1 INTRODUCTION
Biodiversity and the function of ecosystems are threatened by global change drivers such as changes in landuse and climate, as well as biological invasions (Linders et al., 2019; Sala et al., 2000). Invasive species alter a wide range of ecosystem services, including provisioning, regulation, and cultural and supporting functions, and they are particularly hazardous for biodiversity maintenance, human welfare, and the economy (Charles & Dukes, 2007; Chytrý et al., 2009; Hejda, Pyšek, & Jarošík, 2009; Pejchar & Mooney, 2009; Vilà & Ibáñez, 2011). Globalization (e.g., international trade and travel) and climate change (e.g., global warming, droughts, and floods) can interact, which can in turn increase the level of biological invasions (Catford, Jansson, & Nilsson, 2009; Le Maitre, Richardson, & Chapman, 2004; Pino, Font, Carbó, Jové, & Pallarès, 2005; Seebens et al., 2015). As the total number of invasive species increases, some sites may host several alien species (Kuebbing & Nuñez, 2015).
The invasion process is a complex phenomenon, driven by numerous interacting processes, and the effect of this interaction is highly contingent on the context (Chamberlain, Bronstein, & Rudgers, 2014; Frost et al., 2019). Consequently, drivers of plant invasion can vary depending on the specific region and habitat (Taylor et al., 2016). Nevertheless, invasions have a common pattern, which can be summarized as the joint effect of propagule pressure, abiotic characteristics of the environment, and biotic characteristics of both the invader and recipient vegetation (Catford et al., 2009), the so-called PAB framework. Propagule pressure (P) includes dispersal and geographical constraints, while abiotic characteristics (A) comprise environmental and habitat constraints and biotic characteristics (B) describe the internal dynamics of the vegetation and community interactions (Catford et al., 2009). All these factors operate at different spatial scales (Czarniecka-Wiera, Szymura, & Kącki, 2020; Milbau, Stout, Graae, & Nijs, 2009) and are influenced by human activity (Essl et al., 2011). In practice, different indices can be applied as proxies of propagule pressure and abiotic and biotic conditions in modelling plant invasion process (Bazzichetto et al., 2018; Beaury, Finn, Corbin, Barr, & Bradley, 2020; Chytrý et al., 2008; Szymura, Szymura, Zając, & Zając, 2018).
Related to the propagule pressure, the biological invasion correlates with many anthropogenic factors, such as density of the communication network, percentage of urban areas, gardening, and the fragmentation of natural habitats. Such factors can serve as a proxy of propagule pressure (Foxcroft, Pickett, & Cadenasso, 2011; Pollnac, Seipel, Repath, & Rew, 2012; Štajerová, Šmilauer, Brůna, & Pyšek, 2017; Szymura et al., 2018; Vilà & Ibáñez, 2011). In addition, economic and demographic variables reflect the intensity of human activities; therefore, socioeconomic factors such as gross domestic production and human population density can be important in predicting the invasion level (Essl et al., 2011; Hulme, 2017; Pino et al., 2005; Pyšek & Richardson, 2010) because they correlate with trade intensity and communication network density (Hulme, 2009). Among the abiotic interactions with the greatest impact on a large spatial scale (continental, regional), climate is considered the most critical in limiting the geographic distribution of species (Hulme, 2017; Thuiller, Richardson, & Midgley, 2007). In terms of resource availability, invasive species usually prefer productive habitats where they are able to achieve competitive dominance (Czarniecka-Wiera et al., 2020; Peltzer, Kurokawa, & Wardle, 2016; Perkins, Leger, & Nowak, 2011). In addition, environments with high variability in resource availability, resulting from periodic external supply (e.g., surface runoff) or destruction of local vegetation that previously used the resources (e.g., human disturbances, abandonment of agricultural crops), are more susceptible to invasions than habitats with stable availability of resources (Davis, Grime, & Thompson, 2000; Kulmatiski, Beard, & Stark, 2006; Rejmánek, 1989). Given the biotic characteristics of the invader and receipt communities, the limiting similarity hypothesis proposes that the invasion by alien species will be successful if the native species of the recipient community differ from the invader in terms of functional traits and resource requirements (MacArthur & Levins, 1967), which decreases competition for resources (Funk, Cleland, Suding, & Zavaleta, 2008). Thus the functional traits of the invader should not overlap with traits of native plants occurring in the invaded community, which will allow it to occupy an empty niche and successfully invade the community (Funk et al., 2008; Hejda & de Bello, 2013). Because some sites can be invaded by several species simultaneously, determining the interaction between invaders is critical for understanding their distribution (Kuebbing & Nuñez, 2015). For example, the local species assemblage can be driven by a priority effect, and the effect is particularly strong when interacting species have similar use of resources (Vannette & Fukami, 2014). In practice, the abundance and composition of invasive species are also related to landscape characteristics (e.g., habitat fragmentation, patch size, shape, and connections), habitat type, landuse, and the composition of the surrounding landscape because these factors correlate with propagule pressure and habitat quality and availability (Basnou, Iguzquiza, & Pino, 2015; Chytrý et al., 2009; González-Moreno et al., 2013; Štajerová et al., 2017; Szymura, Szymura, & Świerszcz, 2016).
Because of the complexity of biological invasion, better understanding of the underlying factors and their management is challenging. As tools for obtaining reliable and repeatable information for biological analyses as well as nature conservation and management of the invaders, invasive species distribution models (iSDMs) are considered useful (Lozano et al., 2020; Zurell et al., 2020). Modelling species’ environmental requirements and mapping their distributions through space and time help to identify the main introduction pathways and secondary spread and the areas and landuse types that are more prone to invasion. These various threads could be woven into a strategy of prevention and elimination of invasive plant species on a regional scale (Lozano et al., 2020). The iSDMs are especially useful in the face of accelerating global changes and data deficiencies, as well as limited research funding (Yates et al., 2018). The PAB approach, despite its obvious advantages for selection of explanatory variables and model results interpretation, has rarely been used within an invasive species distribution modelling framework (but see Bazzichetto et al., 2018; Czarniecka-Wiera et al., 2020; Lozano et al., 2020).
Goldenrod species from North America represent successful invaders in Europe, Asia, Australia, and New Zealand (Gusev, 2015; Szymura & Szymura, 2013; Ye, Yan, Wu, & Yu, 2019; Zhang & Wan, 2017). In Central Europe, two invasive Solidago species occur, S. giganteaAiton (giant goldenrod) and S. canadensis L. (Canadian goldenrod). Due to their high environmental impact, wide range of distribution, and locally high abundance, invasive Solidagospecies have to be controlled in Europe (Fenesi et al., 2015; Sheppard, Shaw, & Sforza, 2006; Skórka, Lenda, & Tryjanowski, 2010). They have been proposed for addition to the list of hazardous alien species that threaten ecosystems, habitats, or other species in European Union countries (CABI, 2018; EPPO, 2020; Tokarska-Guzik et al., 2015). Unfortunately, the eradication of widely established invasive plant species, such as Solidago , is not feasible. The management strategies need to integrate different options that account for the distribution and abundance of the invader, its environmental niche, and the areas that are likely to experience high impacts (Nagy et al., 2020; Shiferaw et al., 2019; Woodford et al., 2016). Management needs to consider intrinsic factors related to the biology and ecology of the invader, as well as extrinsic environmental factors, such as dispersal vectors and invasion pathways (Shiferaw et al., 2019).
Solidago canadensis and S. gigantea differ with regard to ecological niche in their native range and the time of introduction into Europe. However, previous studies suggest that these two species do not differ regarding their habitat preferences in Central Europe, and observed differences in their spatial distribution patterns emerge from historical contingency and limitation in long-range dispersal. The twoSolidago species occupy different areas and rarely form mixed-species stands (Szymura & Szymura, 2016). In this study, we aimed to find the main drivers of Solidago species’ invasion at a regional scale, using a species distribution model and applying PAB framework for selection of adequate explanatory variables and for ecological interpretation of the models. The distribution models can be used for mapping of invasion probability at a regional level to facilitate invasion control at a macroecological scale.