormann , 2007 ). This is a commonly used, relatively simple and straightforward method ( Diniz‐Filho , 2008 ) that allowed us to correct for spatial autocorrelation within our multimodal‐inference approach with comprehensive randomizations. We calculated the SAC term for each grid cell following Augustin (1996) , using 10 km as neighbourhood size and weighting by inverse distance among neighbouring observations. Alien–Native relationship We investigated the relationship between alien and native richness by testing the correlation among the two variables. Finally, to highlight processes that have a direct impact on alien patterns, not mediated by the influence that variables may have on natives, we fitted separate GAMs after controlling for native richness, i.e. using the residuals of simple linear regression of alien on native richness (not correlated with sampling area; r = 0.035, P = 0.82). Spatial and statistical analyses were carried out with Arc Gis 9.2 ( ESRI, 2006 ) and R 2.10.0 ( R Development Core Team, 2008 ) respectively. Results We found a total of 550 species on the entire Tyrrhenian coast. Alien species make up 8% of the total plant species richness in the study area (see Table S1). Moreover, alien and native richness are positively related ( r = 0.50, P < 0.001; Fig. 1 ), but different external correlates determine their spatial pattern. 1 Positive relationship between alien and native species richness (Pearson correlation r = 0.50, P < 0.001). For natives, three variables had absolute weights greater than zero ( Fig. 2a ; Adj. R ‐square: 0.464, P < 0.001). Human population density (correlated with degree of urbanization) was the most important factor and was negatively related to richness ( Fig. 3a ). The negative effect of human impact on native species richness seems to level off at population densities greater than 40 residents per ha. Numbers of natives were also decreased by high proportions of natural land cover and, to a lesser degree, affected by the progress of the coastline (higher percentages of advancing coastline favour greater numbers of native dune species) ( Fig. 3a ; Fig. S1). 2 Absolute weights of evidence (Δ p ) of independent variables for explaining from left to right patterns of (a) native and (b) alien richness, and of (c) residuals of alien richness after simple linear regression on native richness. 3 Response curves for the three most important independent variables of generalized additive models for native (a) and alien (b) richness, and for residuals of alien richness variability after simple linear regression on native richness (c). Additional response curves for the other independent variables in Supplementary material. No spatial autocorrelation of native richness was detected by examining Moran Is (st. deviate = 0.937, P ‐value = 0.1744). However, we highlighted a certain degree of spatial clustering for alien richness (Moran I st. deviate = 1.608, P ‐value = 0.05), especially after controlling for native richness (Moran I st. deviate = 3.653, P ‐value < 0.001). The model for alien richness included five variables with positive absolute weights, with precipitation and natural surfaces as most important variables, followed by progress of the coastline, recent development and the spatial autocovariate ( Fig. 2b ; Adj. R ‐square = 0.373, P < 0.001). More aliens are found in sectors that receive greater precipitation and have lower winter temperatures (as they were negatively correlated) ( Fig. 3b ). After controlling for natives, the SAC was the most important variable for alien patterns ( Fig. 2c ; Adj. R ‐square = 0.587, P < 0.001). As for the influence of human factors, alien introduction appears to be related not to overall degree of urbanization but only to recent urban development after the 70s ( Fig. 2c ). Alien richness tends to increase with density of recent buildings, but evens out at higher densities ( Fig. 3c ). Annual precipitation and advancing coast were again important variables for alien richness also after removing variability explained by native richness. Discussion Tyrrhenian sandy shores sectors richer in native plant species harbour more introduced species as has been previously reported at comparable scales of analysis in other areas. Bruno (2004) in particular observed this relationship even on coastal environments in North America. Several authors have observed that the same external factors were independently driving a similar response of both alien and indigenous species ( Richardson , 2005 ; Kumar , 2006 ; Stohlgren , 2006 ). In particular, Deutschewitz (2003) found at a regional scale of analysis comparable to our study that similar factors were influencing both alien and native species, even if in different order of importance. The situation appears to be more complex in the coastal environment, where for a large part different and uncorrelated drivers regulate richness of native plants and abundance of alien species. While we assumed that physical environmental factors would be important determinants of the distribution of native plant species, we were only able to show a strong response to human pressure. Human population density (positively related to overall level of urbanization) negatively impacts native plant richness and is the predictor variable with the highest weight of evidence. This result is in contrast to the theory of coincidence of biodiversity with human population ( Araújo, 2003 ). However, studies supporting this theory have generally been conducted at rather coarse scales, while it has been noted that at finer spatial resolutions, the relationship between species richness and human density may turn negative ( Gaston, 2005 ; Pautasso, 2007 ). Nevertheless, Deutschewitz (2003) still found a positive correlation with urbanization with the same grain used in our study. However, human pressure on Mediterranean shores is often quite extreme ( Acosta , 2006 ), and other authors have suggested that very intense human disturbance may also cause an inversion of the correlation with human population density ( Koh , 2006 ). Moreover, coastal environments are particularly sensitive to anthropogenic changes ( Defeo , 2009 ). In fact, there are a number of ways in which human disturbances can affect sandy shore habitats. A direct impact on the vegetation is related to exploitation for tourism (which in our study is correlated with population density; e.g. for bathing establishments r > 0.5), leading to trampling and destructive beach management practices. Intensive management typical of many touristic beaches of Lazio includes mechanical cleaning and dune flattening and affects primarily mature stands of the mobile dune but eventually also pioneer communities of the beach which is the most heavily exploited zone. Such a simplification of the community gradient in dunes inevitably leads to lower overall species richness at the grain examined in this study. Moreover, human disturbances can also have other more indirect effects (coastal erosion, infrastructural barriers landwards blocking the retreat of the vegetation, pollution) on diversity by altering the overall dynamic equilibrium of the coastal ecotone and the successional processes of the vegetation. Richness of native plants was also inversely related to the percentage of natural land cover in the surrounding landscape, a fact which may appear somewhat surprising at first sight. However, natural cover is negatively correlated with agricultural land use, which may support the introduction of native weeds and ruderal species directly favoured by farming activities ( Prach , 2001 ). Moreover, private farmed lands represent quite effective barriers for touristic exploitation since they offer few public accesses to the beach. Conversely, Tyrrhenian natural areas are very fragmented and have often been inappropriately reforested ( Bellarosa , 1996 ) representing in this way rather an incentive for summer tourists, and consequently disturbance, with negative effects on native diversity. Physical environmental factors appear to influence patterns of native species only moderately. In fact, only an effect of the movement of the coastline is supported by a small positive absolute weight of evidence. The advancing coastline, which is inversely related to coastal erosion, appears to favour native richness, probably by influencing the overall community zonation. The continuous deposition of sand is fundamental for the formation and conservation of a wide beach area, embryo dunes and mobile dunes ( Packham & Willis, 1997 ). Several seaward plant communities (on the beach, embryo dunes and mobile dunes) are therefore more common on progressing coasts with more complex dune profiles leading to a more complete vegetation zonation and greater habitat heterogeneity. For example, progressing coastlines offer greater availability of beach habitat which is colonized by typical pioneer communities of specialized halophylous and psammophylous native species such as Cakile maritima and Salsola kali (e.g. Packham & Willis, 1997 ; Debez , 2004 ). Moreover, many typical species of the mobile dune, Ammophila arenaria (L.) Link being the prime example, are very specialized and have particular adaptations for sand burial and accumulation ( Maun, 1998 ) and are thus less frequent or declining in those coastal sectors where sand build up is not important. In contrast to patterns observed for native species, the key driver of alien plant richness was not anthropogenic but climatic. We found more introduced species in more mesic and cooler sectors of Tyrrhenian sandy shores. A strong positive effect of precipitation on the number of naturalized aliens has been reported even at the European level ( Lambdon , 2008 ). Even though the opposite trend has been reported as well, with more non‐native species in warmer and drier sectors ( Gasso , 2009 ), it is a widely shared idea that alien species are more successful invaders where environmental conditions are not excessively harsh and resources are not limiting (e.g., light, nutrients and water). This is in relation to theories that suggest that competition intensity and thus biotic resistance are inversely correlated with the amount of resources available ( Alpert , 2000 ; Davis , 2000 ; Richardson & Pyšek, 2006 ). From this point of view, in xeric environments such as coastal dune systems, the increase in water availability caused by higher precipitation may play an important role on alien plant abundance. Indeed, in other physically challenging semiarid environments, for example of North America, observational ( Hobbs & Mooney, 1991 ; Hunter, 1991 ) as well as experimental ( Milchunas & Lauenroth, 1995 ) studies have also found a positive influence of higher annual precipitation or water availability on invasion by aliens ( Dukes & Mooney, 1999 ). Two factors with influence on alien abundance (natural cover and advancing coastline) show similar relationships as with native species, a fact which may partially help explain the positive correlation among the two groups of species. Firstly, more neophytes are associated with fewer natural areas, probably again rather in relation to agricultural lands which may represent introduction sources for some species. However, other factors are also probably at play here, since farming does not seem to be a likely invasion pathway for many non‐native species we recorded (Table S1). Secondly, progressing coastlines as highlighted for natives increase habitat heterogeneity and thus provide more opportunities for alien species as well as for natives ( Deutschewitz , 2003 ; Kumar , 2006 ). On the other hand, population density, although strongly correlated with several urbanization, tourism and beach exploitation indicators (all sources of propagule pressure in coastal environments), does not have a positive effect on numbers of aliens. Chytrý (2008) focusing on many different habitats and multispecies assemblages also found that the effect of propagule pressure (human‐related proxy variables) on the abundance of alien species was very small relative to the effect of local habitat properties. Indeed, the relative contribution of propagule pressure to abundances of alien species may vary widely depending on the habitat examined. Under the extreme conditions typical of coastal habitats, propagule pressure appears to play a less important role for the abundance of non‐native species, in contrast to trends observed in other environments for several alien species ( Rouget & Richardson, 2003 ; Lockwood , 2005 ). In our study, the only strong exception is recent coastal suburbanization after controlling for native richness, which appears to be the most effective source of propagules for aliens in the study area (besides perhaps a minor role of farming). Garden areas associated with urban sprawl and with summer structures built after the 70s, often without clear urban plans and with gardens directly facing the beach, support the continuous introduction of alien plants on the neighbouring sand dune system. In fact, for a noticeable proportion of the alien flora found on natural dunes, including the invasives Carpobrotus acinaciformis and Agave americana , gardening is the most likely invasion pathway. Alien richness thus increases with coastal urban sprawl up to a certain point, levelling off only when extreme suburbanization removes out a number of propagule sources for alien species, like gardens and intensive crops, leading to a decrease of the exotic flora. While no spatial autocorrelation could be detected for native richness, we highlighted the spatial clustering of alien richness, in particular after removing variability explained by natives (in which case spatial patterning was the most important explanatory variable). Autocorrelation is especially interesting in the case of alien plants because it suggests a spreading trend related to one or few specific source sites that account for the majority of introductions ( Pino , 2005 ). For instance, hotspots of alien richness may be related to larger urban nuclei in the inland (e.g. Rome; Fig. 4 ). 4 Map of residuals of alien richness after simple linear regression on native richness along the coast of the Region Lazio (42°23′ N, 11°39′ E to 41°11′ N, 13°20′ E). Conclusions On Tyrrhenian Italian sandy coasts, while native and alien richness are positively correlated, by and large different variables influence their spatial patterns. Contrary to what we might have expected, native plants are more affected by human factors than alien species, which at this resolution are strongly influenced by environmental conditions. The physical harshness of the coastal environment may play an important role in explaining this apparent paradox on multiple grounds. Specialized and well‐adapted native species in a fragile equilibrium with their natural environment are extremely sensitive to environmental modifications brought about by humans. On the other hand, for more generalist alien species known to naturalize across a wide range of niches in Mediterranean systems, the availability of the limited resources plays a predominant role. In this context, quite surprisingly, overall human development appears to be less important for the abundance of introduced species than may have been anticipated. From a conservation perspective, these findings can offer important guidance for regional‐level managers of coastal sandy systems in the Mediterranean. First of all, there appears to be a strong conflict between human activities close to the coastal landscape (residential and related to tourism, which is highly important for the local economy of many Mediterranean areas) and the preservation of native plant biodiversity on dunes. The importance of more sustainable beach management practices on Mediterranean coastal dunes cannot be overemphasized. On the other hand, exotic species control measures in these physically challenging semiarid environments need to be carried out keeping in mind that sectors with greater availability of limiting resources (e.g. water) may be more exposed to alien species establishment. On the introduction front, urban sprawl and garden areas should be closely monitored, as they appear to be the key sources of propagules in these environments. Acknowledgements We thank N. Zimmermann for his help with climatic data. We also thank D.M. Richardson, M. Chytrý and two anonymous reviewers for their constructive comments that helped improve an earlier version of the manuscript. We also acknowledge the financial support of the Italian Ministry of Education (MIUR), PRIN‐COFIN 2005–2007. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Diversity and Distributions Wiley

Disentangling the relative effects of environmental versus human factors on the abundance of native and alien plant species in Mediterranean sandy shores

Loading next page...
 
/lp/wiley/disentangling-the-relative-effects-of-environmental-versus-human-03x39P0k8k

References (82)

Publisher
Wiley
Copyright
© 2010 Blackwell Publishing Ltd
ISSN
1366-9516
eISSN
1472-4642
DOI
10.1111/j.1472-4642.2010.00677.x
Publisher site
See Article on Publisher Site

Abstract

Introduction Many factors govern plant community assembly and invasibility ( Richardson & Pyšek, 2006 ). Understanding which specific features affect native plant assemblages is ever more important in light of the increasing global loss of biodiversity. In general, environmental factors that have been found to influence patterns of species richness at a regional scale include climate, landscape heterogeneity, spatial patterns, geomorphologic processes and level of protection ( van Rensburg , 2002 ; Hawkins , 2003 ; Davies , 2005 ; Moser , 2005 ; Thuiller , 2006a ). Moreover, human influence is nowadays one further important agent of change for species richness and diversity patterns ( Maestre, 2004 ; Gaston, 2005 ). Human population density and proxies thereof have often been shown to correlate positively with species richness in response to similar factors, such as productivity ( Balmford , 2001 ; Araújo, 2003 ; Chown , 2003 ), a congruence that suggests a marked conflict between conservation and development. However, depending on spatial scale and overall level of urbanization, intense human disturbance can finally lead to an inversion of this tendency with troubling effects on native biodiversity (e.g. Koh , 2006 ). At the same time, the abundance of alien species is also strongly affected by features of the recipient environment influencing community invasibility ( Rejmánek , 2005 ; Thuiller , 2010 ). Pinpointing abiotic and socio‐economic factors associated with greater abundances of neophytes and with higher levels of invasion can thus help elucidate the threat of biotic exchange ( Dark, 2004 ; Richardson , 2005 ). Human‐related factors tend to be particularly important ( O’Shea & Kirkpatrick, 2000 ; McKinney, 2002 ; Pyšek , 2002 ; Liu , 2005 ; Thuiller , 2006b ). This is consistent with theories that relate urbanization and other human activities to higher levels of propagule pressure (a measure of the number of individuals released into a region to which they are not native; Lockwood , 2005 ; Wilson , 2007 ; Britton‐Simmons & Abbott, 2008 ) which increase chances of alien species establishment ( Alston & Richardson, 2006 ). Studies considering both introduced and native species have also generally found a greater importance of factors associated with human disturbance for alien species ( Deutschewitz , 2003 ; Arévalo , 2005 ). Regional‐scale observational studies consistently find a positive native–alien species richness correlation. Such emergent patterns have been partially explained in terms of ecological processes including species interactions and comparable responses to the same or covarying external factors ( Stohlgren , 1999, 2003 ; Sax, 2002 ; Fridley , 2004 ; Davies , 2005 ; Richardson , 2005 ). However, how environmental and human determinants differentially influence alien species distribution patterns in contrast to native ones has not yet been convincingly clarified, particularly in very stressful ecosystems. Extrinsic factors that favour high numbers of native species may also directly increase niche opportunity for invaders or different specific determinants may facilitate the establishment of alien species ( Davis , 2000 ; Thuiller , 2010 ). We use Mediterranean coastal sand dunes as target ecosystems to tease apart environmental determinants from human‐related parameters underlying patterns of richness for native and alien species. Sandy shores are a typical example of a severely constrained ecotone in which both kinds of determinants are particularly intense. Extreme physical stress and disturbance factors (e.g. sand burial, wind or salinity) act at very small spatial scales shaping community zonation along the dune profile in a gradient from pioneer communities on the beach to shrublands and macchias on fixed dunes ( Wilson & Sykes, 1999 ; Acosta , 2006, 2008 ). The total local species pool is moreover filtered by severe environmental conditions acting at the broad scale including summer droughts and coastline movements ( Forey , 2008 ). At the same time, human activities in littoral areas, including suburban development and touristic exploitation, have been historically widespread ( Brown & McLachlan, 2002 ; Defeo , 2009 ). The aim of this work is therefore to explore how the interplay of environmental and human factors determines native–alien patterns of plant richness at a regional scale but focusing specifically on coastal dune habitats. Considering the level of transformation of Mediterranean coasts, we hypothesize a strong influence of human activity on both the native and alien flora. In particular, for alien species we may expect to find a strong role of human‐mediated introduction and propagule pressure as the main drivers of invasions, whereas environmental factors, generally more important determinants of patterns of native species richness, should play a secondary role. To address this objective, we used an explicit treatment of multicolinearity together with a multimodal inference generalized additive framework accounting for spatial structures in the data. Randomization procedures were then used to estimate the absolute contribution of each selected variable to pattern of richness. Methods Study area The study was conducted along 250 km of the Italian Tyrrhenian coast within the limits of the region Lazio (42°23′ N, 11°39′ E to 41°11′ N, 13°20′ E). More than 90% of this coastline is composed of sandy beaches, although some rocky promontories and river outlets can be found. In this region, the coastal dune system has a relatively simple structure, consisting generally of embryo dunes followed by a main mobile dune ridge and a fixed dune ridge. This area is characterized by a Mediterranean climate, with high summer temperatures, dry summers and most of its annual rainfall distributed in autumn–winter ( Carranza , 2008 ). Plant richness data Plant richness data was obtained from a survey of the vascular flora of central Italian coastal dunes in 3′ by 5′ grid‐cells (about 35 km 2 ) which followed the European Cartographic Project protocol ( Ehrendorfer & Hamann, 1965 ; Acosta , 2008 ). Since the focus was on coastal dune vegetation, sampling of the grid was limited to the geologic class of holocenic dunes, as outlined from a geologic map ( Accordi & Carbone, 1988 ) and from orthophotographs. Fifty‐six grid cells fall within the limits of the region Lazio; however 10 were removed because they either did not contain holocenic dunes or were too strongly urbanized. Within each grid cell, all vascular plant species (natives and aliens) were recorded wherever they occurred on recent dunes ( Acosta , 2008 ). Only neophytes (introduced after the 15th century) were classified as aliens in this study. Considering that the more non‐native species are introduced in a region, the higher is the probability that some of them will establish and become invasive, patterns revealed by analysing the overall number of introduced species are relevant for gaining a broad‐scale view of invasion success. In fact, several authors have explicitly relied on overall numbers of exotic species recorded in floras or in field sampling (irrespective of their estimated naturalization degree) as an indication of the threat of biological invasions ( Pyšek , 2002 ; Sobrino , 2002 ; Deutschewitz , 2003 ). We thus included in our classification of alien species both naturalized species (those with establishing, sustainable populations) and ‘casuals’ (those relying on repeated introduction by humans ( Pyšek , 2004 ) (see Table S1). After verifying dependence on sampled holocenic dune area (A), we standardized richness (S) based on the logarithmic species–area relationship derived from the Arrhenius power function S = cA z (1921 – where c and z are constants varying among taxa and types of ecosystems), which best fitted our data ( Lepš & Štursa, 1989 ; Lomolino, 2000 ). We therefore used residuals of LogS vs. LogA regressions for further analyses, since these represent variability unexplained by sampled dune surfaces. External correlates To analyse the main determinants of richness of both natives and neophytes in sandy shores, we calculated candidate variables within buffers of sampled areas. In a GIS environment, we created separately in each grid cell 1.5‐km buffer areas surrounding sampled holocenic dunes, which were mapped along the coast as outlined previously. We chose 1.5‐km buffers to consider only effects of local neighbouring factors. Firstly, we classified the main possible factors in environmental/physical drivers (climate and coastline movements) and in human drivers (land cover, urbanization and tourism). For each driver, we selected a first set of variables of interest for a total of 25 predictors (see Table S2). For climatic variables (average annual temperature, minimum January temperature, maximum July temperature and annual precipitation), we used mean climate grids derived from the statistical downscaling to 100‐m resolution (as described in Zimmermann , 2007 ) of 1‐km Worldclim monthly maps ( Hijmans , 2005 ). The movement of the coastline between 1977 and 1998 (percentage eroding, stable or advancing) was estimated from recent updates of the atlas of Italian beaches ( CNR, 1999 ; Aucelli , 2006 ). Percentages of urban, agricultural and natural cover were calculated from the CORINE 2000 land cover map, which was obtained from the Italian Institute for Environmental Research and Protection (ISPRA – available at http://www.pcn.minambiente.it/ ) as was the percentage of protected areas. In addition, we considered several variables related both to propagule pressure for alien introduction and to disturbance for native species. Population density and a series of indicators of urban development (total density of buildings, density of buildings built before 1945 and density of buildings built after 1971; density of employees of construction activities) and of tourism pressure (density of employees of hotels, bars, restaurants, campsites, travel and tourism agencies and bathing establishments) were obtained from the 2001 census by the Italian Central Institute of Statistics (ISTAT – available at http://www.istat.it ) at the level of census subunits, which represent the finest spatial resolution of Italian demographic surveys. Density of roads was calculated from a planar graph of the road network of the region. To reduce colinearity, we only kept for model fitting the most meaningful and least interrelated variables identified by calculating a Pearson’s correlation matrix. Variables related to population density and tourism pressure were strongly interrelated ( r > 0.5) so we used only population density (per ha) as a proxy of overall level of human pressure and urban development as a whole. Only the density of buildings built after 1971 (per 10 m 2 ) indicating the degree of recent urban development was independent and was treated separately. This variable in the coastal context is a good proxy of coastal urban sprawl involving the proliferation of garden areas with many planted exotic species. Natural and agricultural land cover (%) were negatively correlated, so we only included natural cover ( r = −0.71). Protected areas (%) were not correlated with natural areas. To represent the progress of the coastline, we included the percentage of advancing coast (%– inversely related to retreating coastline; r = −0.4). We included annual precipitation (mm – inversely related to minimum winter temperature; r = −0.86) to represent climatic influences. We thus used a total of six variables (population density, natural cover, protected areas, recent development, precipitation and advancing coast) for model fitting. Model fitting To explain the spatial pattern of native and alien richness as a function of environmental and human variables, we fitted generalized additive models (GAM) with two degrees of smoothing in a multimodal‐inference framework, making inference from more than one single ‘optimal’ model ( Link & Barker, 2006 ; Dormann , 2008 ). Model weighting ( w i ) was based on small‐sample bias adjustment of Akaike Information Criteria (AICc). We estimated the weight of evidence with which each predictor explains the response variable ( w pi ) as the sum of the AICs weights ( w i ) over all models in which the selected predictor appeared. The predictor with the highest w pi (the closest to 1) gets the highest weight of evidence to explain the response variable (the highest relative importance). Since spurious correlations between a predictor and a response variable may occur because of random or unexplained noises, a permutation procedure was then carried out to test the absolute weight of evidence of each presupposed important variable. The absolute weight of evidence in favour of a given variable was calculated by subtracting the median value of 1000 randomized w pi from the original w pi . Only determinants with positive values of absolute weights were considered important drivers ( Thuiller , 2007 ). We evaluated the overall explanatory power of our analysis by regressing observed richness values against fitted values and extracting the R ‐square of the relationship. To derive fitted values from our models, we averaged the predictions from each submodel weighted by the model w i . Spatial autocorrelation Given our sampling design and the fact that invasions often start from few introduction sites, we tested for spatial autocorrelation (SA) using a Moran’s I test (calculated in relation to the first nearest neighbour of each observation). In case of significant autocorrelation in the residuals of native and alien richness, we added to the model a spatial autocovariate term (SAC) which expresses the potential influence of the response variable at neighbouring sampling sites on local response values ( Dormann , 2007 ). This is a commonly used, relatively simple and straightforward method ( Diniz‐Filho , 2008 ) that allowed us to correct for spatial autocorrelation within our multimodal‐inference approach with comprehensive randomizations. We calculated the SAC term for each grid cell following Augustin (1996) , using 10 km as neighbourhood size and weighting by inverse distance among neighbouring observations. Alien–Native relationship We investigated the relationship between alien and native richness by testing the correlation among the two variables. Finally, to highlight processes that have a direct impact on alien patterns, not mediated by the influence that variables may have on natives, we fitted separate GAMs after controlling for native richness, i.e. using the residuals of simple linear regression of alien on native richness (not correlated with sampling area; r = 0.035, P = 0.82). Spatial and statistical analyses were carried out with Arc Gis 9.2 ( ESRI, 2006 ) and R 2.10.0 ( R Development Core Team, 2008 ) respectively. Results We found a total of 550 species on the entire Tyrrhenian coast. Alien species make up 8% of the total plant species richness in the study area (see Table S1). Moreover, alien and native richness are positively related ( r = 0.50, P < 0.001; Fig. 1 ), but different external correlates determine their spatial pattern. 1 Positive relationship between alien and native species richness (Pearson correlation r = 0.50, P < 0.001). For natives, three variables had absolute weights greater than zero ( Fig. 2a ; Adj. R ‐square: 0.464, P < 0.001). Human population density (correlated with degree of urbanization) was the most important factor and was negatively related to richness ( Fig. 3a ). The negative effect of human impact on native species richness seems to level off at population densities greater than 40 residents per ha. Numbers of natives were also decreased by high proportions of natural land cover and, to a lesser degree, affected by the progress of the coastline (higher percentages of advancing coastline favour greater numbers of native dune species) ( Fig. 3a ; Fig. S1). 2 Absolute weights of evidence (Δ p ) of independent variables for explaining from left to right patterns of (a) native and (b) alien richness, and of (c) residuals of alien richness after simple linear regression on native richness. 3 Response curves for the three most important independent variables of generalized additive models for native (a) and alien (b) richness, and for residuals of alien richness variability after simple linear regression on native richness (c). Additional response curves for the other independent variables in Supplementary material. No spatial autocorrelation of native richness was detected by examining Moran Is (st. deviate = 0.937, P ‐value = 0.1744). However, we highlighted a certain degree of spatial clustering for alien richness (Moran I st. deviate = 1.608, P ‐value = 0.05), especially after controlling for native richness (Moran I st. deviate = 3.653, P ‐value < 0.001). The model for alien richness included five variables with positive absolute weights, with precipitation and natural surfaces as most important variables, followed by progress of the coastline, recent development and the spatial autocovariate ( Fig. 2b ; Adj. R ‐square = 0.373, P < 0.001). More aliens are found in sectors that receive greater precipitation and have lower winter temperatures (as they were negatively correlated) ( Fig. 3b ). After controlling for natives, the SAC was the most important variable for alien patterns ( Fig. 2c ; Adj. R ‐square = 0.587, P < 0.001). As for the influence of human factors, alien introduction appears to be related not to overall degree of urbanization but only to recent urban development after the 70s ( Fig. 2c ). Alien richness tends to increase with density of recent buildings, but evens out at higher densities ( Fig. 3c ). Annual precipitation and advancing coast were again important variables for alien richness also after removing variability explained by native richness. Discussion Tyrrhenian sandy shores sectors richer in native plant species harbour more introduced species as has been previously reported at comparable scales of analysis in other areas. Bruno (2004) in particular observed this relationship even on coastal environments in North America. Several authors have observed that the same external factors were independently driving a similar response of both alien and indigenous species ( Richardson , 2005 ; Kumar , 2006 ; Stohlgren , 2006 ). In particular, Deutschewitz (2003) found at a regional scale of analysis comparable to our study that similar factors were influencing both alien and native species, even if in different order of importance. The situation appears to be more complex in the coastal environment, where for a large part different and uncorrelated drivers regulate richness of native plants and abundance of alien species. While we assumed that physical environmental factors would be important determinants of the distribution of native plant species, we were only able to show a strong response to human pressure. Human population density (positively related to overall level of urbanization) negatively impacts native plant richness and is the predictor variable with the highest weight of evidence. This result is in contrast to the theory of coincidence of biodiversity with human population ( Araújo, 2003 ). However, studies supporting this theory have generally been conducted at rather coarse scales, while it has been noted that at finer spatial resolutions, the relationship between species richness and human density may turn negative ( Gaston, 2005 ; Pautasso, 2007 ). Nevertheless, Deutschewitz (2003) still found a positive correlation with urbanization with the same grain used in our study. However, human pressure on Mediterranean shores is often quite extreme ( Acosta , 2006 ), and other authors have suggested that very intense human disturbance may also cause an inversion of the correlation with human population density ( Koh , 2006 ). Moreover, coastal environments are particularly sensitive to anthropogenic changes ( Defeo , 2009 ). In fact, there are a number of ways in which human disturbances can affect sandy shore habitats. A direct impact on the vegetation is related to exploitation for tourism (which in our study is correlated with population density; e.g. for bathing establishments r > 0.5), leading to trampling and destructive beach management practices. Intensive management typical of many touristic beaches of Lazio includes mechanical cleaning and dune flattening and affects primarily mature stands of the mobile dune but eventually also pioneer communities of the beach which is the most heavily exploited zone. Such a simplification of the community gradient in dunes inevitably leads to lower overall species richness at the grain examined in this study. Moreover, human disturbances can also have other more indirect effects (coastal erosion, infrastructural barriers landwards blocking the retreat of the vegetation, pollution) on diversity by altering the overall dynamic equilibrium of the coastal ecotone and the successional processes of the vegetation. Richness of native plants was also inversely related to the percentage of natural land cover in the surrounding landscape, a fact which may appear somewhat surprising at first sight. However, natural cover is negatively correlated with agricultural land use, which may support the introduction of native weeds and ruderal species directly favoured by farming activities ( Prach , 2001 ). Moreover, private farmed lands represent quite effective barriers for touristic exploitation since they offer few public accesses to the beach. Conversely, Tyrrhenian natural areas are very fragmented and have often been inappropriately reforested ( Bellarosa , 1996 ) representing in this way rather an incentive for summer tourists, and consequently disturbance, with negative effects on native diversity. Physical environmental factors appear to influence patterns of native species only moderately. In fact, only an effect of the movement of the coastline is supported by a small positive absolute weight of evidence. The advancing coastline, which is inversely related to coastal erosion, appears to favour native richness, probably by influencing the overall community zonation. The continuous deposition of sand is fundamental for the formation and conservation of a wide beach area, embryo dunes and mobile dunes ( Packham & Willis, 1997 ). Several seaward plant communities (on the beach, embryo dunes and mobile dunes) are therefore more common on progressing coasts with more complex dune profiles leading to a more complete vegetation zonation and greater habitat heterogeneity. For example, progressing coastlines offer greater availability of beach habitat which is colonized by typical pioneer communities of specialized halophylous and psammophylous native species such as Cakile maritima and Salsola kali (e.g. Packham & Willis, 1997 ; Debez , 2004 ). Moreover, many typical species of the mobile dune, Ammophila arenaria (L.) Link being the prime example, are very specialized and have particular adaptations for sand burial and accumulation ( Maun, 1998 ) and are thus less frequent or declining in those coastal sectors where sand build up is not important. In contrast to patterns observed for native species, the key driver of alien plant richness was not anthropogenic but climatic. We found more introduced species in more mesic and cooler sectors of Tyrrhenian sandy shores. A strong positive effect of precipitation on the number of naturalized aliens has been reported even at the European level ( Lambdon , 2008 ). Even though the opposite trend has been reported as well, with more non‐native species in warmer and drier sectors ( Gasso , 2009 ), it is a widely shared idea that alien species are more successful invaders where environmental conditions are not excessively harsh and resources are not limiting (e.g., light, nutrients and water). This is in relation to theories that suggest that competition intensity and thus biotic resistance are inversely correlated with the amount of resources available ( Alpert , 2000 ; Davis , 2000 ; Richardson & Pyšek, 2006 ). From this point of view, in xeric environments such as coastal dune systems, the increase in water availability caused by higher precipitation may play an important role on alien plant abundance. Indeed, in other physically challenging semiarid environments, for example of North America, observational ( Hobbs & Mooney, 1991 ; Hunter, 1991 ) as well as experimental ( Milchunas & Lauenroth, 1995 ) studies have also found a positive influence of higher annual precipitation or water availability on invasion by aliens ( Dukes & Mooney, 1999 ). Two factors with influence on alien abundance (natural cover and advancing coastline) show similar relationships as with native species, a fact which may partially help explain the positive correlation among the two groups of species. Firstly, more neophytes are associated with fewer natural areas, probably again rather in relation to agricultural lands which may represent introduction sources for some species. However, other factors are also probably at play here, since farming does not seem to be a likely invasion pathway for many non‐native species we recorded (Table S1). Secondly, progressing coastlines as highlighted for natives increase habitat heterogeneity and thus provide more opportunities for alien species as well as for natives ( Deutschewitz , 2003 ; Kumar , 2006 ). On the other hand, population density, although strongly correlated with several urbanization, tourism and beach exploitation indicators (all sources of propagule pressure in coastal environments), does not have a positive effect on numbers of aliens. Chytrý (2008) focusing on many different habitats and multispecies assemblages also found that the effect of propagule pressure (human‐related proxy variables) on the abundance of alien species was very small relative to the effect of local habitat properties. Indeed, the relative contribution of propagule pressure to abundances of alien species may vary widely depending on the habitat examined. Under the extreme conditions typical of coastal habitats, propagule pressure appears to play a less important role for the abundance of non‐native species, in contrast to trends observed in other environments for several alien species ( Rouget & Richardson, 2003 ; Lockwood , 2005 ). In our study, the only strong exception is recent coastal suburbanization after controlling for native richness, which appears to be the most effective source of propagules for aliens in the study area (besides perhaps a minor role of farming). Garden areas associated with urban sprawl and with summer structures built after the 70s, often without clear urban plans and with gardens directly facing the beach, support the continuous introduction of alien plants on the neighbouring sand dune system. In fact, for a noticeable proportion of the alien flora found on natural dunes, including the invasives Carpobrotus acinaciformis and Agave americana , gardening is the most likely invasion pathway. Alien richness thus increases with coastal urban sprawl up to a certain point, levelling off only when extreme suburbanization removes out a number of propagule sources for alien species, like gardens and intensive crops, leading to a decrease of the exotic flora. While no spatial autocorrelation could be detected for native richness, we highlighted the spatial clustering of alien richness, in particular after removing variability explained by natives (in which case spatial patterning was the most important explanatory variable). Autocorrelation is especially interesting in the case of alien plants because it suggests a spreading trend related to one or few specific source sites that account for the majority of introductions ( Pino , 2005 ). For instance, hotspots of alien richness may be related to larger urban nuclei in the inland (e.g. Rome; Fig. 4 ). 4 Map of residuals of alien richness after simple linear regression on native richness along the coast of the Region Lazio (42°23′ N, 11°39′ E to 41°11′ N, 13°20′ E). Conclusions On Tyrrhenian Italian sandy coasts, while native and alien richness are positively correlated, by and large different variables influence their spatial patterns. Contrary to what we might have expected, native plants are more affected by human factors than alien species, which at this resolution are strongly influenced by environmental conditions. The physical harshness of the coastal environment may play an important role in explaining this apparent paradox on multiple grounds. Specialized and well‐adapted native species in a fragile equilibrium with their natural environment are extremely sensitive to environmental modifications brought about by humans. On the other hand, for more generalist alien species known to naturalize across a wide range of niches in Mediterranean systems, the availability of the limited resources plays a predominant role. In this context, quite surprisingly, overall human development appears to be less important for the abundance of introduced species than may have been anticipated. From a conservation perspective, these findings can offer important guidance for regional‐level managers of coastal sandy systems in the Mediterranean. First of all, there appears to be a strong conflict between human activities close to the coastal landscape (residential and related to tourism, which is highly important for the local economy of many Mediterranean areas) and the preservation of native plant biodiversity on dunes. The importance of more sustainable beach management practices on Mediterranean coastal dunes cannot be overemphasized. On the other hand, exotic species control measures in these physically challenging semiarid environments need to be carried out keeping in mind that sectors with greater availability of limiting resources (e.g. water) may be more exposed to alien species establishment. On the introduction front, urban sprawl and garden areas should be closely monitored, as they appear to be the key sources of propagules in these environments. Acknowledgements We thank N. Zimmermann for his help with climatic data. We also thank D.M. Richardson, M. Chytrý and two anonymous reviewers for their constructive comments that helped improve an earlier version of the manuscript. We also acknowledge the financial support of the Italian Ministry of Education (MIUR), PRIN‐COFIN 2005–2007.

Journal

Diversity and DistributionsWiley

Published: Jul 1, 2010

There are no references for this article.