Global economic costs of aquatic invasive alien species

duceknowledgegaps.Few costsareproactiveinvestments; increased managementspending isurgently needed to prevent and limit current and future aquatic IAS damages. © 2021 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license


H I G H L I G H T S
• Aquatic invasions have cost the global economy US$345 billion. • Most costs are caused by invertebrates, in North America and damages to resources. • Costs have increased exponentially over time, to at least US$23 billion in 2020. • Aquatic invasion costs are underrepresented compared to terrestrial invasion costs. • Taxonomic, geographic and temporal gaps make these costs severely underestimated.

G R A P H I C A L A B S T R A C T
Science of the Total Environment 775 (2021) 145238

Introduction
The impacts of invasive alien species (IAS) on biodiversity (Mollot et al., 2017;Spatz et al., 2017;Shabani et al., 2020), ecosystem services (Vanbergen, 2013;Blackburn et al., 2019) and human wellbeing (Pejchar and Mooney, 2009) are well recognized (Pyšek et al., 2020). Accordingly, there are numerous national and international policies, regulations and mandates in place to prevent new introductions and limit the geographic spread of IAS [e.g. Convention on Biological Diversity (UNEP, 2011); European Union Regulation 1143/2014 on IAS]. However, records of IAS are continuously increasing, owing to factors such as habitat disturbance, climate change, and an increasing diversity, frequency and intensity of anthropogenic vectors associated with globalising trade and transport networks (Capinha et al., 2015;Seebens et al., 2017Seebens et al., , 2018Turbelin et al., 2017;McGeoch and Jetz, 2019). Alien species numbers are burgeoning across geographical regions and habitat types, with the number of established alien species expected to increase by 36% in the next three decades (Seebens et al., 2020).
Aquatic ecosystems can be severely threatened by IAS, which contribute to extinctions of individual species, substantially change the structure of native communities, and alter ecosystem functioning (Vitousek et al., 1997;Ricciardi and MacIsaac, 2011;Jackson et al., 2017). Aquatic ecosystems provide numerous services to people, from food provision to flood protection and recreation; these services can also be critically altered by the presence of IAS (e.g. Katsanevakis et al., 2014). The vulnerability of aquatic ecosystems to invasions is increased by high interconnectedness among habitats, specifically man-made waterways and shipping, as well as other anthropogenic pressures (Strayer and Findlay, 2010;Poulin et al., 2011;Darwall et al., 2018) and climate shifts (Woodward et al., 2010).
In recent years there have been significant advances across habitat types in understanding ecological impacts of IAS (Kumschick et al., 2015;Dick et al., 2017;but see Crystal-Ornelas and Lockwood, 2020) and the drivers of invasion success (Cuthbert et al., 2019(Cuthbert et al., , 2020Fournier et al., 2019;van Kleunen et al., 2020), as well as methodological advances in assessing the economic dimensions of IAS and their management (Lovell et al., 2006;Hanley and Roberts, 2019). However, studies of economic aspects of IAS have been limited to certain taxonomic groups (Bradshaw et al., 2016), communities, or regions (Pimentel et al., 2000;2005;Kettunen et al., 2009;Cuthbert et al., 2021;Haubrock et al., 2021). In particular, costs of aquatic IAS are generally less well understood than costs of terrestrial IAS, despite some estimates indicating high costs (Lovell et al., 2006;Aldridge and Oreska, 2011). Comprehensive and systematically-assembled data on the costs of aquatic IAS would greatly help planning and prioritisation for their management, in the context of limited resources (McGeoch et al., 2015). Such data would also provide a useful resource for communications with policymakers and the general public: impacts expressed in economic terms are more tangible and comprehensible than complex ecological impacts (Diagne et al., 2020a). This paper is the first systematic effort to describe global patterns and trends in reported costs of aquatic IAS. Our analysis, based on the recently developed InvaCost database (Diagne et al., 2020b), allows us to synthesise standardised costs, identify knowledge gaps and provide recommendations for management and further research. We describe the global monetary costs associated with aquatic IAS based on taxonomic, geographic and temporal descriptors, as well as between fully aquatic and semi-aquatic taxa. In doing so, we examine (i) how costs are structured by implementation method (i.e. observed vs. potential/expected), (ii) reliabilities of cost estimates and (iii) their typology, i.e., whether costs result from damages and losses or management expenditure. Further, we model the yearly and cumulative dynamics of costs and investigate whether they are likely to saturate in the near future. Finally, we assess potential biases between aquatic and terrestrial cost reporting. These biases are then used to identify gaps in management spending between habitats.

Original data
For the purpose of quantifying global costs of aquatic IAS, we used the most comprehensive and up-to-date dataset of costs caused by alien species globally, assembled by the InvaCost project (Diagne et al., 2020a(Diagne et al., , 2020b. At time of writing, this includes 9823 entries in various languages from systematic and opportunistic literature searches (Diagne et al., 2020b;Angulo et al., 2021; full database version 3 at https://doi. org/10.6084/m9.figshare.12668570). This database captures any reported economic costs associated with IAS in their novel range, including those for species that have already become invasive (e.g. management, damages and losses) and species that may become invasive in the future (e.g. prevention and rapid eradication). The InvaCost version 3 database contains a column ("Environment_ IAS") which classifies species as either aquatic (species with a close association with aquatic systems at any life stage, including for reproduction, development and/or foraging; n = 2317 cost entries after our below corrections) or terrestrial (n = 6433 cost entries after our below corrections), independently of where costs occurred. For some analyses, we split out costs for semi-aquatic species: the subset of aquatic species with a looser association with aquatic systems (see Supplementary Material 1). Remaining entries, linked to species from diverse habitats (i.e. a mixture of aquatic and terrestrial) or unspecified habitats, were excluded from analyses. We also carefully screened the published database, removing clear duplicates and correcting clear mistakes. All modifications made in our dataset were sent to updates@invacost.fr as recommended by the database managers.
Briefly, costs in InvaCost are standardised against a single currency for comparability (2017 US$); costs in the database may be 'expanded' so that entries can be considered on an annual basis. That means that costs spanning multiple years (e.g. $10 million between 2001 and 2010) are divided according to their duration (e.g. $1 million for each year between 2001 and 2010); we considered this expanded database version in all analyses (Supplementary Materials 1; n = 5682 aquatic entries). Expansion was done using the expandYearlyCosts function of the 'invacost' R package (R Core Team, 2020; Leroy et al., 2020). The final, unexpanded dataset used in our analyses is provided as Supplementary Material 2. We note that 1 billion = 1 × 10 9 .

Cost descriptors
To obtain a general overview of the costs associated with IAS, we first illustrated them across a number of key database descriptors (see Supplementary Material 1 and https://doi.org/10.6084/m9. figshare.12668570 for complete details). These included (1) broad taxonomic grouping of species presenting costs (invertebrates, vertebrates, plants, other), (2) perceived reliability of each cost entry ("High" vs. "Low"), (3) cost implementation type ("Observed" vs. "Potential"), (4) geographic region in which the cost occurred (within continent-and country-scales) and (5) cost type ("Damage" vs. "Management"). We summed the expanded entries (see above) to quantify cost totals among these descriptors.

Spatial and taxonomic connectivity
We investigated spatial and taxonomic patterns in costs of aquatic IAS with a network analysis (see Supplementary Material 1). Here, we created a bipartite network composed of two types of nodes: countries and IAS. When a species had a reported economic impact in a country, a link was drawn between the two nodes. The weight of the link was equal to the cumulative cost, since 1960. We defined the size of nodes on the network as proportional to their total costs with a log spline, such that country or species nodes with higher costs are easier to distinguish from those with lower economic impacts.

Prediction of annual costs for aquatic IAS
To examine the most appropriate temporal relationship for the accumulation of costs over time, we used the modelCosts function of the 'invacost' package (Leroy et al., 2020). We fitted multiple models to the data and identified the best model(s) by quantitative and qualitative criteria (see Supplementary Material 1). As we were dealing with econometric data, we selected models that were robust to issues of heteroskedasticity, temporal autocorrelation and outliers. We examined the long-term trend of annual costs worldwide between 1960 and 2020, i.e., we predicted costs as a function of years. First, owing to time lags in cost reporting, we corrected the data by removing the most recent, thus incomplete years; not making this correction would result in an inherent underestimation of costs (Supplementary Material 1). Second, we employed and compared a range of statistical techniques on the resulting data: ordinary least squares regression (linear and quadratic), robust regression (linear and quadratic), multivariate adaptive regression splines, generalised additive models (GAMs) and quantile regression [0.1 (lower boundary of cost), 0.5 (median cost value), 0.9 (upper boundary of cost)].

Trend in cumulated costs for aquatic IAS
In addition to modelling annual costs, we mathematically described temporal changes in cumulated costs of aquatic IAS. We chose to rely on a variation of the functional form proposed by Yokomizo et al. (2009) for density-impact curves, where we considered the cumulative cost C in terms of population density u. By assuming logistic growth in the population, C can then be expressed as a function of time and therefore serves as a model for the cumulative temporal cost of impacts (Supplementary Material 1). We used a non-linear regression curve-fitting tool to estimate the best fit parameters, such as cost saturation C max , carrying capacity K and intrinsic growth rate α. We quantified the fit by computing the squared correlation coefficient (r 2 ) and root mean square error (RMSE).

Reporting of invasion costs from aquatic IAS compared to terrestrial IAS
We obtained known numbers of established alien species (n = 13,867) in aquatic and terrestrial habitats globally, using databases such as the inventory of IAS in Europe (DAISIE; see Supplementary Material 1 for full list of sources). Categorising entries originating from either aquatic or terrestrial species, we then counted for the two habitats in InvaCost: the numbers of species having costs (excluding unspecific entries), the number of documents reporting these costs, total costs, and costs only reporting management actions (not reporting damage). Then, we compared these numbers to the proportions of known established IAS between habitats. Finally, we predicted the expected costs of management actions for aquatic IAS, under the hypothesis of an unbiased expenditure between aquatic and terrestrial habitats (based on the known proportion of global aliens that are aquatic).

Global costs and taxonomic groupings
Global costs of aquatic IAS summed to US$345 billion, based on 5682 records from the expanded InvaCost database. These were all published since 1971. Semi-aquatic species cost US$185 billion (n = 2971 records) and fully aquatic species US$149 billion (n = 2518 records), with diverse costs (that spanned semi-aquatic and fully aquatic species) comprising the remaining US$11 billion (n = 193 records). Only 1% of the cost was from fully marine species (US$3.6 billion; n = 234 records).
Most (65%, US$224 billion) of the costs were derived from empirical observations, rather than predictions (Fig. 1b). The majority of costs for aquatic invertebrates were derived from empirical observations (92%). However, just 17% of the costs for aquatic vertebrates and 42% of plant costs, were based on empirical observations. The 10 aquatic genera with the highest documented costs accounted for US$304 billion (88%) of total costs (Fig. 2). These taxa included four invertebrates, three vertebrates and three plants. Mosquitoes belonging to three species of the Aedes genus caused 50% (US$153 billion) of the total top 10 cost. These were followed by ruffes Gymnocephalus cernua (18%, US$53 billion), mussels Dreissena spp. (two species, 16%, US$50 billion), coypus Myocastor coypus (6%, US$19 billion) and primroses Ludwigia spp. (three species, 3%, US$8 billion). Contributions from the remaining genera were relatively small. For all genera, excepting Lithobates, damages outweighed reported management spending (Fig. 2).

Geographic regions
Reported economic costs of aquatic IAS were unevenly distributed across geographic regions (Fig. 3). North America, owing to costs primarily from the United States of America (USA), reported the highest costs (48%, US$166 billion), followed by costs that were not attributed to specific regions (26%, US$91 billion) and costs from Asia (13%, US $45 billion). The costs in Europe and South America accounted collectively for 12% (US$41 billion) of total reported costs, whilst Africa, Oceania-Pacific Islands and the Antarctic-Subantarctic, combined accounted for 0.6% (US$2.1 billion) (Fig. 3). Regarding cost types, 74% (US$256 billion) of global costs were driven by damages, whereas only 6% (US$21 billion) consisted of management-related expenditure (Fig. 3b). Mixed spending (i.e. combined records of damage and management-related spending) comprised 20% of global costs (US$68 billion). Further information on taxonomic and cost typology breakdowns per region is provided in Supplementary Material 1.
At the country level, the USA had the highest recorded cost for aquatic IAS, followed by Brazil, India and France (Fig. 4a); other  Total monetary costs of the top 10 costly aquatic invasive alien genera, alongside species-specific information of underlying data pertaining to each genus. Unspecified species within each genus were also included. Fills illustrate cost type contributions per genus. Note that "Management" corresponds to expenditure related to activities such as prevention, control, eradication and research, whilst "Mixed" is a mixture of cost types.
R.N. Cuthbert, Z. Pattison, N.G. Taylor et al. Science of the Total Environment 775 (2021) 145238 countries were relatively similar in costs. The USA also had the largest number of studies. Other countries that reported costs generally had similar numbers of studies, with numbers from, for example, Spain, Brazil and Australia relatively high (Fig. 4b). We found no reported costs for aquatic IAS from the majority of African and Asian countries.

Spatial and taxonomic connectivity
We found eight clusters of IAS costs that were composed of at least five nodes (coloured clusters in Fig. 5), and eleven minor clusters that were composed of only two nodes (grey nodes in Fig. 5). We found two types of clusters. First, most clusters were composed of one or a few countries and a unique combination of IAS. This was the case, for example, for countries with the highest costs (mainly European and North American countries), which often had clusters of their own. Among these unique country clusters, the USA example was pervasive, with highest reported costs for Dreissena spp., G. cernua and Melaleuca quinquenervia, alongside many other IAS that were unique to that country. Second, there was one cluster that was driven by one genus, Aedes, which had pantropical economic impacts as well as impacts in temperate countries. For most of the Southern Hemisphere countries, Aedes was the only genus for which costs were reported. Despite these specific clusters of costly IAS per country, several IAS taxa had widespread economic impacts on multiple countries, such as Lithobates catesbeianus, M. coypus, Neovison vison, Dreissena spp., Hydrocotyle ranunculoides and Eichhornia crassipes. Interestingly, we found no strong biogeographical structure in the network. Australia, for example, shared costly IAS with geographically disparate regions such as European countries, South Africa, Argentina and Chile.

Prediction of annual costs for aquatic IAS
The linear models projected the highest costs of aquatic IAS in the year 2020 (since 1960; data from years 2013 to 2020 were removed owing to <75% completeness), however they had a relatively poor fit, with high RMSE (Fig. 6; Supplementary Material 1). The quadratic robust regression was removed owing to cost reductions in recent years, but it also had the highest RMSE (0.63). The GAM approach thus provided the best fit to the data (ΔRMSE ≥0.08; Fig. 6). This model indicated a rapid increase in costs by three orders of magnitude between 1970 and 2000, followed by a relatively gradual increase within a further magnitude since 2000 (Fig. 6c). Overall, the best-fitting GAM predicted a cost of aquatic IAS of US$23 billion globally in the year 2020.

Trend in cumulated costs for aquatic IAS
We found that the linear curve and high threshold curve models performed well, with the former providing a slightly better fit (Table S1; Supplementary Material 1). In the long term, the cumulative cost saturates to a fixed value C max (i.e. maximum cumulative cost of impact), where the invasion is completely controlled and no further impact costs are incurred (Fig. 7). A clear saturation in costs (i.e. carrying capacity) was not reached for either the full or adjusted dataset (with outliers removed), indicating that costs will continue to increase in the near future. The reduction in rate of cost increases over recent years was likely an artefact of time lags in cost reporting versus occurrence.

Reporting of invasion costs of aquatic IAS compared to terrestrial IAS
Of 13,867 known established alien species worldwide (see Supplementary Material 1), 26% are associated with aquatic habitats, compared    Fig. 5. Global network of aquatic invasive alien species costs per country. This bipartite network is composed of both species and country nodes. Links indicate the cumulative costs of species in countries. The thicker the link, the higher the cost. Likewise, node size is proportional to the total cumulative cost, with a log spline. 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. to 74% associated with terrestrial habitats (Fig. 8). Although in InvaCost the number of aquatic species and the number of documents reporting their costs constituted relatively similar percentages (20% and 28%, respectively), the value of their reported cost comprised just 5% of the global total. This increased to just 9% when considering only costs reported from management strategies. If management expenditure was unbiased between habitat types according to numbers of known established aliens, we estimated that a further US$39 billion should have been spent on aquatic species to date.

Discussion
Our study reveals that aquatic IAS have likely cost the global economy at least US$345 billion. This estimate is probably highly conservative as it only includes costs that have been documented and captured in the InvaCost database. Moreover, the taxonomic, geographic, temporal and habitat trends among these costs suggest that cost reporting is very uneven, with many IAS and countries entirely lacking reported costs. Most costs were attributed to aquatic invertebrates (US$214 billion), with lower costs for vertebrates (US$97 billion) and plants (US$20 billion). Our estimate of the costs of aquatic IAS globally in the year 2020 -US $23 billion, much higher in magnitude than the cost, for example, of managing global marine protected areas (US$5-19 billion; Balmford et al., 2004) calls for increased investments in management of IAS.

Cost distributions across taxa
Globally, mosquitoes are major contributors to the burden of diseases, with vector-borne pathogens and parasites causing over one billion infections and one million deaths annually (Kilpatrick and Randolph, 2012;Campbell-Lendrum et al., 2015). The massive costs attributed to vector-borne diseases from invasive mosquitoes are thus not surprising, given the costs to healthcare systems worldwide. In Brazil, for example, the government invested approximately US$48 million per year from 2015 to 2017 for limiting population outbreaks of A. aegypti (Bueno et al., 2017). In Columbia, total medical costs for the treatment of dengue-infected patients reached US$3 billion between 2010 and 2012 , and the recent chikungunya outbreak cost about US$76 million to the healthcare system (Cardano et al., 2015). Mosquitoes can also lead to economic losses associated with recreation and tourism, as they discourage people from carrying out certain activities or visiting certain sites (Claeys-Mekdade and Morales, 2002). In the present study, damages to sectors such as health comprised 73% of mosquito costs, with just 4% spent on management. Future range expansions of invasive mosquitoes are expected to increase their economic impact (Iwamura et al., 2020). Although mosquitoes vector diseases in their terrestrial-based adult life stage, where most costs are incurred, larval and pupal life stages are invariably spent in water where management is often targeted, with the characteristics  and distribution of aquatic habitat patches determining mosquito distributions at various scales via key trait-and density-mediated processes (Pintar et al., 2018;Cuthbert et al., 2019).
The Eurasian ruffe (G. cernua) was second most costly and has caused declines of native fish by predation and competition, with considerable economic impacts through reductions in commercially-and recreationally-valuable fish species (Leigh, 1998). In turn, the zebra and quagga mussels (Dreissena polymorpha and Dreissena bugensis) are hyper-successful macrofouling freshwater bivalves, which are highly costly to infrastructure through impeding navigation structures, obstruction of water flow in pipes and occlusion of water filters (Sousa et al., 2014). The coypu (M. coypus) has caused substantial economic losses through agricultural impacts, as well as infrastructural damage (Panzacchi et al., 2007). The primroses (Ludwigia spp.) are known to reduce water quality that can affect economically important taxa such as fish, and can be extremely costly to control (Williams et al., 2010).
Costs attributed to invasive aquatic invertebrates such as the zebra and quagga mussels were deemed highly reliable and mostly based on empirical observations rather than extrapolations. In contrast, a large share of vertebrate costs was potential costs, as in the case of the three most costly vertebrate taxa, Eurasian ruffe G. cernua, coypu M. coypus and American bullfrog L. catesbeianus. Therefore, realised vertebrate costs require improved validation and reporting to the extent possible from their actual invaded habitat. Similarly, reported costs of plants, including the highly damaging Ludwigia species and broadleaved paper bark M. quinquenervia, were primarily potential costs, not incurred at the time of estimation. Although ecological impacts of aquatic plants have been well-studied by invasion scientists (Pyšek et al., 2008;Gallardo et al., 2016), there is scope for more thorough recording of realised economic impacts.

Cost distributions across geographic regions and types
The costs of aquatic IAS were also unevenly distributed across geographic regions, with particularly high reported costs in North America (US$166 billion) and Asia (US$45 billion). In turn, a substantial proportion (26%) of the costs were unattributed to specific geographic regions. Moreover, most costs were driven by damages (74%), whilst management (principally control-related) costs were just 6%. It may be expected that management costs are lower than damage or loss costs: if the inverse were true, management would not be economically justifiable. However, the InvaCost search strategy may have exacerbated this difference. Reports of management costs may have been disproportionately missed by the systematic literature searches because management studies often do not mention costs, economics or other InvaCost search terms (Diagne et al., 2020b) in their title, abstract or keywords (e.g. Sandodden and Johnsen, 2010).
At the country scale, the USA exhibited both the highest magnitude of costs and the greatest number of studies compared to all other countries. The high degree of cost reporting in the USA is unsurprising given that early estimates of costs focused on this country (Pimentel et al., 2000, Pimentel et al., 2005, which sparked research efforts to better understand costs and provide a more refined spatial and temporal description for those costs. The USA also scores highly on several socioeconomic variables that have been found to correlate positively with reported costs of IAS Kourantidou et al., 2021), such as GDP (1st in world), human population (3rd), international tourism arrivals (3rd) and research expenditure (9th).
In contrast, the InvaCost database contains no aquatic IAS costs at all for many countries, particularly in Asia and Africa. However, even in countries such as South Africa, where research on biological invasion is leading , large gaps in our knowledge of economic costs are evident. For example, South Africa is a global invasion hotspot for freshwater fish and has been invaded by numerous invertebrate taxa in freshwater, estuarine and marine environments, with well-known impacts on human wellbeing (Appleton et al., 2009;Ellender and Weyl, 2014;Weyl et al., 2020;Robinson et al., 2020). However, we captured no monetary costs for such taxa. Similarly, in other African countries, IAS without formally documented or quantified costs are known to affect human societies via impacts to biological communities, local fisheries and water storage infrastructures (e.g. Nile perch in East Africa; Harris et al., 1995;Kwena et al., 2012;Aloo et al., 2017, and crayfish in Lake Naivasha, Kenya; Kafue River, Zambia; Madzivanzira et al., 2020). Limited cost reporting in Africa and Asia is likely reflective of a low priority given to IAS research, or capabilities (Pyšek et al., 2008), despite high levels of introduction via, for example, aquaculture (Lin et al., 2015). However, there may have been some bias introduced by the original InvaCost search string, as no currencies from these continents were explicitly included as search terms (even if searches were performed in 15 non-English languages, Angulo et al., 2021).
Nonetheless, limited cost reporting in Africa and Asia is alarming given that invasions in these countries may disproportionately impact livelihoods, given high levels of poverty, limited resources for research and management, and an overall limited preparedness to meet challenges brought by IAS (Early et al., 2016). Limited cost reporting also hinders management actions, as the extent of IAS cost is not fully realised by managers.
Network analyses additionally revealed a distinct lack of global structuring of costs, whereby clustering appeared disparate across taxa and countries, insinuating a largely random distribution of costs and vast gaps in cost reporting of well-known aquatic IAS. That is, for many countries, there was generally only one cluster, indicating unique combinations of economic impacts associated with particular species, despite some of these species being highly widespread. One example of an exception to this is Aedes spp., which had a consistent and pantropical impact, resulting in a distinct cost cluster. Nonetheless, other context-dependencies, such as differences in climate and pathways, likely also influence IAS compositions.

Temporal trends in costs
The majority of fitted models indicated exponentially increasing annual costs of aquatic IAS since 1960 over several magnitudes, to a best-fit extrapolated annual global cost of US$23 billion in 2020. Model differences in recent years likely reflect differential sensitivities to time lags in cost reporting. Model predictions of cost increases over time align with increasing rates of biological invasions worldwide (Seebens et al., 2017), as globalisation and intensification of trade and transport networks result in high propagule and colonisation pressures from novel source pools (Seebens et al., 2018). Given that invasion rates will increase further in future (Seebens et al., 2020), we can expect further increases in economic costsalthough investments in management, especially prevention and rapid eradication, could limit realised costs (Leung et al., 2002). Moreover, these results align with the findings of Bradshaw et al. (2016) who have suggested, specifically for invasive insects such as mosquitoes, that costs are generally largely underestimated and are expected to increase through time. Our mathematically-modelled density-impact curves also suggest that costs of IAS to the global economy will continue to increase, as they were far from an asymptotic plateau, even where extreme values were removed and time lags not incorporated. Moreover, this population-level approach does not account for unreported costs or those arising from future IAS spread, and this likely results in further underestimation.

Reporting of invasion costs of aquatic IAS compared to terrestrial IAS
Despite over one quarter of known alien species using aquatic environments, only 5% of the total cost in the InvaCost database was attributed to aquatic species. Further, the majority (54%) of these costs were from semi-aquatic rather than fully aquatic species. On one hand, this finding potentially reflects a bias in cost reporting towards terrestrial systems, in line with ecological research in general (Menge et al., 2009;Richardson and Poloczanska, 2008). With respect to management costs of IAS, if investments of equivalent magnitude to terrestrial were made for aquatic systems, one would anticipate a further US $39 billion to have been spent to date. Note that this extrapolation does not consider potentially lower costs in aquatic ecosystems (i.e. less infrastructure to damage) or differences in management efficiencies between terrestrial and aquatic environments. On the other hand, the disparity between aquatic and terrestrial costs may thus reflect genuinely lower costs of aquaticparticularly marine -IAS relative to terrestrial IAS. There are limited human assets and infrastructures in aquatic systems, limiting the scope for easily-quantifiable damages and resulting in minimal investments in prevention and management. For example, terrestrial agricultural practices are heavily impacted by crop pests (Paini et al., 2016;Ahmed and Petrovskii, 2019), whereas agricultural activities in aquatic systems (e.g. rice fields) are relatively scarce. However, aquatic systems do offer highly valuable ecosystem services that could be affected by IAS, such as aquaculture, and often through cascading effects that are difficult to predict (Walsh et al., 2016). Thus, we encourage investment in management of IAS in aquatic systems to limit future costs that stem from damage and loss (Leung et al., 2002).

Conclusions
Urgent and coordinated management actions are required globally to reduce economic and ecological impacts from aquatic IAS. Whilst costs of aquatic IAS are escalating, knowledge of impacts across major taxonomic groupings, geographic regions and habitat types remains diffuse. These knowledge gaps suggest costs of aquatic IAS are underestimated, particularly relative to their ecological impacts and to the more intensivelystudied terrestrial species. Equally, geographical biases in reported costs highlight the need for increased and improved cost reporting, given that allocation of finite resources to manage IAS is underpinned by adequate understandings of costs. We urge our results to be applied as an incentive for managers, stakeholders and scientists to increase and improve cost reporting and invest in a more adequate protection of aquatic ecosystems.
CRediT authorship contribution statement