Biodiversity monitoring is a difficult and expensive activity that is chronically underfunded. Visual Encounter Surveys (VES) are a common monitoring tool for poikilothermic organisms in streams and rivers, but many species are challenging to detect with this method. Environmental DNA (eDNA) detection methods have been growing in popularity as a supplement or replacement for VES for aquatic species, but they are not yet widely adopted, in part due to perceived costs, a lack of understanding about their efficacy, and a lack of technical expertise. We implemented a paired VES and eDNA survey of 13 species (6 native and 7 invasive) in three rivers within and around Sequoia and Kings Canyon National Parks (SEKI). We found that species detection using eDNA methods was consistently higher compared to traditional snorkel VES, and eDNA was an accurate, cost-effective method for detecting biodiversity. Using eDNA and VES techniques, we were able to conduct a survey of aquatic biodiversity in areas within and neighboring the SEKI boundary. Our work highlights the potential for eDNA methods to be used in conjunction with traditional VES to minimize costs and improve capacity for resource management.
INTRODUCTION
Effective biodiversity monitoring is key to the successful management of resources. It can be used to understand trends, detect early warning signs, and quantify conservation success. However, like most natural resource programs, biodiversity monitoring is a persistently difficult and expensive activity that is chronically underfunded. The United States National Park Service (NPS) has the responsibility for managing unique landscapes that serve as refuges for biodiversity (Coetzee et al. 2014). This mission compels the NPS to conserve and maintain species richness and biodiversity, an edict that is challenging due to a lack of historical species data, rapidly changing conditions, and lack of funding (Leopold et al. 1963; Brody and Tomkiewicz 2002; Colwell et al. 2012; Haefele and Loomis 2016). The NPS, as well as other resource managers, often lack comprehensive surveys of even the most common species, and therefore species documentation becomes almost exclusively reliant on unverified records, incomplete museum archives, community science based observations, and coarse modeling of species ranges (Stohlgren et al. 1995; Willis and Birks 2006; Austin et al. 2013; Kyle 2014; Santos et al. 2014). Such is the case for Sequoia and Kings Canyon National Parks (SEKI).
The majority of SEKI's freshwater vertebrate biodiversity assessment has relied on historical observations (NPS 2013) and broad descriptions of regional species ranges (Santos et al. 2014), which are often poor predictors at the local scale (Maréchaux et al. 2017). Missing or inaccurate baseline data causes species with limited occurrence data to end up on the list of species extirpated from SEKI. Mischaracterizing species as extirpated or present with little evidence can bias condition assessments of habitats. Time and funding constraints make carrying out the basic biological surveys required to document the presences or absences of species a challenge for NPS personnel. There is a need for accurate survey methods for baseline detections to inform management decisions that can be deployed with limited resources.
Recent technical advances make environmental DNA (eDNA) a promising approach as an efficient method for assessing baseline biodiversity data (Coble et al. 2019; Lecaudey et al. 2019). Environmental DNA surveys consist of the identification of genetic material collected from environmental samples like water or other substrates. These methods are increasingly being demonstrated as effective for biodiversity monitoring (Bohmann et al. 2014; Thomsen and Willerslev 2014). Aquatic eDNA sampling has been used in the detection of native aquatic species (Goldberg et al. 2011; Pilliod et al. 2013a, 2013b; Wilcox et al. 2016; Raemy and Ursenbacher 2018; Adams et al. 2019) as well as surveillance and early detection of invasive species (Dejean et al. 2012; Mahon et al. 2013) and pathogens (Hall et al. 2016; Kamoroff and Goldberg 2017; Peters et al. 2018). Often, eDNA-based strategies provide rapid and cost-effective alternatives to traditional sampling methods (Jerde et al. 2011) and have promising applications for monitoring programs with limited resources.
Figure 1.
Study and opportunistic site along the Kings and Kaweah River watersheds, Tulare County, California. The six primary study site sampling locations included the Upper North Fork (1), Lower North Fork (2), Upper Middle Fork (3), Lower Middle Fork (4), Upper South Fork (5), and Lower South Fork (6) in the Kaweah River watershed.
To better understand the aquatic biodiversity of SEKI along its border with non-NPS lands, we initiated multi-species eDNA surveys paired with traditional snorkel visual encounter surveys (VES). We targeted rare native species, common species (native and nonnative), and highly invasive nonnative species with unknown occurrence. Using eDNA techniques, we also targeted pathogens that could not be easily detected using visual methods. Our primary goals were to (1) better document the aquatic species composition in the Kaweah River watershed along the foothill region inside and neighboring the SEKI boundary, and (2) determine the efficacy of multiple species detection using eDNA techniques. Our study lays a framework for eDNA surveys to accurately establish baseline data of species occurrence and to inform adaptive management strategies.
MATERIALS AND METHODS
Site Description
Sequoia and Kings Canyon National Parks, located in the southern Sierra Nevada Range in California, are co-managed as a single entity and encompass nearly 1 million acres and an elevational gradient of nearly 4000 m. The study took place within three tributaries of the Kaweah River (North, Middle, and South Fork) in the foothill region of SEKI nearby the community of Three Rivers, California (Figure 1). This area is characterized by its Mediterranean climate, with large perennial snowmelt rivers and small and largely ephemeral spring-fed streams. In the Kaweah River watershed, peak runoff typically occurs in late May to early June. During surveys and water sampling, bankful width was within 25–50% for all tributaries and water visibility ranged from clear to partial visibility.
Target Species
Our target species were composed of (1) rare native species including species on the SEKI extirpated list: foothill yellow-legged frog (Rana boylii; Baird 1854), riffle sculpin (Cottus gulosus; Girard 1854), and the Sacramento pikeminnow (Ptychocheilus grandis; Ayres 1854); (2) common native species: western toad (Anaxyrus boreas; Baird and Girard 1852), northwestern pond turtle (Actinemys marmorata; Baird and Girard 1852), and rainbow trout (Oncorhynchus mykiss; Evermann 1908); (3) common nonnative species: American bullfrog (Lithobates catesbeianus; Shaw 1802) and brown trout (Salmo trutta; Linnaeus 1758); (4) highly invasive species with unknown occurrences: New Zealand mudsnail (Potamopyrgus antipodarum; Gray 1853), quagga mussel (Dreissena bugensis; Andrusov 1897), and zebra mussel (Dreissena polymorpha; Pallas 1771); and (5) pathogens Bd (Batrachochytrium dendrobatidis) and Bsal (B. salamandrivorans), which were investigated using eDNA techniques only. Other, nontarget, species were noted during VES.
Sampling Protocol
We collected aqueous eDNA samples and conducted snorkel VES once a month from July to September 2018 along the North, Middle, and South Fork tributaries of the Kaweah River. We sampled and surveyed paired sites in each tributary, one site within the SEKI boundary and one site ∼5–12 km outside the SEKI boundary (n = 6; Figure 1). During each sampling period at each site, we collected two replicate eDNA samples as well as conducted snorkel VES in 100 m reaches beginning at the eDNA sampling location. We also opportunistically collected one or two eDNA samples at eight additional sites in SEKI and two additional sites outside of SEKI in Three Rivers, California, from August 2017 to December 2018 (n = 10; Figure 1). There were either no or incomplete VES conducted at the opportunistic sites and the data collected were used to document species occurrences; these data were not used in any of the statistical analyses (Appendix). We opportunistically collected additional eDNA samples at the paired sites (Table 1); these data were also not used in any of the statistical analyses.
eDNA Field Methods
We collected two 1-L water samples from the surface of a well-mixed (i.e., flowing water) area of the stream using disposable Whirlpaks. We then filtered each water sample through a 0.45 µm cellulose nitrate filter membrane in a single-use 250 mL funnel (Nalgene 147-2045) at the site using a hand vacuum pump with silicon tubing into a 2 L vacuum filter flask. For each sample, we used disposable nitrile gloves and single-use tweezers to transfer the filter into a 2 mL vial with 1 mL 95% molecular grade ethanol. To prevent sample contamination, we used single-use, disposable materials for all samples across all sampling locations, including between duplicate and blank samples. We decontaminated all field gear between sites using 1:10 ratio of commercial strength bleach (6.125% sodium hypochlorite) for 15 min to prevent contamination. To detect any contamination, we collected one negative control sample by filtering 1 L of distilled water on site after sample collection for ∼1/3 of sampling events.
eDNA Assays
To analyze the eDNA samples, we used species-specific quantitative PCR (qPCR) assays (Table 2). We used previously published assays for Bd (Boyle et al. 2004), Bsal (Blooi et al. 2013), New Zealand mudsnail (Goldberg et al. 2013), American bullfrog (Strickler et al. 2015), foothill yellow-legged frog (Bedwell and Goldberg 2020), brown trout (Carim et al. 2016), rainbow trout (Brandl et al. 2015), and Sacramento pikeminnow (Brandl et al. 2015). We developed and validated assays for quagga mussel, zebra mussel, riffle sculpin, western toad, and northwestern pond turtle (Table 2). Assays were developed using available sequence data from Genbank (Clark et al. 2016) and validated in silico using Primer Blast (Ye et al. 2012; Table 1). Each assay was then validated against tissue samples from the target and co-occurring nontarget species (Table 2). An internal positive control (IPC, ThermoFisher; or IC, Qiagen) was included in each reaction to test for inhibition.
eDNA Sample Analysis
We extracted and prepared eDNA samples for qPCR in a dedicated eDNA lab at Washington State University, Pullman, Washington. To extract samples, we followed the DNeasy/Qiashredder protocol described by Goldberg et al. (2011). We created an extraction negative control with each extraction set (∼24 samples) and an additional qPCR negative control with each qPCR set.
Samples were analyzed in triplicate and compared with a standard curve created from a 10-fold serial dilution in duplicate of tissue sample (diluted 10–3 to 10–6) or custom gblock DNA (10,000 – 10 copies per reaction; Integrated DNA Technologies, Coralville, Iowa). Samples were considered inhibited if the Cq value of the internal positive control was ≥3 cycles later than those from the standard curve. We ran each sample in triplicate and a positive sample was defined as any sample that showed exponential amplification in all three replicate wells. If a sample was inhibited, it was cleaned using a OneStep PCR Inhibitor Removal Kit (Zymo, Inc., Irvine, California) and rerun. We reanalyzed samples if the first uninhibited amplification yielded inconsistent results (i.e., one or two wells amplifying). If one or more replicates from the second run amplified, we considered the sample positive. Reactions were run on a Bio-Rad CFX96 Touch Real-Time PCR Detection System (Bio-Rad, Hercules, California). Each qPCR included 3 µL of DNA extract in a total volume of 15 µL. Cycling conditions for new assays can be found in Table 2.
Visual Encounter Survey Methods
At all of our main survey sites (n = 6), we paired eDNA sampling with snorkel VES. We started each VES at the location of eDNA sampling, following collection to prevent contamination, and extended 100 m upstream (Wilcox et al. 2016). Each 100 m transect was split into four 25 m sections. Two experienced snorkel surveyors surveyed the entire stretch, simultaneously and independently documenting the species detected underwater in all wetted areas of the survey segments. Throughout the survey, the shoreline was scanned to account for terrestrial stages of amphibious species. At the end of the stream segment, the surveyors convened to record the occurrence of each species in each 25 m segment.
Statistical Methods
To determine the efficacy of eDNA techniques and the detection of multiple species across the Kaweah watershed, we analyzed (1) how eDNA detection techniques compared to traditional snorkel VES, (2) how eDNA detection varied temporally (across sampling periods), and (3) how eDNA detection varied spatially depending on the location of an observed target species. To compare the effectiveness of eDNA techniques and paired VES we used a χ2 test and the average binomial detection across observable target species (n = 11), sites (n = 6), and site visits (n = 3). For VES and eDNA comparison, we only included species that could be detected via VES (i.e., we did not include Bd or Bsal in the analysis). We used generalized linear mixed-effects models (glmer) in the lme4 package in R (Bates et al. 2015) with binomial regression to determine if there was a change in eDNA detection across sampling periods (n = 3) for all the target species (n = 13) and sites (n = 6). We defined the predictor variable as the sampling period (July, August, or September), and the response variable was eDNA detection at the site (0 = species was not detected in either eDNA sample, 1 = species was detected in ≥1 eDNA sample). We used site as the random variable for the glmer models. All statistical analysis was conducted in R 4.2.0 (R Core Team 2022). Lastly, we used a binomial regression model to determine if eDNA detection was influenced by the location of a known target species (i.e., are you more likely to detect a species using eDNA if the species was observed near the eDNA sampling location within the 100 m transect?). Surveyors recorded the occurrence of each target species in four 25 m segments above the eDNA sampling point. For this model, we only included eDNA observations with paired VES observation and analyzed each eDNA sample separately. We defined the predictor variable as the location of the first observation of a target species (0–25 m, 25–50 m, 50–75 m, or 75–100 m upstream from the eDNA sampling location) and the response variable was eDNA detection in each sample (0 = species was not detected in the sample, 1 = species was detected in the sample).
RESULTS
Of our target species, we detected the following using eDNA and/or VES: Bd, brown trout, American bullfrog, rainbow trout, riffle sculpin, Sacramento pikeminnow, northwestern pond turtle, and western toad (Table 1). We detected these species at sites outside of the park, but we did not detect riffle sculpin, a rare native species, or Bd, an invasive pathogen, at any site inside of the park (Table 1). We only detected riffle sculpin using eDNA techniques. At varying sites, we detected target species in eDNA samples but not during snorkel surveys. Specifically, we detected northwestern pond turtle at three sites, brown trout and Sacramento pikeminnow at four sites, American bullfrog at five sites, and rainbow trout at seven sites using eDNA techniques and not during VES (Figure 2). We also routinely detected common, nontarget species via VES, including Sacramento suckers (Catostomus occidentalis occidentalis; Ayres 1854), California roach (Hesperoleucus symmetricus symmetricus; Baird and Girard 1854), Sierra and California newts (Taricha sierrae; Twitty 1942, T. torosa; Rathke 1833), Sierran treefrogs (Pseudacris sierra; Jameson, Mackey and Richmond 1966), and smallmouth bass (Micropterus dolomieu; Lacepède 1802).
Environmental DNA techniques were more effective at detecting species compared to traditional snorkel VES (χ2 test = 11.931, df = 1, P < 0.001; Figure 2). A positive eDNA detection was paired with 15% (n = 29) of negative VES observations, while only 0.5% (n = 1) of negative eDNA detections was paired with a positive VES observation (one occurrence of a single western toad larvae during visual survey) across all sites (n = 6) and sample periods (n = 3) (Figure 2). We found no evidence of a difference in eDNA detection for all target species across the sampling periods (z = –0.387, P = 0.6985). We also found no evidence that the location of an observed target species affected eDNA detection within the 100 m reaches (z = –0.003, P = 1). All negative control samples, extraction negatives, and PCR negatives tested negative.
Table 2.
Previously unpublished qPCR assays used in this project. PCR condition 1: 1X QuantiTect Multiplex PCR Mix (Qiagen Inc., Hilden, Germany), 0.2 µM of each primer and probe. Cycling protocol: 15 min at 95 °C then 50 cycles of 94 °C for 60 s followed by 60 °C for 60 s. PCR condition 2: 1X QuantiNova Pathogen Master Mix (Qiagen Inc., Hilden, Germany), 0.2 µM of each primer and probe. Cycling protocol: 2 min at 95 °C then 50 cycles of 94 °C for 5 s followed by 60 °C for 30s.
Figure 2.
(A) The average number of species detected for all target species (n = 10) with standard deviation error bar across all primary sites (n = 6) and sampling periods (n = 3) using both environmental DNA (eDNA) and snorkel visual encounter surveys VES). (B) The average positive detection for each individual species across all primary sampling sites (n = 6) and sampling periods (n = 3) for eDNA and VES methods. No bars indicate that the species was not detected.
All previous unpublished eDNA assays validated successfully against tissue samples. We were able to validate the eDNA assays for northwestern pond turtle using field eDNA samples and paired VES observations. For riffle sculpin, we detected the species using eDNA samples but not during the VES observations. We did not detect zebra or quagga mussels using eDNA or VES techniques nor did we detect western toad DNA using eDNA techniques although a single western toad larva was observed at one site.
DISCUSSION
We found eDNA sampling to be a successful survey strategy for aquatic species in the foothills of the Kaweah River. Our eDNA approach proved to (1) accurately detect a target species, especially cryptic and invasive species, across temporal and spatial scales (see Figure 2), and (2) better detect multiple target species compared to VES techniques. Environmental DNA was effective at detecting abundantly common species such as rainbow trout, Sacramento pikeminnow, and brown trout; rare and cryptic species such as the riffle sculpin and western pond turtle; and species not detectable using traditional VES techniques such as Bd. Our eDNA methodology (i.e., collecting two 1-L samples at a single location downstream of a 100 m reach) was consistent at detecting our target species over time (July–September) and across the entire survey reach (100 m). Species-specific studies have shown temporal variation in eDNA detection likely linked to changes in biological conditions (e.g., breeding status) or environmental conditions (e.g., water temperature) (Janosik and Johnston 2015; Bedwell and Goldberg 2020). However, in our study, we were able to consistently detect a range of taxa across temporal and spatial scales, highlighting eDNA as a promising tool for assessing biodiversity in this and similar systems during the season when they are most accessible.
Our work is consistent with other studies that have found aqueous eDNA detection to be more sensitive compared to traditional survey methods (e.g., Jerde et al. 2011; Smith and Goldberg 2020). Our study confirms previous work that eDNA is a valid alternative to VES for detecting presence of vertebrates and microorganisms in river or stream surveys (reviewed in Rees et al. 2014; Coble et al. 2019) and determining detection probability, and can be used as a cost-efficient method for biodiversity surveys (Smart et al. 2016; Lugg et al. 2018). Moreover, these methods are exceedingly simple to implement in the field and can be executed with limited training. However, there are tradeoffs to using eDNA survey methods and its utility is dependent on a host of abiotic and biotic factors (Curtis et al. 2021; Goldberg et al. 2018; Goldberg et al. 2016; Halstead et al. 2017; Kessler et al. 2020). As a result, although eDNA techniques may be cost effective and with higher sensitivity, pairing eDNA with VES may be most accurate when the detection of many species is required or when a target species is very challenging to detect using other methods. Additionally, paired VES and eDNA surveys allow for the possibility of detecting terrestrial life stages of amphibious species as well as other nontarget species.
The paired VES and eDNA surveys were effective at describing aquatic biodiversity in the foothill region of the Kaweah River tributaries within and neighboring the SEKI boundary. A primary product of this work was an up-to-date species list for SEKI, limiting the reliance on outdated or poorly supported data on species occurrences. The last natural resource condition assessment investigated museum records and observation databases to determine species presence (Thorne et al. 2013; York et al. 2013). The museum records and databases indicated that native fish in the Kaweah River drainage in SEKI were declining, based, in part, on historical observations of several species, like riffle sculpin and Sacramento pikeminnow, and lack of current observation or records. Our current assessment supports the assumption that native fish along the Kaweah River within the park are increasingly rare. We did not detect riffle sculpin within the park and had only one detection of Sacramento pikeminnow at a site in a single eDNA sample. However, riffle sculpin and Sacramento pikeminnow were detected at two and five locations outside of the park, respectively. Similarly, we did not detect foothill yellow-legged frogs at any of the sites (inside or outside of the park boundary), further supporting evidence of their regional extirpation (Hayes et al. 2016). The foothill yellow-legged frog southern Sierra Distinct Population Segment that encompasses our study area was recently listed as Endangered under California's Endangered Species Act and is currently proposed for federal listing. There were no detections of any common invasive aquatic species not known to occur in the area such as the zebra mussel, quagga mussel, or New Zealand mudsnail. Such species are particularly cryptic and devastating once established in a watercourse. Therefore, accurate and regular surveillance is critical to preventing establishment of such undesirable species. We found American bullfrogs were prevalent in all three tributaries outside the SEKI boundary and only in one tributary within the SEKI boundary. We detected Bd at three sites outside of the park (and no sites within the park), likely due to co-occurrence with invasive American bullfrogs (Yap et al. 2018).
Accurate inventories of regional biodiversity are important to understanding range expansion, reduction, or extirpation of species of special concern, nonnative invasive species, and pathogens. The inventory of the Kaweah River is critical to help park managers understand the vulnerabilities of park aquatic ecosystems. Our study can help SEKI land managers assess the allocation of resources based on empirical documentations rather than sparse historical records or broad-scale distribution models. Our work highlights an approach for assessing biodiversity using eDNA techniques, an accurate and cost-effective method that can be employed by SEKI as well as other land and resource managers.
ACKNOWLEDGMENTS
We thank Danny Boiano, Tyler Coleman, Shelby Moshier, Robert Seward, and Gabrielle Ruso for reviewing this manuscript. We also thank NPS employees Kelly Martin, Joshua Flickinger, and Colin Bloom for field work. We thank the SEKI Interpretation Division for contributing to the Community Science portion of this study, and Southern California Edison and N. and E. Leedy for private land access. Tissue samples for assay validation provided by Christopher Yee, Julia Burco, Cheron Ferland, Jamie Bettaso, Katherine Haman, Susan Barnes, Jason Baumsteiger, Steven Wells, National Park Service at Lake Mead, Glen Holmberg, and Brad Brandewie. This research was funded by the Sequoia Parks Conservancy through a donation from the Easter Day Foundation and a Science Learning Center grant. Funding for assay development provided by US Deptartment of the Interior, US Bureau of Land Management, US Fish and Wildlife Service, US Forest Service, and the Bonneville Power Administration. Laboratory analyses were conducted by Mary Sterling. This work was supported in part by the USDA National Institute of Food and Agriculture, McIntireStennis project 1018967. Funding for manuscript writing came in part from Stillwater Sciences.
