Economic costs of invasive alien species across Europe

Biological invasions continue to threaten the stability of ecosystems and societies that are dependent on their services. Whilst the ecological impacts of invasive alien species (IAS) have been widely reported in recent decades, there remains a paucity of information concerning their economic impacts. Europe has strong trade and transport links with the rest of the world, facilitating hundreds of IAS incursions, and largely centralised decision-making frameworks. The present study is the first comprehensive and detailed effort that quantifies the costs of IAS collectively across European countries and examines temporal trends in these data. In addition, the distributions of costs across countries, socioeconomic sectors and taxonomic groups are examined, as are socio-economic correlates of management and damage costs. Total costs of IAS in Europe summed to US$140.20 billion (or €116.61 billion) between 1960 and 2020, with the majority (60%) being damage-related and impacting multiple sectors. Costs were also geographically widespread but dominated by impacts in large western and central European countries, i.e. the UK, Spain, France, and Germany. Human population size, land area, GDP, and tourism were significant predictors of invasion costs, with management costs additionally predicted by numbers of introduced species, research effort and trade. Temporally, invasion costs have increased exponentially through time, with up to US$23


Introduction
Despite an increasing number of indicators and alarming reports on the rapid decline of biodiversity globally (Díaz et al. 2020;Haubrock et al. 2021b), efforts to halt biodiversity losses have remained insufficient (Hulme 2009;Scalera 2010;Rayment et al. 2018).Notwithstanding the multiple signals of the rapid decline of natural capital worldwide, global economic resources allocated to prevent and mitigate such losses have not proven adequate to meet conservation management goals, or have been designated inefficiently (Murdoch et al. 2007;Underwood et al. 2008;Stokstad 2010;McCarthy et al. 2012;Waldron et al. 2013Waldron et al. , 2017)).In a highly connected world, with escalating trade and demand for resources, the number of invasive alien species (IAS) is rapidly in-creasing (Seebens et al. 2017).In fact, biological invasions are one of the most eminent global threats to biodiversity, ecosystem services and livelihoods (Bellard et al. 2016;Pysek et al. 2020).Whilst much effort has been directed to improve understanding of the ecological impacts of IAS, knowledge about their economic impacts is limited to a few species, habitats, and/or regions, and often only to direct costs that are straightforward to properly quantify or estimate (Kettunen et al. 2009;Bradshaw et al. 2016).
As a historic epicenter of migration, tourism and trade, Europe represents a hub for alien species introductions (Turbelin et al. 2017).Although several studies have attempted to assess the environmental and socio-economic impacts of IAS in Europe (Weber and Gut 2004;Vilà et al. 2009Vilà et al. , 2010;;Keller et al. 2011), only a few have quantified them in monetary terms (Gren et al. 2009;Kettunen et al. 2009).Pimentel et al. (2000Pimentel et al. ( , 2005) ) and Kettunen et al. (2009) were among the first to attempt to summarize the economic impact of IAS on a continental scale, raising awareness of the actual and potential costs associated with IAS (Hensley 2012).However, due to limited availability of published data at the time, they had to rely heavily on personal communications and technical reports.Kettunen et al. (2009) reported total annual costs of IAS of ~€12 billion across Europe, although given the scarcity of data available at this time, sources and methods used were generally scant (Bradshaw et al. 2016;Diagne et al. 2020aDiagne et al. , 2020b)).Other publications have attempted to collectively assess the costs of IAS (Hoffmann and Broadhurst 2016), for different organism types (Lovell et al. 2006;Van der Veer and Nentwig 2015;Bradshaw et al. 2016;Barbet-Massin et al. 2020;Cuthbert et al. 2021b), and for different countries (e.g.Great Britain: Williams et al. 2010).Scalera (2010), for example, reviewed EU-funded projects on IAS and reported an investment of more than €132 million between 1992 and 2006.Substantial variation in estimations of management and damage costs of IAS and the methodologies used, due to many sources being somewhat scattered and providing only anecdotal information at local, regional and national scales, have limited the estimation of IAS costs so far (e.g.Britton et al. 2010;Oreska and Aldridge 2011).Importantly, in several cases, data reporting the costs of IAS are often found in the grey literature (IUCN 2018), not easily accessible, sometimes not publicly available and not written in English (Angulo et al. 2021b).
This lack of reliable, readily-available data on IAS costs remains a critical knowledge gap in assessing the diversity of impacts associated with biological invasions.Its absence can give the false impression that this information is limited, as costs may be rarely reported in a systematic manner.In addition, the lack of reliable and comprehensive quantification of IAS costs leads to an absence of an economic rationale serving as a solid basis for decision-making by policy makers and other stakeholders.A robust and transparent assessment of costs of IAS at the scale of continents, European states, or trading blocs is currently lacking.While cost estimates are useful at a national scale, their calculation at broader scales may be even more crucial.For example, within both the European Union (EU) and European Economic Area (EEA), where trade agreements encourage the free movement of goods and potentially facilitate the spread of IAS, information on the economic impact of each species could demonstrate the requirements for a greater or lower emphasis on continent-wide biosecurity and control measures.Such an evidence base would also indicate the extent to which different countries are investing into relevant actions, and where funds or political pressure may be targeted to enhance the economic security of both nations and wider trading blocs.
In this context, the InvaCost initiative (Diagne et al. 2020a(Diagne et al. , 2020b) ) tackles this lack of collated data, presenting a comprehensive and urgently-needed database that can be used to thoroughly investigate the costs of IAS at a range of scales, from subnational to continental.Here, we use the InvaCost database to (i) describe Europe-wide impacts of IAS among countries, cost types and economic sectors, (ii) investigate the causes for differences in these costs among European countries, and (iii) examine the temporal trends in costs of IAS in recent decades.

Data compilation and extraction
IAS in InvaCost represent those which have established and spread in novel ranges and have reported socioeconomic impacts (i.e.monetary costs).To estimate the cost of biological invasions on the European economy, we used the InvaCost database (InvaCost v.1.0;Diagne et al. 2020a and subsequent additions, see below).The InvaCost v.1.0database comprises 2,419 entries of reported economic costs of IAS retrieved from published peer-reviewed and grey literature (as of December 2017).Data in InvaCost v.1.0were retrieved from publications in English identified in the Web of Science platform (https://webofknowledge.com/), Google Scholar (https://scholar.google.com/),Google (https://www.google.com/),and through direct contacts with regional experts.InvaCost is a living database for which correction of potential errors and addition of new cost entries are further expected (Diagne et al. 2020a).The InvaCost v.1.0database has been extended recently with 5,212 data entries from non-English sources (Angulo et al. 2020).This dataset was derived from a search in fifteen languages, including languages relevant for Europe: French, Spanish, Portuguese, German, Greek, Dutch, Ukrainian, and Russian (as of May 2020).The cost search protocol was similar to the original InvaCost protocol (Diagne et al. 2020a); however, the majority of these entries resulted from targeted searches, i.e. via searching web pages and directly contacting IAS experts and stakeholders to request for potentially unpublished/publicly unavailable documents containing cost information.We further added supplementary cost data from new references containing cost information, obtained through the same search protocol as used for InvaCost v.1.0(2,374 entries; Ballesteros-Mejia et al. 2020).Individual cost records were standardized to a common currency: 2017 US$ (see Diagne et al. 2020a for detailed information on conversion; exchange rate for 2017: US$1 = €0.8852;World Bank 2020).

Data processing
First, we cleaned the raw data in the InvaCost database.We removed obvious duplicate or overlapping costs, identified through chains of citations or identical cost details.
Where necessary, we split aggregated costs (e.g. if the InvaCost database contained a single cost for Europe but the original source contained costs for each individual country).The period of estimation across reported costs varied considerably, spanning periods of several months to several years.For the purpose of the analysis, and in order to obtain comparable IAS costs, we considered all costs for a period of less than a year as annual costs, and re-calculated costs covering several years on an annual basis.This was performed using the "expandYearlyCosts" function of the 'invacost' package version 0.3-4 (Leroy et al. 2020) in R version 4.0.2(R Core Team 2020).We thus estimated average annual costs represented in the InvaCost database.Deriving the total cumulative cost of invasions over time requires consideration of the probable duration time of each cost occurrence.The duration consisted of the number of years between the probable starting ("Probable_starting_year") and ending ("Probable_ending_year") years of the costs reported by each publication included in the InvaCost database (Diagne et al. 2020a).When information was missing for the starting year, we conservatively considered the publication year of the original reference.For the ending year of costs, however, information was missing only for costs likely to be repeated over years (i.e."potentially ongoing", contrary to "one-time" costs occurring only once along a specific period).Therefore, we considered that these costs might still occur until 2020: the last year from which publications were included in InvaCost and in the non-English dataset.Subsequently, to obtain a comparable total cumulative cost for each estimate over each defined invasion period, we multiplied each annual estimate by the respective duration (in years).All analyses were performed for the period from 1960 to 2020, as monetary exchange rates could not be obtained from official institutions (e.g.World Bank) prior to 1960.The overall number of cost entries before expansion was 4867 and 7461 after expansion, whereby "expansion" refers to the process of annualising cost data of different durations using the aforementioned "expandYearlyCosts" function.

Economic cost descriptors
To examine the costs of IAS incurred within Europe, we filtered the full dataset based on the geographic region "Europe".We provide our final dataset used as a supplement (Suppl.material 1).Naturally, these analyses include species which are native in some European countries, but invasive in others (e.g.European rabbit), but invasion costs are only documented in novel ranges.Costs that were incurred from multiple or unspecified taxa were included in analyses but categorised as "Diverse/Unspecified".The resulting invasion cost totals were examined according to different descriptive fields of the most up-to-date database available when writing this manuscript: i. Official_country: describing the national origin of the listed cost for European countries only.For technical reasons, Kosovo and Serbia were considered as one country, while Turkey was excluded entirely as costs were not clearly attributable to Europe.For transcontinental Russia, we considered and presented only the European part for the total cost, while not considering it for further analyses which were based on fully European countries.As such, Turkey and Russia were excluded from detailed analyses to avoid am-biguities given their transcontinental nature, whereby there was a lack of European-scale indicators that would permit comparison with other European states.Moreover, the underlying spatial resolution of data often precluded determination of European and Asian contributions as costs were presented at national, not regional, scales.Overseas territories (e.g.French Guiana, Reunion, Pitcairn and the Canary Islands) were also excluded; ii. Method_reliability: illustrating the perceived reliability of cost estimates based on the type of publication and method of estimation.Estimates in peer-reviewed publications or official reports, or with documented, repeatable and/or traceable methods were designated as "High" reliability (hereafter, "reliable"); all other estimates were designated as "Low" reliability (Diagne et al. 2020a); iii.Implementation: referring to whether the cost estimate was actually realised in the invaded habitat ("Observed") or whether it was expected ("Potential"); iv.Type_of_cost_merged: grouping of costs according to the categories: (a) "Damage-Loss" referring to damages or losses incurred by invasion (e.g.costs for damage repair, resource losses, medical care), (b) "Management" comprising control-related expenditure (for example monitoring, prevention, management, eradication, research, communication) and money spent on education and maintenance costs, (c) "Diverse/ Unspecified" including mixed damage-loss and management costs (cases where reported costs were not clearly distinguished among cost types); v. Impacted_sector: the activity, societal or market sector that was impacted by the cost (Suppl.material 2); note that individual cost entries not allocated to a single sector were classified under "Mixed" in the "Impacted_sector" column.

Economic cost correlations
We first explored whether the two main types of costs, "Management" and "Damage-Loss", can be explained by country-specific factors.To do so, we calculated the cumulative reliable observed costs for 1960-2020 of each type of cost at the country level and selected a range of socio-economic variables that we hypothesize could be linked to biological invasions (Suppl.material 3).Then, we calculated Spearman rank correlations (r s ) between the country-level expenditures and damage costs and the selected socio-economic variables using the R package 'ggpubr' (Kassambara 2017).Further, we also explored correlations between country-level expenditures and damage costs.

Spatial and taxonomic connectivity of costs
To examine the spatial and taxonomic connectivity of invasion costs in Europe, we constructed a bipartite network composed of two types of nodes: (1) countries and (2) taxonomic groupings (excluding studies reporting costs on diverse taxonomic groups).For taxonomic groupings, we also captured habitat types of each taxon (e.g."terrestrial arthropod" instead of "arthropod").When an IAS group economically impacted a given country, a link was drawn between the associated nodes with a weight proportional to the economic impact.As such, the size of the nodes, and thickness of the links, correspond to the magnitude of cumulative economic costs incurred for the 1960-2020 period.To investigate spatial and taxonomic patterns of costs in Europe, we applied the Map Equation community-detection algorithm (version 0.19.12,www.mapequation.org;Rosvall and Bergstrom 2008).This approach groups nodes into clusters with high intragroup connectivity, enabling clusters of similar costs to be established (i.e.countries sharing costs from the same invasive taxa) (Leroy et al. 2019).Network analyses were performed with the 'biogeonetworks' R package version 0.1.2(Leroy 2020), and the network was represented with Gephi 0.9.2 using the ForceAtlas2 algorithm (Bastian et al. 2009).

Temporal dynamics of accumulated costs
For the temporal estimation of the average annual costs, we used the 'invacost' package in R (Leroy et al. 2020).This package allows modelling the trend of costs over time with an array of linear and non-linear model types and enables a summary and comparison of their respective outputs.Given the evidence that numbers of IAS show no sign of saturation (Seebens et al. 2017), we expected their associated costs to be stable or increasing.In addition, we can expect a time lag between the occurrence of costs, their publication, and their reporting in InvaCost (Leroy et al. 2020).Therefore, as per Seebens et al. (2017), we excluded recent years from model calibration.The last eight years appear to have less than 75% completeness within the global InvaCost database (Leroy et al. 2020); therefore, we chose to exclude them from model calibration (i.e. years post-2013).
A range of modelling techniques were then applied to model the temporal dynamics of reported costs ("modelCosts" function): ordinary least squares regressions (linear, quadratic), robust regressions (linear, quadratic -R package 'robustbase'; Maechler et al. 2020), multivariate additive regression splines (MARS -R package 'earth'; Milborrow et al. 2018), generalised additive models (GAM -R package 'mgcv'; Wood et al. 2016) and quantile regressions (quantiles 0.1, 0.5, 0.9 -R package 'quantreg'; Koenker 2020).These approaches enabled quantification of average annual costs, measurements of variation in cost estimates over time and assessment of predictive performance across models (based on RMSE).Model selection was also performed on the basis of techniques that are relatively robust to issues of heteroskedasticity, outliers and temporal autocorrelation that are common in econometric data (Leroy et al. 2020).Moreover, the diverse modelling approach enabled potential generalities in trends to be determined, such as whether all models were consistent in projecting cost increases through time.
As a separate analysis, we further used the aforementioned combination of approaches to examine temporal trends in economic costs, based on the GDP-qualified economic costs of the European countries from the year the cost occurred (i.e., costs divided by GDP per year), elucidating whether invasion costs are still increasing relative to economic growth.For this, we utilized robust regressions modelling as implemented in the 'invacost' package, since those are based on iteratively reweighted least squares, which makes them less sensitive to outliers compared to ordinary least square regressions (Yohai 1987; Koller and Stahel 2011).

Method reliability % cost % entries
Implementation %cost % entries High Low Observed Potential 0% 25% 50% 75% 100% 0% 25% 50% 75% 100% Figure 1.Nature of reported costs (monetary totals and numbers of database entries) for IAS across European countries according to percentages considering method reliability (high vs. low) and implementation type (potential vs. observed).Highly reliable figures are from peer-reviewed, official and/or reproducible sources; observed costs have been empirically realised (i.e.excluding expected cost estimations).
A list of the costliest invasive alien species in Europe can be found in Table 1.Considering all costs, five invertebrates, three vertebrates, and two plants were present in the top 10.When considering only reliable observed costs, three invertebrates, four vertebrates, two plants and one fungi genera were included in the top 10.Rattus species had the highest reliable observed costs (4 th highest when considering all costs) (reliable: US$6.60 billion; all: US$6.67 billion) spanning across 2 countries.Hereafter, all analyses are performed with Russia omitted.

Economic cost correlations
Figure 4 highlights the geographical variations in the total cost of invasions throughout Europe, without and with standardization by GDP.There is a positive relationship between the total cost of invasions and country GDP, i.e. countries with a higher GDP tend to have higher reported observed costs (Figure 4c).High costs of invasion compared to GDP were observed in eastern European countries such as Ukraine, Serbia, Romania, Moldova and Hungary, suggesting that this trend may also change when more studies are undertaken or translated (Suppl.material 4).
We found significant positive correlations between damage-loss and management costs with the following socio-economic variables of the considered countries: human population size, land area, GDP, international tourism as expenditures and as number of arrivals.We also found significant positive correlations between management costs and the number of introduced alien species, research effort as the number of papers on the topic of biological invasions and expenditure in R&D, number of researchers, and imports of goods and services, with other tested socio-economic variables showing no significant correlations (Table 2).Moreover, the EU country-specific expenditure in IAS management and in damages-losses induced by IAS were not significantly correlated (r s = 0.10, p = 0.560).

Spatial and taxonomic connectivity of costs
Eight distinct clusters of nodes were found to be strongly interconnected across taxa and countries (Figure 5).These clusters comprised assemblages of typically one or two countries, alongside one or more groups of organisms.The UK was primarily highly impacted by terrestrial mammals, birds, forbs and aquatic organisms; the Netherlands and Finland by terrestrial arthropods; Norway by aquatic microorganisms; Germany and Estonia by semi-aquatic mammals; Sweden by microorganisms, molluscs and aquatic arthropods/plants; Spain by a diverse array of groups, excepting taxa such as macroalgae and nematodes; and Belgium by semi-aquatic amphibians and terrestrial plants.In turn, the main impacts in France, Italy, as well as in multiple eastern European countries, were caused by terrestrial forbs which turned out to be the costliest group in Europe.Nevertheless, the substantial array of inter-cluster links suggested that European states were each impacted by a diverse array of invasive alien taxa (Figure 5).The larger the link, the higher the cost.Likewise, node size is proportional to the total cumulative cost.For species nodes, node size represents the total cost they had over all countries.For country nodes, the node size represents the total cost of all species in that country.Note that studies reporting costs on 'diverse' groups of organisms rather than specific species were excluded from this network.
Figure 6.Temporal trend of total annual invasion costs recorded in Europe according to multivariate adaptive regression splines (MARS) (a red) and quantile regressions; from bottom to top: 0.1: light grey, 0.5: grey, 0.9: dark grey (b) between 1960 and 2020, as well as reliable observed costs, MARS (c red) and quantile regressions; from bottom to top: 0.1: light grey, 0.5: grey, 0.9: dark grey (d) between 1970 and 2020.Error bands on MARS represent prediction intervals (i.e. the interval of cost that any individual year can have).Error bands on quantile regressions represent 95% confidence intervals.Yearly data are triangles (until 2013) and circles (after); only the former are used in the models.
between 1960-1969 was below US$0.16 billion, it increased to an average annual cost of US$6.35 billion per year in 2010-2020.Considering only reliable, observed costs, the first database entry occurred a decade later than when considering all costs, totalling at an average annual cost of US$963.9 million per year (€802.9 million annually).Reliable costs between 1970-1979 averaged US$26.1 million per year, increasing to US$3.75 billion per year in 2000-2010 before dropping to US$944.3 million in 2010-2020, likely due to lags between costs and their reporting.However, averaging across such long time periods may not clearly demonstrate temporal trends.As such, the best fitting models of temporal cost trends (MARS and quadratic OLS, see Suppl.materials 5, 6) both predict a steep linear increase on a log-scale in IAS driven costs to Europe over the 1960-2013 period (Figure 6).Considering all costs, the best model (MARS: predicted 2013 costs of US$23.58 billion / €19.64 billion; OLS: 0.1 st quantile: US$3.62 billion; 0.5 th quantile: US$15.57billion; 0.9 th quantile: US$59.02 billion) indicated a 12.6 to 14.1-fold increase every ten years of costs incurred from IAS (Figure 6a, b), while considering only reliable costs (MARS: predicted 2013 costs US$4.07 billion / €3.39 billion; OLS: (0.1 st quantile: US$133.18million; 0.5 th quantile: US$172.52 million; 0.9 th quantile: US$27.68 billion) suggested a 10.7-fold increase every ten years of reliable observed costs inferred from IAS (Figure 6c, d).If these trends were to continue over the most recent years for which data is incomplete, then extrapolations in 2020 based on MARS models would yield US$139.56 billion / €116.24 for all costs and US$21.98 billion / €18.31 billion for reliable observed costs only.Considering GDP-qualified economic costs, monetary impacts continued to significantly increase in recent decades (model coefficients shown in Suppl.material 7), irrespective of concurrent economic growth in Europe (Figure 7).Accordingly, the proportional share of GDP devoted to invasion costs has been increasing through time, with invasion costs rising at a greater rate than the rate of economic growth, as evidenced by the steep increase in recent years.

Discussion
The total cumulative cost of IAS in Europe between 1960 and 2020 was estimated at US$140.20 billion.We identified an exponential increase in the costs of IAS over the studied time period, with costs increasing at least ten-fold every decade.Invasion costs reached US$24 billion in 2013 alone (the last year with 'complete' data), and our model extrapolated 2020 costs of up to US$140 billion.While the reported annual cost of IAS in Europe represented < 0.01% of the European Union (EU) GDP (2017 US$15.3 trillion), it was considerably larger than the annual GDP of national economies such as Malta -in recent years (US$12.8billion).
While this total may overestimate some individual costs (e.g. in those cases where reported timelines of expenditure for a specific project were unclear in the literature), it remains a highly conservative value given the many challenges attached to assigning costs to IAS impacts.For the purposes of this analysis, we have considered reported costs and expenditure.However, we note that costs of IAS are generally not restricted to directly quantifiable damages or expenditure on management, but also include various indirect costs that are not always easily quantifiable, and therefore not as commonly reported in the literature.For example, many IAS have substantial impacts on human health, native species or ecosystem services that indirectly harm ecosystems and undermine human wellbeing, yet these costs are not easy to capture or quantify (Medlock et al. 2012;Hamaoui-Laguel et al. 2015;Ogden et al. 2019).A striking illustration has been published by Walsh et al. (2016) who reported a significant decrease in the biomass of the grazer Daphnia pulicaria in lakes invaded by the spiny water flea Bythotrephes longimanus, in turn causing a substantial decrease in water quality by affecting its clarity and total phosphorus content.Other examples include biting nuisances by invasive mosquito species (e.g.Aedes albopictus) or invasive ant species (e.g.Solenopsis invicta) which can negate recreational activities (e.g.Angulo et al. 2021c); and adverse impacts by invasive tree-boring insects (e.g.Agrilus planipennis) on trees that could be costly for the respective economy, although these costs are seldom quantified.Indirect costs are often overlooked or at best underestimated, resulting in minimal investments for alleviation (Rogers et al. 2017;Linders et al. 2019).Although our cost estimations cover 410 species (340 species when considering only reliable observed costs), there remain over ~4,000 IAS in Europe without reported costs (Pagad et al. 2018), indicating that our estimates are highly conservative.Moreover, often costs such as salaries of invasion researchers or managers are not published or accounted for.
Marked differences in cost reporting and totals were found among European countries, with impacts to the UK, Spanish, French, Russian and German economies being most pervasive considering all costs (see Cuthbert et al. 2021a;Angulo et al. 2021a;Renault et al. 2021;Kirichenko et al. 2021;and Haubrock et al. 2021a, respectively).The highest observed costs were found in the UK (Cuthbert et al. 2021a), a country with a long colonial history highly reliant on trade (Clark et al. 2014) and previously identified as a "receiver and donor" country (e.g. for aquatic invasions see García-Berthou et al. 2005).Similar to the UK, the rest of the aforementioned countries with the highest total costs have large economies and most of them were colonial powers, all factors that putatively contribute to high levels of invasions and impacts (Hulme 2009;Hulme et al. 2009).However, the west-European dominance in IAS costs may also be explained by the limited reporting of costs for Eastern European, and potentially also some Nordic, countries.Additionally, the limited reporting of the invasion costs may partly be attributed to the gap of the InvaCost database in sources/documents in languages other than English.The non-English data were collected for only a subset of European languages (Angulo et al. 2021b), leaving aside several languages from Eastern and Northern Europe (e.g.Romanian, Hungarian, Serbian, Polish, and Nordic languages -Finnish, Swedish, Danish etc.).For Eastern European countries, e.g.those of the former communist bloc, one reason for their low reported costs may be that up until 1990, there was little documentation of monetary impacts or, if there was, this information was not made publicly available.Further, differences in societal norms, awareness or regulations may contribute to the lower reported costs for Eastern European countries.However, we note that, considering highly reliable observed costs only, Eastern Russia, Ukraine and Romania exhibited relatively high costs.Regardless of the drivers of this limited reporting, it is a concern, considering that coordinated responses and cooperation are key to efficiently managing invasions and mitigating their impacts (Kark et al. 2015;Latombe et al. 2017;Ogden et al. 2019).
Cultural differences among countries, regional perceptions and national priorities may also influence the level and way of reporting, for example through perceived country-specific sectors of economic importance e.g.forestry and agriculture.In some countries, alien taxa such as trees have been perceived to provide cultural heritage services, particularly in areas with lower levels of development and life satisfaction (Vaz et al. 2018), which might influence cost reporting.Our results also reflect the difficulties of identifying how different sectors may have been impacted -a substantial share of reported costs (29%; US$41.17 billion) were not attributed to a single affected sector.Another important driver of differences in reporting across European countries may lie in differences in perceptions of the severity of IAS impacts.For example, a Europeanwide survey on attitudes towards biodiversity indicated substantial differences between citizens of different countries in their perceptions towards newly introduced plants and animals.Residents of Spain, Portugal and Slovenia were most likely to view them as a great threat to biodiversity, while those from Finland, the Netherlands and Eastern European countries were less likely to be concerned about the threats of introduced species (European Commission 2013Commission , 2015)).For Eastern European countries, initiatives during the Soviet Union times to increase production (i.e. in agriculture, fisheries etc.) and support regional employment may have contributed to the view that new species introductions hold large positive economic potential, which later on may have shaped public views and research agendas towards favoring and/or accepting these species (Kourantidou and Kaiser 2019).Furthermore, in European aquatic systems, alien taxa were reportedly introduced to improve yields from fish farming historically, and particularly in human-altered waterbodies (Arbačiauskas et al. 2010).Although the reasons for the differences in perception of IAS as a threat are not well understood, with perception and values attributed to biodiversity being complex but consistent among social categories, gender and age (Atlan and van Tilbeurgh 2019), higher levels of awareness of their harmful impacts can help support more management actions, research investments and increased efforts to document and report costs.However, these also depend on public support, and this may also vary across specific actions or environments (e.g.Perry and Perry 2008;Crowley et al. 2017).Ultimately, the differences in perceptions of IAS among European states could be a major driver in unevenness of cost reporting among nations, as well as through differences in national-scale policy frameworks.A lack of reporting from many states likely renders our totals as underestimates, but the extent of this underestimation probably differs among countries.
Despite this variability in reported economic costs among European countries (in France, for example, <1% of total reported costs were associated with management as compared with 86% in Germany or 92% in the Netherlands; see e.g.Renault et al. 2021;Haubrock et al. 2021a), the majority of costs (US$84.18billion; 60%) comprised expenditure on damages and losses, while control-related expenditure represented only 20% of all costs (US$28.17 billion).This dominance of damage costs over management investments is paralleled in other regions, such as Asia (Liu et al. 2021), Africa (Diagne et al. 2021b), North America (Crystal-Ornelas et al. 2021), Central/ South America (Heringer et al. 2021), and Australia (Bradshaw et al. 2021); but some individual countries appear to have more management costs (Angulo et al. 2021a for Spain;Ballesteros-Mejia et al. 2021 for Ecuador;Watari et al. 2021 for Japan).Similar to Kourantidou et al. (2021), a number of socio-economic factors significantly correlated with both the reported damages and management costs of IAS, namely: human population size, land area, GDP, and international tourism of the studied countries.These predictors help explain some of the discrepancies in shares of IAS management and damages cost across European countries.First, in countries with higher population, larger land areas, and more international tourism, new species are more likely to be introduced, propagate and invade, while higher human population may also result in increased awareness of specific damage types, e.g. to infrastructure (Mooney and Cleland 2001;Hulme et al. 2009;Hall 2015).This might lead to an increased willingness to pay for managing them.On the other hand, higher GDP might lead to higher resources (e.g.funding and capacity) available to understand and manage IAS.Indeed, the strong relationship found between research effort and numbers of researchers and management cost magnitudes exemplifies this point: greater research investments align with higher reporting of management costs.Our results also indicate that increasing imports of goods and services are associated with greater management spending.It may be assumed that money spent on IAS management would be at least a partial reflection of the total damages incurred.However, there was no significant relationship between reported damage-loss and management costs (Table 2).If management expenditure is largely independent of the number of IAS present and their negative economic impacts, this may reflect a fixed budgetary availability (i.e. the funding available for IAS management is independent of the number of IAS and their impacts in the country).Moreover, the overall three-fold difference in damage-related compared to management costs (eight-fold for observed reliable costs) is alarming, particularly given that preventative measures for invasions (which are classified under management in this study) are shown to be effective at reducing costs than longer-term interventions (Leung et al. 2002;Ahmed et al. 2021), and that countries with a higher proportion of money spent on biosecurity experience generally lower damage costs (Jay et al. 2003;Kritikos et al. 2005).
The InvaCost data also indicate more than a 10-fold increase every ten years in costs associated with IAS since 1960.This finding is likely a result of several trends: foremost the increasing number of IAS in Europe (Seebens et al. 2017), global cost trends (Diagne et al. 2021a;Cuthbert et al. 2021c) and the increasing number of publications within the field of invasion science (Richardson and Pyšek 2008).This is followed by the increase in the GDP of most European countries; and the increasing awareness and number of legislative instruments (at national and EU levels) adopted to tackle IAS (Garcıa de Lomas and Vilà 2015; Turbelin et al. 2017, but see Coughlan et al. 2020).These factors likely contribute to a growth in reported costs and also to an increase in budgets over time.With several thousand alien species established in Europe (Dawson et al. 2017) and legislation in place to tackle IAS throughout the continent, it is somewhat surprising that management and mixed costs (which comprise some management component) represent a small proportion of the total.However, this disconnect between resources made available to mitigate invasion impacts and the large number of IAS worldwide is not a trend unique to Europe (Andreu et al. 2009).Management of IAS can be compromised by a range of factors including insufficient knowledge of species origin and biology, lack of appropriate management strategies, societal ignorance, and lack of resources (Sharp et al. 2011;Courchamp et al. 2017;Kirichenko et al. 2019).Financing provided for biomonitoring and/or eradication plans is frequently of insufficient length, compromising outcomes while simultaneously increasing both management and damage costs (Sutcliffe et al. 2018;Pergl et al. 2019).Further, the insufficient cooperation among and within countries, for example in implementing risk assessments and management planning for IAS, can result in ineffective management strategies (Sharp et al. 2011;Keller et al. 2011).Even if such planning deficiencies are specifically considered, as in the framework proposed by the Convention on Biological Diversity (CBD 2020), the feasibility of management actions remains impaired by the paucity of resources (Heink et al. 2018).

Conclusion
The cost estimations presented in this publication synthesize the state of knowledge on economic costs associated with IAS at the European level.Such cost information on biological invasions at regional scales is especially important for planning coordinated responses, cooperative action but also for interaction at multiple levels among European countries within the EU or EEA and with non-European countries through e.g. trade agreements.Further, we identified significantly higher costs in recent years than previous estimates of ~€12 billion (Kettunen et al. 2009), despite the identified knowledge gaps for various IAS.This becomes particularly important in light of the effects of past agreements such as the freedoms guaranteed by Article 21 of the Treaty on the Functioning of the EU, with the freedom of movement being linked to the enhanced displacement of various species within Europe (de Sadeleer 2014).From a management co-operation standpoint, whether within the EU or between trading partners within Europe, the economic burden imposed by IAS becomes particularly relevant, given that increasing costs burden certain countries disproportionately, likely putting monetary strain on economically weaker countries.A comprehensive appraisal of costs would ultimately contribute to well-targeted investments into conservation measures on an EU and continental scale.innovation programme under the Marie Skłodowska-Curie grant no.747120.MG and CD were funded by the BiodivERsA-Belmont Forum Project "Alien Scenarios" (BMBF/PT DLR 01LC1807C).NK was partially supported by the Russian Foundation for Basic Research (grant no.19-04-01029-A) [national literature survey] and the basic project of Sukachev Institute of Forest SB RAS (project no.0287-2021-0011) [InvaCost database contribution].DR thanks InEE-CNRS who supports the network GdR 3647 'Invasions Biologiques'.Funds for AJT, EA and LBM contracts come from the AXA Research Fund Chair of Invasion Biology of University Paris Saclay.BL, DR and FC are French agents (affiliated, respectively, to the Muséum National d'Histoire Naturelle, University of Rennes and Centre National de la Recherche Scientifique); their salaries, for which they are grateful, are typically not accounted for in assessment of costs on biological invasions.At last, the authors want to express their thanks for the translation of the abstract to other European languages, namely to Paride Balzani, Antonin Kouba, Sandra Hodic, and ROS Educational Consultancy Ltd & Garnock Media Ltd.

Figure 2 .
Figure 2. Distribution of IAS costs in Europe by a type of cost b cost type (left half ) and impacted sector (right half ) and c impacted sector.Panel b highlights linkages between cost types and impacted sectors, for example 5% (US$2.76/50.97 billion) of total costs were attributed to management, and 64% (US$1.76/2.76 billion) of these costs were incurred in the Authorities and Stakeholders sector, representing 81% (US$1.76/2.17 billion) of costs incurred by the Authorities and Stakeholders sector.Only reliable observed costs are considered (i.e.excluding irreproducible cost estimations and expected costs).

Figure 4 .Figure 5 .
Figure 4. Maps showing for each European country where data were available: a total reliable observed costs of IAS for the period 1960-2020 in million US$ (i.e.excluding irreproducible cost estimations and expected costs) b total reliable observed costs of IAS standardised by GDP (US$), and c scatter plot of total cost of IAS against GDP.Data are from a-c InvaCost (Ballesteros-Mejia et al. 2020; Diagne et al. 2020a; Angulo et al. 2021b) b, c WorldBank (2020).Countries in white located in Europe did not have reported costs in the InvaCost database, or in the case of Russia and Turkey were excluded from this analysis due to their transcontinental nature.

Figure 7 .
Figure 7. Temporal trend of costs considering the GDP-standardized average decadal costs (black bars) and total annual GDP-standardized invasion costs (triangles until 2013, circles after) recorded in Europe (on a log scale).Robust regression analysis between 1970 (the first year of documented reliable observed costs) and 2019 (last year with available GDP data) is overlaid, showing linear regression in orange and quadratic regression in blue.Error bands on robust regressions represent 95% confidence intervals.Model coefficients are presented in Suppl.material 7.
highlights linkages between cost types and impacted sectors, for example 5% (US$2.76/50.97 billion) of total costs were attributed to management, and 64% (US$1.76/2.76 billion) of these costs were incurred in the Authorities and Stakeholders sector, representing 81% (US$1.76/2.17 billion) of costs incurred by the Authorities and Stakeholders sector.Only reliable observed costs are considered (i.e.excluding irreproducible cost estimations and expected costs).Percentage contributions of different impacted sectors and cost types according to country.Only reliable observed costs are considered (i.e.excluding irreproducible cost estimates and expected costs).

Table 1 .
Top 10 cost-contributing genera considering (a) total and (b) reliable observed costs (i.e.excluding irreproducible cost estimations and expected costs), illustrating species taxonomy, total costs and numbers of database entries.Numbers of impacted countries per genus are also shown.Note that costs and entries are pooled across the entire genus (i.e. for all species), with constituent species listed therein.

Table 2 .
Relationships of cost of IAS in European countries with country-specific factors.Two types of costs are included: cost of "Damage-Loss" and cost of "Management".Country-specific factors are presented in Suppl.material 3. Statistics shown are Spearman correlation coefficients (p-values associated).Bold numbers indicate significance at the 0.05 level.