Introduction

Globally, translocations and reintroductions are increasingly used to counteract biodiversity losses1,2,3,4,5,6. Conservation translocations are often helpful for recovery of species following extirpations or genetic bottlenecks; however, attempts to implement conservation translocations have resulted in mixed success—ranging from recovery to reduced survival and repatriation failure7,8,9,10,11. The current patchwork of trial-and-error is inefficient, and a structured approach to refining reintroduction methods and planning would improve translocation outcomes8,12,13,14,15.

Approaching restoration and reintroduction with scientific methods and data—such as efficient survey techniques with sample sizes large enough to determine meaningful baselines—ensures effectual use of limited conservation resources and improves decision making16. Management decisions are frequently shrouded in uncertainty; for example, it is not always apparent why species have been extirpated from an area17. To reduce or explicitly incorporate uncertainties into decision-making, a well-structured, data-driven approach to planning reintroductions (and other conservation actions) is needed16,18,19,20.

Collaboration between managers and conservation scientists to incorporate conservation science and site-specific knowledge into the planning process increases the likelihood of achieving conservation goals21. For example, prioritizing and choosing locations for reintroductions and associated restoration efforts—such as invasive species removal—is informed by reintroduction biology22 as well as managers’ understanding of site-specific conditions23,24 and the feasibility of various management alternatives23,25,26. Prioritizing management alternatives based on site conditions facilitates interventions while minimizing waste and “analysis paralysis” (i.e., when problems go unaddressed while more data are collected and analyzed)27. Tools that explicitly facilitate managers’ decision-making process by directly addressing uncertainties and comparing alternative actions increases the likelihood of reintroduction success19,23,26.

Prioritizing management actions under uncertainty is paramount in the southern Sierra Nevada, USA. The Sierra Nevada encompass a biodiverse and dynamic landscape, with numerous endemic species subjected to frequent disturbances, such as flooding, landslides, drought, and wildfire, many of which are now exacerbated by climate change28. Thus, it is important to understand which recovery actions in this region would help meet conservation goals.

Many amphibians in the region are declining, including the stream-breeding foothill yellow-legged frog (Rana boylii), which is listed as endangered in the southern Sierra Nevada29,30. Conservation planning for R. boylii is complex because disease and non-native aquatic predators are among the leading causes of decline31,32,33,34. Non-native aquatic predators and competitors of R. boylii in the southern Sierra Nevada include American bullfrogs (Lithobates catesbeianus; hereafter “bullfrogs”) and multiple species of crayfish (Pacifastacus leniusculus, Procambarus clarkii, Orconectes rusticus, and Faxonius virilis; hereafter “crayfish”). Bullfrogs and crayfish may act as reservoir hosts for the amphibian pathogen Batrachochytrium dendrobatidis (Bd)35,36,37. Bullfrog presence has also been associated with increased Bd infection burdens, and in some cases, die-offs, in R. boylii35. Chytridiomycosis—the disease caused by Bd—may have contributed substantially to R. boylii’s historical decline, and chytridiomycosis-induced mortality events have been observed contemporarily in the species17,35,38. Non-native predator eradication programs—such as those removing non-native fish and bullfrogs—have been successful in restoring some Sierra Nevada aquatic habitats39,40. These eradication initiatives have laid the groundwork for a surge in ranid reintroduction efforts4,7,24,41,42,43,44, presenting an important opportunity to closely examine restoration methods and prioritization.

Many tools are available to help plan management actions. Although geographical information system tools and remote sensing can assist in habitat evaluation at a coarse scale, developing information about resource quality at the local scale requires time in the field45—whether through field surveys, direct observation, and/or sample collection—which may include environmental DNA (eDNA). Environmental DNA is trace DNA found in the water column and other substrates that can be used to detect species46,47. The relatively low cost, minimal invasiveness, simple field collection, and rapid processing of samples make eDNA detection methods well-suited for monitoring aquatic wildlife species39,48,49. The high sensitivity of eDNA analysis also allows for detection of invasive, cryptic, rare, or even presumed-extinct species50,51,52,53 that may escape other methods of detection, such as visual encounter surveys (VES). Environmental DNA methods can be used to test for the presence of R. boylii, and therefore help designate sites as either occupied and a potential source population, or as having restoration and recipient potential54. The sensitivity and geographic specificity of eDNA also make it an effective tool for focusing targeted efforts on invasive aquatic species eradications and monitoring for their reestablishment39,55.

Rana boylii are highly detectable in streams using eDNA methods51, as are the invasive bullfrogs and crayfish56,57 affecting R. boylii in the southern Sierra Nevada35,58,59. Invasive species removal efforts and conservation translocations of R. boylii to areas from which they have been extirpated is urgent in the region. In particular, the Tuolumne and Merced River watersheds were chosen as high priority conservation areas by management agencies due to the opportunities for restoration with an abundance of publicly accessible land and climate refugia24,60. Genetic variability of remaining R. boylii populations in the region is high61, providing good potential sources for translocations and reintroductions. Much uncertainty remains, however, in the distribution of R. boylii and invasive species in the southern Sierra Nevada. Therefore, combining site-specific species occurrence data with an explicit method for prioritizing management actions at sites would improve recovery of R. boylii in the region.

In this study, we applied a mixed methods approach to addressing uncertainties in the habitat restoration and reintroduction of R. boylii in the southern Sierra Nevada. Specifically, we addressed occurrence uncertainties for R. boylii, bullfrogs, crayfish, and Bd at the landscape scale by combining eDNA sampling with VES. Results of these surveys were then used to determine appropriate recovery actions through the development of a decision framework in concert with a prioritized menu of site-specific restoration actions. Our approach combines multiple species with multiple methods (i.e., eDNA, VES, and decision framework development) to directly inform decision making. The questions we addressed were (1) What is the distribution of R. boylii in the study area; (2) What is the distribution of biological threats to R. boylii (i.e., non-native aquatic predators and a pathogen) in the study area; and (3) How can this information be used to discern and prioritize restoration actions for R. boylii in a dynamic stream system?

Methods

Study area

The southern Sierra Nevada presents unique challenges and opportunities for prioritizing R. boylii recovery. Encompassing approximately 9540 km2, the Tuolumne and Merced River watersheds (Fig. 1) fall under primarily federal jurisdiction, facilitating the ongoing implementation of recovery actions and long-term monitoring as well as presenting opportunities to leverage professional expertise and funding. Conventional VES for R. boylii are challenging in this region’s rugged landscapes, where the terrain prohibits or reduces access to, and visibility in, many streams.

Figure 1
figure 1

Map of survey transects in the southern Sierra Nevada, California, USA, conducted June – September 2020 (green) and 2021 (blue).

Full size image

Site selection

To facilitate broad-scale sampling across the Tuolumne and Merced River watersheds, we selected focal sites (i.e., stream locations for which individual management action decisions can be made) to survey based on historical and contemporary R. boylii occurrence data (e.g., the California Natural Diversity Database62). We obtained historical and contemporary R. boylii and non-native species occurrence information from local, state, and federal wildlife and land management agencies (including the Bureau of Land Management, U.S. Forest Service, U.S. National Park Service, California Department of Fish and Wildlife, and U.S. Geological Survey). Using the R. boylii localities, we identified the sites that had the highest restoration potential, based on (1) accessibility, (2) habitat suitability, and (3) most recent observations. All sites needed to be accessible for surveys, future monitoring, and restoration activities. Therefore, we only considered sites if they were accessible via road or trail or if the landscape was amenable to cross-country travel. We also only considered restoration sites that were publicly owned, as private land ownership could hinder future monitoring efforts or restoration activities. We chose sites that were considered suitable based on a R. boylii-specific habitat model of connections between blocks of protected lands in the Sierra Nevada63. Of the sites that were considered suitable and accessible, we prioritized sites with recent observations, wherein sites where frogs were presumed extant (post-2006) were prioritized over historical (prior to 2006) sites. We identified 57 sites within 9 HUC-8 level sub-watersheds: North Fork Merced River (N = 5); South Fork Merced River (N = 12); main stem Merced River (N = 16); South Fork Tuolumne River (N = 5); Middle Tuolumne River (N = 4); North Fork Tuolumne River (N = 2); and main stem Tuolumne River (N = 13).

Sample collection and analysis

We conducted VES and eDNA surveys June–September 2020 and 2021 to target when flows are low and eDNA from the focal species is more likely to be concentrated and therefore more detectable51,64. Because we designed the study to be comprehensive (i.e., for maximum detection of the target species in specific management areas of interest), we surveyed throughout all sites of interest rather than conducting random sampling. Water samples were collected using 5-µm polyethersulfone self-desiccating filters and a backpack-based pump system (Smith-Root, Vancouver, Washington, USA)65,66. The 5-µm pore size was selected to allow for large volumes of water to be filtered continuously. Because eDNA has limited transport in small streams51, we surveyed by sampling continuously while walking slowly upstream in the water, with the filter extended out on a telescoping pole (≥ 2 m), collecting water from pools and the main channel. Filter packs were sealed back in their packaging and stored in the dark at ambient temperature. Pump sample volume was set to 0.5 L/min, using a 190-mL offset in the sample volume quantification to account for the water held in the input tube and pump, per the manufacturer’s instructions67. Filter samples were collected as a single sample per transect, and transects were made continuous by replacing each filter when it clogged or the pump broke prime (i.e., lost effective vacuum pressure), whichever came first. Multiple filters were frequently collected within each stream (mean = 4.3; range = 1–22 filters). Fresh nitrile gloves were worn when handling each filter housing. Field blanks were collected by filtering 1 L of distilled water at the end of each field day prior to decontamination of equipment. All external gear (i.e., waders, boots) was decontaminated with a 10% bleach solution when moving between streams to avoid spreading microorganisms (such as Bd) between sites. The backpack sampler was also decontaminated internally with a 2% bleach solution, per the manufacturer’s instructions67. Two independent-observer visual encounter surveys were also conducted within each eDNA stream transect, so that three total replicate surveys (one eDNA and two VES) were conducted within each stream reach. The VESs were conducted while wading in-stream wherever possible, because this is the best vantage point from which to see R. boylii basking on rocks or the edges of rocks in the stream. The rugged terrain in many areas also makes the banks impassable by foot. The VESs occurred at least 30 min after eDNA sample collection68,69,70, either the same day or within 48 h of eDNA sampling. Stream water temperature was measured using a hand-held thermometer at the beginning of each survey. At two sites in 2021, we collected Bd swabs from the skin of early post-metamorphic R. boylii following standardized procedures71.

Because the primary objective of this study was to determine which actions would be most beneficial in specific areas, we collected and analyzed the data in a way that could inform these decisions. Previous studies have demonstrated very high probabilities for eDNA detection of our target species39,51,72,73. We paired eDNA with VES for the additional ability to detect species that could be missed with eDNA methods. Given the purpose of the study, we did not choose sites to sample randomly or attempt to draw inference to unsampled areas. Instead, we aimed to survey primary places of interest that had the highest conservation and restoration potential (refer to Site Selection, above), in order to inform which recovery actions would be the most beneficial to attempt at specific sites.

We extracted DNA from samples using the Qiashredder/DNeasy protocol of Goldberg et al.53 in a restricted access laboratory dedicated to low-quantity samples, and analyzed samples using published quantitative polymerase chain reaction (qPCR) assays for R. boylii51, American bullfrog (L. catesbeianus74), and the amphibian chytrid fungus (B. dendrobatidis; Bd75). We also developed and applied an assay for the signal crayfish (P. leniusculus; Supplementary Molecular Methods; Supplementary Table S1). Because Bd was only detected in one filter in 2020, we did not assay for Bd in all filter samples in 2021. In 2021, we only tested the water filters for Bd in the streams where we were collecting Bd swabs at the same time. We collected the eDNA samples immediately before collecting the Bd swabs. In 2020, we included surveys and assays for northwestern pond turtle (Actinemys marmorata) and three trout species to determine whether A. marmorata and native trout may be positive ecological indicators for R. boylii, and the extent to which non-native trout may be R. boylii predators. However, due to funding constraints, we did not test for A. marmorata or trout species in the 2021 samples and they were therefore not included in this analysis (refer to Supplementary Information).

Environmental DNA and Bd swab samples were initially analyzed in triplicate and considered positive if they tested consistently positive across the three wells. Samples that produced inconsistent results (1 or 2 positive) were rerun in triplicate and considered positive if they tested positive in ≥ 1 wells in both triplicate runs. An internal positive control (ThermoFisher Scientific, Waltham, Massachusetts, USA) was included with each assay, and samples were cleaned with a OneStep™PCR Inhibitor Removal Kit (Zymo Inc., Irvine, California, USA) if the Cq for the internal control was delayed > 3 cycles. If samples continued to test as inhibited after treatment, they were further diluted 1:10 and then 1:100. Any sample needing 1:100 dilution that tested negative was considered inconclusive and excluded from the study. Quantitative standards were run in duplicate for each species, consisting of either DNA samples derived from tissue from external skin, diluted 1:1000 through 1:1,000,000 in QuantiTect Nucleic Acid Dilution Buffer (Qiagen®), or gblock™ standards (Integrated DNA Technologies, Coralville, Iowa, USA), in a tenfold series dilution 10,000 to 10 copies in QuantiTect Nucleic Acid Dilution Buffer (Qiagen; refer to Supplementary Molecular Methods). All eDNA and VES data were managed, analyzed, or visualized using Microsoft Excel, ArcGIS Pro, and R76,77,78.

Decision support tool development

At the request of the U.S. Fish and Wildlife Service, we conducted an assessment of feasible R. boylii management actions based on the VES and eDNA survey results. After evaluating sites of interest for the threat of non-native predators with eDNA and VES, we identified and prioritized feasible R. boylii recovery actions for each site. This assessment was done in consultation with management agency stakeholders, including U.S. National Park Service staff that have conducted California red-legged frog (Rana draytonii) translocations to Yosemite Valley4,24,44, as well as U.S. Forest Service staff that have conducted monitoring at R. boylii-occupied sites.

To assess the suitability of specific management options at different sites, we first placed the species detection data in a site-by-species matrix, focusing specifically on R. boylii and non-native predator occurrence, as these are the primary species of concern for restoration decision making (Fig. 2). Each site was then assigned a status category based on the species detected at that site: R. boylii present without non-native predators (category “A”); R. boylii present with non-native predators (“B”); R. boylii absent without non-native predators (“C”); and R. boylii absent with non-native predators (“D”) (Supplementary Box 1).

Figure 2
figure 2

Environmental DNA (eDNA) and visual encounter survey (VES) detections for the foothill yellow-legged frog and non-native aquatic predator species that threaten its survival and recovery. LICA = American bullfrog (Lithobates catesbeianus); PALE = signal crayfish (Pacifastacus leniusculus); RABO = foothill yellow-legged frog (Rana boylii); UNCR = unknown crayfish species. Signal crayfish were detectable only with eDNA; “unknown crayfish” were detected only with VES. Numbered sites represent streams where R. boylii are present, to protect the exact locations of this endangered species. SF = South Fork; MS = main stem; Ck. = creek.

Full size image

At the site level, status categories were then applied to a menu of research needs and plausible actions under consideration by management agencies (“potential management actions”). Potential management actions included: habitat assessments; demographics and baseline monitoring; bullfrog and crayfish control; species detections/surveys; in situ rearing/captive rearing and release; and reintroductions or translocations (Supplementary Information). The menu, informed by the restoration categories, was placed within the context of a decision tree (Fig. 3) to provide a stepwise framework to evaluate site-specific management actions for R. boylii conservation, depending on priority and threat. For example, at a site where R. boylii is present without non-native predators (category A), the management actions could include collecting demographic baseline monitoring data to determine the status of the R. boylii population, and determining whether captive propagation (whether in situ [i.e., in-stream] captive propagation and reintroduction) or translocations from another site or ex situ captivity are appropriate (Supplementary Table S2).

Figure 3
figure 3

Decision tree for potential Rana boylii conservation actions.

Full size image

Finally, sites were assigned to priority classifications based on both status category and site-specific constraints (Table 1; Supplementary Table S2). The highest-priority sites are those where potential near-term management actions are expected. This may include sites where bullfrogs and crayfish currently threaten existing R. boylii populations, or where these species are within dispersal distance of R. boylii populations and therefore may pose an imminent invasion threat. High priority sites are also those where R. boylii occurs and demographic data would be needed to determine whether the population may serve as a source population, or whether captive propagation would be needed to bolster existing populations. One high-priority site (Merced 03; refer to Supplementary Table S2) has suitable habitat for R. boylii and is therefore a suitable translocation recipient site. Moderate-priority sites are those where no bullfrogs or crayfish were detected, but further evaluation would be needed for R. boylii reintroduction feasibility. Low priority sites are those where bullfrogs, crayfish, or both were detected and R. boylii are absent, and the site is not currently within dispersal distance of suitable translocation sites. In addition, a site may be regarded as low priority even if bullfrogs and crayfish were not detected if other factors make the site unsuitable for any of the potential management actions. For example, Fern Spring is currently unsuitable for reintroductions because it is in an extremely high-traffic area; Tuolumne 03, although occupied by R. boylii, is a low priority for any of the potential actions because of its remote location, which both protects it from the threat of non-native predators being introduced but also makes most of the potential management actions infeasible (refer to Supplementary Table S2).

Table 1 Sample management actions and prioritization of sites by status for the North Fork Merced River.
Full size table

Results and discussion

Species detections

In total, we surveyed 57 sites and 186 km of stream (Fig. 1), filtering 2690 L of water through 367 eDNA filters (mean 47.2 L per site; mean 6.4 filters per site) in addition to 76 field blanks. All field blanks were assayed for all species associated with the year of sampling and tested negative. Rana boylii were detected at 11 sites and non-native predators (bullfrogs, crayfish, or both) co-occurred with R. boylii at five of these sites (Fig. 2). Neither R. boylii nor non-native aquatic predators were detected at 31 out of the 57 sites (54%). Mean stream temperature during eDNA sample collection was 18.16 °C ± 4.83 standard deviation (SD).

We detected Bd in a single filter (two out of three replicates, > 33.7 internal transcribed spacer [ITS] copies/L) in 2020 (site Tuolumne 02). We detected R. boylii in this stream as well, but not from the same filter. It is unlikely that the non-detection of Bd eDNA at other sites represents the absence of Bd at these sites, as Bd has been detected in R. boylii swabs from other sites in the study area (AJA, unpublished, 2020–2021). In 2021, we detected Bd in six swabs collected from 11 R. boylii metamorphs in two different streams (one Bd positive in Tuolumne 08 and five in Merced 01). However, no Bd eDNA was detected from either of these streams. Swab-detected Bd prevalence and mean infection burden were high in R. boylii in Merced 01 (100% and 1.54 × 106 copies, respectively). The corresponding eDNA extract from the filter collected in the same stream immediately prior to swabbing was still inhibited after treatment for inhibitors and a 1:10 dilution of those treated extracts. Therefore, only a very strong signal would have been able to be detected from this sample. The high level of inhibition is likely because of the low streamflow when this sample was collected, increasing the likelihood of PCR-inhibitory compounds getting into the filter. It is not clear why we did not detect Bd in the broader sampling of these stream systems, but it seems unlikely that these results reflect true Bd prevalence. In a nearby lentic system, Bd eDNA detection was sensitive enough to predict chytridiomycosis outbreaks79, and Bd was detected in R. boylii eDNA samples from streams in the central Sierra Nevada54. Stream eDNA does not spread far from a source54 and detection requires continuous input80. If Bd shedding is temporally patchy (e.g., if many individuals in a population respond to temperature conditions with certain basking behaviors81,82), detection in lotic systems may be rare when infected animals are at low densities. Bd did not play a major role in the decision support tool criteria because the data for it were so sparse; however, the high Bd loads on swabs collected at site Merced 01 have been communicated to the relevant management agencies.

Actinemys marmorata (the native turtle that we sampled for in 2020) and R. boylii frequently co-occur throughout their range83, and we detected A. marmorata with either eDNA or VES in four of eight R. boylii-occupied streams. This indicates that some of these sites are more intact systems and that the two species have overlapping ecological requirements, further helping to prioritize sites for reintroduction. Therefore, restoring habitat for R. boylii may also benefit the conservation of A. marmorata, which is proposed for listing under the U.S. Endangered Species Act84.

The benefits of eDNA sampling in conjunction with VES varied by species. In one instance, the survey crew identified a small frog as R. boylii while eDNA analysis only detected bullfrogs at that site. After the coauthors—who have decades of field experience with the species in question—examined the photographs taken by the surveyors, it became clear that the frog was indeed a juvenile bullfrog, not R. boylii. These two species can be easily mistaken when the individual is small, exemplifying the advantages to having a multi-method design that combines natural history knowledge and observation with assay-driven eDNA techniques. In addition, we detected A. marmorata at four more streams with eDNA than with VES (eDNA = 9; VES = 5), indicating that eDNA is a much more sensitive method for A. marmorata detection than VES. This result is consistent with recent findings from farther north in California85. Only 20% of visual “unknown crayfish” detections and P. leniusculus eDNA detections overlapped (Fig. 2), indicating that there is at least one additional crayfish species present in the study area that was not assayed for, and that P. leniusculus is more difficult to detect visually than the other species. The other crayfish species known to occur in the Merced and Tuolumne River watersheds include the red swamp (P. clarkii), rusty (O. rusticus), and virile (F. virilis) crayfish, all of which are outside of their native range in the study area86. These individual species are not identifiable in the field during VES surveys, often requiring gonopod examination under a microscope87. Finally, eDNA could likely provide information on pathogens that is not available from field observations; however, in this case we had almost no eDNA detection of Bd, which seems unlikely to reflect true Bd prevalence in these streams and differs from what we know of Bd prevalence in local lakes79.

Advancing the use of eDNA to landscape-scale decision making

Through the process of ascertaining site-appropriate restoration actions based on the eDNA and VES data, we identified 15 high-priority sites, 30 moderate-priority sites, and 12 low-priority sites in the study area (Supplementary Table S2). We detected bullfrogs and crayfish successfully using both eDNA and VES, and the use of eDNA provided an early warning system for bullfrog and crayfish invasions into R. boylii-occupied streams. At an R. boylii-occupied site where bullfrogs had previously been eradicated (Tuolumne 04), bullfrogs were not detected during routine VES monitoring and eDNA sampling conducted from 2017 to 2019 (R. Grasso, Yosemite National Park, written communication, 2021). We did not detect bullfrogs with eDNA or VES at this site in 2020; however, in 2021, we detected bullfrogs there with eDNA, but not VES. Bullfrogs were detected with eDNA, but not VES, at site Tuolumne 08 for the first time in 2021—in a portion of the stream where only R. boylii were detected in 2020. At Jordan Creek—a site with previously known bullfrog populations—bullfrogs were detected with eDNA but not VES (Fig. 2). Similarly, P. leniusculus eDNA was detected at six sites where the species was not detected visually (Fig. 2). These results highlight the importance of repeated, multi-year sampling for non-native species invasions in dynamic stream landscapes, for which eDNA sampling appears to be well-suited.

Using an effective monitoring strategy is critical when managing species in dynamic systems faced with ongoing threats compounded by climatic extremes. In drought years, reduced streamflows facilitate bullfrog breeding in reaches of stream that they have not previously been able to occupy in wetter years35. In the Sierra Nevada, signal crayfish populations are also partly regulated by streamflow: artificial impoundments provide refugia that facilitate upstream repopulation after scouring from high-flow events88. Similar phenomena have been observed with bullfrogs in central coastal California, where drought facilitated the movement of Bd-carrying bullfrogs to R. boylii-occupied streams, likely triggering observed chytridiomycosis outbreaks and the first record of chytridiomycosis-induced mortality in R. boylii35.

We leveraged a fine-scale understanding of the dynamics between R. boylii and non-native aquatic predators to inform reintroduction decision making and management action prioritization. Without field observations and survey efforts, and only using existing range maps and knowledge, we would not have had the information necessary to inform the planning and prioritization of on-the-ground management actions. For example, the identification of new bullfrog invasions into R. boylii-occupied sites has clarified dynamics between drought and potential invasive species control actions when R. boylii and bullfrogs occur in proximity. In addition, identification of suitable sites for R. boylii demographic studies where its most pressing threats are absent has increased the likelihood that these areas can be prioritized by conservation managers.

Conclusion

We applied a multi-method approach to a data-poor, landscape-scale conservation challenge: identifying and prioritizing potential management alternatives for an endangered species in a dynamic stream system. Our research advances the integration of eDNA technology directly into conservation decision-making. Our continuous in-stream sampling technique was used for the first time at a landscape scale, and enabled eDNA detection of sparsely distributed species. This new sampling technique is powerful, both for uncovering small remnant populations of rare, imperiled species and for early eDNA detection of invasive species so they can be mitigated early in the invasion process. Our combined approach—which incorporated eDNA-derived results with a decision framework—could culminate in better-informed reintroduction decisions that streamline efforts, reduce costs, and improve the success of reintroductions and associated management efforts in the southern Sierra Nevada.

The decision support tool additionally gives managers the ability to prioritize actions based on species occupancy and site-specific criteria that can be easily altered to meet shifting management needs as more information becomes available.

To guide additional landscape-scale management actions in the future, our occurrence results can be incorporated into species distribution models to predict community responses to climate change89. Because the Sierra Nevada region is increasingly recognized as an important area for climate change refugia90,91,92,93, some portions of the ecoregion may provide future habitat for species needing cooler and wetter conditions60. This study provides a pathway for informed management of a declining freshwater species in an uncertain future and demonstrates how fine-scale, multi-species data obtained with eDNA can be used to prioritize actions at a landscape scale.