Taxonomic and geographical representation of freshwater environmental DNA research in aquatic conservation

s, or keywords. Overall, 13% of the articles retrieved were published in concurrence with the keyword ‘conservation’, and, on average, 400 articles (34%) of all publications targeted or included aquatic environments (14% ‘freshwater’, 16% ‘marine’; Supporting information Table S1). Surprisingly, only one publication (Harper, Buxton, et al., 2019) was returned from Scopus and none from either Web of Sciences or PubMed when querying ‘environmental DNA’ (eDNA) AND ‘aquatic conservation’. Likewise, only one publication (Bellemain et al., 2016) was returned from Scopus and Web of Sciences, and none from PubMed, when querying ‘environmental DNA’ (eDNA) AND ‘freshwater conservation’. After skimming all abstracts, 272 publications were collated on metazoan freshwater eDNA from 2005 to 2018 that met the selected criteria, covering primary research articles, reviews, letters, opinions, notes, and comments (Figure 1, Supporting information Tables S2, S3). The total number of publications since 2005 increased exponentially (y = 0.3367^e0.5118x; Pearson R = 0.98), with a maximum of 77 papers in 2017 on eDNA in freshwater research, and an average increase in the annual publication rate of 51% between 2005 and 2018 (Figure 1a, b). This pattern was even more pronounced for freshwater fishes, the taxonomic group most studied. The research output on freshwater fish eDNA also increased exponentially (y = 1.0866^e0.7185x; Pearson R = 0.99) with an average increase in the annual publication rate of 72% (Figure 1b). For the two other major subgroups— amphibians (annual average increase 45%) and ‘non‐primary research articles’ (56%) —the pattern was less pronounced, but the scientific output on these also increased significantly over the years (all Mann–Kendall tests P < 0.001). By far the majority (87%) of all publications were primary research articles, whereas non‐primary research articles (reviews, letters, comments, opinions, views, and introductions) comprised only 13% of the dataset (Supporting information Table S3). FIGURE 1 (a) Cumulative number and (b) annual trend curves of freshwater environmental DNA (eDNA) studies using metazoan organisms (‘macrobial freshwater eDNA’) from 2005 to 2018 3.2 | Objectives of the studies The most common objective of all primary research publications in the dataset (57%) was conservation science, in terms of (a) invasive species (86 articles) and (b) endangered or rare species (67 articles) (Figure 2a, Supporting informationTable S3). Primary research output with a focus on advancing or elucidating freshwater eDNA methodological development comprised 61 studies (23% of the total). Methods included the development of filtering, storage, and extraction methods (Minamoto, Naka, Moji, & Maruyama, 2016), or identifying possible sources of field cross‐contamination (Merkes, McCalla, Jensen, Gaikowski, & Amberg, 2014). Of the primary research articles, more than 80% (203 studies) used a target species detection approach, whereas a small but annually increasing proportion of studies used eDNA metabarcoding (30 studies, 11%; Figure 2a). The objectives of the remaining 24 articles comprised eDNA techniques as diverse as food web analyses (‘molecular scatology’, Koizumi et al., 2016; Thalinger et al., 2016), or the detection of ‘surface water contamination’ (Martellini, Payment, & Villemur, 2005; Schill & Mathes, 2008). 3.3 | Representation of habitats Of the 238 field studies, 77% were performed in lentic (35%) and lotic (42%) habitats (Figure 2b, Supporting information Table S3). The lotic environments investigated included habitats as diverse as small headwater streams (Baldigo, Sporn, George, & Ball, 2017) and small brooklets (Katano, Harada, Doi, Souma, & Minamoto, 2017), up to complete river catchments, such as the Cuyahoga River in Ohio, USA (Cannon et al., 2016), the River Glatt catchment in Switzerland (Deiner, Fronhofer, Machler, Walser, & Altermatt, 2016), or the Mekong River in South East Asia (Bellemain et al., 2016). The lentic habitat diversity ranged from phytothelmata (Brozio et al., 2017) and small ponds (Mauvisseau et al., 2018) to the North American Great Lakes in the USA and Canada (Jerde et al., 2013), and artificial and highly impaired freshwater reservoirs in Singapore (Lim et al., 2016). Artificial ecosystems like mesocosms were used in 45 studies (16%). Various sediments as an additional ‘habitat’ for eDNA in mesocosms FIGURE 2 The percentage of primary research articles using macrobial environmental DNA to detect the presence of freshwater species according to (a) objective, (b) habitats (ecosystem types), (c) geographical region (geographical distribution of ecological studies per continent), and (d) target species (order/class/phylum level) used. If more than one objective, continent, freshwater ecosystem type, or target species was included in an article, the respective variable was counted as many times as it occurred to produce these charts. For the true number of articles, see Supporting information Table S3 BELLE ET AL. 2000 and lentic or lotic systems were investigated in seven (2%) of the studies (Figure 2b, Supporting information Table S3). Subterranean or groundwater habitats (the overall predominant freshwater habitats worldwide) were investigated in two studies in the dataset (Niemiller et al., 2018; Vörös, Marton, Schmidt, Gal, & Jelic, 2017; see also Supporting information Table S2). 3.4 | Representation of geographic regions More than 90% of all primary research papers described work undertaken in North America (52%, 123 studies), Europe (20%, 48 studies), and Asia (19%, 44 studies) (Figure 2c, Supporting information Table S2). With almost 70% (30 studies) of the Asian studies performed, Japan is the most investigated geographical area of this continent (Table S2). Australasia (17 studies) and South America (six studies) contributed 9% to the research output (e.g. Brozio et al., 2017; Lopes et al., 2017; Robson et al., 2016; Simpfendorfer et al., 2016) (Figure 2c), but there were no contributions from Africa (Figure 2c). There is great potential for further expansion of eDNA studies in the tropics, as more than 75% (179 studies) of the research papers concerned temperate to subtropical zones either in the Northern or Southern Hemisphere (Table S2). The exceptions are the studies conducted in northern Australia (e.g. Robson et al., 2016), as well as a few in South East Asia (Bellemain et al., 2016; Wilson, Sing, Chen, & Zieritz, 2018), and the first example of eDNA metabarcoding in monitoring tropical freshwater reservoirs in Singapore (Lim et al., 2016). 3.5 | Taxonomic group composition and potential taxonomic bias Fish species were the most frequently investigated taxa (54%, 128 studies; Figure 2d). Studies were focused either on detecting particular species (Boothroyd, Mandrak, Fox, & Wilson, 2016; Clusa & García‐Vázquez, 2018) or on analysing taxon richness (Hänfling et al., 2016), or they used fish as model organisms to address methodological issues (Barnes et al., 2014; Stoeckle et al., 2017). Amphibians (40 articles; e.g. Biggs et al., 2015; Ficetola et al., 2008) and freshwater arthropods (25 studies; e.g. Doi et al., 2017; Mächler, Deiner, Steinmann, & Altermatt, 2014) were the second‐ and third‐most investigated taxa, with a surprisingly high percentage of the studies (10%) targeting freshwater molluscs (e.g. Currier, Morris, Wilson, & Freeland, 2018; Stoeckle, Kuehn, & Geist, 2016) (Figure 2d, Supporting information Table S3). Each of the other groups contributed less than 10% to the research output (Figure 2d). The primary research articles using a target species approach investigated 123 freshwater fish species, 29 amphibian species, and FIGURE 3 (a) Taxonomic representation of environmental DNA (eDNA) studies (light blue bars) in different taxonomic groups in relation to their relative abundance in freshwater environments (dark blue bars). The overall freshwater species richness and relative abundance follow Balian et al. (2010). The relative abundance in eDNA studies considers the number of species investigated in target‐species freshwater eDNA studies per taxonomic group. (b) Ratio of the number of species considered in target‐species freshwater eDNA studies to the overall number of known species for different BELLE ET AL. 2001 28 mollusc species. Freshwater mammals (four species), trematodes (two species), and fungi (two disease species) were the least investigated groups (Supporting information Table S2). With 23 articles published on each of the invasive common and Asian carp species, Cyprinus carpio Linnaeus, 1758 and Hypophthalmichthys molitrix (Valenciennes, 1844)/Hypophthalmichthys nobilis (Richardson, 1845) (e.g. Mahon et al., 2013; Wilson et al., 2014), these were the best investigated research organisms, followed by the bluegill sunfish (Lepomis macrochirus Rafinesque, 1819; e.g. Evans et al., 2016; Takahara, Minamoto, & Doi, 2013). Apart from amphibians, where the red‐listed great crested newt, Triturus cristatus (Laurenti, 1768), was investigated in six studies (e.g. Rees et al., 2014), invasive species were the research organisms most frequently targeted in all other major taxonomic groups: for molluscs, the zebra mussel, Dreissena polymorpha (Pallas, 1771) (eight studies), quagga mussel, Dreissena bugensis (Andrusov, 1897) (seven studies), and Asian clam, Corbicula fluminea (O.F.Müller, 1774) (three studies); and for arthropods, red swamp crayfish, Procambarus clarkii (Girard, 1852) (six studies), signal crayfish, Pacifastacus leniusculus (Dana, 1852) (four studies), and the amphipod Gammarus pulex (Linnaeus, 1758) (three studies) (e.g. Ardura, Zaiko, Borrell, Samuiloviene, & Garcia‐Vazquez, 2017; Clusa, Miralles, Basanta, Escot, & Garcia‐Vazquez, 2017; Deiner, Walser, Machler, & Altermatt, 2015; Harper, Turnbull, Bean, & Leaver, 2018; Mauvisseau et al., 2018). Taking into account the relative estimated number of freshwater species in relation to the number of eDNA studies (Figure 3a), and the investigated number of species in each taxonomic group (Figure 3b), eDNA in different vertebrates—especially reptiles, mammals, and fish species—was more often investigated than in all other taxonomic groups (Figure 3a, b). For invertebrate species, molluscs were the most investigated group in relation to their estimated freshwater species richness, whereas aquatic insect species were the least represented and investigated taxonomic group (Figure 3a, b). The number of eDNA studies of freshwater crustaceans was more balanced compared with their relative species richness in freshwater ecosystems (Figure 3a). In contrast, the number of species investigated in relation to the total estimated crustacean species richness was low (Figure 3b). Freshwater macrophytes showed the same investigated ratio of species/total species number as the amphibian species investigated (Figure 3b). taxonomic groups. If the ratio approaches 1, the relative abundance of species studied equals the relative abundance of the respective taxonomic group. Note that the y‐axis is log‐transformed. Only two species of freshwater fungi and two species of parasitic trematodes were investigated using eDNA techniques. Both were excluded from the diagram owing to uncertain freshwater species estimates in those two taxonomic groups 4 | DISCUSSION The increasing popularity of eDNA in biodiversity monitoring demonstrated in this study calls for a critical review on the trends, applications, and geographical and taxonomic representation of this technique in the context of freshwater conservation. The study showed a strong bias towards specific species and taxonomic groups, such as freshwater fish, both monitoring rare and endangered species as well as invasive alien species, but a more balanced representation of studies from lentic and lotic habitats. The relatively large number of studies addressing methodological issues and ‘artificial ecosystems’, such as mesocosms, reveals the initial need for tackling fundamental challenges such as inhibition and contamination before reaching more harmonized and standardized approaches. The results showed a distinct gap in studies in the tropics in general, and on the African continent in particular, suggesting that regions with exceptionally high freshwater biodiversity have not yet been sufficiently targeted by eDNA monitoring.


| INTRODUCTION
Freshwater ecosystems are considered the most imperilled habitat type worldwide (Boon & Pringle, 2009;Geist, 2011;Reid et al., 2019;Strayer & Dudgeon, 2010;Vörösmarty et al., 2010). In temperate zones, they are also one of the most intensively monitored habitats owing to legal requirements such as the European Water Framework Directive (WFD; Council of the European Communities, 2000) or the Habitats Directive (Council of the European Communities, 1992). Most of the monitoring approaches apply classical field-based techniques (Geist, 2015;Hering, Moog, Sandin, & Verdonschot, 2004; reviewed by Birk et al., 2012). However, classical monitoring methods for biodiversity assessment and freshwater conservation, such as electrofishing, gill netting, or kick sampling, often include expensive and labourintensive capturing (Mueller, Pander, Knott, & Geist, 2017) that can adversely affect the target organisms that are being surveyed, either by direct impact on these species through sampling or by disturbance of their habitats (Hering et al., 2018).
'Macrobial eDNA' techniques refer to various molecular approaches in which shed intra-or extracellular metazoan DNA (eDNA) is extracted from soil, air, or water. Subsequently, the genetic material obtained is amplified and then analysed, or analysed directly.
Analogously, for detecting species richness, with its associated sampling costs and survey time, eDNA metabarcoding can be superior over classical methods, sometimes by several orders of magnitude (Boussarie et al., 2018;Sigsgaard et al., 2017). Part of the reason is that only water samples need to be collected for analyses instead of directly sampling or observing the target species. Owing to the sensitivity of eDNA detection and the greater chance of overlooking rare species in their habitat by classical surveys, it also typically results in greater species richness estimates than conventional surveys. eDNA methods as an addition to classical sampling help to obtain more accurate estimates relevant for species conservation; for example, by elucidating 'dark' diversity (or 'hidden species richness'), microhabitat use, or the influence of human disturbance on dispersal (Boussarie et al., 2018).
Some even suggest entirely replacing classical monitoring approaches by the sole use of eDNA techniques for biomonitoring (Bourlat et al., 2013;Leese et al., 2016). This approach led to a highly controversial discussion, especially concerning quantitative estimates derived from eDNA techniques to supplement or to replace classical monitoring Rees, Gough, Middleditch, Patmore, & Maddison, 2015;Roussel et al., 2015). The strong and increasing interest in eDNA is also reflected in the large number of reviews, opinion papers, and letters published in the field of freshwater eDNA (collated by Coble et al., 2019). To date, most of these papers address how eDNA can be used to detect biodiversity throughout different environments (Bohmann et al., 2014;Pedersen et al., 2015;. There is a wealth of systematic reviews that summarize and focus on the methodology of eDNA techniques in fresh water , eDNA persistence and transport in aquatic environments (Barnes et al., 2014), the possible use of eDNA techniques in detecting aquatic invasive species (Darling, 2015;Darling & Mahon, 2011), the potential of eDNA techniques in ecology and wildlife monitoring (Barnes & Turner, 2016;Bohmann et al., 2014), in palaeolimnology (Domaizon, Winegardner, Capo, Gauthier, & Gregory-Eaves, 2017), in the use of eDNA metabarcoding in surveying plant and animal communities (Deiner et al., 2017), as well as in determining the environmental factors that affect the success of eDNA detection (Stoeckle et al., 2017). Other pieces of work outline the potential use of eDNA metabarcoding in biological assessments or biomonitoring of aquatic ecosystems (Macher et al., 2018;Mächler et al., 2019;Pawlowski et al., 2018), and possibilities for its application to ecological status assessment under the WFD (Hering et al., 2018).
Fewer reviews or comments specifically address critical issues concerning the implementation of eDNA methods in general (Roussel et al., 2015), and particularly the standardization of sampling designs (Dickie et al., 2018;Wilcox et al., 2018), as well as downstream laboratory and data analysis methods (Goldberg et al., 2016).
Considering the many reviews already published about the applicability of eDNA techniques, it is most surprising that, to the best of our knowledge, the aspect of the general applicability of eDNA techniques in the context of freshwater conservation has rarely been addressed (but see Harper, Buxton et al., 2019;Harper, Lawson et al., 2019).
Knowledge of the current application and implementation of eDNA techniques in fresh water is crucial to assess potential gaps and biases, and to recommend future development in the context of biological conservation. Obtaining important knowledge and insights in assessing the representation of specific research output (taxonomically and in terms of geographical representation) has been shown already in other areas-for example, invasion biology (Jeschke et al., 2012;Pysek et al., 2008), biodiversity sciences and taxonomy (Cardoso, Erwin, Borges, & New, 2011;Clark & May, 2002;Troudet, Grandcolas, Blin, Vignes-Lebbe, & Legendre, 2017), or for other specific molecular techniques (e.g. conservation genetics; Pérez-Espona, 2017). The online publication databases Web of Science Core Collection (www.webofknowledge.com), Scopus (www.scopus.com), and PubMed (www.ncbi.nlm.nih.gov/pubmed) were initially searched using the keywords and search strings given in Supporting informationTable S1. Publications solely considering microbial organisms or prokaryotic biodiversity were subsequently excluded, as well as those articles on marine and terrestrial environments. Other papers not covered by the database queries were identified from screening the reference lists of publications found through the keyword literature search (Supporting information Table S2). Some studies may not have been identified because of the use of different discipline-specific keywords-for example, 'ecogenomics' for elucidating micro-and macrobial communities under stress , 'surface water contamination' (Schill & Mathes, 2008), or 'molecular scatology' (Thalinger et al., 2016). Some new studies may also have been missed owing to the time lag in uploading to literature databases.
The following information was extracted from each publication: authors, title, year of publication, journal, type of article (e.g. review or primary research), objective(s), geographical region, and taxonomic group(s) investigated. A trend analysis of the annual increase in publication numbers and their focus (taxonomic group and target [i.e. original paper or review]) was conducted using the R package bibliometrix version 1.9.3. (Aria & Cuccurullo, 2017). Mann-Kendall trend tests, in the R package 'trend' (Pohlert, 2018), were used to test for significance.
Subsequently, the collated articles were read in full and initially classified as 'primary research articles' or 'non-primary research articles'. 'Non-primary research articles' comprised reviews, letters, comments, opinions, views, and introductions to special journal issues that summarized, discussed, or analysed data from studies already published on aquatic macro-organismic eDNA research. To avoid data duplication, 'non-primary research articles' were therefore excluded from further analysis.
For evaluating the research content, the primary research publications were categorized according to their general aim addressing: (a) biodiversity assessments using the 'eDNA metabarcoding' approach (sensu Taberlet et al., 2012;Deiner et al., 2017) with universal or group-specific primers targeting the expected overall metazoan biodiversity or a specific taxonomic group in the respective habitat;  Table S3).
To assess the taxonomic groups investigated, and whether there is any taxonomic bias in the eDNA methods applied to freshwater conservation and research, the major taxonomic groups targeted in the current eDNA literature were classified according to species/order/class/phylum. To elucidate the taxonomic bias in eDNA primary research, the freshwater species in research papers using a target species approach were recorded. Subsequently, the number of eDNA studies per taxonomic group and the proportion of the number of freshwater species (species richness) per taxonomic group covered by eDNA research were compared with the freshwater species richness described for the respective taxonomic group and the groups' relative abundances. Both values were extracted from Balian et al. (2010).

| Number of publications and trends
The literature database keyword search resulted in 1,480 (Web of Sciences Core Collection), 1,214 (Scopus), and 864 (PubMed) articles that indexed 'environmental DNA' or 'eDNA' either in the title, abstracts, or keywords. Overall, 13% of the articles retrieved were published in concurrence with the keyword 'conservation' , and, on average, 400 articles (34%) of all publications targeted or included aquatic environments (14% 'freshwater', 16% 'marine'; Supporting information Table S1). Surprisingly, only one publication (Harper, Buxton, et al., 2019) (Figure 1a, b). This pattern was even more pronounced for freshwater fishes, the taxonomic group most studied. The research output on freshwater fish eDNA also increased exponentially (y = 1.0866^e0.7185x; Pearson R 2 = 0.99) with an average increase in the annual publication rate of 72% (Figure 1b). For the two other major subgroups-amphibians (annual average increase 45%) and 'non-primary research articles' (56%) -the pattern was less pronounced, but the scientific output on these also increased significantly over the years (all Mann-Kendall tests P < 0.001).
By far the majority (87%) of all publications were primary research articles, whereas non-primary research articles (reviews, letters, comments, opinions, views, and introductions) comprised only 13% of the dataset (Supporting information Table S3).

| Objectives of the studies
The most common objective of all primary research publications in the dataset (57%) was conservation science, in terms of (a) invasive species
Various sediments as an additional 'habitat' for eDNA in mesocosms  Table S3). Subterranean or groundwater habitats (the overall predominant freshwater habitats worldwide) were investigated in two studies in the dataset (Niemiller et al., 2018;Vörös, Marton, Schmidt, Gal, & Jelic, 2017; see also Supporting information Table S2).

| Representation of geographic regions
More than 90% of all primary research papers described work
The primary research articles using a target species approach investigated 123 freshwater fish species, 29 amphibian species, and FIGURE 2 The percentage of primary research articles using macrobial environmental DNA to detect the presence of freshwater species according to (a) objective, (b) habitats (ecosystem types), (c) geographical region (geographical distribution of ecological studies per continent), and (d) target species (order/class/phylum level) used. If more than one objective, continent, freshwater ecosystem type, or target species was included in an article, the respective variable was counted as many times as it occurred to produce these charts. For the true number of articles, see Supporting information Table S3 28 mollusc species. Freshwater mammals (four species), trematodes (two species), and fungi (two disease species) were the least investigated groups (Supporting information  Mahon et al., 2013;Wilson et al., 2014), these were the best investigated research organisms, followed by the bluegill sunfish

| DISCUSSION
The increasing popularity of eDNA in biodiversity monitoring demonstrated in this study calls for a critical review on the trends, applications, and geographical and taxonomic representation of this technique in the context of freshwater conservation. The study showed a strong bias towards specific species and taxonomic groups, such as freshwater fish, both monitoring rare and endangered species as well as invasive alien species, but a more balanced representation of studies from lentic and lotic habitats. The relatively large number of studies addressing methodological issues and 'artificial ecosystems' , such as mesocosms, reveals the initial need for tackling fundamental challenges such as inhibition and contamination before reaching more harmonized and standardized approaches. The results showed a distinct gap in studies in the tropics in general, and on the African continent in particular, suggesting that regions with exceptionally high freshwater biodiversity have not yet been sufficiently targeted by eDNA monitoring.
FIGURE 3 (a) Taxonomic representation of environmental DNA (eDNA) studies (light blue bars) in different taxonomic groups in relation to their relative abundance in freshwater environments (dark blue bars). The overall freshwater species richness and relative abundance follow Balian et al. (2010). The relative abundance in eDNA studies considers the number of species investigated in target-species freshwater eDNA studies per taxonomic group. (b) Ratio of the number of species considered in target-species freshwater eDNA studies to the overall number of known species for different taxonomic groups. If the ratio approaches 1, the relative abundance of species studied equals the relative abundance of the respective taxonomic group. Note that the y-axis is log-transformed. Only two species of freshwater fungi and two species of parasitic trematodes were investigated using eDNA techniques. Both were excluded from the diagram owing to uncertain freshwater species estimates in those two taxonomic groups The finding of only a single paper for both search terms 'environmental DNA' and 'freshwater conservation' (Bellemain et al., 2016) indicates that researchers publishing papers on eDNA may fail to recognize the significance of their work to freshwater conservation.
In addition, some of the topics reviewed, such as the methodological development in invasive species research, may not always be carried out with regard to their impact on native species and their conservation as the primary motivation. However, there are a few studies where the use of eDNA was clearly related to meeting a conservation objective (Bellemain et al., 2016;Harper, Buxton, et al., 2019). Harper, Buxton, et al. (2019) recommended a broad standardization of eDNA workflows in order to ensure more robust, comparable, and ecologically meaningful data for pond biodiversity. The development of standard protocols is essential for including eDNA in effective management and conservation yet is also needed across habitat types. eDNA methods have been used most frequently for invasive species research, followed by endangered or rare species monitoring.
The observed focus on endangered and rare species can be explained by the fundamental interest in mapping and understanding distribution patterns as basic information in conservation sciences (Stewart, Ma, Zheng, & Zhao, 2017;Tucker et al., 2005;Yoccoz, Nichols, & Boulinier, 2001). Knowledge about the distribution and abundance of a species is a prerequisite for conservation management (Gaston, 2010;Heywood, 2011), yet endangered species are often rare, with small population sizes and patchy distribution patterns. They may also react sensitively to human disturbance, and therefore are difficult to detect or monitor with classical survey methods (Schill & Galbraith, 2019;Sigsgaard et al., 2017;Sigsgaard, Carl, Møller, & Thomsen, 2015). Given the sensitivity of species detection using eDNA, this technique can be an important supplementary tool for endangered species monitoring (Biggs et al., 2015;Itakura et al., 2019;Stoeckle et al., 2016;Thomsen et al., 2012), especially in deep and turbid aquatic habitats (Bellemain et al., 2016).
The value of eDNA techniques in endangered species and freshwater conservation can be illustrated using the example of freshwater mussels, as these belong to the most threatened groups in fresh water (Lopes-Lima et al., 2017) and are considered target species for conservation (Geist, 2010(Geist, , 2015. Even monitoring the adults in populations of endangered European freshwater pearl mussel, Margaritifera margaritifera (Linnaeus, 1758), can be extremely difficult, as this species also occurs in large and fast-flowing rivers where access is difficult (Boon et al., 2019). With many captive breeding programmes for the species being in place in Europe (Gum, Lange, & Geist, 2011) checks are needed to determine whether the release of juvenile mussels into stream sections previously without mussels has been successful. As juvenile mussels burrow into the substrate, their detectability with classical habitat survey methods is low and in certain countries such as Sweden is not permitted owing to the potential destruction of their habitat by monitoring (Boon et al., 2019). eDNA protocols are available for this species that can be particularly useful in screening larger rivers for the occurrence of mussels as well as for the detection of juveniles (Stoeckle et al., 2016), both providing useful information for conservation. On the other hand, limitations of eDNA methodology must be considered, such as the lack of information that it provides on population size and demography (Stoeckle et al., 2016) as well as the possible release of DNA from dead shell material (Geist, Wunderlich, & Kuehn, 2008). Consequently, a combination of eDNA as a screening tool with detailed field survey work can contribute to more effective monitoring and conservation in freshwater mussels.
There are several possible reasons for the large number of eDNA studies focusing on invasive species research. First, as well as the known ecosystem impacts, invasive species can cause severe economic damage, as evident, for example, from work on zebra mussels and quagga mussels (Connelly, O'Neill, Knuth, & Brown, 2007;Vanderploeg et al., 2002) and on the Asian carp species in the Mississippi and Laurentian Great Lakes (Pimentel, 2005;Pimentel, Zuniga, & Morrison, 2005). Second, the control and mitigation of biological invasions are most effective during the early stages of invasion, especially before establishment, during transport, introduction, and naturalization (Hulme, 2006;Simberloff et al., 2013). In these early stages, the abundance of invasive species is low; and because of small population sizes, these species are often not detectable with classical methods. Because of the high sensitivity, and subsequently the possibility to detect low individual abundances, or even single individuals, eDNA techniques are increasingly suggested and used as an important tool to monitor biological invasions (Darling & Mahon, 2011;Dejean et al., 2012;Egan et al., 2015).  1973;US Congress, 1990), make regular biological monitoring a necessity. Environmental agencies and government bodies often rely on well-established monitoring methods and, to date, mostly use and accept eDNA approaches for specific targets such as the monitoring of invasives only (Laramie, Pilliod, Goldberg, & Strickler, 2015). The relatively large proportion of studies addressing methodological development highlights a continuing need for further adjustment, validation, and optimization of eDNA techniques, from sampling through to laboratory detection, in the great variety of lentic and lotic habitats that all differ in their environmental conditions, such as water chemistry.
Although eDNA provides an efficient tool for detecting species composition, it has its limitations, such as the lack of reliable abundance assessments Stoeckle et al., 2016). In the context of most monitoring programmes, such as the European WFD (Council of the European Communities, 2000), both species composition and abundance need to be recorded, limiting the universal applicability of using eDNA assessments exclusively .
An additional challenge is the use of eDNA in rivers as opposed to lakes, where it is more difficult to know the location from which the detected eDNA originated Macher & Leese, 2017;Song, Small, & Casman, 2017;Wilcox et al., 2018).
The observed geographic bias towards Northern Hemisphere temperate zones can be partly explained by the greater number and execution of existing legal frameworks in these regions requiring aquatic biomonitoring programmes. The eDNA research need in the tropics and African countries probably results from socioeconomic constraints and the lack of the required infrastructure in many of these countries, which makes it difficult to implement eDNA studies, or biodiversity studies in general (Di Marco et al., 2017;Leese et al., 2018;Sodhi, Koh, Brook, & Ng, 2004;Waldron et al., 2013). So far, most of the eDNA research conducted in the tropics is restricted to areas where the necessary infrastructure is in place-for example, northern Australia (Robson et al., 2016) and Singapore (Lim et al., 2016). The lack of taxonomic resolution (Gaston, 1994) or expertise (Gaston & May, 1992), and therefore incomplete species or DNA reference databases, also hampers the development of eDNA techniques in most biodiversity-rich tropical countries in Africa and South America. The under-representation of eDNA studies from areas that are likely hotspots of aquatic biodiversity (and at the same time of aquatic habitat change and biodiversity decline) is alarming and should be given higher priority in future international research and development programmes. In addition, a systematic assessment of diversity between morphotaxonomists and DNA barcoders is needed here .
In line with the geographical bias, the taxonomic group representation and the research bias towards iconic vertebrate species evident from this study is pervasive in conservation sciences (Clark & May, 2002;Di Marco et al., 2017), and thus not surprising. It has previously been described by several other workers in associated research fields -for example, conservation genetics (Pérez-Espona, 2017) and invasion biology (Pysek et al., 2008). Following this general bias, freshwater fish are the taxa investigated most, as they represent iconic and socioeconomically important flagship species, many globally invasive species, and keystone species for ecosystem services (Geist, 2015).
Given the importance of macroinvertebrates in biomonitoring (Hering et al., 2004), as invasive species (Connelly et al., 2007;Vanderploeg et al., 2002), and as emerging infectious disease vectors in a changing world climate (Conn, 2014;Schneider et al., 2016), it is surprising that more macroinvertebrate species are not investigated using eDNA methods. Biodiverse groups such as aquatic insects appear severely under-represented, also in regions where the legal framework requires monitoring of macroinvertebrates. For example, the saprobic system is based on tolerance values of certain known macroinvertebrate species (Rolauffs, Hering, Sommerhäuser, Rödiger, & Jähnig, 2003), and, not just in Europe, freshwater macroinvertebrates are very important indicators in assessing water quality (e.g. as biological quality elements in the WFD; Council of the European Communities, 2000). Given the great richness of aquatic insect species, for many groups there is still the need to resolve the classical, integrative, or molecular taxonomy (e.g. aquatic Diptera/Psychodidae; Kvifte, 2018), and to enhance the respective DNA barcode reference databases (Elbrecht, Vamos, Meissner, Aroviita, & Leese, 2017;Leese et al., 2018). Furthermore, invertebrate species complexes that underwent rapid radiations resulting in incomplete genetic lineage sorting pose challenges to eDNA detection if using traditional mitochondrial DNA barcodes (Meier, Shiyang, Vaidya, & Ng, 2006). In general, cryptic species, a lack of alpha taxonomy, and insufficient taxonomic knowledge may hamper the current progress, as well as the fact that genetic reference databases are exhaustive only for certain aquatic invertebrate groups-for example, freshwater bivalves or European Ephemeroptera, Plecoptera, and Trichoptera taxa (Hendrich et al., 2015;Moriniere et al., 2017;Zhou et al., 2016). The example of 85% coverage of North American freshwater invertebrate genera (biological quality elements) in public barcoding databases (Curry, Gibson, Shokralla, Hajibabaei, & Baird, 2018) shows the great potential for next-generation biomonitoring if public barcode libraries are exhaustive enough. In addition, the high genetic resolution obtained offers great potential. If standardized molecular operational taxonomic units (sensu Blaxter et al., 2005) are used for analysis, the effects of single or multiple stressors can be detected even for cryptic or undescribed species (Beermann et al., 2018;Macher et al., 2016).

| CONCLUSIONS
Analyses of eDNA are of increasing importance and relevance for freshwater monitoring and conservation. To date, there is no universally applicable methodology for freshwater eDNA techniques. Many taxonomic groups and large geographical areas have not yet been investigated or assayed with eDNA techniques. Nevertheless, in conservation, biodiversity assessment, and management programmes, the application of 'traditional' genetic methods is already widely accepted, and genetic information is explicitly included in conservation decision making (Geist, 2011), as suggested by the Convention on Biological Diversity, Article 2 (United Nations, 1992). The increasing development and recent technical progress in the field of eDNA research is likely to lead to the increasing use of eDNA techniques in biomonitoring, invasion biology, and aquatic conservation.
Harmonization of eDNA techniques and data analyses are needed if this method is to be applied routinely and reproducibly for the different taxonomic groups in different geographical regions, and in the various habitats encountered in freshwater environments worldwide.
This becomes most important if eDNA methods are integrated within existing legal monitoring frameworks. In regions that are considered biodiversity hotspots and where rapid habitat change and biodiversity loss are likely, biological monitoring may benefit from using eDNA approaches as an initial and comparatively rapid step for determining species occurrences and distributions instead of laborious and expensive classical monitoring (Bálint et al., 2018). To help with this, the development of new European standards produced by the European