Alternative futures for global biological invasions

Scenario analysis has emerged as a key tool to analyze complex and uncertain future socio-ecological developments. However, currently existing global scenarios (narratives of how the world may develop) have neglected biological invasions, a major threat to biodiversity and the economy. Here, we use a novel participatory process to develop a diverse set of global biological invasion scenarios spanning a wide range of plausible global futures through to 2050. We adapted the widely used “two axes” scenario analysis approach to develop four families of four scenarios each, resulting in 16 scenarios that were later clustered into four contrasting sets of futures. Our analysis highlights that socioeconomic developments and technological innovation have the potential to shape biological invasions, in addition to well-known drivers, such as climate and human land use change and global trade. Our scenarios partially align with the shared socioeconomic pathways created by the climate change research community. Several factors that drive differences in biological invasions were underrepresented in the shared socioeconomic pathways; in particular, the implementation of biosecurity policies. We argue that including factors related to public environmental awareness and technological and trade development in global scenarios and models is essential to adequately consider biological invasions in global environmental assessments and thereby obtain a more integrative picture of future social–ecological developments.


Introduction
Biological invasions are recognized as a major driver of biodiversity loss (Pyšek et al. 2020). Invasive alien species (IAS) can threaten native biota and alter ecosystem functioning, disrupt the delivery of ecosystem services and cause numerous social and economic impacts (Diagne et al. 2020;Simberloff et al. 2013). The number of alien species continues to increase globally with no sign of saturation despite efforts to halt invasions (Seebens et al. 2017). Advances have been made in understanding the direct drivers of biological invasions, such as ecological determinants (Simberloff et al. 2013) and anthropogenic factors, including climate change (Hulme 2017), global trade (Early et al. 2015) and human disturbance (Spear et al. 2013). However, it is still unclear how these direct drivers are shaped by social developments and how indirect social drivers may determine the future influence of biological invasions (e.g. Lotz and Allen 2013). A better understanding of how social change may determine biological changes is a prerequisite to understand and effectively manage biological invasions in the Anthropocene (Essl et al. 2020).
Scenario analysis provides a systematic method to assess how complex interactions among many drivers of change may produce multiple plausible futures (Peterson et al. 2003). Scenario analysis has been increasingly used to analyze likely outcomes of global and regional environmental developments (Oteros-Rozas et al. 2015;Pereira et al. 2020;Spangenberg et al. 2012;van Vuuren et al. 2014). Scenarios are neither predictions nor forecasts, rather they are descriptions and/or qualitative explorations of alternative paths along which the future might unfold (Van der Heijden 2005). The comparative analysis of a set of scenarios can be used to identify key uncertainties and allows the incorporation of alternative or competing perspectives and theories into analysis of potential futures (Peterson et al. 2003). Scenarios are qualitative in nature, although they can be combined with models to produce quantitative estimates of future changes (Lenzner et al. 2019).
We used scenario analysis to investigate the complex and uncertain interactions underlying biological invasions and to capture a variety of expert knowledge on how biological invasions interact with other relevant drivers of global change. The development of scenarios allowed us to explore a wide range of potential variations in the number of IAS that are likely to become established through to 2050. Although analyses of biological invasions have been undertaken at regional scales (e.g. Chytrý et al. 2012;Le Maitre et al. 2004;Roura-Pascual et al. 2011;Wilson et al. 2020), the last global scenario analyses incorporating biological invasions were completed over two decades ago (Carpenter et al. 2005;Sala et al. 2000) and focused on drivers of biodiversity change rather than invasions themselves. Therefore, undertaking an analysis of global trajectories of biological invasions both addresses a critical research gap in environmental science (IPBES 2019), by developing a set of global IAS scenarios and linking those to existing global scenarios. Furthermore, these scenarios can contribute to the ongoing Intergovernmental Science Policy Platform on Biodiversity and Ecosystem Services (IPBES) assessment of Invasive and Alien Species (Stoett et al. 2019), as well as provide a starting point for future quantifications of the effects of biological invasions on the environment and human livelihoods (Lenzner et al. 2019).
We developed new global scenarios for biological invasions to avoid being constrained by pre-existing scenarios that were not created with a focus on the drivers of IAS change, such as the widely used Shared Socioeconomic Pathways (SSPs). The SSPs scenarios were developed by the climate change research community and include both qualitative descriptions and quantifications of broad trends in socioeconomic developments (van Vuuren et al. 2014). These scenarios are used to produce alternative sets of future drivers for integrated assessment models simulating the evolution of land use, energy consumption and resulting greenhouse gas emissions under different climate policy assumptions . However, the development of these scenarios has primarily focused on climate change and they only offer one set of scenarios, framed by a specific set of assumptions, excluding many other framings of potential futures. We compared the SSPs against our biological invasion scenarios, to advance the integration of invasion science into global environmental assessments, a topic that has been identified as a research priority (CBD 2010; IPBES 2019).

Scenarios development
We adapted the common "two axes" scenario analysis approach (Van der Heijden 2005). Although the two-axes scenario approach has strong benefits in its ability to communicate a set of scenarios quickly and in a transparent manner, the use of a single set of scenarios framed through this approach has been criticized, because it limits the exploration of futures through a single set of assumptions and drivers . There are alternative methods that integrate many drivers in a single scenario set, but such approaches end up with a limited set of plausible futures and run the risk of being less transparent in their assumptions .
Our approach was to construct multiple two-axis scenario sets, i.e. four families of four scenarios each. This approach allowed us to investigate potential futures through multiple framings based on the different driver combinations. This, in turn, enabled us to explore a much more multidimensional possibility space for plausible trajectories of biological invasions. We then clustered the 16 scenarios we had developed into a reduced and manageable set of futures (representing archetypes of scenarios) to facilitate comparisons with the single set of widely used SSPs (Fig. 1).
The scenario analysis was conducted using a six-step participatory process, combining expert-based online discussions with a two-day workshop (Vienna, 6-7 October 2016) (Fig. 1). The participants were mainly invasion biologists (24 out of 30), but four experts on topically relevant research such as global change biology and environmental economics and two experts on scenario development were also present (Table S1). Attendees to the workshop (19) were mainly from Europe, but there were also three non-European experts. The remaining participants that contributed online before and after the workshop were four European and seven international collaborators (Table S1). All participants are authors of the manuscript.
Step 1. Specify objective. All participants (30) agreed upon the objective of the scenario analysis through online discussions, which was: "Exploring different plausible futures concerning biological invasions at the global level through to 2050". Multiple indicators are available to assess  (Wilson et al. 2018), but participants used the number of established IAS as the most relevant indicator when thinking about the future. It was preferred over other indices (i.e. rate of change or IAS impact) because of its simplicity and direct link with drivers influencing invasions.
Step 2. Identify drivers. Scenario development experts (L. Rutting and J. Vervoort) reviewed a wide range of existing scenario exercises (Alexandratos and Bruinsma 2012;CCAFS 2014;Gallopin et al. 1997;IPCC 2000IPCC , 2013Mora et al. 2016;OECD 2009;Palazzo et al. 2016;Rockefeller Foundation and GBN 2010;Vervoort et al. 2016;Vervoort et al. 2013) and identified the drivers considered in these future scenarios. This preliminary list of drivers was refined by all participants prior to the workshop and also during the scenario workshop. During this process participants modified, added and removed drivers based on the published literature and their expertise to create a set of drivers influencing biological invasions through to 2050.
Step 3. Identify relevant and uncertain drivers. Participants at the workshop selected the most relevant (i.e. important for biological invasions) and uncertain drivers (i.e. the range of plausible different directions this driver may take is largely unknown) from the list drawn in the previous step based on a voting system. Each participant had ten points to use to rank the most relevant drivers influencing the future status of biological invasions and an additional ten points to rank the most uncertain drivers based on their expertise. Scores were distributed freely among the drivers, with the possibility of assigning multiple points to a single driver. The drivers were then plotted in a coordinate system on the basis of their relevance and uncertainty. Workshop participants examined the position of each driver in the coordinate system and selected through group discussion the highest ranking drivers.
Step 4. Identify key pairs of drivers. This step involved the selection of pairs of drivers that led to the most challenging, diverse and relevant scenarios for exploring the future of biological invasions. Instead of choosing one single pair of drivers (Van der Heijden 2005) or combining a larger number of drivers ) into one overarching, small set of scenarios, workshop participants discussed and voted on the most useful (i.e. challenging, diverse and relevant) pairs of two drivers. First, participants voted for all possible pairs of drivers retained in the previous step and identified the pairs with the most votes, and then voted and selected the four most useful pairs among those with the most votes obtained in the previous voting round. In each voting round, each participant at the workshop had five points to distribute freely among the pairs of drivers; it was possible to assign multiple points to a single pair or to leave some points unassigned. These four pairs of drivers were developed into four scenario families (composed of four scenarios each) and each pair of drivers determined the framing of each future in the family in the following steps.
Step 5. Framing key pairs of drivers. Examination and framing of each of these four pairs of drivers was carried out by breakout groups composed of 4-5 workshop participants, who discussed the different possible states of the driver axes. Drivers can be conceptualized in different ways, for instance, economic development can be defined along extremes of high or low, stable or unstable, equal or unequal, etc. The participants in the workshop defined which driver states were most appropriate for exploring futures of biological invasions. They did so by examining combinations of driver states and what types of scenarios these combinations yield-if a combination did not yield a useful set of scenarios (i.e. challenging, diverse and relevant), then the participants redefined these driver states to produce a more useful set of scenarios.
Step 6. Develop plausible scenarios. Scenarios for biological invasions were developed by the same breakout groups that in the previous step examined and framed each pair of drivers. Each pair of drivers formed a pair of axes, which yielded four scenarios. The group developed narratives for these scenarios (i.e. descriptions of how the future may unfold under each scenarios), by examining both general, contextual developments (i.e. politics and demographic developments; socioeconomic and trade developments; lifestyle and values; technological developments; and environmental developments and natural resources) and specific details on biological invasions. It is important to note that while selected pairs of drivers served as starting points and a way to frame the scenarios, the other drivers from step 2, which did not define the scenario axes, could also be taken into consideration in the development of the scenario narratives themselves. Since each group worked on one pair of drivers, the workshop resulted in four families of four scenarios for a total of 16 different scenarios.

Scenarios comparison
To identify similarities and differences among the scenarios for biological invasion created during the scenario analysis, but also between these scenarios and the widely used SSPs, we clustered all scenarios based on a set of variables. To select these variables, first we coded the scenarios narratives by means of qualitative content analysis. We identified terms that symbolically captured the essence of the different portions of the narratives and organized them in categories (Saldaña 2013). These terms (hereafter referred to as variables) are affected by variations in scenario assumptions, but do not necessarily correspond to the drivers used to frame the scenarios. We identified variables that appeared in three or four families of scenarios and selected them as the variables of interest to compare the scenarios.
Then, we qualitatively assessed how these variables of interest were likely to change under different scenarios for biological invasions by 2050, as well as the resulting total number of established IAS. The assessment of these variables and the number of established IAS was standardized using a scoring rubric, where the magnitude of change was measured using a 5-level Likert scale ranging from + 1 (high increase) to − 1 (high decrease) (Table S3). A value of 0 (no change) designates the current rate of change, while "increase" implies an acceleration and "decrease" implies a slowing of the current rate of change. This scoring rubric was created by the study participants with the specific purpose to compare the scenarios. To facilitate the assessment and to ensure equivalencies among the different ratings, each level had a description associated with it. We attempted to link these descriptions to existing publications assessing and/or projecting changes in future trends related to each variable (i.e. publications embedded in Table S3); the use of absolute values did not intend to be exact, but served to characterize the magnitude of change of each variable and impacts across each five-level category (Table S3). Workshop participants were asked to assess only the scenarios that they contributed to create, by rating the variables considered in the scoring rubric based on the descriptions of the scenario narratives. This process was conducted online by the four breakout groups that created the scenarios during the workshop (in previous step 6). In addition to the scenarios for biological invasion, we conducted a similar assessment of the SSPs. The assessment of SSPs narratives based on the variables of interest considered in the scoring rubric was based on the scenario descriptions provided in several publications Dellink et al. 2017;Fujimori et al. 2017;Jiang and O'Neill 2017;Kriegler et al. 2017;O'Neill et al. 2017O'Neill et al. , 2016Riahi et al. 2017;van der Mensbrugghe 2015;van Vuuren et al. 2017).
These expert-based assessments of future changes in variables of interest under the different scenarios were then used to ordinate and cluster the scenarios/variables. First we applied a principal component analysis (PCA) using the prcomp() function from the 'stats' package in R v4.0.3 (R Core Team 2020) and then used the first two components to visualize the distribution of scenarios in a bivariate plot formed by these two new linear variable combinations. Additionally, we also applied a hierarchical cluster analysis on the scenario and variable coordinates in the main two axes of the PCA. We used the hclust() function from the 'stats' package in R v4.0.3 (R Core Team 2020) and the complete linkage algorithm (the Ward algorithm gave similar results-not shown here). Scenarios for biological invasions were characterized by the number of established IAS expected to have become established by 2050 (i.e. last row in Table S3); this variable was not considered in the PCA, but it was used to differentiate scenarios likely to result in a low or high number of established IAS, respectively, and to assess the coherence of the scenarios within each cluster. These analyses enabled us to identify the relationships between the scenarios for biological invasions and those of SSP narratives, to ensure the consistency of our scenario narratives but also to contribute to the work of other research programs assessing the effects of global change on biodiversity.

Scenarios development
The workshop participants identified 24 drivers as potentially suitable for building the scenarios for biological invasions, which were grouped into five categories: (1) politics and demography (6 drivers); (2) economy and trade (3 drivers); (3) lifestyle and values (4 drivers); (4) technology (2 drivers); and (5) environment and natural resources (9 drivers) (Fig. 2a). During the workshop, 10 of these 24 drivers were classified as being both relevant and uncertain for the storyline development (Fig. 2b). Although climate change (driver 16) was not among the most uncertain drivers, participants decided to include it in the scenario analysis because of its potential to exacerbate invasions (Hulme 2017). Other drivers presenting a high relevance and uncertainty were also combined into a single driver because of their strong relationships. For example, drivers related to the category lifestyle and values (i.e. drivers 11-13) were grouped together, and economic growth was merged with global trade (drivers 7 and 9, respectively). Seven drivers were finally considered as the most relevant and uncertain ones (Fig. 2b).
We evaluated all pairwise combinations of these seven drivers and used voting to select the four pairs of drivers to develop the scenarios. Among all possible pairs of drivers (i.e. 21 pairs), the four pairs considered the most diverging by the participants were: (1) international politics, governance and legislation vs. global trade; (2) global trade vs. social norms (composite of drivers 11-13); (3) social norms vs. technology and innovation; and (4) climate change vs. land use change/development (Figs. 2c and S1). These pairs of drivers included six unique drivers. These drivers correspond to well-known drivers of biological invasions, such as global trade, land use change/development and climate change, as well as drivers associated with societal variables (usually called indirect drivers, e.g. IPBES (2019)) that have been largely ignored in the invasion literature: international politics, governance and legislation; lifestyle and social norms; and technological development and innovations (Fig. 2c). The latter group of drivers could have an enormous influence on the number of IAS in the future, but they 1 3 are difficult to quantify. Each of these pairs of drivers was used to create a family of four scenarios ( Fig. 3; Text S1).

Scenarios comparison
The sixteen scenarios for biological invasions identified using our novel multi-set scenario exploration span a wide range of futures, while also sharing similarities ( Fig. 3; Text S1). Examining the scenarios narratives by means of qualitative content analysis, we found that the variables (i.e. terms of the content analysis) that received the highest attention across the different scenario families were those related to politics and demographic developments, as well as environmental and natural resources. Variables related to technological developments were the least assessed despite their potentially substantial impact on biological invasions (Table S2). Overall, we identified 17 variables of interest to assess the similarities/differences between scenarios using a scoring rubric developed for such purpose (Table S3).
The first two components of the PCA, based on the expert-based assessments of changes in these key variables (Fig. 4), explained 79% of the variance (Fig. 5). The first component (explaining 48% of variance) had positive associations with variables showing the impact of human activities on the environment and negative associations with variables related to political and social responsibility; hence, this component primarily measured public environmental awareness. The second component (32% variance) had negative associations with variables related to technological and trade developments, so this component measured the implementation of technology and trade expansion worldwide (Fig. 5b). Scenarios were ordered along these two main axes and covered the entire scenarios' space (Fig. 5a).
Using the scenario coordinates in the main two axes of the PCA, we clustered the scenarios into four groups (representing the four corners of a scenarios' space) to have a manageable set of scenarios. These groups were coherent with respect to the level of biological invasions of the scenarios they included. Two groups corresponded to futures with higher level of biological invasions (A and B) and two others were characterized by lower levels of invasions (C and D) (Fig. 4, upper horizontal bar; Fig. 6). In addition to scenarios, we also clustered the variables based on how they co-varied across scenarios (Fig. 4, left vertical bar). Futures with rising numbers of established IAS were associated with increasing levels of human pressure on natural environments (cluster IV), while declining trends in biological invasions tended to occur in futures with increasing levels of sustainability policies and lifestyles (cluster I) and social and economic stability (cluster II). Increasing levels of global governance, technological development and transport (cluster III) were indistinctly associated with scenarios with both a high and low level of biological invasions (Fig. 4). Variables within the same cluster appeared highly correlated  (Fig. 2b). Red shades correspond to expert-based assessments of changes in the number of established invasive alien species under each scenario (Fig. 4)  (Fig. S2). These scenario/variable clusters are also evident from the PCA (Fig. 5).
Some of our results do not necessarily follow the common perception that biological invasions are only associated with increasing trade and economic growth, rather we found that how society develops (e.g. economic and social development, IAS mitigation strategies) is relevant to increase or reduce the risk posed by IAS. For example, the scenarios in cluster A present high levels of biological invasions despite declining trends in trade and transport (Fig. 6). Expert assessments suggest an increasing number of established IAS might also result from lack of global political and social coordination and a decline in sustainability policies and lifestyles, together with high levels of human modification of natural environments. Although recent analyses of biological invasions have incorporated surrogates of human modification and economic wealth, few analyses have included variables related to sustainability, such as levels of environmental awareness or development of sustainability policies (Sardain et al. 2019;Seebens et al. 2017). Our results suggest that understanding the future of biological invasions requires more interdisciplinary research that analyses how social and ecological drivers interact to shape biological invasions and in particular more attention should be paid to how emerging or novel technologies could reshape the introduction and spread of IAS (Kueffer 2017;Ricciardi et al. 2017).

Comparison with shared socioeconomic pathways
Following the development of our qualitative scenarios for biological invasions, we also used PCA and cluster analysis to compare them to the shared socioeconomic pathways (SSPs) ). This comparison revealed clusters of similar scenarios from both sets of scenarios, but differences among SSPs did not always correspond to differences in IAS levels among our scenarios (Figs. 4 and  5a). Clusters A and B which have higher levels of biological invasions appear well represented by the SSPs. The scenarios in cluster A share features with the relatively pessimistic SSP3 (named Regional rivalry-a rocky road) in which the world disintegrates politically and economically into smaller  (Table S3). Dendrograms represent the similarity of scenarios/variables based on the hierarchical cluster analysis. Variables (coded as V#) are described in Table S3, while scenarios for biological invasions (S#) in Text S1 regions, while most scenarios in cluster B would be more similar to the high economic growth pathway SSP5 (Fossilfueled development-taking the highway) reliant on very high levels of fossil fuel use (Figs. 4 and 5a). Two scenarios in cluster B (S31 and S44) would, however, share many elements with SSP4 (Inequality-a road divided) describing a highly unequal world and SSP2 (The middle-of-the-road) where trends do not shift markedly from historical patterns (Fig. 4). Although SSP2 is associated with cluster B, it is placed between clusters B and D (Fig. 5a).
Conversely, futures characterized by lower levels of IAS (C and D) are not well captured by the SSPs. The scenarios in cluster D are associated with the relatively optimistic SSP1 (Sustainability-taking the green road) that is oriented towards sustainability, but the IAS outcomes vary from low decrease (S22) to low increase (S43). While cluster D shares features with SSPs, scenarios in cluster C describing a world characterized by regional sustainable developments that present the lowest levels of biological invasions do not. Scenarios in this cluster could be similar to a SSP1 variant with rapid shift to lower consumption lifestyles (Figs. 4 and 5a).
The comparison of SSPs to our scenarios reveals that existing environmental scenarios do not represent key variables that shape biological invasions. The SSPs were created to represent different combinations of socioeconomic challenges for the mitigation and adaptation to climate change ), but they lack influential biologically oriented variables responsible for major changes in future biodiversity. Some variables appear to co-vary with variables in the SSPs, but others, such as biosecurity, do not (Figs. 4 and 5). We suggest that including biological invasions in global environmental scenarios requires: (1) consideration of a broader range of sustainability oriented scenarios (such as those included in cluster C) in order to capture a wider possible range of biological responses and (2) incorporation of key drivers/variables relevant for biological invasions that are currently missing from SSPs, such as how the geographic patterns of trade influences the spread of IAS and the implementation of biosecurity policies. Further understanding of these drivers/variables and their influence on other relevant variables is essential to enable quantitative analysis of how alternative future societal dynamics could shape biological invasions (Lenzner et al. 2019).

Conclusions
Ecological and a few economic factors have captured most of the attention of invasion science, but our expert-based analysis signals the primary role of socioeconomic developments and technological innovation as important drivers of biological invasions. Analyses of how trade and transport dynamics shape biological invasions have provided a starting point, but our study indicates that understanding the future of biological invasions requires analyzing how variables such as technological innovation, urbanization, wealth inequality, social stability, biosecurity and sustainability policies interact with one another to determine biological invasions. Realistic assessments of future biological invasions can only be achieved by examining a broader diversity of factors than are currently addressed and considering drivers and responses of biological invasions explicitly (rather than implicitly by relevant covariates).
Biological invasions are the result of interlinked and complex social-ecological dynamics, so further interdisciplinary research is required to broadly examine how people  (Fig. 4, central matrix) with the first two components of the principal component analysis. The dashed circles and letters/ roman numerals correspond to the clusters found in Fig. 4 for both scenarios and variables, respectively. The codes for variables (V#) and scenarios for biological invasions (S#) are listed in Table S3 and Text S1, respectively and societies interact with biological invasions (Vaz et al. 2017). The expertise of workshop participants undoubtedly influenced the selection of factors explaining biological invasions. However, the diverse range of factors identified highlights the participants' breadth of expertise. Future scenario analyses would likely be improved if they were able to better represent the global variety of experiences surrounding biological invasions, for example by ensuring a diversity of experts with respect to age, gender, cultural background, education and expertise .
Our novel scenario development method has proven to be an important component in allowing us to engage with a multi-dimensional scenario space beyond any one scenario set, while our aggregation method allowed for a comparison across this larger set. We captured different potential future trajectories of biological invasions, focusing on a large variety of interacting drivers and then grouped these multiple scenarios into four clusters presenting divergent futures. Clusters can be considered as scenario archetypes that encapsulate the variability embedded within a certain future. Thus, they can be used as incipient or preliminary formulations to develop new global scenarios or to downscale global scenarios to finer spatial scales.
Further developments of these global scenarios and their refinement into regional or local contexts are needed to better understand the synergies between drivers/variables shaping the future of biological invasions across spatial scales. This work provides a first global analysis of alternative future trajectories of IAS and a basis for the future quantification of the effects of biological invasions on biodiversity, human livelihoods and well being. Having identified the primary importance of variables related to public  (Fig. 4). The description associated with each cluster has been elaborated based on the characteristics of the scenarios included in each group (Text S1). Red shades are proportional to the number of established invasive alien species expected for each cluster by 2050 environmental awareness and technological and trade development as key determinant of biological invasions will also help to shape how to address future policy and management needs that are required to mitigate the impact of this major driver of biodiversity loss.

Supplementary Information
The online version contains supplementary material available at https:// doi. org/ 10. 1007/ s11625-021-00963-6. . We highly appreciate the constructive comments of three anonymous reviewers and the handling editor, Iris Bohnet. This manuscript is based on the work done by BvH while serving at the U.S. National Science Foundation, but the views expressed in the paper do not necessarily reflect those of the U.S. National Science Foundation.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.