How to learn to stop worrying and love environmental DNA monitoring

Abstract

Environmental DNA is one of the most promising new tools in the aquatic biodiversity monitoring toolkit, with particular appeal for applications requiring assessment of target taxa at very low population densities. And yet there persists considerable anxiety within the management community regarding the appropriateness of environmental DNA monitoring for certain tasks and the degree to which environmental DNA methods can deliver information relevant to management needs. This brief perspective piece is an attempt to address that anxiety by offering some advice on how end-users might best approach these new technologies. I do not here review recent developments in environmental DNA science, but rather I explore ways in which managers and decision-makers might become more comfortable adopting environmental DNA tools—or choosing not to adopt them, should circumstances so dictate. I attempt to contextualize the central challenges associated with acceptance of environmental DNA detection by contrasting them with traditional “catch-and-look” approaches to biodiversity monitoring. These considerations lead me to recommend the cultivation of four “virtues,” attitudes that can be brought into engagement with environmental DNA surveillance technologies that I hope will increase the likelihood that those engagements will be positive and that the future development and application of environmental DNA tools will further the cause of wise management.

“Every revolutionary idea seems to evoke three stages of reaction. They may be summed up by the phrases: (1) It’s completely impossible. (2) It’s possible, but it’s not worth doing. (3) I said it was a good idea all along.”

― Arthur C. Clarke

What is environmental DNA, and what’s all the fuss about?

The term “environmental DNA” (eDNA) has been around for some time and has been used in multiple subfields of ecology and environmental science to refer generally to any DNA that is extracted from an environmental sample. Perhaps due to this extensive history, usage of the term is still multivalent, with different people using the term to mean different things in different contexts. For the purposes of this paper, I will define eDNA specifically as, DNA collected from an environmental sample without any attempt to isolate the organism(s) from which the DNA derives. This definition is consistent with much current usage (e.g. Cristescu and Hebert, 2018), and it is also specific enough to enable a clear discussion of the issues associated with adoption of eDNA in decision-making contexts. In keeping with the theme of this special issue, my discussion will focus primarily on aquatic habitats—although eDNA tools have been applied in other systems—and I will pay special attention to issues associated with biological invasions and biosecurity.

eDNA has laid claim to a very particular place in the aquatic biodiversity monitoring toolkit. Given its potential for extremely high sensitivity, eDNA is frequently offered as a solution to the problem of detecting populations at low densities. This feature is especially appealing to managers aiming to mitigate risks associated with invasive species. It is generally acknowledged that the return on investment associated with management action decreases dramatically throughout the course of invasion, such that a dollar spent on prevention of a new incursion delivers substantially higher return than a dollar spent on control of widespread and abundant invasive populations (Leung et al., 2002). As a result, this management community is ceaselessly seeking tools that allow detections of target species with higher sensitivity, with more rapid turnaround times, with lower costs, and in ways that enable easy deployment in a greater variety of habitats. Of course, these are attractive features of any biodiversity surveillance tool. Yet many of the traditional tools we employ for these tasks rely on what I will refer to as “catch-and-look” technologies. These technologies are often quite old (fishing nets, for example, have been around for literally thousands of years), very stable, and have the obvious benefit of allowing close morphological investigation of captured individuals. But the effort required to both capture a target organism and analyze it morphologically often renders these tools relatively ineffective at detecting rare individuals.

The detection of a target species without the need to capture any individual of that species thus clearly confers an enormous advantage. Many species, after all, are very hard to catch. They live in habitats that are difficult to access, or they exhibit annoying capture-avoidance behaviors. It is generally much easier, faster, and less expensive to obtain a sample of water than it is to catch a fish (or any other target, for that matter). eDNA approaches offer the additional benefit of decoupling taxonomic identification from direct morphological observation of individuals. Again, this is a significant advance given that many individuals lack the traits on which morphological species identifications are based (e.g. pre-adult forms of most species) and that certain groups still stubbornly refuse to provide sufficient morphological clues for taxonomic identification at useful resolution.

But these clear benefits of “capture-free” approaches such as eDNA also present significant challenges. To put it bluntly: If you’re not actually capturing the target organism, how do you know that it’s really there? This problem was brought into sharp focus by the case of Asian carps in the Great Lakes region of the United States. Fear that these introduced species might enter the Great Lakes and threaten its highly profitable sport-fishing industry prompted attempts to better delineate invasion fronts using eDNA. Surprisingly, eDNA was detected in areas where fish had never been found by traditional capture techniques, suggesting that the species were much closer to entering the Lakes than had been assumed (Jerde et al., 2011). This triggered considerable anxiety among managers tasked with preventing further advance of the invasion, and it also prompted a rather public cry from some stakeholders for researchers to “show us the carp” (Dahlman, 2010). In one dramatic instance, a 6 mile stretch of the Chicago Sanitary and Ship canal was flooded with 2,300 gallons of piscicide in response to detections of Asian carp eDNA. Among the tens of thousands of bodies scooped from the poisoned river was a single bighead carp (Shaper, 2009). Never was a group of ecologists so relieved to find a single dead fish.

This, I believe, is the central challenge for managers considering adoption of eDNA tools. Given that eDNA is a capture-free method, these managers need to know if it is possible to infer an underlying distribution of actual target organisms from a pattern of eDNA detections. Fundamentally, the question facing these managers is this: Can eDNA can be trusted to tell us where a target organism is?

I wouldn’t be writing this paper if I didn’t believe that there is strong and mounting evidence for an affirmative answer to this question. Nevertheless, I recognize that there are a host of caveats, uncertainties, and unknowns that must be attached to that answer. And I also appreciate that distrust of eDNA by the management community has, in some cases, been earned through missteps in early adoption of eDNA tools and imperfect communication of those caveats, uncertainties, and unknowns. My goal in what remains is not to provide a technical defense of eDNA methods or even to urge adoption of eDNA tools. Instead, I will offer my own perspective on how interested end-users might engage with eDNA tools in a way that, I hope, will facilitate informed and judicious decisions about adoption of these tools in appropriate management contexts.

The four virtues

I plan to frame this advice as an appeal for what I am calling “virtues.” Virtue may seem like an odd terminological preference in this context. It has an archaic ring to it, and its association with philosophical and religious traditions may strike a discordant note in a paper aimed at a scientific audience. But I’d argue that virtue can do quite a bit of work for us in the current setting; in fact, it carries precisely the meaning that I want to convey in providing these recommendations. A virtue is, firstly, a commendable quality or trait. If I am correct, the following represent qualities or traits that are most likely to facilitate positive engagements between potential end-users and eDNA tools—in my mind, a most commendable outcome. But virtue also has a secondary meaning: it reflects a capacity to act appropriately in a given circumstance, a meaning perhaps revealed most clearly in the phrase “by virtue of.” Thus, the virtues outlined below are meant to spur, or at least enable, action. Their cultivation should encourage end-users to overcome paralysis in the face of uncertainty regarding the usefulness of eDNA methods and should empower them to make decisions regarding the appropriate implementation of new monitoring tools.

Knowledge

I will begin with the most obvious virtue, knowledge. It probably goes without saying that engagement with eDNA technology should be approached with some fundamental working knowledge of how eDNA tools work, what their limitations are, and what they can offer the manager that is not already offered by other tools. This might seem a daunting task, since technological advances associated with eDNA monitoring have exploded in recent years. This literature aims primarily to address questions related to the ability of eDNA detection to provide actionable information on the presence of target organisms. These include, but are not limited to: Which genetic markers are most appropriate for particular end uses? What is the fate and transport of eDNA in various environments? What are the detection limits associated with eDNA, and how do they compare with those of alternative methods? What are potential sources of error and how can they be reduced, eliminated, or at least quantified? What are the best practices—from field sampling through data interpretation—to ensure accurate, robust, and reproducible results of surveillance efforts? What methods exist for translating patterns of eDNA detections into underlying distributions of target organisms, and how can they be improved? Additional complexity derives from the possibility of applying both species-specific and community-profiling approaches to analyzing eDNA samples, both of which are attended by their own sets of specific research challenges.

Keeping abreast of these developments is a trial even for active researchers. And at the risk of disappointing the reader I will make no attempt to address any such technical questions in this brief perspective piece. I do, however, wish to strike two notes of encouragement. First, despite the highly technical nature of some aspects of eDNA monitoring, I believe that a basic understanding of the science behind these approaches should be well within the grasp of even non-scientists. It is entirely possible for a potential end-user to rapidly acquire sufficient knowledge to understand why eDNA tools are powerful in some ways and limited in others. Fortunately for these end-users, we all live in an era during which scientists are keenly sensitive to the importance of communicating their work to partners and stakeholders with widely varying backgrounds and are far more cognizant of their responsibility to ensure that research products are utilized appropriately. This is perhaps reflected in the second point I wish to make, which is that there do exist a number of recent reviews that have admirably summarized technical advancements in the field of eDNA research, in a variety of contexts. I provide a completely non-comprehensive and nonsystematic selection of such reviews in Table 1, and I would direct the interested reader to these as a useful starting point. A number of these reviews, though not all of them, have even been written with nontechnical audiences in mind. The rapid emergence of this literature—several of these reviews have come across my desk in the few months that I have been busy preparing this manuscript—attests to the remarkable eagerness of scientists to critically assess advancements in this field and communicate that assessment to potential end-users.

Table 1.

Non-systematic and non-comprehensive collection of critical reviews on eDNA-related methods published in the past 5 years.

Author(s)	Year	Journal	Brief summary
(Alberdi et al., 2018)	2018	Methods in Ecology and Evolution	Analysis of methodological pipelines for metabarcoding (community profiling)
(Barnes and Turner, 2015)	2015	Conservation Genetics	Assessment of interactions between eDNA and the environment and influence on detection probability
(Bohmann et al., 2014)	2014	Trends in Ecology and Evolution	Very broad overview, survey of applications, sources of uncertainty and recommendations for future research/application
(Coble et al., 2019)	2019	Science of the Total Environment	Specific applications in forestry management, assessment of state of eDNA science with relevance to that field
(Cristescu and Hebert, 2018)	2018	Annual Review of Ecology, Evolution, and Systematics	Broad methodological review with focus on uncertainties in data interpretation
(Deiner et al., 2017)	2017	Molecular Ecology	Focus on metabarcoding (community assessment) using eDNA, recommendations for study design and methodological workflow from sampling to bioinformatics
(Diaz-Ferguson and Moyer, 2014)	2014	International Journal for Tropical Biology	Focus on history of applications in marine and freshwater environments, with comments on experimental design
(Evans and Lamberti, 2018)	2018	Fisheries Research	History of eDNA application in fisheries research, with overview of limitations on application in that field
(Freeland, 2017)	2017	Genome	Extensive assessment of issues surrounding choice of molecular markers and primer design, for either targeted detection or community profiling via metabarcoding
(Goldberg et al., 2016)	2016	Methods in Ecology and Evolution	Guidelines for eDNA quality control from sample collection and handling through interpretation of results
(Harper et al., 2018)	2018	Hydrobiologia	Methodological considerations specific to eDNA collection and analysis in ponds
(Lear et al., 2018)	2018	New Zealand Journal of Ecology	Extensive assessment of technical issues related to extraction, storage, and amplification of eDNA, with recommendations for best practices
(Rees et al., 2014)	2014	Journal of Applied Ecology	Survey of applications for aquatic populations with assessment of potential future applications
(Shaw et al., 2017)	2016	Marine and Freshwater Research	Focus on eDNA-based community profiling in aquatic biodiversity studies, aims to provide guidance on adoption by nongeneticists
(Xiong et al., 2016)	2016	Marine Biology	Methodological overview of technical challenges associated with community profiling methods in marine systems, focus on detection of invasive species

Self-Awareness

People adopt new technologies for a wide variety of reasons. Sometimes those technologies seem exciting, or they offer some vague promise of improvement. They might even promise the benefits of membership in an exclusive users’ club. Often these reasons can feel quite compelling. But I would caution those considering adoption of eDNA tools not to approach the decision the same way I approached the purchase of my first smartphone. eDNA may be the shiniest new tool in the biodiversity surveillance toolkit, but it is not necessarily the best tool, or even an especially useful tool, in all circumstances. Whereas knowledge entails understanding the tool itself, the virtue of self-awareness entails understanding the needs that the tool is meant to fill. What does the current monitoring toolkit look like? What are the limitations of that toolkit, what gaps need to be filled, and what are the desired characteristics of a tool meant to fill those gaps? If those new tools were to provide new information, what options exist to act on that information? If the answer to any of these questions is “I don’t know,” then it may be too soon to consider adoption of eDNA tools.

I would like to emphasize two points here. First, it is critically important for managers to recognize both the power and the limitations of their existing tools. Researchers developing eDNA tools will, at times, experience frustration with the scrutiny leveled at their method. This is not because they are opposed to scrutiny in principle (in fact they welcome it!), but because it seems that the same critical eye is not often trained on the traditional catch-and-look methods already in use. Managers rightfully demand information on specificity and sensitivity of eDNA methods and wish to understand uncertainties and errors associated with those methods. But are those same demands made of the traditional tools already being employed in decision-making contexts?

There is, in fact, a substantial body of literature identifying the limitations of morphological taxonomy in various biodiversity monitoring contexts. For example, bioassessments for water quality monitoring rely extensively on morphological identification of macroinvertebrate taxa. Nevertheless, a 2010 audit of an official European freshwater monitoring program found that greater than 30% of species level identifications differed between primary analysts and auditors, resulting in significant discrepancies in final ecological assessments (Haase et al., 2010). Errors in identification of these taxa are so widely recognized and unavoidable that error rates of 10–15% at the genus level (and considerably higher at the species level) are considered typical (Stribling, 2006). In one study, even after a detailed reconciliation process between primary and auditing taxonomists, taxonomic disagreement could only be reduced to 14% (Stribling et al., 2008). These error rates increase dramatically when taxonomists are asked to assign identities to specimens known to present challenges. For instance, an assessment of morphological assignments for larval fish revealed accuracy as low as 13.5% at the species level; even at the family level, only 80% of individuals could be accurately identified (Ko et al., 2013). These problems persist even in cases of well-studied and commercially important taxa (Puncher et al., 2015).

Recognition of these limitations, of course, is one of the driving pressures behind incorporation of molecular identification methods in various monitoring contexts (Packer et al., 2009; Pfrender et al., 2010). This is not to disparage traditional morphology, but to provide a simple reminder that eDNA methods are not uniquely prone to error. It is relatively easy to recognize the biases and errors associated with a novel technology, because that’s what everyone is doing—trying to figure out how it works, what it does well, what it does poorly. But it may be much harder to see the biases and errors associated with the tools that we use every day. Self-awareness demands critical evaluation of all tools, not just the new ones being considered but the ones that have been relied upon for years. Only with that awareness can the relative value of novel tools be fairly assessed.

The second point I want to make is about error. Error is unavoidable. Every surveillance tool has a nonzero rate of both false positive and false negative errors. A major part of self-awareness is understanding one’s attitude to such error. This “error tolerance profile” can be framed most easily as a straightforward question: What are you more afraid of, a false positive (you claim the target is there when it isn’t) or a false negative (you claim the target is absent when it’s there)? In the context of invasive species management, there is strong theoretical support for placing a premium on false negative avoidance. And yet decision-makers frequently express grave concern over a possible increase in false positive errors. This is problematic precisely because highly sensitive tools such as eDNA shift the likelihood of error away from false negatives and toward false positives. The error tolerance profile of managers is, of course, driven by perceived costs associated with public reaction to management decisions. In principle, it seems logical that the costs of missing a new invasion might outweigh the costs of false positive detections. But it is easy enough to imagine a public more outraged at the unnecessary closure of a popular water body (a closure that can be attributed unambiguously to the acts of the decision-maker) than at the failure to prevent a new invasion (which might be attributed more vaguely to various forces, some beyond our control).

What is critical, I think, is that decision-makers understand and are honest about their own error tolerance profile and the costs on which it is based. And it might be worth noting that a public well educated about the risk of various unwanted outcomes can be remarkably tolerant of false positive detections. This is well-established in the field of medicine. For instance, one survey of American women found surprisingly high acceptance of false positives in breast cancer screening: 37% of all women surveyed were willing to accept a false positive rate as high as 10,000 per life saved, and tolerance was even higher among women who had already experienced a false positive test result (Schwarz et al 2000). I am not aware of any similar surveys of attitudes toward false positive detections in the context of invasive species management. However, I would wager that public tolerance of false positives might be higher than imagined, particularly in places where that public is familiar with the costs associated with successful invasions.

Preparedness

The example of Asian carp eDNA detections in the Great Lakes region is well-known not only because it was one of the earliest examples of molecular surveillance influencing management, but also because the incorporation of eDNA into the carp monitoring toolkit was a very public, complex, and rather painful process. There are plenty of lessons to be learned from this high-profile example, but one of the most important is that embarking on eDNA monitoring without a clear plan for interpreting and utilizing eDNA detection data can only lead to great unhappiness. Being pioneers in this field, managers involved with early eDNA monitoring for Asian carp were at a considerable disadvantage. They had little way of knowing how challenging it could be to translate a pattern of positive detections into a straightforward conclusion on whether the target species was present. And though it was clear from the start that eDNA could add value, it wasn’t clear exactly how the tool should best be utilized, or how the results of eDNA monitoring should influence decisions.

With the benefit of hindsight and armed with lessons learned from the Asian carp situation and others, today’s aquatic resource managers should be better prepared. The term “preparedness” connotes readiness in the face of potential disaster, and while this may be somewhat overstating the case the term nevertheless speaks to the unpleasantness that may face decision-makers if eDNA monitoring is conducted without prior agreement among stakeholders regarding how to understand and utilize detection data. Of central importance is development of a plan for interpreting genetic monitoring data. Given that eDNA is a capture-free surveillance tool, this is a decidedly non-trivial matter. And there are various ways of addressing it, some more technically demanding than others. In the past several years a number of research groups have developed sophisticated site occupancy models that utilize various statistical methods to infer underlying target species distributions from a pattern of positive and negative detections (Chen and Ficetola, 2019). Such models emphasize the critical role that sampling strategy plays in determining the power of eDNA studies (Davis et al., 2018), again highlighting just how important it is to consider how one will interpret results even before a single sample is taken. Of course, complex statistical tools are not the only way to interpret eDNA detection data. Even a single positive detection may be sufficient to trigger action, depending on the manager’s error tolerance profile and the availability of response options. But even in the most straightforward case, it is minimally necessary to determine what will be considered a “positive” detection—and it would be best to make that decision before data is already in hand.

In addition to having a plan for how to interpret eDNA detection data, it would also be advisable to have a plan for how to use that data. What management options are available to respond to eDNA detections, and when exactly should action be triggered? Answers to this question will depend in large part on issues discussed above. A manager’s error tolerance profile, combined with the availability and effectiveness of alternative monitoring tools, will clearly drive decisions about how to employ eDNA tools (Figure 1). When the perceived cost of false negative detections is high (as would be expected in the case of invasive species management), extremely sensitive tools such as eDNA are likely to be very attractive. And if the corresponding cost of a false positive is low, managers may even be willing to adopt eDNA as a sole or primary detection method, despite the risk of false positives. If both types of error carry high costs, then utilization of eDNA will likely depend on the degree to which traditional catch-and-look methods might be used in a confirmatory role. It is assumed that such methods will lack the sensitivity of eDNA, but if the cost of their deployment is not prohibitive (i.e. if catch per unit effort is sufficiently high) it may be reasonable to envision eDNA as a front-line early warning surveillance method, to be followed by more rigorous (and likely more expensive) confirmation by catch-and-look. However, if there is very high cost associated with achieving adequate sensitivity using traditional methods, the manager may need to confront the possibility that eDNA represents the best and maybe only tool available (Darling and Mahon, 2011). It is also worth noting that we now live in a world in which more dramatic actions—e.g. restrictions on usage of an aquatic resource, quarantine, etc.—may defensibly be triggered directly by positive eDNA detections (Rees et al., 2017). The point here is obviously not to provide any prescriptions for action, as that can only be done with full knowledge of the decision-making context. I instead want to emphasize the importance of knowing, even before sampling begins, what various monitoring outcomes will mean in terms of management action. The alternative—waiting until you get the positive detection to decide what to do about it—is very likely to place unpleasant burdens on the decision-maker.

Figure 1.

Schematic illustrating the potential value of eDNA in different contexts. The flow chart shows outcomes (dark grey) based on answers to questions regarding management circumstances (light grey). Methods such as eDNA, with high expected sensitivity (low false negative rates) are likely to be extremely valuable when the perceived cost of a false negative detection is very high. In these cases, the role of eDNA in a monitoring program may depend on the relative cost of eDNA and alternative traditional monitoring tools. Note that there are circumstances under which eDNA may be the best or only available tool, despite potential for false positive error (boxes marked with check marks, ✓), and that when the perceived cost of a false negative is low eDNA may not be a particularly attractive option (boxes marked with an ✗). In other cases, the use of eDNA may be worth considering depending on availability of resources (boxes marked with question marks).

Patience

Patience—the capacity to tolerate delays, setbacks, or obstacles without distress or disappointment—is perhaps the most familiar of the virtues. And so it is worth remembering that eDNA is a young technology. Contemporary popularity of the tool can reasonably be traced to a manuscript published in 2008 by Ficetola et al, describing eDNA detection of amphibians in European ponds (Ficetola et al., 2008). The method isn’t even a teenager yet, and growing pains are to be expected.

We are now beyond the proof of concept phase. This does not mean that there are no longer any questions left to answer, only that nobody is surprised anymore that eDNA can be broadly applied across aquatic (and, in some cases, terrestrial) systems and taxa, and that it can provide highly sensitive detections in various biodiversity assessment and management contexts. Much has already also been accomplished in establishing appropriate data quality and reporting standards and identifying crucial quality assurance criteria that will render eDNA studies fit-for-purpose in decision-making contexts. Yet this work remains unfinished. And the difficult task of establishing trust and acceptance in those contexts has begun but will be ongoing for the foreseeable future.

It is also important to emphasize that we have barely started to realize certain potential benefits of eDNA monitoring. eDNA has been cast not only as a more sensitive option for detecting difficult-to-capture targets, but also as a means to do so more rapidly, less expensively, and in environments that may be inaccessible to traditional methods. Though we’ve come a long way in demonstrating the former, we have yet to make much progress toward the latter. One problem is that we still have not leveraged the economies of scale inherent to molecular methods. Modern molecular laboratories, equipped with highly sophisticated robotic sample handling systems, are capable of processing samples with extremely high throughput. This means that per sample analytical costs should plummet with increasing sample sizes. Unfortunately, with very few exceptions, we have not taken advantage of this efficiency. In most monitoring programs there are simply too few samples available to process. This is (in large part) because it is still too difficult to collect samples and conduct the necessary pre-molecular processing steps (typically filtration and other tasks prior to DNA extraction). Advances in these areas would not only dramatically increase the attractiveness of eDNA monitoring, but could create positive feedback loops, with new markets created for molecular laboratories which in turn would increase analytical efficiencies and decrease costs.

A second area that we are only beginning to explore is the development of automated, remote, and in situ eDNA detection platforms. Capture-free surveillance tools may form the basis for grand visions of globally distributed and automated biodiversity monitoring networks (Bohan et al., 2017), but the realization of their full potential will require more concerted exploration of advancements in engineering, microfluidics, robotics, and other fields. Such work has already begun. Early studies establish promise in the application of aerial drones and swimming robots for sampling, and multiple groups across numerous ecological disciplines are actively pursuing the use of “ecogenomic sensors” for monitoring aquatic biodiversity (Danovaro et al., 2016; Doi et al., 2017; Ottesen, 2016). It will nevertheless be some time before the true promise of eDNA—better, cheaper, faster, and more easily deployable—will come to fruition. And it will doubtless come piecemeal, with advances in some ecological contexts, and for some taxa, coming more rapidly than for others.

It may also be enlightening here to recognize that eDNA is not the only capture-free approach to biodiversity surveillance currently vying for a place in the manager’s toolkit. Passive acoustic monitoring of both terrestrial and aquatic habitats is being adopted to assess bird, fish, and even macroinvertebrate diversity (Darras et al., 2017; Linke et al., 2018; Venier et al., 2017); remote camera imagery is similarly employed in terrestrial and marine habitats to detect the presence of target species (Bicknell et al., 2016; Steenweg et al., 2017; Wearn and Glover-Kapfer, 2019); and tools capable of reading chemical and spectral signatures from satellite or aerial sensors are being developed to monitor taxonomic groups as diverse as forest canopy trees and large mammals (Asner and Martin, 2009; Cavender-Bares et al., 2016; Leblanc et al., 2016; Möckel et al., 2016). These tools exist currently in various states of maturity and exhibit different levels of specificity, accuracy, and robustness. We can thus locate eDNA research within a broader effort to explore the universe of potential tools beyond the constraints of catch-and-look monitoring. Importantly, all such tools will face similar challenges: whatever data are generated, there will be questions about how much they can be trusted to reflect the actual presence of target taxa. I do not think it is overstating the case to suggest that we may be witnessing something of a revolution in biodiversity monitoring, and that it is too early to see plainly what new world awaits. In the meantime, patience may help those interested in adopting these new tools to soberly distinguish hype from reality in the present, to take inevitable mis-steps in stride, and to develop a clear vision of how best to realize the future benefits of these innovative tools.

Conclusions

I think it is safe to say that we are well beyond phase one of Arthur C. Clarke’s famous formulation, quoted in the above epigraph. No one seriously doubts that it is at least sometimes possible to glean management-relevant information from eDNA monitoring. But, for many observers, the jury is still out on whether or not eDNA surveillance is worth doing. And though I am obviously a proponent of eDNA tools, I will refrain from predicting if or when we may reach the point when everyone can agree that it was a good idea all along. It is far easier to predict that the trajectory of progress in this field will be clearer, more deliberate, and more rewarding if it is a truly collaborative effort between researchers and end-users. In conversations with end-users who have considered eDNA methods I have encountered attitudes ranging from enthusiastic embrace to hesitancy, discouragement, and resistance. My hope is that the ideas outlined above may help to bring this engagement to a more even keel. While much of what I have said here is intended to inspire investment from end-users who may have doubts about the value of eDNA tools, I am equally concerned to recognize that eDNA is not the right tool in all management contexts, and that application of eDNA should be pursued only with knowledge, self-awareness, preparedness, and patience. Cultivation of these virtues will take time and effort. But rather than impeding progress, I believe that this considered approach will reduce anxiety on the part of all stakeholders and will ultimately lead to more effective tools better tailored to management needs.