Quantifying the Sensitivity of Targeted eDNA Surveys to Improve Detection of Invasive Cane Toads (Rhinella marina)

Abstract

ABSTRACT

Environmental DNA (eDNA) surveys are increasingly used to monitor biodiversity because they are often more sensitive (have higher detection probability) than conventional monitoring methods. Sensitivity is a key consideration in designing monitoring programs because it determines the survey effort (e.g., number of samples per site) required to achieve a given likelihood of detecting a species. However, assessing the sensitivity of eDNA surveys and examining the factors influencing this in the field remain understudied. Here, we quantify the importance of key factors likely to influence eDNA sensitivity and compare the results of eDNA surveys to conventional visual surveys for detecting invasive cane toads (Rhinella marina) in northern Australia. We sampled waterbodies across the invasion front and showed that both eDNA and visual surveys had similar performance in detecting cane toads. Environmental DNA sensitivity varied predictably across waterbodies as a function of several factors. Sensitivity was higher: (1) when a greater volume of water was sampled at a water body; (2) at waterbodies with higher toad densities; (3) at smaller waterbodies; and (4) when cane toad tadpoles were present. We show how these findings can be used to tailor survey effort to ensure a specified level of detection probability at individual waterbodies, for example, by scaling the number of samples taken to water body size and tadpole presence/absence. Our study highlights the value of quantitatively assessing the sensitivity of eDNA surveys in the field and understanding the factors influencing sensitivity to achieve monitoring objectives.

Introduction

Targeted environmental DNA (eDNA) surveys are being used increasingly to detect and monitor species because the method is often highly sensitive and can be time and cost-effective relative to conventional surveys (e.g., visual, auditory, catch-based; Fediajevaite et al. 2021; Takahashi et al. 2023). This is especially the case for species and/or environments that are difficult to survey using conventional methods, such as aquatic systems (Fediajevaite et al. 2021; Takahashi et al. 2023). Sensitivity, or the probability of detecting a species when it is present at a site, is a fundamental consideration in designing robust monitoring programs. In the context of biosecurity and invasive species, early detection is often a priority, and ensuring that monitoring has sufficient sensitivity to detect new incursions is critical to ensuring effective management responses (Bylemans et al. 2016). Despite extensive research comparing the relative sensitivity of eDNA to traditional surveys, the absolute sensitivity of eDNA surveys in field conditions remains understudied for many species and systems (Fediajevaite et al. 2021). Quantifying sensitivity is important because it allows us to determine if a survey method is likely to achieve its monitoring aims. Additionally, if we understand how variation in key factors influences eDNA sensitivity, we could ensure a given level of detection probability at survey sites by adjusting sampling effort to site conditions, thereby optimizing the use of time and resources (Furlan et al. 2016, 2019).

The sensitivity of eDNA surveys depends in part on aspects of survey design that are under surveyor control, including the number of samples taken per site, sample volume, and the number of PCR replicates analyzed per sample (i.e., survey effort; Furlan et al. 2016). However, sensitivity also varies as a function of species and site-level characteristics outside the surveyor's control, and understanding this variation could help refine survey design. For some characteristics, we can make a priori predictions regarding how they should affect sensitivity. For example, in water sampling, we would expect to capture more eDNA, and hence to achieve greater sensitivity, where target species density is higher (Doi et al. 2015; Takahara et al. 2012; Yates et al. 2019), and/or when individuals are more active (Everts et al. 2021). We might also expect greater sensitivity when sampling small relative to large waterbodies due to a greater concentration of eDNA (Furlan et al. 2019), and greater sensitivity where the target species or its life stages are fully aquatic as opposed to partially aquatic or terrestrial (Buxton et al. 2017; Everts et al. 2021; Takahara et al. 2020). Additionally, if we collect more water per sample, we expect to capture more eDNA and hence have higher sensitivity (Mirimin et al. 2020; Muha et al. 2019; Schultz and Lance 2015). However, for many species and systems, our understanding of how such site and species-level characteristics affect eDNA sensitivity in field surveys remains limited, despite its importance in determining suitable sampling schemes and making quantitative assessments of species presence or absence.

In this study, we aim to bridge this gap by quantifying how key characteristics, including target animal density, size of sampled waterbodies, sample volume, and presence of an aquatic life stage affect eDNA sensitivity in the field. We did this by surveying sites for the invasive cane toad (Rhinella marina) in northern Australia. Cane toads were introduced to Australia in 1935 and have steadily expanded their range (Doody et al. 2018), causing severe population declines in native predators due to lethal poisoning (Doody et al. 2009, 2021; Jolly et al. 2015; Shine 2010) with subsequent ecosystem disruptions (Doody et al. 2017, 2023, 2015). At the time of this study, cane toads were advancing across northern Australia in the Kimberley region, more than 2000 km from their introduction site, with the invasion front advancing approximately 50 km per year (DBCA 2024). Being able to detect cane toads at low densities when they initially colonize new areas could assist in managing or preventing new incursions into sensitive sites (Everts et al. 2023).

Two species-specific eDNA assays have been developed for cane toad detection and validated in the field, with eDNA detected at all field sites where toads were observed (Edmunds and Burrows 2019; Tingley et al. 2019). For both assays, the authors suggested that cane toad eDNA surveys of waterbodies were highly sensitive and encouraged their use for monitoring incursions. Both studies, however, were carried out in sampling environments where we would expect sensitivity to be high, having predominantly sampled waterbodies with well-established populations where aquatic tadpoles were present. In contrast, surveys of more recently invaded waterbodies at the southern invasion front failed to detect cane toads using eDNA at several sites where toads were present based on visual and acoustic surveys (Macgregor et al. 2021). Here, we aimed to quantify survey sensitivity by sampling waterbodies across the northern invasion front, including recently invaded sites where toad density should be low. We assess how site conditions, including the density of toads and water body characteristics, affect sensitivity, and thus evaluate the potential for eDNA surveys to detect early incursions of cane toads. Specifically, our aims were to:

• Quantify the absolute sensitivity of targeted eDNA surveys for detecting cane toads across the current invasion front in northern Australia.
• Quantify how species and site-level characteristics, including cane toad density, sample volume, perimeter of sampled water body, and presence of cane toad tadpoles, affect eDNA sensitivity.
• Design eDNA sampling schemes that account for species and site-level characteristics to ensure 95% confidence in detection if cane toads are present at sites.
• Compare the relative sensitivity of visual and eDNA surveys for detecting cane toads at the invasion front.

Methods

Study Species

Cane toads are a large anuran, native to Central and South America (Zug and Zug 1979). They are a generalist species that thrive in a variety of climatic and environmental conditions (Urban et al. 2007), including disturbed habitats, which they exploit for feeding opportunities (González-Bernal et al. 2016) and dispersal (Brown et al. 2006). Cane toads are highly fecund, with a single female producing up to 30,000 eggs (Zug and Zug 1979), with spawning typically linked to rainfall (Yasumiba et al. 2016). They can reproduce year-round, and multiple times per year, in any standing water (Zug and Zug 1979), though they prefer shallow, warmer areas (Hagman and Shine 2006). Eggs hatch within two or three days, and aquatic tadpoles metamorphose after 4 to 8 weeks (Raven et al. 2017).

Cane toads were introduced to northeastern Australia in 1935, and have since expanded their range over 2000 km westward, advancing approximately 50 km per year (DBCA 2024). Range expansion occurs primarily during the wet season (November–April), when abundant water in the landscape facilitates dispersal (Brown et al. 2011). During the dry season (May to October), when rainfall is scarce, cane toads reduce movement and retreat to areas with access to moisture, such as permanent waterbodies (Brown et al. 2011). Toad densities can be very high in invaded regions of Australia, particularly when concentrated around permanent waterbodies in the dry season (up to 2138 ha⁻¹; Freeland 1986). Because of their high densities and conspicuous behavior, citizen-led ‘toad busting’ groups can collect hundreds of individuals per night (Greenlees et al. 2020; Shine et al. 2018).

Study Area

We conducted eDNA and visual surveys of 35 waterbodies (sites) in the West Kimberley region of Western Australia, part of the wet-dry tropics, over approximately 3 weeks in June and July of 2023 (dry season). We surveyed waterbodies such as wetlands and farm dams, where toads are concentrated during the dry season. We selected 30 sites across the current invasion front to capture a range of site characteristics, including variation in toad densities and water body sizes (Figure 1). Five uninvaded sites were also included as negative controls, located approximately 100 km ahead of the invasion front in 2022/2023.

[IMAGE OMITTED. SEE PDF]

Waterbodies were identified as potential sites using Google Maps satellite imagery. Only self-contained waterbodies with clearly delimited boundaries were selected for sampling to minimize the chances of cane toad movement between waterbodies and to allow us to estimate water body size. We calculated water body size by measuring the length and width of each water body in the field at its maximum extent and estimated the perimeter as 2 × (length + width). We used this perimeter estimate as an ecologically relevant measure of water body size, given that cane toads spend most of their time around water body edges (Mayer et al. 2015).

Visual Surveys

We conducted visual surveys for cane toads at each water body at night. Nocturnal visual surveys have been traditionally used to monitor cane toads because they are active at night and conspicuous around the edges of waterbodies, particularly during the dry season (Greenlees et al. 2018; Phillips et al. 2007). Each survey was conducted by a pair of observers walking together around the perimeter of the water body at a steady pace and searching for toads or their eyeshine with the aid of a headtorch. Each survey was timed, and the total number of toads was recorded. We converted the absolute count into an encounter rate by dividing the total number of toads observed by the number of minutes it took to complete the survey, so that the encounter rate was the number of toads observed per minute. We used this as a proxy for toad density that accounted for variation in survey effort. Visual surveys were conducted once per site because estimates of cane toad population metrics based on single-visit encounter/capture rates are repeatable during the dry season (Freeland 1986).

We also recorded whether cane toad tadpoles were observed during the nocturnal visual surveys or while collecting water samples for eDNA during the day (see below). Cane toad tadpoles are easily identifiable by their small black bodies and strong tendency to form large, conspicuous aggregations in the shallow margins of waterbodies (Raven et al. 2017).

Targeted eDNA Surveys

Field Sampling

Field sampling was conducted during daylight hours using standard eDNA water sampling kits (EnviroDNA, ). Five water samples were taken per water body at approximately equidistant locations around the perimeter. Following the manufacturer's sampling guidelines, water was drawn through a 50 mL or 60 mL syringe and pushed through a 1.2 μm enclosed filter to trap eDNA. This process was repeated until either 500 mL of water was filtered or until the filter had clogged (e.g., by sediment or algae) and water could no longer be pushed through, at which point the water volume filtered was recorded. The 500 mL target was designed to maximize the chances of capturing eDNA while remaining practical to filter (Takahashi et al. 2023). A Tris-EDTA-based DNA preservative provided with each kit, was flushed through the filter and the enclosed filter was then capped and sealed in an individually labeled bag for transport to the laboratory at ambient temperature. At each site, an additional kit was used to sample 50 mL of MilliQ water as a negative field control. The MilliQ water was kept in a sealed container (one per site) and opened at the water body's edge for approximately 30 s before being filtered. The control was treated as a sixth sample in the laboratory analysis. Kit components, including sterile gloves, were used for only one sample to avoid cross-contamination.

Environmental DNA Extractions

All eDNA extractions were carried out in a designated trace DNA laboratory at the University of Canberra, approximately 5 to 7 weeks after sampling. We extracted DNA from each EnviroDNA filter using a modified Qiagen DNeasy Blood & Tissue kit (spin column protocol). First, the preservative was flushed out and the filter was left to dry for 15 min. 720 μL of buffer ATL and 80 μL of proteinase K were then introduced to each filter before incubation at 65°C for 45 min with constant agitation. The lysis solution was transferred into 10 mL conical tubes, before 800 μL of cold ethanol (100%) and 800 μL of buffer AL were added to each sample. 600 μL of the sample solution was then centrifuged in a spin column at 6797 g for 1 min; this step was repeated once more to process a total of 1200 μL of sample. We then employed a two-step cleaning process on samples: (1) 500 μL of buffer AW1 was added and centrifuged at 6797 g for 1 min, and (2) 500 μL of buffer AW2 was added and centrifuged at 20817 g for 3 min. Finally, all samples were eluted with 100 μL of ultrapure water. A negative control was also introduced to the laboratory workflow during each batch of extractions.

Tissue Sample gDNA Extractions

We extracted genomic DNA (gDNA) from cane toad toe webbing tissue samples for positive controls in subsequent laboratory analysis. Tissue samples were collected in 2022 from one of the long-invaded sites in the present study. To extract gDNA, we followed the Qiagen DNeasy Blood and Tissue Kit manufacturer's protocol (spin column protocol).

Quantitative PCR

All eDNA extracts were analyzed with a quantitative PCR (qPCR): QuantStudio 7 Real-Time PCR in a 96 well plate format. We prepared all plates in a designated PCR setup room at the University of Canberra. The assay we used, designed by Tingley et al. (2019), targets an 80 base pair (bp) region covering part of the mitochondrial tRNA-Gly and NADH dehydrogenase subunit (ND3) genes. Custom forward and reverse primers from Invitrogen, and a custom probe from TaqMan were used for analysis. To determine whether to use M13 or non-M13 tailed primers for sample analysis, we ran a preliminary standard curve plate comparing their performance with the cane toad gDNA. Both primers were highly efficient (99.638% and 100.394% for M13 and non-13, respectively), with high R-squared values (0.989 and 0.985 for M13 and non-M13, respectively). However, the M13 primers had steeper amplification curves and facilitated Sanger sequencing in fewer steps, so they were used for subsequent analysis:

Forward: R.Marina_ND3_F TGTAAAACGACGGCCAGTACCCCAGGAGAAAATAATGTCTCT.

Reverse: R.Marina_ND3_R CAGGAAACAGCTATGACCACCAGAAGCTAACAGTGGCTAAAAT.

In the above sequences, bold denotes M13 tails, and regular text denotes the cane toad Tingley et al. (2019) primers.

As per the Tingley et al. (2019) study, we used 10 μL qPCR reactions consisting of: 2 μL of sample eDNA extract, 5 μL TaqMan Environmental Master Mix 2.0, 0.9 μL M13 (−21)_R.marina_ND3_F primer [10 μm], 0.9 μL M13_R.marina_ND3_R primer [10 μm], 0.25 μL of R.marina_ND3_P_FAM probe [10 μm], and 0.95 μL of UltraPure DNase/RNase-Free Distilled Water.

Amplification was carried out under conditions of 10 min at 95°C, followed by 50 cycles of 15 s at 95°C and 1 min at 60°C. Each eDNA sample was run in triplicate at a 1:1 concentration and 1:10 concentration (diluted with UltraPure water) in case there were PCR inhibitors in the sample (i.e., three 1:1 replicates and three 1:10 replicates per sample). Because the 1:1 replicates amplified at a higher rate for all samples, we used these for all subsequent data analyses. Negative controls were introduced to the workflow for each qPCR plate. Each plate also included a positive control of 0.1 ng/μL cane toad gDNA run in triplicate. For samples where at least one qPCR replicate amplified, the replicate that amplified earliest (indicating a higher concentration) was purified and sent for bidirectional Sanger sequencing to verify the presence of cane toad eDNA.

Data Analysis

Occupancy Modeling

We fitted a hierarchical occupancy model to estimate sensitivity and explore the factors hypothesized to influence it. We used this approach because the model separates the field sampling (probability of capturing eDNA in a sample given it is present at a site) from the laboratory (probability of amplifying eDNA in a sample containing eDNA) components of sensitivity. We could therefore quantify the effect of factors hypothesized to influence sensitivity through their influence on eDNA capture probability (i.e., toad density, presence of tadpoles, volume of water filtered per sample, and water body perimeter), separate from the sensitivity of the laboratory methods we used. The occupancy model captured the three-level hierarchical structure of the eDNA survey data (i.e., sites, samples, and qPCR replicates) in terms of successive Bernoulli trials (Schmidt et al. 2013):1$$$ {z}_i\sim \mathrm{Bernoulli}\ \left(\varPsi \right) $$$2$$$ {a}_{ij}\mid {z}_i\sim \mathrm{Bernoulli}\ \left({z}_i{\theta}_{ij}\right) $$$3$$$ {y}_{ij k}\mid {a}_{ij}\sim \mathrm{Bernoulli}\ \left({a}_{ij}\omega \right) $$$where $$$ {z}_i $$$ is the latent presence ($$$ {z}_i=1 $$$) or absence ($$$ {z}_i=0 $$$) of cane toads at site $$$ i $$$ given the overall probability of site occupancy $$$ \varPsi $$$ ($$$ {z}_i $$$) is a latent variable because occupancy is inferred from eDNA detection rather than directly observed; $$$ {a}_{ij} $$$ is the presence ($$$ {a}_{ij}=1 $$$) or absence ($$$ {a}_{ij}=0 $$$) of cane toad eDNA in sample $$$ j $$$ collected from site $$$ i $$$, which depends on whether the site was occupied by cane toads ($$$ {z}_i=1 $$$) or not ($$$ {z}_i=0 $$$) and the probability $$$ {\theta}_{ij} $$$ that cane toad eDNA was captured in the sample given that cane toads were present; $$$ {y}_{ijk} $$$ is the observed data describing whether qPCR replicate k from sample j at site i amplified ($$$ {y}_{ijk}=1 $$$) or not ($$$ {y}_{ijk}=0 $$$), which depends on whether the sample contained cane toad eDNA ($$$ {a}_{ij}=1 $$$) or not ($$$ {a}_{ij}=0 $$$) and the probability $$$ \omega $$$ that eDNA amplified, given it was present in a sample.

The site characteristics hypothesized to influence eDNA capture probability $$$ {\theta}_{ij} $$$, and by extension, survey sensitivity, were included in the model as covariates:4$$$ {\displaystyle \begin{array}{cc}{\log}_{it}\left({\theta}_{ij}\right)\hfill & =\kern4pt {\beta}_0+\left({\beta_1}^{\ast }{\mathrm{tadpole}\ \mathrm{presence}}_i\right)+\left({\beta_2}^{\ast }{\mathrm{toad}\ \mathrm{density}}_i\right)\hfill \\ {}\hfill & +\left({\beta_3}^{\ast }{\mathrm{sample}\ \mathrm{volume}}_{ij}\right)+\left({\beta_4}^{\ast }{\mathrm{waterbody}\ \mathrm{perimeter}}_i\right)\hfill \end{array}} $$$

The probability that eDNA was captured in a sample was therefore allowed to vary for every sample $$$ j $$$ collected from site $$$ i $$$ as a function of the volume of water filtered for each sample, and to vary between sites as a function of the toad density, the presence/absence of toad tadpoles, and the perimeter of the water body. To compare the effect sizes of covariates, continuous covariates (toad density, sample volume, and water body perimeter) were standardized to mean zero and standard deviation one. The presence of tadpoles was coded as a binary variable (present = 1; absent = 0) and was not standardized.

Model Fitting

To fit the occupancy model, we removed the five control sites where cane toads were absent and estimated occupancy at the sites potentially occupied by toads. The data we analyzed comprised five samples collected from each of 30 sites (150 samples) with three qPCR replicates per sample, giving a total of 450 qPCR replicates coded as amplified for cane toad DNA ($$$ {y}_{ijk}=1 $$$) or not ($$$ {y}_{ijk}=0 $$$). We fitted the model to the data in a Bayesian framework using Markov Chain Monte Carlo (MCMC) methods to sample from the posterior distributions. Seven parameters were estimated in fitting the model to the data: the overall occupancy probability ($$$ \varPsi $$$); the probability eDNA amplified in a qPCR replicate given it was present in a sample ($$$ \omega $$$); and the parameters that specified the effect of covariates on $$$ {\theta}_{ij} $$$: $$$ {\beta}_0 $$$, $$$ {\beta}_1 $$$, $$$ {\beta}_2 $$$, $$$ {\beta}_3 $$$, $$$ {\beta}_4 $$$. Because we used a Bayesian model, we had to specify priors, using uninformative priors for $$$ \varPsi $$$ and $$$ \omega $$$ by specifying a uniform probability between 0 and 1. For the parameters associated with the covariates, we specified flat normal priors with mean 0 and variance 1000. We fitted the model in JAGS called through R using the package jagsUI (Kellner 2021). We ran three MCMC chains with 30,000 iterations following a burn-in of 20,000 iterations. We evaluated model convergence using R hat values, which were ~1.0, meaning the model successfully converged (Gelman and Rubin 1992).

Estimating Survey Sensitivity

The absolute sensitivity of an eDNA survey can be expressed in terms of the sensitivity of the field and laboratory components described in the above model (Furlan et al. 2016):5$$$ {\displaystyle \begin{array}{c}\Pr \left(\mathrm{detect}\ \mathrm{target}\ \mathrm{eDNA}\kern0.5em \mathrm{a}\mathrm{t}\ \mathrm{a}\ \mathrm{site}\right|\mathrm{eDNA}\ \mathrm{present}\ \mathrm{a}\mathrm{t}\ \mathrm{t}\mathrm{hat}\ \mathrm{site}\Big)\hfill \\ {}\kern0.5em =1-{\left[\left(1-\theta \right)+\theta {\left(1-\omega \right)}^{n_a}\right]}^{n_s}\hfill \end{array}} $$$where $$$ \theta $$$ is the probability that eDNA is captured in a sample, given eDNA is present at the site (field sensitivity), $$$ \omega $$$ is the probability eDNA is amplified in a qPCR replicate given a sample contains eDNA (laboratory sensitivity), $$$ {n}_a $$$ is the number of qPCR replicates per sample, and $$$ {n}_s $$$ is the number of samples collected at a site. In our study, the value of $$$ \theta $$$ varied between samples and sites, being dependent on covariate values (Equation 4). To accommodate this, we calculated the value within the square brackets of Equation (5) for each sample, and then obtained the product of these values across samples at each site to calculate site-level sensitivity.

Sampling Schemes

Survey sensitivity should increase with increasing survey effort. For eDNA surveys, key components of effort include the number of samples collected per site ($$$ {n}_s $$$), the number of qPCR replicates run per sample ($$$ {n}_a $$$), and the volume of water collected or filtered per sample (Schultz and Lance 2015). We explored how changing survey effort should affect survey sensitivity by calculating the effort required to achieve 95% sensitivity under a range of conditions. To do this, we simulated different conditions by selecting a range of values for toad encounter rate, water body perimeter, and volume of water filtered, and then calculated the expected values of $$$ \theta $$$ for the different sets of conditions using Equation (4) and the medians of the posterior distributions for parameters $$$ {\beta}_0 $$$, $$$ {\beta}_1 $$$, $$$ {\beta}_3 $$$, and $$$ {\beta}_4 $$$ obtained from the model fit. We thus simulated situations where tadpoles were absent ($$$ {\beta}_2 $$$ = 0), which is the likely scenario when toads first invade a site. The number of samples required for 95% confidence in cane toad detection (i.e., 0.95 sensitivity) can be calculated by rearranging Equation (5):6$$$ {n}_s=\frac{\log (0.05)}{\log \left[\left(1-\theta \right)+\theta {\left(1-\omega \right)}^{n_a}\right]} $$$

Results

Comparison of eDNA and Visual Surveys

Neither visual nor eDNA surveys detected cane toads at any control sites located approximately 100 km ahead of the invasion front (Table 1). All eDNA controls performed as expected; cane toad eDNA was not detected in any field, extraction, or qPCR negative controls, and all positive gDNA controls amplified, suggesting effective qPCR reactions occurred. All 74 putative cane toad eDNA detections were also confirmed by Sanger sequencing (only the replicate with the highest Ct threshold per sample was sequenced). 69 of 74 bidirectional sequences matched 93.6% to 100% pairwise for cane toad (NCBI accession no. NC_066225). The remaining five sequences were of very poor quality, so only short regions could be used for comparison (30–47 bp). However, all five matched 100% pairwise for cane toad (NCBI accession no. NC_066225), and only other true toads were similar, so these were also considered true positive detections, as cane toads are the only toads in Australia. Therefore, all positive cane toad detections from either survey method were considered true positives.

TABLE 1 Number of sites where cane toads were detected (Yes) or not (No), with each survey method.

	Invasion front sites	Control sites
Yes	No	Yes	No
eDNA survey	23	7 (4)	0	5
Visual survey	23	7 (4)	0	5

The combination of eDNA and visual surveys detected cane toads at 27 of 30 sites, comprising 19 sites where cane toads were detected with both methods, as well as four sites where cane toads were only observed during visual surveys (eDNA false negatives), four sites where toads were only detected through eDNA (visual survey false negatives), and three sites where neither method detected cane toads (possible true negatives; Table 1).

Survey Site Characteristics

There was substantial site-to-site variation in the species and site characteristics measured. Where cane toads were observed in visual surveys, the mean encounter rate was 2.7 toads per minute (range = 0.07 toads per minute to 8.82 toads per minute; Figure 2B). The mean water body perimeter was 212 m (range = 40 m to 560 m; Figure 2F), and the mean water sample volume was 307 mL (range = 15 mL to 500 mL; Figure 2D). Although we aimed to filter 500 mL per sample, the sample volume varied because some filters became clogged by sediments or algae. At sites where tadpoles were present, the average proportion of qPCR replicates that amplified was 0.80 (range = 0.40 to 1), while for sites where tadpoles were not observed, the average was less than half this (0.34; range = 0.07 to 0.73) despite often high toad encounter rates (Figure 2A).

[IMAGE OMITTED. SEE PDF]

Sensitivity Estimates and the Effect of Species and Site Factors

From the fitted occupancy model, the mean probability that eDNA was captured in a sample given it was present at a site ($$$ {\theta}_{ij} $$$) was 0.546 (range across sites = 0.0931 to 1.0). The probability of capture varied substantially among sites, due in part to variation associated with the covariates we measured. In all cases, the covariate parameter estimates had 95% credible intervals (CrI) that did not overlap zero, suggesting they all affected survey sensitivity (Figure 3A). The probability eDNA was captured in a sample increased when the toad encounter rate was higher, when tadpoles were present at a site, and when more water was filtered per sample, but decreased when the perimeter of the water body was larger. The presence of tadpoles at a site had the strongest effect (note the different x-axis scales in Figure 3A), with the wide posterior distribution probably reflecting the small sample size (tadpoles were observed at five sites).

[IMAGE OMITTED. SEE PDF]

The probability that eDNA captured in a sample was amplified in a qPCR replicate $$$ \omega $$$ was 0.670 [95% CrI 0.603 to 0.735]. Using the sample-level estimates of eDNA capture probability and the probability of amplification, mean absolute survey sensitivity (the probability of detecting cane toad DNA at a site given cane toads were present and taking five samples each with three qPCR replicates) was 0.897 but varied among sites from 0.381 to 1 (Figure 3B). That is, on average, our eDNA survey had a ~90% chance of detecting cane toads at occupied sites. While at most sites the sensitivity was > 0.8, low-sensitivity sites were characterized by an absence of tadpoles, low toad densities, low water volumes filtered due to turbid conditions, and large perimeters. Absolute eDNA sensitivity was high at all sites where cane toads were detected with eDNA surveys but not visual surveys (black bars in Figure 3B). However, eDNA sensitivity varied, and was sometimes low, at sites where toads were detected with visual surveys but not eDNA surveys (white bars in Figure 3B).

Sampling Schemes to Achieve 95% Sensitivity

Increasing the number of qPCR replicates per sample led to only a slight improvement in survey sensitivity if more than one sample was taken per site, and more than one replicate was analyzed per sample (Figure A1). We thus focused on how variation in sample numbers per site affects sensitivity. Furthermore, tadpole presence strongly influenced sensitivity (Figure 2A,C,E; Figure 3A), such that 95% survey sensitivity was achieved with one to two samples if tadpoles were present, regardless of other site factors. We therefore focused on situations where tadpoles were absent, which is likely to be the case at the invasion front when adult cane toads first colonize an area.

Figure 4 shows the number of samples per site required to achieve 95% sensitivity given variation in site characteristics due to differences in toad encounter rate, water body perimeter, and volume of water that can be sampled. When site conditions are unfavorable, many more samples than the five taken in this study are required to ensure a 95% detection probability. For example, for a recent incursion where toad encounter rate is low (1 toad per minute), the water body is large (perimeter = 500 m) and the water is turbid so only a small volume can be filtered (10 mL), we estimate that 69 samples would be required to achieve 95% sensitivity (Figure 4 bottom left). In contrast, at a water body with a high toad encounter rate ($$$ 10 $$$ toads per minute), a small perimeter (10 m), and where a large volume of water is filtered per sample (1000 mL), only two samples should achieve 95% sensitivity (Figure 4 top right).

[IMAGE OMITTED. SEE PDF]

Discussion

Quantifying survey sensitivity and understanding the conditions that affect sensitivity is an important step in designing and evaluating any monitoring program. Survey sensitivity will determine the survey effort (e.g., number of samples) required to achieve a specified level of detection probability. Environmental DNA surveys are often considered more sensitive than conventional visual surveys, but for many species and systems, field testing to support this remains limited. Furthermore, the effect of species and site-level characteristics on survey sensitivity is not well understood for many applications. In our study, targeted cane toad eDNA surveys were generally highly sensitive at detecting toads (mean sensitivity = 90%). However, conventional visual surveys were comparably sensitive, identifying the same number of occupied sites as eDNA. Nevertheless, an advantage of the eDNA survey was that we could measure how site and species-level factors influenced sensitivity and thus quantify sensitivity at individual sites. All factors we considered affected the probability of eDNA capture and, hence, survey sensitivity. Specifically, sensitivity increased with cane toad density, the presence of cane toad tadpoles, and the volume of water filtered per sample, but decreased with increasing water body perimeter. By quantifying these relationships, we were able to calculate the number of samples required to achieve 95% sensitivity given different site conditions. In the context of cane toad monitoring, this means eDNA sampling efforts can be tailored to achieve a specified level of sensitivity to detect recent incursions, where a water body may initially be occupied by only one or a few adult toads. Early detection of cane toads should facilitate timely and effective interventions at sensitive sites.

The factors we found associated with greater sensitivity should all act by increasing the amount of eDNA in samples, either by influencing the concentration of eDNA available for capture at a water body (higher toad density, presence of tadpoles, smaller water body perimeter), or the amount of eDNA captured (greater sample volume). These factors have been identified as increasing sensitivity for other species/systems (Buxton et al. 2017; Doi et al. 2015; Everts et al. 2021, 2022; Furlan et al. 2019; Mirimin et al. 2020; Muha et al. 2019; Schultz and Lance 2015; Takahara et al. 2020, 2012; Yates et al. 2019), although our quantitative understanding of their effect on survey sensitivity remains limited. In our study, the presence of tadpoles was particularly influential, such that when tadpoles were present, just one or two samples would provide 95% sensitivity, regardless of other factors. This is perhaps expected given that, for other amphibian species, the presence of aquatic larvae is often strongly associated with high eDNA concentrations (Buxton et al. 2017; Everts et al. 2021, 2022; Takahara et al. 2020). Tadpoles are continuously in contact with the water, are active and growing, and may therefore continuously shed more eDNA than partially aquatic or terrestrial, post-metamorphic life stages. Cane toad tadpoles also aggregate at the shallow, warmer margins of waterbodies (Raven et al. 2017), where eDNA samples were collected. We thus highlight the importance of accounting for factors likely to influence eDNA sensitivity in most studies (e.g., target species density, sample volume, water body size) along with target species and life-history specific factors. Identifying such factors a priori should assist in designing surveys with sufficient sensitivity given the conditions likely to be encountered in the field.

Prior studies that developed eDNA assays to detect cane toads have suggested the method is highly sensitive (Edmunds and Burrows 2019; Tingley et al. 2019), but have not explicitly identified factors affecting sensitivity in the field. Tingley et al. (2019) fitted the occupancy model we used (Equations 1–3) to their data, allowing us to compare estimates of the two probabilities that comprise survey sensitivity ($$$ \theta $$$ and $$$ \omega $$$). In the laboratory, both studies were similarly likely to amplify cane toad eDNA once captured in samples [Tingley et al. (2019) $$$ \omega $$$ = 0.731; this study $$$ \omega $$$ = 0.670]. However, we were less likely to capture cane toad eDNA in samples compared to Tingley et al. (2019) [Tingley et al. (2019) $$$ \theta $$$ = 0.857; this study $$$ \overline{\theta} $$$ = 0.546]. Edmunds and Burrows (2019) also detected cane toad eDNA in every sample they took, which, although not explicitly estimated, indicates a high capture probability ($$$ \overline{\theta} $$$). The higher capture probability in prior studies relative to ours is likely a consequence of our sampling recently invaded sites, whereas prior studies sampled areas where cane toads were well established and tadpoles were present at all or most sample sites. Once established, cane toad populations often reach high densities (up to 2138 ha⁻¹; Freeland 1986), such that citizen-led ‘toad busting’ groups can collect hundreds of animals per night (Greenlees et al. 2020; Shine et al. 2018). It is likely that sites with well-established cane toad populations will have high eDNA concentrations, which probably explains the high sensitivity of prior studies. Nevertheless, similar to our study, Macgregor et al. (2021) did not detect eDNA at several sites at the southern invasion front where cane toads were observed with conventional survey methods. Our study emphasizes to the importance of sampling across a range of site conditions, and varying eDNA concentrations to evaluate sensitivity and identify the factors influencing this.

There is potential to extend the modeling approach to account for additional factors that could influence sensitivity. These include environmental variables that affect rates of eDNA degradation (e.g., UV-B, temperature, pH; Strickler et al. 2015) or transport (e.g., flow rate; Harrison et al. 2019), factors that can inhibit PCR reactions (e.g., humic substances, algae, sediment; Stoeckle et al. 2017), and other aspects of species life-history, such as phenology (e.g., seasonal variation in activity; Everts et al. 2021). These additional factors might explain the lack of eDNA detection at two sites in our study where cane toads were abundant and estimated sensitivity was high (Figures 2A and 3B). Incorporating such additional factors as covariates in the model would allow us to evaluate their influence on sensitivity, and may help explain discordance between eDNA and conventional survey methods.

The high sensitivity of targeted eDNA and comparable sensitivity of visual surveys suggest both are useful for monitoring cane toads. Relative to eDNA surveys, visual surveys are cheap and provide immediate evidence of toad presence. Visual surveys are successful in this case because cane toads are large, conspicuous animals that are often abundant at invaded waterbodies (e.g., we observed 97 toads at a single water body in this study). For other more cryptic species, conventional survey methods may be less sensitive or more difficult and expensive to undertake, making eDNA surveys an attractive option (e.g., Platypus Ornithorhynchus anatinus McColl-Gausden et al. 2023; Northern corroboree frog Pseudophryne pengilleyi Rojahn et al. 2018; Largetooth sawfish Pristis pristis Simpfendorfer et al. 2016). There is also potential to increase the sensitivity of eDNA surveys, for example, by applying multiple markers for the same species in multiplex (Brys et al. 2023). As we have shown, a further advantage of eDNA surveys is that we can leverage the replicate nature of sampling to quantify survey sensitivity at each site, allowing survey effort to be tailored to achieve a desired detection probability. Visual surveys, in contrast, require repeat site visits to estimate sensitivity. Nevertheless, the choice of survey method will depend on a range of factors in addition to sensitivity, including cost and practicality. For cane toad management, the potential to tailor eDNA sensitivity to individual sites could assist in protecting high-value toad-free areas from incursion by ensuring adequate sampling to achieve a desired detection probability when cane toad numbers are low. In contrast, visual surveys may be a more cost-effective way to detect or confirm cane toad presence across wide areas where toads are well established and sensitivity is high. Alternatively, initially undertaking a visual survey followed by eDNA sampling at sites without visual detections could combine the benefits of both survey methods. Such complementary multi-method approaches can significantly improve survey sensitivity (Kyle et al. 2022).

Our study evaluated the sensitivity of eDNA surveys for detecting cane toads across a range of conditions, including early incursions where densities are low. We demonstrated how targeted eDNA surveys can be tailored to achieve a specified detection probability at individual sites, given variation in factors that affect sensitivity. This understanding should assist in determining whether eDNA surveys are feasible for monitoring a target species and to assist in optimizing survey design to achieve monitoring outcomes. Our study contributes to the growing body of evidence that eDNA surveys can be a powerful biomonitoring method, but emphasizes that careful consideration needs to be given to the environmental and species-specific context in which it is used.

Author Contributions

Conception and design of the study: E.K.L., S.C., A.T.G., P.G.N., and R.P.D.; Funding acquisition: S.C. and R.P.D.; Field surveys and laboratory analysis: E.K.L.; Data analysis: E.K.L. and R.P.D.; Data interpretation: E.K.L., S.C., A.T.G., P.G.N., and R.P.D.; Writing and editing the manuscript: E.K.L., S.C., A.T.G., P.G.N., and R.P.D.

Acknowledgments

We would like to thank Mount Gibson Iron Ltd. for funding this project. We would also like to extend our appreciation to the Australian National eDNA Reference Centre, EcoDNA group at the University of Canberra for providing state-of-the-art facilities for laboratory analyses. Open access publishing facilitated by University of Canberra, as part of the Wiley - University of Canberra agreement via the Council of Australian University Librarians.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

The R code and data required to reproduce the analysis in this manuscript are available at: .

Appendix

Appendix - A

[IMAGE OMITTED. SEE PDF]

Appendix

Appendix - B

[IMAGE OMITTED. SEE PDF]

References

Brown, G. P., C. Kelehear, and R. Shine. 2011. “Effects of Seasonal Aridity on the Ecology and Behaviour of Invasive Cane Toads in the Australian Wet–Dry Tropics.” Functional Ecology 25, no. 6: 1339–1347. https://doi.org/10.1111/j.1365‐2435.2011.01888.x.

Brown, G. P., B. L. Phillips, J. K. Webb, and R. Shine. 2006. “Toad on the Road: Use of Roads as Dispersal Corridors by Cane Toads (Bufo marinus) at an Invasion Front in Tropical Australia.” Biological Conservation 133, no. 1: 88–94. https://doi.org/10.1016/j.biocon.2006.05.020.

Brys, R., D. Halfmaerten, T. Everts, C. Van Driessche, and S. Neyrinck. 2023. “Combining Multiple Markers Significantly Increases the Sensitivity and Precision of eDNA‐Based Single‐Species Analyses.” Environmental DNA 5, no. 5: 1065–1077. https://doi.org/10.1002/edn3.420.

Buxton, A. S., J. J. Groombridge, N. B. Zakaria, and R. A. Griffiths. 2017. “Seasonal Variation in Environmental DNA in Relation to Population Size and Environmental Factors.” Scientific Reports 7, no. 1: 46294. https://doi.org/10.1038/srep46294.

Bylemans, J., E. M. Furlan, L. Pearce, T. Daly, and D. M. Gleeson. 2016. “Improving the Containment of a Freshwater Invader Using Environmental DNA (eDNA) Based Monitoring.” Biological Invasions 18: 3081–3089. https://doi.org/10.1007/s10530‐016‐1203‐5.

DBCA. 2024. “Cane Toads.” https://www.dbca.wa.gov.au/wildlife‐and‐ecosystems/animals/cane‐toads.

Doi, H., K. Uchii, T. Takahara, S. Matsuhashi, H. Yamanaka, and T. Minamoto. 2015. “Use of Droplet Digital PCR for Estimation of Fish Abundance and Biomass in Environmental DNA Surveys.” PLoS One 10, no. 3: e0122763. https://doi.org/10.1371/journal.pone.0122763.

Doody, J. S., B. Green, D. Rhind, C. M. Castellano, R. Sims, and T. Robinson. 2009. “Population‐Level Declines in Australian Predators Caused by an Invasive Species.” Animal Conservation 12, no. 1: 46–53. https://doi.org/10.1111/j.1469‐1795.2008.00219.x.

Doody, J. S., C. McHenry, M. Letnic, C. Everitt, G. Sawyer, and S. Clulow. 2018. “Forecasting the Spatiotemporal Pattern of the Cane Toad Invasion Into North‐Western Australia.” Wildlife Research 45, no. 8: 718–725. https://doi.org/10.1071/WR18091.

Doody, J. S., C. McHenry, D. Rhind, C. Gray, and S. Clulow. 2021. “Impacts of Invasive Cane Toads on an Endangered Marsupial Predator and Its Prey.” Endangered Species Research 46: 269–277. https://doi.org/10.3354/esr01158.

Doody, J. S., D. Rhind, B. Green, C. Castellano, C. McHenry, and S. Clulow. 2017. “Chronic Effects of an Invasive Species on an Animal Community.” Ecology 98, no. 8: 2093–2101. https://doi.org/10.1002/ecy.1889.

Doody, J. S., D. Rhind, C. M. McHenry, and S. Clulow. 2023. “Invasional Meltdown‐Under? Toads Facilitate Cats by Removing a Naïve Top Predator.” Wildlife Research 51: WR22177. https://doi.org/10.1071/WR22177.

Doody, J. S., R. Soanes, C. M. Castellano, et al. 2015. “Invasive Toads Shift Predator–Prey Densities in Animal Communities by Removing Top Predators.” Ecology 96, no. 9: 2544–2554. https://doi.org/10.1890/14‐1332.1.

Edmunds, R. C., and D. Burrows. 2019. “Development of an eDNA Assay for Cane Toad (Rhinella marina) (Report 19/08).”

Everts, T., D. Halfmaerten, S. Neyrinck, N. De Regge, H. Jacquemyn, and R. Brys. 2021. “Accurate Detection and Quantification of Seasonal Abundance of American Bullfrog (Lithobates catesbeianus) Using ddPCR eDNA Assays.” Scientific Reports 11, no. 1: 11282. https://doi.org/10.1038/s41598‐021‐90771‐w.

Everts, T., C. Van Driessche, S. Neyrinck, et al. 2022. “Using Quantitative eDNA Analyses to Accurately Estimate American Bullfrog Abundance and to Evaluate Management Efficacy.” Environmental DNA 4, no. 5: 1052–1064. https://doi.org/10.1002/edn3.301.

Everts, T., C. Van Driessche, S. Neyrinck, H. Jacquemyn, and R. Brys. 2023. “The American Bullfrog Exposed: Distribution, Invasion Fronts, and Spatial Configuration of Invasion Hubs Revealed by eDNA‐Based Monitoring and Environmental Assessments.” Management of Biological Invasions 14, no. 2: 201–220. https://doi.org/10.3391/mbi.2023.14.2.02.

Fediajevaite, J., V. Priestley, R. Arnold, and V. Savolainen. 2021. “Meta‐Analysis Shows That Environmental DNA Outperforms Traditional Surveys, but Warrants Better Reporting Standards.” Ecology and Evolution 11, no. 9: 4803–4815. https://doi.org/10.1002/ece3.7382.

Freeland, W. 1986. “Populations of Cane Toad, Bufo‐Marinus, in Relation to Time Since Colonization.” Wildlife Research 13, no. 2: 321–329.

Furlan, E. M., D. Gleeson, C. M. Hardy, and R. P. Duncan. 2016. “A Framework for Estimating the Sensitivity of eDNA Surveys.” Molecular Ecology Resources 16, no. 3: 641–654. https://doi.org/10.1111/1755‐0998.12483.

Furlan, E. M., D. Gleeson, C. Wisniewski, J. Yick, and R. P. Duncan. 2019. “eDNA Surveys to Detect Species at Very Low Densities: A Case Study of European Carp Eradication in Tasmania, Australia.” Journal of Applied Ecology 56, no. 11: 2505–2517. https://doi.org/10.1111/1365‐2664.13485.

Gelman, A., and D. B. Rubin. 1992. “Inference From Iterative Simulation Using Multiple Sequences.” Statistical Science 7, no. 4: 457–472. https://doi.org/10.1214/ss/1177011136.

González‐Bernal, E., M. J. Greenlees, G. P. Brown, and R. Shine. 2016. “Toads in the Backyard: Why Do Invasive Cane Toads (Rhinella marina) Prefer Buildings to Bushland?” Population Ecology 58: 293–302. https://doi.org/10.1007/s10144‐016‐0539‐0.

Greenlees, M., G. P. Brown, and R. Shine. 2020. “Pest Control by the Public: Impact of Hand‐Collecting on the Abundance and Demography of Cane Toads (Rhinella marina) at Their Southern Invasion Front in Australia.” Global Ecology and Conservation 23: e01120. https://doi.org/10.1016/j.gecco.2020.e01120.

Greenlees, M. J., S. Harris, A. W. White, and R. Shine. 2018. “The Establishment and Eradication of an Extra‐Limital Population of Invasive Cane Toads.” Biological Invasions 20: 2077–2089. https://doi.org/10.1007/s10530‐018‐1681‐8.

Hagman, M., and R. Shine. 2006. “Spawning Site Selection by Feral Cane Toads (Bufo marinus) at an Invasion Front in Tropical Australia.” Austral Ecology 31, no. 5: 551–558. https://doi.org/10.1111/j.1442‐9993.2006.01627.x.

Harrison, J. B., J. M. Sunday, and S. M. Rogers. 2019. “Predicting the Fate of eDNA in the Environment and Implications for Studying Biodiversity.” Proceedings of the Royal Society B 286, no. 1915: 20191409. https://doi.org/10.1098/rspb.2019.1409.

Jolly, C. J., R. Shine, and M. J. Greenlees. 2015. “The Impact of Invasive Cane Toads on Native Wildlife in Southern Australia.” Ecology and Evolution 5, no. 18: 3879–3894. https://doi.org/10.1002/ece3.1657.

Kellner, K. 2021. “jagsUI: A Wrapper Around Rjags to Streamline JAGS Analyses.” In (Version R package version 1.5.2). https://CRAN.R‐project.org/package=jagsUI.

Kyle, K. E., M. C. Allen, J. Dragon, et al. 2022. “Combining Surface and Soil Environmental DNA With Artificial Cover Objects to Improve Terrestrial Reptile Survey Detection.” Conservation Biology 36, no. 6: e13939. https://doi.org/10.1111/cobi.13939.

Macgregor, L. F., M. Greenlees, M. de Bruyn, and R. Shine. 2021. “An Invasion in Slow Motion: The Spread of Invasive Cane Toads (Rhinella marina) Into Cooler Climates in Southern Australia.” Biological Invasions 23, no. 11: 3565–3581. https://doi.org/10.1007/s10530‐021‐02597‐2.

Mayer, M., G. P. Brown, B. Zimmermann, M. J. Greenlees, and R. Shine. 2015. “Habitat Use of the Introduced Cane Toad (Rhinella marina) and Native Frog Species in Tropical Australia.” Journal of Tropical Ecology 31, no. 6: 531–540. https://doi.org/10.1017/S0266467415000474.

McColl‐Gausden, E. F., J. Griffiths, L. Collins, A. R. Weeks, and R. Tingley. 2023. “The Power of eDNA Sampling to Investigate the Impact of Australian Mega‐Fires on Platypus Occupancy.” Biological Conservation 286: 110219. https://doi.org/10.1016/j.biocon.2023.110219.

Mirimin, L., A. Hickey, D. Barrett, et al. 2020. “Environmental DNA Detection of Arctic Char (Salvelinus alpinus) in Irish Lakes: Development and Application of a Species‐Specific Molecular Assay.” Environmental DNA 2, no. 2: 221–233. https://doi.org/10.1002/edn3.60.

Muha, T. P., C. V. Robinson, C. Garcia de Leaniz, and S. Consuegra. 2019. “An Optimised eDNA Protocol for Detecting Fish in Lentic and Lotic Freshwaters Using a Small Water Volume.” PLoS One 14, no. 7: e0219218. https://doi.org/10.1371/journal.pone.0219218.

Phillips, B. L., G. P. Brown, M. Greenlees, J. K. Webb, and R. Shine. 2007. “Rapid Expansion of the Cane Toad (Bufo marinus) Invasion Front in Tropical Australia.” Austral Ecology 32, no. 2: 169–176. https://doi.org/10.1111/j.1442‐9993.2007.01664.x.

Raven, C., R. Shine, M. Greenlees, T. M. Schaerf, and A. J. Ward. 2017. “The Role of Biotic and Abiotic Cues in Stimulating Aggregation by Larval Cane Toads (Rhinella marina).” Ethology 123, no. 10: 724–735. https://doi.org/10.1111/eth.12645.

Rojahn, J., D. Gleeson, and E. M. Furlan. 2018. “Monitoring Post‐Release Survival of the Northern Corroboree Frog, Pseudophryne pengilleyi, Using Environmental DNA.” Wildlife Research 45, no. 7: 620–626. https://doi.org/10.1071/WR17179.

Schmidt, B. R., M. Kéry, S. Ursenbacher, O. J. Hyman, J. P. Collins, and N. Yoccoz. 2013. “Site Occupancy Models in the Analysis of Environmental DNA Presence/Absence Surveys: A Case Study of an Emerging Amphibian Pathogen.” Methods in Ecology and Evolution 4, no. 7: 646–653. https://doi.org/10.1111/2041‐210x.12052.

Schultz, M. T., and R. F. Lance. 2015. “Modeling the Sensitivity of Field Surveys for Detection of Environmental DNA (eDNA).” PLoS One 10, no. 10: e0141503. https://doi.org/10.1371/journal.pone.0141503.

Shine, R. 2010. “The Ecological Impact of Invasive Cane Toads (Bufo marinus) in Australia.” Quarterly Review of Biology 85, no. 3: 253–291. https://doi.org/10.1086/655116.

Shine, R., C. Everitt, D. Woods, and D. J. Pearson. 2018. “An Evaluation of Methods Used to Cull Invasive Cane Toads in Tropical Australia.” Journal of Pest Science 91, no. 3: 1081–1091. https://doi.org/10.1007/s10340‐018‐0975‐z.

Simpfendorfer, C. A., P. M. Kyne, T. H. Noble, et al. 2016. “Environmental DNA Detects Critically Endangered Largetooth Sawfish in the Wild.” Endangered Species Research 30: 109–116. https://doi.org/10.3354/esr00731.

Stoeckle, B. C., S. Beggel, A. F. Cerwenka, E. Motivans, R. Kuehn, and J. Geist. 2017. “A Systematic Approach to Evaluate the Influence of Environmental Conditions on eDNA Detection Success in Aquatic Ecosystems.” PLoS One 12, no. 12: e0189119. https://doi.org/10.1371/journal.pone.0189119.

Strickler, K. M., A. K. Fremier, and C. S. Goldberg. 2015. “Quantifying Effects of UV‐B, Temperature, and pH on eDNA Degradation in Aquatic Microcosms.” Biological Conservation 183: 85–92. https://doi.org/10.1016/j.biocon.2014.11.038.

Takahara, T., N. Iwai, K. Yasumiba, and T. Igawa. 2020. “Comparison of the Detection of 3 Endangered Frog Species by eDNA and Acoustic Surveys Across 3 Seasons.” Freshwater Science 39, no. 1: 18–27. https://doi.org/10.1086/707365.

Takahara, T., T. Minamoto, H. Yamanaka, H. Doi, and Z. i. Kawabata. 2012. “Estimation of Fish Biomass Using Environmental DNA.” PLoS One 7, no. 4: e35868. https://doi.org/10.1371/journal.pone.0035868.

Takahashi, M., M. Saccò, J. H. Kestel, et al. 2023. “Aquatic Environmental DNA: A Review of the Macro‐Organismal Biomonitoring Revolution.” Science of the Total Environment 873: 162322. https://doi.org/10.1016/j.scitotenv.2023.162322.

Tingley, R., M. Greenlees, S. Oertel, A. van Rooyen, and A. Weeks. 2019. “Environmental DNA Sampling as a Surveillance Tool for Cane Toad Rhinella marina Introductions on Offshore Islands.” Biological Invasions 21: 1–6. https://doi.org/10.1007/s10530‐018‐1810‐4.

Urban, M. C., B. L. Phillips, D. K. Skelly, and R. Shine. 2007. “The Cane Toad's (Chaunus [Bufo] Marinus) Increasing Ability to Invade Australia Is Revealed by a Dynamically Updated Range Model.” Proceedings of the Royal Society B: Biological Sciences 274, no. 1616: 1413–1419. https://doi.org/10.1098/rspb.2007.0114.

Yasumiba, K., R. A. Alford, and L. Schwarzkopf. 2016. “Seasonal Reproductive Cycles of Cane Toads and Their Implications for Control.” Herpetologica 72, no. 4: 288–292. https://doi.org/10.1655/Herpetologica‐D‐15‐00048.1.

Yates, M. C., D. J. Fraser, and A. M. Derry. 2019. “Meta‐Analysis Supports Further Refinement of eDNA for Monitoring Aquatic Species‐Specific Abundance in Nature.” Environmental DNA 1, no. 1: 5–13. https://doi.org/10.1002/edn3.7.

Zug, G. R., and P. B. Zug. 1979. “The Marine Toad, Bufo Marinus: A Natural History Resume of Native Populations.” Smithsonian Libraries 284: 1–58.