The current state of eDNA research in freshwater ecosystems: are we shifting from the developmental phase to standard application in biomonitoring?

Environmental DNA (eDNA) has seen a massive increase in application in freshwater systems with a concurrent growth in protocol developments and a drive to gain a better understanding of the ‘ecology’ of eDNA. This raises the question of whether we are currently still in an early, developmental phase of eDNA-based assessments or already transitioning into a more applied stage for biomonitoring. I conducted a systematic literature review on 381 eDNA-focused studies in freshwater systems targeting macro-organisms over the last 5 years, assessing study goals, methods, target systems and taxa and study design aspects. The results show an increase of biomonitoring-focused studies throughout the years, while the fraction of studies investigating the ‘ecology’ of eDNA decreased. The application of metabarcoding significantly increased while studies applying qPCRs tentatively declined. A geographic inequality was observed concerning study numbers and study goals biased towards the global North. Descriptive studies increased, but the fraction of in-field studies and studies applying eDNA and conventional methods combined revealed no trend. These results show a shift towards application-focused work for eDNA-based assessments but also reveal this field to still be developing. In this transitional phase, practitioners need to ensure consistency and data comparability for long-term monitoring programmes.


Introduction
In the ongoing global biodiversity crisis, freshwater habitats have been classified as one of the most vulnerable ecosystems (Dudgeon et al., 2006;WWF, 2020). In order to halt the decline in biodiversity and ecosystem functioning of these habitats, various legislative frameworks, such as the EU Water Framework Directive (WFD), the Clean Water Act of the USA (CWA) or the Water Act 2007 of Australia have been installed. Hereby, all these legislative bodies require solid biodiversity assessment methods in order to ensure the biological integrity of the protected systems (Birk et al., 2012;Bohan et al., 2017;Leese et al., 2018).
The utilization of environmental DNA (or eDNA) has been announced to revolutionize the way we Abstract Environmental DNA (eDNA) has seen a massive increase in application in freshwater systems with a concurrent growth in protocol developments and a drive to gain a better understanding of the 'ecology' of eDNA. This raises the question of whether we are currently still in an early, developmental phase of eDNA-based assessments or already transitioning into a more applied stage for biomonitoring. I conducted a systematic literature review on 381 eDNA-focused studies in freshwater systems targeting macro-organisms over the last 5 years, assessing study goals, methods, target systems and taxa and study design aspects. The results show an increase of biomonitoring-focused studies throughout the years, while the fraction of studies investigating the 'ecology' of eDNA decreased. The application of metabarcoding significantly increased while studies applying qPCRs tentatively declined. A geographic inequality was observed concerning study numbers and study goals biased towards the global North. Descriptive 1 3 Vol:. (1234567890) monitor biodiversity, particularly in aquatic systems (Deiner et al., 2017;Cristescu & Hebert, 2018;Pawlowski et al., 2018;Ruppert et al., 2019;Beng & Corlett, 2020). Environmental DNA is defined as DNA that is captured from an environmental sample without pre-isolating any target organisms . The sample can be anything from water, soil, sediments, air, faeces or gut contents or even ancient sources like cores from ice, sediments or permafrost Thomsen & Willerslev, 2015). With respect to macrobial organisms, this DNA captured from the environment stems from the organisms interacting with their surroundings and permanently shedding genetic material into the environment through, e.g. sloughing of skin cells, hairs or scales, excretion of body fluids or defecation (eDNA sensu stricto Pawlowski et al., 2020).
For freshwater assessments, the striking advantages of eDNA-based methods are among others (1) non-invasiveness-causing little to no disturbance to the target species and target system (when targeting macro-organisms and when sampling eDNA sensu strictu, see above), (2) time and cost efficiency allowing to monitor biodiversity over large spatial and temporal scales, (3) high detection sensitivity, allowing reliable detection of rare or elusive species or early stages of biological invasions and (4) universality of sampling, which allows to assess biodiversity on a wide taxonomic range using the same eDNA sample (Goldberg et al., 2016;Deiner et al., 2017;Shaw et al., 2017;Pawlowski et al., 2018;Beng & Corlett, 2020).
In the case of water samples, eDNA-based approaches first capture and concentrate this shed DNA from the water via filtration, centrifugation or precipitation and then purify the DNA from this concentrate through DNA extraction (Shaw et al., 2017;Ruppert et al., 2019). The next step in the workflow selectively targets the DNA of the investigator's taxon (or taxon group) of interest. The taxonomic level can hereby range from individual species to whole domains or even 'total biodiversity' (Compson et al., 2020) and also dictates the specific method applied for the detection of the target DNA in the 'mixture' of an eDNA sample: The first of the two main avenues in this regard encompasses single-species or 'targeted' eDNA approaches. These aim to detect DNA of individual (often threatened or invasive) species. A widely applied method for this targeted approach is the usage of real-time or quantitative polymerase chain reaction (qPCR). Hereby, species-specific primers (and sometimes hydrolysis probes) that can be fluorescently labelled are used during PCR amplification. The presence of the target DNA in the eDNA extract is inferred by an increase of the fluorescence signal during PCR (Logan et al., 2009;Doi et al., 2015). In the last few years, digital droplet PCR (ddPCR) has emerged as an alternative targeted DNA detection method. It generates thousands of nanoliter droplets during PCR, ideally each of them containing one or no copy of the target DNA. PCR is carried out in each droplet in parallel using fluorescently labelled probes and the presence of the target PCR is inferred by an end-point measure of the fluorescence signal in all droplets in parallel (Hindson et al., 2011;Doi et al., 2015). ddPCR seems to surpass qPCR concerning sensitivity, inhibitor resistance and DNA quantification (Doi et al., 2015;Hunter et al., 2017), but currently requires higher acquisition and running costs than qPCR. The second avenue encompasses community approaches. Hereby, metabarcoding is the most commonly applied method if multiple species, whole species groups or even complete biodiversity is targeted in an eDNA survey. Hereby, general or universal primers are used to amplify a target marker of the target taxon (group) from the mix of DNA in the eDNA extract, followed by a high-throughput sequencing (HTS) step. The obtained sequences are then usually compared to reference sequences in order to identify the species composition of the target group in the eDNA extract Shaw et al., 2017;Ruppert et al., 2019). Other methods to characterize the biodiversity contained in an eDNA sample, though less frequently applied, include conventional PCRs (e.g. Wilson et al. 2018;Xu et al. 2018;Cardas et al. 2020), CRISPR-Cas (Williams et al., 2019a), hybridization capture (e.g. Wilcox et al., 2018) and metagenomics (e.g. Bista et al., 2018;Cowart et al., 2018).
The compelling assets of eDNA-based approaches have led to an intense drive to add eDNA-based methods to the standard toolbox for ecosystem monitoring (particularly for aquatic systems) and to include them in the monitoring schemes of the abovementioned legislative bodies (e.g. Leese et al., 2016;Deiner et al., 2017;Hering et al., 2018) but also to continuous research on how to improve workflows and correctly interpret the data. This resulted in an impressive growth of scientific publications on eDNA-based applications in ecology and conservation throughout the last two decades (Beng & Corlett, 2020;Compson et al., 2020). Efforts to improve the workflow can focus on eDNA sampling and sample preservation (Hunter et al., 2019;Ladell et al., 2019;Kumar et al., 2021), DNA extraction (Wong et al., 2020;Anderson & Thompson, 2021), the development of new primers or optimization of laboratory protocols for species-specific assays (e.g. Elberri et al., 2020;Schenekar et al., 2020;Ito and Shibaike, 2021) or for species community assays (Miya et al., 2015;Elbrecht & Leese, 2017) or the development of new software or pipelines for sequence data processing (Sato et al., 2018;Carøe & Bohmann, 2020;Diana et al., 2021;Macher et al., 2021). Advances that have been made in the last decade in the field of waterborne eDNA-based assessments include among others the comparison of differing capture methods of eDNA, with filtering slowly emerging as the most efficient method for many research questions , gaining a better understanding of the correlation between sampling effort and DNA recovered, as well as detection probabilities (e.g. Cantera et al., 2019;Furlan et al., 2019), how to identify and deal with inhibited samples (e.g. Bruce et al., 2021) and the development of suitable metabarcoding primers for important bioindicator groups such as diatoms, macroinvertebrates or fishes (e.g. Miya et al., 2015;Elbrecht & Leese, 2017;Vasselon et al., 2017).
eDNA-based assessments have already proven to be effective to answer various ecological questions in freshwater habitats. Such include assessing the presence of elusive, endangered or rare species, like cryptic macroinvertebrates, high-risk fish species or amphibians like the great crested newt Triturus cristatus (Mächler et al., 2014;Biggs et al., 2015;Doi et al., 2017;Roy et al., 2018;Griffiths et al., 2020), monitoring the spread of invasive species in aquatic systems like Asian carp, Dreissenid mussels or the American bullfrog Lithobates catesbeianus (Dejean et al., 2012;Jerde et al., 2013;Ardura et al., 2017), studying spawning ecology of elusive species (Sigsgaard et al., 2015;Maruyama et al., 2018) and species richness assessments of bioindicator taxa Deiner et al., 2017).
While for various taxa and habitats, robust and reliable protocols have been developed and have been proven to be equal or even superior to conventional monitoring (see examples above), there are still aspects in the process of eDNA analyses that are not yet fully understood and for certain target taxa or habitats more research and/or protocol development might be needed before we can include these approaches reliably into monitoring schemes.
Areas requiring further research and development (regarding freshwater eDNA analyses) include standardized sampling protocols for habitats with very differing environmental conditions (such as springs, small creeks, large streams, ponds, lakes, swamps or estuaries) as well as the definition of sampling controls to keep the data among studies consistent and comparable . Concerning laboratory analyses, open questions on remain concerning PCR inhibition, PCR stochasticity or primer bias . Furthermore, the uneven availability of published protocols and/or primer sets for different taxa might pose a significant challenge for eDNA-based assays to be fully integrated into standard biomonitoring schemes (Belle et al., 2019). Welldesigned protocols, reported with defined minimum standards, are needed for every bioindicator group or target taxon of interest for freshwater biomonitoring to enable the full implementation of eDNA-based methods into comprehensive biomonitoring frameworks. Thereby, the standardized minimum standards are required to understand the applicability of the respective assays for certain research questions or biodiversity assessments, as well as the associated uncertainties and limitations (such as detection sensitivities) and comparability with others . Analytical aspects like the incorporation of imperfect detection models (such as occupancy modelling), abundance estimations, taxonomic assignment and the assignment of an ecological status are also still a very active field of research and will require harmonization (Beng & Corlett, 2020) with the aim to improve consistency and comparability among surveys. For metabarcoding, the availability of well-curated reference databases is a fundamental prerequisite. The efforts to achieve this are tightly interwoven with the global barcoding initiative of the international Barcode of Life project (iBOL). Inequalities in barcode availability have been identified and researchers are actively working on closing those gaps (Weigand et al., 2019).
A topic that has become a major focus of investigators in the field of eDNA-based research in the recent years, is the dynamics of eDNA production, distribution, persistence and degradation in the respective study system (Turner et al., 2015;Eichmiller et al., 2016;Tsuji et al., 2017;Allan et al., 2020;Wood et al., 2020). These dynamics, also referred to as the 'ecology' of eDNA (Barnes & Turner, 2016) are of decisive importance for estimating power, sensitivity as well as the spatiotemporal scale of the drawn inferences from any eDNA study. For example, depending on various environmental factors, such as temperature, pH, UV radiation or availability of oxygen, eDNA in aquatic systems can degrade within days to the point of non-detectability (Barnes et al., 2014;Barnes & Turner, 2016;Seymour et al., 2018) or persist up to hundreds or even thousands of years in an analyzable state in sediments of freshwater systems (e.g. Giguet-Covex et al., 2019). Likewise, eDNA in lotic systems might be transported downstream only several hundred metres (Jane et al., 2015;Wilcox et al., 2016) or up to dozens of kilometres Pont et al., 2018). The environmental factors driving these dynamics show great variability among systems, requiring habitat-specific assessments (Harper et al., 2019;Sales et al., 2021), promoting the increase of studies targeting at a better understanding of eDNA 'ecology'.
Because of the rapid development of eDNA-based assessments in the last two decades, with certain research questions being answered, but new important aspects requiring more research being uncovered, eDNA-based methods are still developing and our understanding of how to optimally conduct and interpret the results of such studies is still progressing. Publications on new or modified protocols in the whole workflow still seem to be published at an equal speed as the application of previously published protocols. However, in order to implement a novel methodology in national or international biomonitoring scheme, the methodology needs to be clearly defined. Furthermore, this workflow has to be proven as effective and robust without the need to adapt it frequently, as protocol modification might result in loss of valuable data comparability among different assessments or in long-term biomonitoring studies. Therefore, the concurrent and ongoing expansion in eDNA application and advancements in the optimal execution of such studies raises the question of whether we are still at a very early, developmental phase of eDNA-based monitoring with the majority of the research still focusing on methodological advancements and our fundamental understanding of eDNA dynamics, or whether eDNA-based methods are already transforming into a standard tool for freshwater biomonitoring with increased true application of such methods for biodiversity assessments.
The aim of this study was to assess the current state of eDNA-based research on macro-organisms in freshwater ecosystem, with a special emphasis on the main goals of the individual studies reviewed. In particular, I wanted to evaluate whether a relative shift towards true application of eDNA-based methods as a biodiversity monitoring tool can be observed in the most recent years. Furthermore, an overview of the relative frequencies of target taxa and target freshwater systems in such studies is presented and selected aspects of study designs illuminated.

Methods
A literature search was conducted through the Clarivate Web of Science portal on October 30 th , 2021. The following search terms and search fields were thereby used: eDNA OR 'environmental DNA' (All fields) and freshwater OR river$ OR pond$ OR lake$ (All fields) and English (Language) and Article (Document Type). The search was limited to the years 2017 or newer by setting the search field (Publication Date) to Last 5 years (2017-2021). A second search was conducted on November 21 st , 2021, using the same search terms but not limiting the search field (Publication Date), allowing to return results from all years of the database (dating back to the year 1900). From this second search, only the number of publications per year was recorded.
The search results of the first search were exported from Web of Science and imported into Mendeley Reference Manager 1.19.8 (Mendeley Ltd.), whereby the following metadata of the articles were kept: article title, author list, publication year and month, journal, volume, issue, start and end page (or article number), DOI (if available) and abstract. The titles and abstracts were screened to remove out-of-scope documents. Thereby, the following exclusion criteria were used: (1) Articles focusing solely on marine or terrestrial ecosystems and not targeting freshwater habitats. Articles targeting estuarine habitats were kept for analysis, (2) articles solely focusing on microorganisms such as bacteria, viruses, pathogens, protozoans, prokaryotes or fungi, (3) articles focusing solely on ancient DNA in lake sediments, (4) articles utilizing eDNA from freshwater sampling to infer terrestrial biodiversity only, (5) articles focusing solely on diet analysis, (6) articles focusing solely on barcode generation and (7) articles such as reviews, opinion articles or conference papers.
For publications passing this filter, the full articles were downloaded. If the full article could not be accessed, the article was removed from analysis. The remaining articles were subjected to full analysis and the following information was extracted from each publication: Main goal of the study: The primary purpose for conducting the work, assigned to the following three categories (multiple categories were assigned to a study if it targeted more than one goal with equal emphasis): (1) 'Biodiversity monitoring'-studies collecting field samples from systems for which the presence of the target taxon was unknown. Studies applying metabarcoding to field samples were always assigned to this category. (2) 'Development or optimization of protocols'. Studies developing new or optimizing existing protocols in the complete workflow of an eDNA assessment, including sampling, sample preservation, sample processing and data analysis.
(3) 'Understanding the "ecology" of eDNA'-studies aiming to elucidate the dynamics of eDNA production, distribution, transport, fate and/or degradation of eDNA in the target system.
Target taxon: The species or species group being the main target of investigation in the study. If the target taxon was not unambiguously defined in the publication, the targeted taxa of the utilized primers were noted. The target taxa were then grouped into the following 13 categories: 'fishes' (including lampreys), 'amphibians', 'crustaceans', 'reptiles', 'mammals', 'molluscs', 'macroinvertebrates', 'macrophytes', 'microalgae', 'vertebrates', 'multiple' (of the previously listed), 'metazoa or higher' and 'other'. While these groups are at different taxonomic levels, they were chosen to obtain an as homogeneous and consistent classification as possible.
Target system: the freshwater system is targeted by each study. Usually the system from which the field samples originated. For laboratory-based studies, this was the named target system, if explicitly stated. The target systems were then grouped into the following 10 categories: 'rivers', 'lakes/ponds', 'lakes/ponds and rivers combined', 'artificial channels', 'dam reservoirs', 'estuaries', 'groundwater', 'wetlands', 'multiple' (any combination of the previously mentioned categories except lakes/ponds and rivers combined') and 'other'.
Country: The country of origin from which field samples were taken. If study was solely laboratory based without utilizing field samples, the country of the first author(s)' affiliation was listed. Countries were then grouped into the following seven continental assignments: 'Africa', 'Asia', 'Australia', 'Europe', 'North America', 'South America', 'multiple' (of the previously listed).
Study design aspects: (a) Study type: Studies were either categorized as 'manipulative' (the investigators actively and significantly altered the system of which the eDNA samples originated, with the aim to influence the presence of the target eDNA in this system), or 'descriptive' (no significant alterations of the system by the investigators). (b) Study location: Either categorized as 'in-field' (eDNA samples were collected from an uncontrolled environment, outside of a laboratory setting) or 'solely laboratory-based or theoretical' (eDNA samples were collected from a controlled setting or no eDNA samples were collected at all). If the study was split into two clearly separated sub-studies with different designs, both study types could apply to one study. (c) eDNA versus conventional: For each study targeting biodiversity monitoring, it was listed whether biodiversity monitoring was carried solely using eDNA-based approaches or in combination with alternative or conventional methods.
All data were compiled into a database in Microsoft Excel (Online Resource 1). Statistical analyses and plotting were carried out using R 4.1.2 (R Core Team, 2021) and IBM SPSS v23 (IBM Corp, 2016). Exponential curve fitting of the yearly publication trend was carried out using the nls() function of R and linear modelling to reveal temporal trends of study aspects via the lm() function. Chi-squared tests of homogeneity of methods applied among study goals as well as of study targets and study systems among years were conducted with the chisq.test() function. Figures were finalized using CorelDRAW 2019 (Corel Corporation 2019).

Yearly number of publications
The search encompassing the years 1900-2021 returned 1343 search results. Thereby, a steep increase in publications between 2008 and 2020 can be observed, fitting an exponential curve (Fig. 1). The highest number of studies was published in the year 2020 (221 studies), followed by the year 2021 (196 studies), which was not yet fully completed by the time of the search (November 2021) and 2018 (170 studies).
The search limited to the last 5 years returned 876 search results. After screening title and abstract of these 876 publications, 431 (49.2%) out-of-scope articles were removed and 445 (50.7%) were retained for which full articles were downloaded. A total of 11 articles were removed from analysis because the full article was not accessible. During data extraction, an additional 64 were excluded from analysis because they did not meet the inclusion criteria, resulting in a final dataset of 381 publications (43.5% of initial search results). Thereby, the yearly number of publications steadily increased starting at 51 (year 2017) and ending at 108 (year 2021) (Fig. 1).

Main goals of studies
Overall, 247 of the 381 studies in this meta-analysis (64.8%) featured 'biodiversity monitoring' either as the sole study goal or in combination with other goals. A total of 220 studies (57.7%) included 'development or optimization of a protocol' as a study goal and 76 studies (19.9%) aimed to increase the 'understanding of the "ecology" of eDNA'. Hereby, an increase in percentage of studies targeting 'biodiversity monitoring' was observed between 2017 and 2021 (F-statistic = 39.8, P = 0.008), whereas studies targeting a 'understanding of the "ecology" of eDNA' decreased in percentage (F-statistic = 10.5, P = 0.047). No clear trend regarding the percentage of studies targeting 'protocol development or  (Table 1, Fig. 2).
Geographically, a clear separation could be seen between the global North, namely North America, Europe and Asia, and the global South, namely South America, Africa and Australia (Fig. 3). A total of 337 studies (88.5%) of the 381 assayed studies were conducted in the global North (North America: 136 studies-35.7%, Europe: 110 studies-28.9%, Asia: 91 studies-23.9%) with only a small fraction of the studies (41 studies-10.8%)-targeting systems in the global South (South America: 16 studies-4.2%, Africa: 4 studies-1.0%, Australia 21 studies-5.5%). Three studies (0.8%) were conducted on multiple continents. Within Asia, 19 studies were conducted in China and 58 studies were conducted in Japan; therefore, 84.6% of the studies in Asia were conducted in these two countries (Online Resource

Methods applied
Across all 381 studies, qPCRs were the most frequently applied method throughout the 381 studies (210 studies-55.1%), followed by metabarcoding (139 studies-36.5%), ddPCR (18 studies-4.7%) and other methods (34 studies-8.9%). However, the proportion of studies applying metabarcoding significantly increased throughout the last 5 years (Fig. 4, F-statistic = 11.3, P = 0.0437), whereas the application of qPCR showed a slight trend of decrease (F-statistic = 9.23, P = 0.056). No trend was observed concerning the proportion of studies applying ddPCR or other methods (F-statistic = 2.1 and 3.7, respectively, Biodiversity monitoring was the most frequently addressed study goal in seven out of the 10 target system categories, whereas 'protocol development or optimization' was the most-frequent study goal in three of these categories. 'Understanding the "ecology" of eDNA' was the least frequently listed  study goal for all systems, with three system categories not addressing this goal at all (Fig. 2b, Online Resource 3).

Study design aspects
On average, the vast majority of the surveyed studies were designed in a descriptive manner (334 studies-87.7%) whereas 62 studies (16.3%) had a manipulative design. There was a weak tendency towards an increase of the proportion of descriptive studies throughout the years (Fig. 7a, F-statistic = 6.59, P = 0.083) as well as a decrease in the proportion of manipulative studies (F-statistic = 5.92, P = 0.093).
Most of the 381 studies (344 studies-90.3%) utilized in-field samples whereas only a small fraction (37 studies-9.7%) used samples from a controlled environment or were of purely theoretical nature. No temporal trend could be observed concerning those two proportions (F-statistic = 0.587 and 1.028, respectively, both P > 0.1, Fig. 7b). Finally, out of the 248 studies targeting biodiversity monitoring, 131 studies (52.8%) applied only eDNA-based methods, whereas 117 studies (47.2%) applied both, eDNA-based and conventional or alternative methods in parallel. Again, no temporal trend could be observed concerning those two proportions (both F-statistic = 0.618, both P > 0.1, Fig. 7c).

Discussion
The number of publications meeting the search criteria increased throughout the last 5 years reflecting the well-documented ongoing growth in the field of eDNA research (e.g. Compson et al., 2020;Beng and Corlett, 2020;Mathieu et al., 2020). 'Biodiversity monitoring' was the most-frequent study goal in the publications reviewed. The relative frequency of studies targeting this goal significantly increased throughout the years surveyed, potentially indicating a shift towards more true application of eDNA-based methods in freshwater monitoring. In parallel, the absolute number of studies featuring 'understanding the "ecology" of eDNA' as a study goal remained relatively constant, causing a significant decrease in the relative frequency of this goal. Given that this aspect of eDNA research has gained interest only recently Temporal trends concerning study design aspects in the surveyed studies. Dots represent percentage of total studies in the respective publication year. Lines represent regression curves from linear modelling and symbols to the right-hand side of the lines indicate direction and significance of respective slope coefficients (+ : positive, − : negative, n.s.: not significant, P: P value). Shadings indicate 95% confidence inter-vals of linear models. a Descriptive versus manipulative study design, b samples stem from a laboratory setting versus samples stem from in-field sampling, c within 'biodiversity monitoring' studies only: Methods applied were solely eDNA-based versus a combined approach also including conventional methods and has been advocated as an important issue to be addressed in eDNA-based assessments (Barnes & Turner, 2016;Harrison et al., 2019;Jo et al., 2021a), this decrease should not be interpreted as a loss of interest in this facet of eDNA research but potentially as a risk of this aspect falling behind in the fast-growing field of eDNA-based assessments. Jo et al. (2021b) not only describe an increase of studies investigating eDNA dynamics in their literature review but also highlight the low frequency of such studies compared to eDNA studies in general. As the application of eDNA-based methods continue to expand also into new study systems, the lag of studies investigating eDNA dynamics in these systems might hamper our ability to draw meaningful conclusions in such systems. Therefore, active research on this aspect is still highly needed. Finally, studies targeting 'protocol development or optimization' grew at a similar rate as the total number of publications in general, revealing a constant relative rate of methodological advancements. However, the proportion of studies incorporating protocol development varied greatly among target taxa and study systems (see Online Resource 3). More than half of the studies featuring 'protocol development or optimization' (119 studies-56.7%) combined these efforts with a biodiversity monitoring assay, very often in the form of developing a taxon-specific qPCR protocol and then applying this protocol in a survey to detect the taxon in the field. As the application of eDNA-based methods is still expanding across the tree of life, this trend is expected to continue in the future.
The geographic bias towards the global North of the returned studies of the literature search has been described in similar contexts before (Belle et al., 2019;Coble et al., 2019;Weigand et al., 2019) and was suggested to result from socioeconomic constraints and potentially the lack of the required infrastructure complicating eDNA implementation in global South countries (Belle et al., 2019). Additionally, this study revealed that studies targeting 'protocol development or optimization' or 'understanding the "ecology" of eDNA' are even more severely underrepresented in the global South or tropics. As mentioned before, the dynamics of eDNA might differ drastically among study systems and across climatic zones, and therefore, target-system-specific assessments are required (Harper et al., 2019;Sales et al., 2021). Such studies are more likely to be financed by fundamental research funds, and therefore, a bias for such research towards well-funded research centres in temperate climates in the global North may have evolved. However, it is essential to steer such research also towards yet understudied aquatic systems to better support biodiversity monitoring via eDNA-based assessments in these systems.
Among all eDNA-based approaches, metabarcoding has been particularly emphasized for its immense potential in biodiversity monitoring (Deiner et al., 2017;Ruppert et al., 2019;Compson et al., 2020) and the significant increase of studies utilizing metabarcoding described in this review might already reflect this. Furthermore, HTS costs continue to decrease (Hall, 2013;Wetterstrand, 2021) and new, smaller HTS sequencing platforms (such as the illumina iSeq™ or MiniSeq™ or Oxford Nanopore's MinION) are being developed that could furthermore reduce total costs per survey for end-users. These advancements are likely to push metabarcoding for application in this field additionally.
In parallel, the application of qPCR showed a slight trend of decrease in frequency. Whether this tendency will get statistically better supported in the upcoming years needs to be evaluated. A potential increase in ddPCR application (e.g. if acquisition and running costs become competitive to those of qPCR) might restrain qPCR application in the future. However, in this study, the application of ddPCR did not reveal a statistically significant increase in frequency, yet. Other methods that are just being explored for application in eDNA-based research, such as CRISPR-Cas or metagenomics might need more time to be picked up for true application in biodiversity monitoring, but initial trials reveal already exciting prospects (Won et al., 2017;Garlapati et al., 2019;Williams et al., 2019b;Schroeter et al., 2020;Duran-Vinet et al., 2021).
The relative frequency of methods applied clearly varied among study goals. Metabarcoding was most frequently applied for 'biodiversity monitoring', probably once again underpinning its particular suitability for this goal. In contrast to that, qPCRs were more frequently applied in studies targeting 'protocol development or optimization' and 'understanding the "ecology" of eDNA'. The former simply results from qPCRs protocols being mostly species specific, requiring the development of more protocols, in comparison to metabarcoding protocols, which are designed for whole species communities. Studies investigating the 'ecology' of eDNA often require quantification of DNA. Absolute quantification of target DNA in an eDNA sample still remains a challenging task in eDNA research. However, qPCR and ddPCR approaches are emerging to have a higher potential for reliable quantification, as metabarcoding has additional method-inherent complications for such an estimate Sassoubre et al., 2016;Valentini et al., 2016;Deiner et al., 2017;Thalinger et al., 2018). The higher reliability of ddPCR for quantification might also be reflected in the relatively high frequency of this method in studies targeting 'understanding of the "ecology" of eDNA' compared to the other two study goals.
In this review, fishes were by far the most-frequent study taxa (being targeted by approximately half of the 381 studies surveyed), as described also in other meta-analyses (Belle et al., 2019;Coble et al., 2019). Drivers for this pattern, being observed also in freshwater research and conservation in general (Di Marco et al., 2017;Pérez-Espona, 2017), are undoubtedly socioeconomic interests but also numerous fish species either representing iconic flagship species, globally invasive species or keystone species for ecosystem service (Geist, 2015;Belle et al., 2019). The economic interests have potentially provided extra fuel for the rapid development of protocols targeting fishes (both, for metabarcoding and also speciesspecific assays) leading to more available protocols to date that allow a large application of such assays in biomonitoring studies. Furthermore, abundance estimations of fishes are often key parameters for many ecological indices calculated for freshwater habitats. This presumably led to an additional effort to improve protocols for density or abundance estimations of fishes.
The second most-frequent target taxon was molluscs. For these, the study motivation was either rooted in their susceptibility to habitat deterioration (e.g. Carlsson et al., 2017;Rodgers et al., 2020;Stoeckle et al., 2021) or in their role in ongoing biological invasions (e.g. De Ventura et al., 2017;Pie et al., 2017;Trebitz et al., 2019). The genus Dreissena, encompassing the highly invasive zebra and quagga mussels, was the most frequently targeted genus within molluscs (16 studies-37.2%) highlighting the effort to monitor and detect these invasive neobiota in many systems.
Regarding study systems, there was a clear domination of lotic systems, with rivers being the targeted system in approximately 50% of the surveyed studies. This predominance mirrors the general pattern of riverine systems being more frequently targeted in freshwater research or conservation in general (Wurtsbaugh et al., 2014).
It shall be noted that the trends described in this review are limited to the years 2017-2021, and therefore, previous research on specific taxa or systems (such as protocols being developed already earlier) might have influenced this pattern. However, other reviews encompassing larger timeframes (Beng & Corlett, 2020;Compson et al., 2020) show agreeing patterns in comparable aspects regarding target taxa and systems suggestion similar long-term trends. Fishes continue to be the most researched taxon group. This is expected to continue in the future due socioeconomic interets but also due to their key role in many monitoring schemes. However, as eDNAbased approaches do seem to be expanding more broadly across the tree of life, we might see a flattening of that monopoly in the future (which would be a desireable trend, see below).
The tendency towards more descriptive study designs throughout the last 5 years might also indicate a shift towards more frequent true application of eDNA-based studies is aquatic biomonitoring. Since biodiversity assessments are usually of descriptive design, this would be an expected concurring trend with an increase of studies featuring 'biodiversity monitoring' as a study goal. Interestingly, the frequency of purely laboratory-based or theoretical studies did not significantly decrease which would also be expected in line with this trend.
Among the studies featuring 'biodivesity monitoring' as their study goal, the fraction of studies applying both, eDNA-based and conventional methods did not change over time, remaining at around 50% of the studies. This might suggest a relative high frequency of 'proof-of-concept' studies that verify results retrieved by eDNA methodology using conventional methods (such as electrofishing or kicknet sampling and morphological identification). However, as been highlighted previously, eDNA-based or molecular tools in general are not intended to completely replace established or conventional methods but to complement them Harper et al., 2019;Ruppert et al., 2019;Beng & Corlett, 2020). Both, molecular and conventional methods can elucidate different aspects of biodiversity that the other cannot. For example, eDNA can have higher detection rates for invasive species in biological invasions (Dejean et al., 2012;Bylemans et al., 2016), but so far eDNA-based methods are not able to give meaningful information about important demographic parameters of a population, such as age structure (Stoeckle et al., 2016;Pont et al., 2021). Therefore, a parallel application of molecular and conventional methods for biodiversity assessments is expected to continue also in the future.

Conclusion
This systematic literature review confirmed the ongoing rapid growth of eDNA-focused literature in freshwaters systems. Furthermore, the relative frequency of true application of eDNA-based methods for biodiversity monitoring did increase significantly over the last 5 years, potentially indicating a current phase of transition towards these methods becoming a standard tool for biodiversity monitoring. As fishes and riverine systems are the most frequently target taxa and systems, respectively, it is expected that true application of eDNA-based methods will first happen on a large scale for these systems with the other taxa and systems following after them.
A remaining challenge for broad implementation of these methods into legislative frameworks will be the need of standardization among methods, reliable abundance estimations and incorporation of demographic parameters in such assessments, all of which are currently being worked on actively (Goldberg et al., 2016;Pawlowski et al., 2018;Shu et al., 2020;Bohmann et al., 2021).
A critical pattern revealed in this study is the great variability in research efforts directed at different taxonomic groups, with fishes standing out as the main group of interest. However, for holistic biodiversity assessments, all target taxa need to be detected with equal certainty, and therefore, equal effort should be put into the development of methodology and detection assays. Thus, targeted promotion of the research into less tested, but often more biodiverse bioindicator groups is highly needed.
Finally, resilience is important for any monitoring approach. As shown, eDNA protocol development is still ongoing in all target taxa and systems. There is, thus, a risk that protocols being currently adopted in biomonitoring schemes will potentially be proven inferior to newer protocols at a later stage. As mentioned above, any change in methodology during an ongoing biomonitoring assessment might compromise data comparability. Therefore, managers and researchers should ensure that the chosen protocol meets certain minimum standards required for their research question or monitoring goal and can, therefore, be maintained for longer timeframes.
Taken together, eDNA-based approaches are a powerful addition to the biomonitoring toolbox and their uptake for biodiversity assessments is evident and increasing. However, stakeholders should be aware that new insights are still being gained and advances are continually made. In this transitional phase, it is then particularly critical for practitioners to ensure data consistency and comparability for longterm and/or large-scale biomonitoring assessments.
Combined with the positive assets of established methods, eDNA-based methods will definitely boost our efficiency to monitor freshwater biodiversity, enabling managers to quickly react to negative trends in these systems and to ensure healthy and functioning freshwater habitats also in the future.
Data availability All data generated or analysed during this study are included in this published article and its supplementary information files.

Conflict of interest
There is no known conflict of interest with any organization concerning the content in this manuscript.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.