Massive economic costs of invasive bivalves in freshwater ecosystems

Phillip J. Haubrock (  Phillip.Haubrock@Senckenberg.de ) Senckenberg Research Institute and Natural History Museum Frankfurt, Department of River Ecology and Conservation, Gelnhausen, Germany. https://orcid.org/0000-0003-2154-4341 Ross N. Cuthberg GEOMAR Helmholtz‐Zentrum für Ozeanforschung Kiel, 24105 Kiel, Germany Anthony Ricciardi Redpath Museum and McGill School of Environment, McGill University, Montreal, Canada Christophe Diagne Université Paris-Saclay, CNRS, AgroParisTech, Ecologie Systématique Evolution, 91405, Orsay, France Franck Courchamp Université Paris-Saclay, CNRS, AgroParisTech, Ecologie Systématique Evolution, 91405, Orsay, France


Introduction
Freshwater ecosystems have been identi ed as among the most threatened worldwide, owing to their sensitivity to the effects of climate change (Woodward et al. 2010) and a range of other anthropogenic pressures (Darwell et al. 2018), including invasive species (Strayer 2010;Poulin et al. 2011). Globally, invasive non-native species are a major driver of erosion of native biodiversity and the disruption of ecosystem functioning (Malcolm and Markham 2000;Stigall 2010;Blackburn et al. 2019). Furthermore, they are a burgeoning economic stressor on virtually all resource sectors -especially those associated with inland waters, where they are several times more likely than natives to become socioeconomic pests (Hassan and Ricciardi, 2014). Indeed, invasion rates worldwide have been steadily increasing with no sign of saturation (Seebens et al. 2017), owing to increasing globalization, intensi cation of global transport networks and accessibility of new non-native source pools (Seebens et al. 2018). At present, most countries have limited capacity to manage invasions (Early et al. 2016) and are increasingly forced to make decisions regarding investment in biosecurity versus other societal needs.
In recent years, the ecological impacts of invasive species on recipient ecosystems have been better described (e.g. Kumschick et al. 2014;Dick et al. 2017;Crystal-Ornelas et al. 2020). However, whilst categorizations for invader socioeconomic impacts have been designed , there remains a paucity of quanti ed socioeconomic costs incurred by invasions, constraining effective cost/bene t analysis and rationale for policy makers to invest the sparse available resources toward prevention (but see Lovell et al. 2006;Marbuah et al. 2014). This constraint exists even though preventive measures are generally considered more cost effective than long-term mitigation and control (Keller et al. 2008;Ahmed et al., this issue), and management is less costly than losses from damages (Leung et al. 2002). Pimentel et al. (2000Pimentel et al. ( , 2005, and later Kettunen et al. (2009), were among the rst to attempt to summarize the costs of invasive species on large scales. Despite methodological shortcomings, these pioneer studies had the bene t of raising awareness of the potentially huge costs associated with non-native species (Hoffmann and Broadhurst 2016). Their shortcomings originate from the problem that some categories of costs are di cult to quantify, especially regarding damages to ecosystem services or other indirect effects (Charles and Dukes 2008) and that they are often not comparable and thus summable. The lack of such synthesis, however, is critical because it can give the false impression that costs for invasive species are lower than empirically observed. In turn, this can result in an under-allocation of economic resources to tackle invasive species. Regional or even global estimations of the cost of invasions rely on the resolution of cost estimation at smaller spatial scales and at various taxonomic levels. In particular, it is important to document the economic costs of taxonomic groups known to include damaging invasive species, as it could help to inform decision-making at the national level and thus provide appropriate economic incentives for controlling the arrival and spread of such species.
A group of aquatic invasive species that has caused signi cant ecological and socioeconomic impacts are freshwater bivalves (Sousa et al. 2009;, including, inter alia, several hyper-successful invasive species from the genera Dreissena, Limnoperna, and Corbicula (Boltovskoy et al. 2006;Karatayev et al. 2007;Sousa et al. 2008). These taxa have caused a broad range of impacts (e.g. macrofouling, habitat modi cation, restructuring communities and food webs, nutrient mineralization, contaminant transfer, alteration of oxygen availability and sedimentation rates, and promotion of excessive macrophyte and algal growth; see reviews by Karatayev et al. 1997;Boltovskoy et al. 2006;Ward and Ricciardi 2007). As a result, they affect various sectors of society (e.g. infrastructure, municipal and industrial water supply systems, and sheries; Minchin et al. 2002;Water eld 2009). Arguably, the enormous costs associated with invasions of invasive bivalves such as the Asian clam Corbicula uminea and the zebra mussel Dreissena polymorpha have done more to raise public awareness of aquatic invasions than their respective ecological impacts, although the economic and ecological impacts are often linked (e.g. Kao et al. 2015). On the other hand, invasive freshwater bivalves have, on occasion, been associated with certain perceived bene cial effects for human activities, as with other invaders (Kourantidou et al., this issue). For instance, their ltration capacity can substantially increase water clarity (Phelps 1994;Higgins and Vander Zanden 2010;Boltovskoy et al. 2009), which may bene t certain recreational activities (e.g. scuba diving and angling), while at the same time causing food web disruptions that harm sheries (Kao et al. 2015).
Despite the notoriety of invasive freshwater bivalves in invasion science, information on economic costs for invasive bivalves in freshwater ecosystems is often scanty or anecdotal, which challenges efforts to prioritize management action. To broadly address this pervasive lack of information and provide a basis for quanti cations of costs associated with most invasive species worldwide, the InvaCost database has recently been developed (Diagne et al. 2020). This database contains extensive information on the costs (e.g. cost types, impacted sectors, regional attributes, cost estimation reliability, etc.) associated with ~ 500 invasive species. In the present study, we use a subset of the database to describe global costs associated with invasive freshwater bivalves, anticipating unevenness in cost reporting towards a few regions and a few highly conspicuous invasive species. Moreover, we investigate how these costs are structured, and identify knowledge gaps in cost estimations for key invasive freshwater bivalves.

Original data
To estimate the cost of bivalve invasions of fresh waters on the global economy, we considered cost data from version 3 of the InvaCost database (full database and descriptive les are available at https://doi.org/10.6084/m9. gshare.12668570). This database (9,823 entries; Diagne et al. 2020;Angulo et al. 2021) compiles entries that extensively describe documented costs globally, enabling large-scale cost synthesis associated with invasive species in different spatial and temporal frames. We note that this database only reports monetary values for invasion costs, without considering monetised bene ts of invasive species quantitatively. Therefore, the analyses which follow re ect that scope and only consider costs. Grey and published references were retrieved from standardised searches in online repositories (ISI Web of Science, Google Scholar and Google search engine) and opportunistic collection based on targeted searches. Full information on the search terms (see Supplementary material 1) is provided in Diagne et al. (2020) and Angulo et al. (2021). Gathered references were thoroughly examined to assess relevance, and then scrutinized for collating cost estimates associated with invasive species. Every cost entry was recorded, depicted by 64 parameters, and nally converted to a common and up-to-date currency (US dollars (US$) 2017; see Diagne et al. 2020, for detailed information; Supplementary Material 2). From this full database, 231 cost data entries were identi ed as exclusively belonging to the Bivalvia class using the 'Class' column lter and 226 cost data entries belonging to bivalves which impact freshwaters (see Figure 1). We therefore excluded fully marine species, but focused on various taxa such as D. polymorpha and Mytilopsis spp. that occur in both brackish and freshwater ecosystems (e.g. Leppäkoski et al., 2002).

Estimating the total costs
Deriving the total cumulative cost of invasions over time requires consideration of the probable duration time of each cost occurrence. This duration consisted of the number of years between those mentioned in the 'Probable_starting_year_adjusted' and the 'Probable_ending_year_adjusted' columns. When information was missing for the 'Probable_starting_year_adjusted' column, we conservatively considered the publication year of the original reference. For the 'Probable_ ending_year_adjusted' column, information was missing only for potentially ongoing costs ('Occurrence' column), which are costs likely to be repeated over years (contrary to one-time costs occurring only once along a precise period). We used this temporal information to annualize the invasion cost entries (4th step in Figure 1). This was done by 'expanding' the database via the expandYearlyCosts function of the 'invacost' R package (Leroy et al. 2020) -a process that causes each entry in the database to correspond to a single year, thereby increasing the number of entries beyond that of the original data. For example, an initial single cost between 2000 and 2009 that totalled at US$ 10,000 would become ten entries at US$ 1,000 each after the expansion. All analyses were performed using this version of the database. A full explanation of this and other functions used is available in Leroy et al. (2020). For one cost entry, the probable ending year was presumably after 2020. Hence, all resulting cost estimates projected beyond 2020 were not taken into account. Similarly, costs were not available before 1980. This resulted in a subset of 468 expanded database entries ( Figure 1). The dataset was then reduced to 443 entries by removing entries before 1980 to ensure comparability of currency translations ("recent" in Figure 1) and is provided in Supplementary Material 3.
Finally, the invasion costs were speci cally estimated by summing all entries according to different descriptive columns of the database (see Supplementary Material 2): (i) 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 o cial reports, or with documented, repeatable and/or traceable methods were designated as High reliability; all other estimates were designated as Low reliability (Diagne et al. 2020b). We acknowledge that this approach, which categorises costs as High reliability based on their presence in peer-reviewed material, may not be fully representative of the diverse forms of method reliability of cost estimates. Nevertheless, these criteria provided clear, objective and reproducible means of assessing material, as it was not feasible to assess method reliability on a broader categorical scale; (ii) Implementation: referring to whether the cost estimate was actually realised in the invaded habitat (observed) or whether it was extrapolated (potential), based on the methods reported in the underlying study (i.e., we did not perform extrapolations ourselves); (iii) Geographic_region: describing the geographic origin of the listed cost; (iv) Type_of_cost_merged: grouping of costs according to the categories: (a) Damage costs referring to damages or losses incurred from invasion (e.g. costs for damage repair, resource losses, medical care), (b) Management costs comprising control-related expenditure (for example monitoring, prevention, management, eradication) and money spent on education, research and maintenance costs, (c) Mixed costs including mixed damage and management costs (cases where reported costs were not clearly distinguished among cost types). We note that Management costs include also research spending, irrespective of the ndings, because this work often aims to better understand the ecology of invaders and their impacts, in turn informing management options; (v) Impacted_sector (i.e. the activity, societal or market sector where the cost occurred; see Supplement 4). Individual cost entries not allocated to a single sector were modi ed to "Other".

Temporal dynamics of costs
We analysed the economic costs of invasive macrofouling bivalves over time. For this, we used the calculateRawAvgCosts-function implemented in the R package "invacost" (Leroy et al. 2020). With this method, we calculated the observed cumulative and average annual costs between 1980 -2020, considering 10-year intervals.

Economic costs among method reliability and implementation types
Although constituting the majority of cost entries (n = 328), highly reliable cost estimates comprised only 10 % of the documented total ($ 6.2 billion), with the remaining costs not originating from accessible peer-reviewed or o cial sources. Observed costs accounted for 77 % from freshwater bivalves, whereas other potential costs were derived in the absence of the invader in the study area based on observed costs in other regions (i.e. in the case the species were to be introduced) or based on extrapolated predictions of an existing impact over time (see Diagne et al. 2020 for details). In particular, 72 % of documented Dreissenidae costs, as well as 99% of Cyrenidae costs were observed.

Economic costs among geographic regions and cost types
Approximately 99% of the total costs were incurred in North America (Figure 2a). For Dreissenidae, the single M. trautwineana cost was incurred in South America ($ 0.007 billion), 69 speci c D. polymorpha cost entries were incurred in North America ($ 18.2 billion), 13 in Europe and North America combined ($ 1.10 billion), and 173 in Europe ($ 0.06 billion). No invasive bivalve costs were reported for Africa, Asia or Oceania. All costs of the family Mytilidae (L. fortunei; n = 31; $ 0.012 billion) were incurred in South America, while the two entries of Unionidae (S. woodiana, $ 0.002 million) originated from Europe.
With respect to cost types, 48% of bivalve-related costs were categorised as due to damages or resource losses ($ 30.3 billion), with relatively little (3%; $ 1.7 billion) spent on control singularly (Figure 3b). The largest share of costs (50%; $ 31.6 billion) was, however, categorized as general (mixed) as they contained elements relating to several types and were thus not speci c. For Cyrenidae, the majority of costs were due to damages, whereas the remainder were associated with mixed control and damages exclusively.

Economic costs across North American sectors
In North America speci cally, where the vast majority of bivalve costs were reported, 39% ($ 24.2 billion; n = 51) of bivalve costs was incurred by unde ned or unspeci ed socioeconomic sectors (Figure 4), whilst 19% ($ 11.6 billion; n = 25) impacted public and social welfare directly (e.g. via power/drinking water plant and irrigation system damage), only being surpassed by 41% of costs ($ 26.3 billion; n = 93) attributed to authorities and stakeholders (e.g. public and private sector interventions; see Diagne et al. 2020 for full de nition of each category). Of the remaining sector types, 'Environment' was listed with $ 369.6 million, followed by 'Fisheries' with $ 7.4 million. At the species-level, C. uminea had lower speci c costs to the public and social welfare sector than D. polymorpha ($ 2.2 vs. 9.0 billion).

Economic cost accumulations through time
Cost accumulations between 1980 and 2020 are presented in Figure 5. In total, these costs remained at a consistent magnitude over the past decade and amounted to $ 63.6 billion, with an average annual cost over the entire period of $ 1.6 billion. Whilst the effects of time lags in cost reporting were not incorporated into analyses, average cost estimates became reduced slightly towards the end of the last decade, indicating a gap in cost reporting.

Discussion
The present study demonstrates massive economic costs associated with invasive freshwater bivalves, estimated at a total of $ 63.6 billion USD over the period 1980-2020. The resulting average annual cost of $ 1.6 billion is lower than the previous annual cost estimation ($ 2 billion USD) for the zebra mussel and Asian clam in the United States (Pimentel et al. 2005). However, here we explicitly account for temporal dynamics in costs over a longer period, using a more conservative methodology and more robust data. Within the InvaCost database, Dreissenidae constituted the majority of data sources and costs, while fewer cost entries referred to Cyrenidae and none for other families, excepting minor additions from the Mytilidae and Unionidae. Within these families, D. polymorpha, D. bugensis, and C. uminea were implicated in the vast majority of economic damage, particularly in North America where they are widespread and locally abundant. Nonetheless, species such as C. uminea are global invaders (Sousa et al. 2008), and thus a lack of cost estimation for such taxa on a wide scale is surprising and indicates a profound lack of reporting. Furthermore, few documented costs were reported for the golden mussel L. fortunei, which is invasive in southeast Asia and South America (e.g. Sousa et al. 2014;Boltovskoy and Correa 2015). Accordingly, the current availability of costs identi ed is inherently species-speci c, and thus, costs likely represent a gross underestimation of the full scale of economic impacts across taxonomic groups, given the range of impact types associated with many macrofouling freshwater species and entirely unreported groups (Sousa et al. 2009.
On a taxonomic level, some key species of freshwater bivalves with well-known invasion histories (e.g. the golden mussel L. fortunei, the dark false mussel Mytilopsis leucophaeta, the Chinese pond mussel Sinanodonta woodiana) account for only a few entries in the InvaCost database, owing to a lack of published or traceable cost data. Macrofouling induced by L. fortunei and M. leucophaeta (a predominantly brackishwater species that was not represented in InvaCost), in particular, has been recognized as an economic problem for South America and Europe, respectively, where they foul municipal and industrial water supply systems (Verween et al., 2010). Yet, their invaded regions contributed very little of the total documented costs of freshwater invasive bivalves. Both L. fortunei and M. leucophaeta generate dense colonies causing obstructed water ow in pipes, occlusion of water lters, and corrosion of surfaces that result in system shutdowns, chemical/mechanical treatment, and equipment replacement (Magara et al. 2001;Montalto and De Drago 2003;Rajagopal et al. 2003;, virtually identical to the biofouling impacts associated with Dreissena and Corbicula. In a review of the economic impacts of L. fortunei on man-made structures,  noted that "objective estimates of the economic losses are extremely rare", but nevertheless economic impacts are probably quite substantial. The authors mentioned that the annual costs of maintenance and cleaning tasks owing to Limnoperna biofouling in one pipeline project in China, for example, have been anecdotally reported at over $1 million USD. However, this cost was not included in the analysis as no citable reference could be located. In Brazil, over 30 hydroelectric power plants along the Paraná River and its tributaries have been colonized by L. fortunei; a shutdown of a single 40 MW turbine for servicing as a result of biofouling could cost $6.2 million USD per year in lost power generation (reviewed by . Moreover, the geographic bias of cost estimations towards North America and the complete lack of documented cost estimation within Asia, Africa, and Oceania re ect major knowledge gaps in the economic costs of invasive bivalves spatially. While North America is unique in its cultural history, leading to a substantially higher study effort, it is also possible that actually exerted impacts of invasive bivalves are unevenly distributed owing to differences in economic activity. Further, it may be possible that early estimates for invasion costs in the USA led to greater reporting efforts for invasion economic effects in the last two decades (Pimentel et al. 2000). Indeed, the zebra mussel invaded most of the waterways in central and western Europe well before the mid-20th century (Dediu, 1980). We speculate that this produced a baseline bias in which subsequent costs were not viewed as novel and thus, were not reported -in contrast with the sudden incursion and recognition of massive costs following the more recent invasion of North America.
Moreover, zebra mussel densities in the Great Lakes reached peaks that were 1-2 orders of magnitude larger than what is typically reported in Europe, probably because as invasions progress mussel densities tend to level off at a lower equilibrium density (Burlakova et al., 2006;Jernelöv, 2017). However, this trend was already highlighted more broadly in invasion science (Early et al. 2016). In turn, less than 1 % of the globally reported costs of invasive bivalves were estimated from within Europe or South America; but an absence of evidence is not evidence of absence.
Our analyses indicated that studies reporting invasive freshwater bivalve costs have remained at a similar magnitude in recent decades. Whilst average decadal cost estimates tended to decline slightly in recent years, this is likely to be an artefact of time lags in cost estimation, rather than an empirical reduction in economic impact. The relative stability in cost increases for freshwater bivalves might also relate, in some cases, to improved management e ciencies -in spite of increases in both invasive species numbers (Seebens et al. 2017) and global invasion costs (Diagne et al. 2021, see also Cuthbert et al., 2021 for aquatic IAS) through time. For example, once being initially impacted by pipe-clogging and having to shut down for cleaning, industries will typically bleed chlorine in their water intakes to eliminate further fouling, thus reducing on-going costs. On the contrary, it is also entirely possible that the annual monetary burden actually increased between years owing to new invasions, interventions or damages, leading to a gross underestimation of costs, owing to (i) insu cient reporting (Wakida-Kusunoki et al. 2015; Enders et al. 2019) and / or (ii) the very conservative nature of our approach. An outstanding example of the latter is the impact of biofouling by the Asian clam Corbicula uminea on the operation of power plants in the United States over several decades, compromising fail-safe operations and causing emergency shutdowns of nuclear facilities. The control and mitigation costs, as well as costs related to reduced plant operating e ciencies, were estimated by Isom (1986) to exceed $1 billion USD per year, based on various anecdotal costs recorded primarily before 1980. Our approach led us to ignore all costs prior to 1980, despite C. uminea having invaded the USA and other regions many decades before (Crespo et al., 2015). Further, these costs only pertain to power plants in the US, whereas C. uminea is globally invasive and has fouled water supply systems in other countries. In addition to impacts on technological systems, C. uminea is known to negatively impact native bivalve abundance and diversity (Sousa et al., 2008), and to alter physical habitat structure including water quality, sediment composition, and submerged vegetation (Phelps 1994), thus producing ecosystem impacts that can be di cult to quantify in monetary terms (Darrigan, 2002). It should be emphasised, therefore, that we consider the presented costs to be highly conservative overall, particularly given the prominent cost reporting gaps, both taxonomically and spatially.
Another factor contributing to uncertainty surrounding our estimate is the di culty in quantifying types of economic damage associated with ecosystem services (Spangenberg and Settele 2010). Invasive freshwater bivalves can be ecosystem engineers and keystone species where they have disproportionate effects on ecosystem structure and function-and thus, the various services they provide to humans (e.g. aquaculture).
For instance, dreissenid mussels indirectly stimulate benthic algal growth (Boegman et al. 2008), invasive aquatic weed proliferation (Zhu et al. 2007), and harmful algal Microcystis blooms (Vanderploeg et al. 2001). Furthermore, dreissenid species have been shown to create new pathways for the transfer of contaminants (e.g. Hg, Cd, PCBs, botulism toxin; Hogan et al. 2007;Carrasco et al. 2008). These effects likely result in substantial indirect socioeconomic impacts that are di cult, if not impossible, to evaluate in terms of monetary losses. More directly, costs of invasive macrofouling bivalves incurred for technological systems other than power plants (municipal and industrial water supply systems in general; fouling of lock-and-dam structures and aquaculture equipment) are virtually undocumented for most regions of the world other than the USA and Canada. Research effort into freshwater bivalves is concentrated in North America and Europe (Lopes-Lima et al. 2014), with a consequent lack of detailed reporting of basic aspects of invasions in other regions (Lopes-Lima et al. 2018), where invasive freshwater bivalves have been reported only relatively recently (e.g. Africa; Clavero et al. 2012). In these cases, published documentation of ongoing costs is urged to fully account for monetary aspects of invasion within emerging economies.
The sparse economic data for invasive freshwater bivalves also inhibits recognition of any potential bene ts these species provide to humans, and thus impedes comprehensive cost-bene t analyses which could further inform and direct management actions among different economic sectors or regions. For example, ltration activities of dense populations of Dreissena spp. and C. uminea have been shown to substantially increase water clarity (Phelps, 1994;Higgins and Vander Zanden 2010;Boltovskoy et al. 2009), which (while causing myriad ecological disruptions and harm sheries whose focal species depend on prey that are competing with mussels for resources; Kao et al. 2018) could bene t certain recreational activities such as scuba diving, which in turn could conceivably drive tourist revenue and increase the property value of neighbouring real estate. Conversely, accumulations of sharp shells on beach sands are a hazard to the feet of swimmers. Whilst many bene cial effects are di cult to quantify in monetary terms, or are yet to be shown, it is unlikely that they will outweigh the presently documented (and underestimated) costs of $ 63.6 billion USD.
In conclusion, our study highlights very fragmented data that calls for national and regional authorities to produce more and better structured reporting of invasion costs. Given that many known invasive freshwater bivalve species (such as Batissa violacea, Sphaerium corneum, and Pisidium spp.; see Sousa et al., 2013) and invaded regions completely lacked reported economic costs, our gures are likely gross underestimations. Nonetheless, the monetary costs reported in this study are still very high (e.g., over 1 billion US$ per year) and should provide added incentive to manage invasive bivalves in freshwater systems. When speci c cost types were known, damages and resource losses were an order of magnitude higher than control or management costs, suggesting that more management is needed to prevent the spread and establishment. Given that invasion rates are expected to keep increasing over time (Seebens et al. 2017(Seebens et al. , 2020, we predict that the costs of invasive macrofouling freshwater bivalves will increase substantially in the future.

Declarations Funding
The authors acknowledge the French National Research Agency (ANR-14-CE02-0021) and the BNP-Paribas Foundation Climate Initiative for funding the InvaCost project that allowed the construction of the InvaCost database. The present work was conducted following a workshop funded by the AXA Research Fund Chair of Invasion Biology and is part of the AlienScenario project funded by BiodivERsA and Belmont-Forum call 2018 on biodiversity scenarios. RNC is funded by a research fellowship from the Alexander von Humboldt foundation. CD was funded through the 2017-2018 Belmont Forum and BiodivERsA joint call for research proposals, under the BiodivScen ERA-Net COFUND programme with Project "Alien Scenarios" (BMBF/PT DLR 01LC1807C).

Con icts of interest/Competing interests
No con ict of interest has to be declared.

Availability of data and material
The underlying data was provided as supplementary material.

Code availability
The code required has been referenced in the related sections within the methods.
Authors' contributions PJH and RNC led the writing and analysis. AR provided valuable insights and contributed to the writing. CD and FC provided the database and contributed to all aspects of the manuscript production. Figure 1 successive steps of ltering from the entire InvaCost database to the conservative subset analysed for annualized costs of freshwater bivalves between 1980 and 2020. Each step is detailed in the text.

Figure 5
Annual total and observed costs between 1980 -2020 of invasive macrofouling freshwater bivalves and the number of published cost entries between the same period. Points with bars represent decadal means. Note the broken y-axis scale.