| TIME | DAY 1 – OCTOBER 26, 2026 | PRESENTERS |
| 9:00 AM | Workshop | Audrey Harris (PSMFC/ IDFG), Rebekah Horn (CRITFC) |
| Genetics 101 for Fisheries Professionals |
| Genetic data are widely incorporated into fisheries management, with applications including quantification of adaptive diversity and population resilience to climate change, genetic identification of samples to species and origin in forensic cases, and genetic monitoring of hatchery stocks through pedigree-based analysis. Though managers and biologists may not collect genetic data firsthand, a baseline understanding of the types of genetic data available, their utility, and how results apply to management plans can prove extremely useful. This workshop is designed for non-geneticists within the field of fisheries and covers best practices for DNA sample collection and the types of genetic analyses and results fisheries professionals can expect to encounter, with a specific focus on parentage-based tagging and genetic stock identification. Providing fisheries professionals with the tools and terms to understand the analyses being performed and how to interpret relevant results will empower them to understand how genetics can be used to answer research and management questions in their own study systems and species. Participants will experience a mix of short lectures from experienced geneticists and hands-on activities with workshop material that can be taken home for future reference. The workshop is suitable for fisheries professionals of all levels, and no previous genetic experience is required. |
| 12:00 PM | Check-in Registration | |
| 1:00 PM | Welcome/Orientation | |
| 1:10 PM | Plenary | Scott Heppell (Oregon State University) |
| | Data QA/QC Session | |
| 1:40 PM | Transparent management inputs – data workflows incorporating GitHub | Erik Suring (ODFW) |
| Recent changes in Oregon coast Coho Salmon (Oncorhynchus kisutch) monitoring and downstream fisheries management information inputs gave us the opportunity to review our data and analysis workflows and share them using GitHub. We consolidated publicly available Oregon Department of Fish and Wildlife Salmonid Life Cycle data and published the R code used to process individual site data into a coast-wide marine survival index (MSI) in a GitHub repository. These data also feed into the Oregon Coast Natural marine survival forecast and the forecast code with covariates was also shared in GitHub. The availability of the data and code improved the Pacific Fisheries Management Council review of the new MSI analysis and Oregon Production Index Technical Team (OPITT) review for fisheries management. The GitHub repository could be cloned by reviewers and fisheries managers for assessment prior to meetings; there were 53 repository clones in the days leading up to the 2026 OPITT meeting. |
| 2:00 PM | OXBOW: Wrangling Fish and Environmental Data into a Single SQL Database Using Advanced Claude Models | Derrek Faber (ODFW) |
| Fisheries monitoring generates a sprawl of disconnected data: Passive Integrated Transponder (PIT) tag detections, capture and biological records, and the environmental conditions water temperature, streamflow, weather, and tidal stage, that give them meaning. These live in separate spreadsheets, agency databases, and the regional PTAGIS system, leaving biologists to reconcile fish and environmental records by hand. Outside the Columbia, Snake, and Willamette basins, studies fall beyond the purview of PTAGIS and frequently lack any structured database. OXBOW closes this gap by wrangling these heterogeneous streams into a single relational (SQL) database behind a web portal, so a study’s fish and environmental data are organized, validated, and queryable in one place. |
| Equally notable is how OXBOW was built. Tasks once reserved for development teams like designing a PostgreSQL schema that unifies fish and environmental data, standing up a containerized Django application, wiring the PTAGIS API, and aligning environmental conditions to every detection, were accomplished by describing each problem in plain language to advanced Claude models. The conversational interface closes the gap between domain expertise and working software: a biologist can build and refine an entire data system through dialogue, without formal training in programming or databases. The error-prone work of exporting, time-matching, and merging disparate datasets by hand is gone; fish and environment arrive already joined. OXBOW is offered less as a product than as a proof of concept: with tools like Claude, the people who know the data best can build the integrated databases they need themselves. |
| 2:20 PM | The Data Mindful PDF: How to Make Table-filled Reports More Useful | Brian Maschhoff (Salmonetics) |
| The Portable Document Format (PDF) is a standardized ISO 32000 file type developed by Adobe to present and exchange documents independently of software, hardware, or operating systems. These files encapsulate a fixed-layout description of text, fonts, and graphics to preserve the original formatting across all platforms. Over the last few decades, it has become the de facto standard format for publications and reports, including those by agencies and other organizations supporting salmon and steelhead research in the Columbia Basin. These documents usually embed a lot of data in tables and figures meant to be consumed visually: data in, and meaning oozes out. But if the reader wants to make use of that data in other ways, the PDF format is less than ideal as it is designed around placement of text and graphics, not around the structure of the data itself. If the stars align, one can successfully select and copy data from a table and paste it into a spreadsheet, but that is more often a happy happenstance than the rule. In this talk, I will show some methods and supporting technologies by which report creators can supercharge their PDFs to make data extraction a better experience for curious human and non-human end-users. |
| 2:40 PM | Automating salmon counting and identification with computer-vision deep learning | Dr. William I. Atlas, Sami Ma, Charles Xu, Henry Fang, Thor Veen (Wild Salmon Center; Salmon Vision Collaborative) |
| Around the Pacific Rim, salmon populations are monitored using a variety of technologies including RGB video collected in weirs and fishways, forward looking sonars deployed at a fixed site in the migration corridor, and with overflight counts that are typically conducted from helicopters or fixed wing aircraft. Each of these monitoring approaches has their benefits, but all are characterized by high recurring costs for data collection and review. For the last five years the Salmon Vision collaborative has been working with BC First Nations, Alaska Native Communities, and government management agencies to build computer-vision tools for automated analysis of these data streams. Here we report on our progress, including RGB video models that can count salmon species with <5% error relative to human counts at sites with robust training datasets. |
| 3:00 PM | Break | |
| | Data QA/QC Continued | |
| 3:20 PM | Developing a Fish Age Database: Lessons Learned & Future Considerations | Matt Weeber (ODFW) |
| Reliable estimates of age composition are an integral component of the technical management process in Pacific Salmon Commission fisheries management. These data are used in forecasting stock abundance, assessing freshwater fisheries' impacts, and estimating escapement to the spawning grounds by age. Scale analysis is the preferred method of estimating age composition of natural stocks of Pacific salmon. Each year, several biologists from different projects within the Oregon Department of Fish and Wildlife (ODFW) analyze thousands of lines of biological data from Fall Chinook Salmon monitoring. The data is generated from creel, mark/recapture, trap, and spawning ground surveys. The data stretch back 30 plus years and cover all major coastal basins in Oregon. The challenges are that the data are vast and diffuse; the data are housed within a mixture of Access databases and hundreds of electronic spreadsheets. In 2016, ODFW set out to create and use a centralized, relational database housing all of the biological data from coastal Chinook Salmon monitoring from the West Region. The following presentation covers the objectives of the project, as well as lessons learned and future considerations. |
| 3:40 PM | Structuring Knowledge, Not Just Data: An Ontology Framework for Comparable, Reproducible Conservation Indicators | Elizabeth Ng, Erika Rubenson, Sam Haffey (Spheros Environmental) |
| Conservation and natural-resource organizations must often aggregate heterogeneous monitoring data into comparable indicators, yet biological, ecological, and socio-economic observations can vary widely in source, spatial format, and temporal scale. The burden of synthesizing and contextualizing these data often falls on a few staff, making it difficult to automate workflows or scale up reporting. Working with a national conservation grant-making foundation, we developed a knowledge organization system comprising a controlled vocabulary, taxonomy, and ontology. This framework makes the relationships within the data explicit, machine-readable, and queryable, supporting interoperability and enables automated aggregation of metrics and key performance indicators at multiple scales. Drawing on document review, consultation with staff, and a survey of conservation ontologies, we modeled the organization's structure and reporting needs. We then formalized the taxonomy and ontology using entity-relationship diagrams for encoding in Resource Description Framework and Web Ontology Language. Central to the framework is the separation of three concepts: Conservation Action, Conservation Outcome, and Observation. Conservation Action is defined at the finest observable scale and situated in space and time. This distinction resolved the conflation of action with result and attributes outcomes to the actions that produced them. It also accommodates context- and funder-specific definitions of terms such as "restored" and supports inference of new relationships across projects. The framework was built on the Extensible Observation Ontology and the Conservation Measures Partnership Unified Classification of Conservation Actions. The resulting system is an extensible, reproducible foundation for quantifying conservation impact and sharing comparable indicators across organizations, with applications in data discovery and artificial intelligence. |
| 4:00 PM | Using Manual Text Annotation and Machine-learning-assisted Natural Language Processing (NLP) to Extract, Classify, and Compare Conservation Actions and Strategies Across Management Plans | Jacob Usinowicz, Michelle Stantial, Lauren Kemper, Erika Rubenson, Elizabeth Ng (Spheros Environmental) |
| Conservation planning across public lands involves large volumes of complex, agency-specific language that can obscure shared priorities and coordination opportunities. We present a methodology for systematically extracting, classifying, and comparing conservation actions and strategies across management plans using a combination of manual text annotation and machine-learning-assisted natural language processing (NLP). This approach was applied to 50 federal, state, and regional management plans across the Sierra-Cascade-Inyo and Southern California regions of California, in partnership with the California Wildfire and Forest Resilience Task Force, the Climate and Wildfire Institute, and the National Fish and Wildlife Foundation. Using the Conservation Measures Partnership Unified Classification of Conservation Actions (Version 2.0) as a foundation, we developed an expanded hierarchical taxonomy tailored to California's conservation context. Reviewers annotated plan text using Label Studio to identify conservation actions and outcomes, which were then mapped to the taxonomy using synonym-based matching logic and a custom named entity recognition (NER) model. Relative emphasis of each action and strategy within and across plans was assessed through frequency analysis and quantile-based ranking, enabling standardized comparison despite variation in language and document structure. Applied to this case, the methodology revealed shared priorities such as habitat restoration and fuel reduction, while surfacing gaps in cross-jurisdictional coordination. The resulting framework and data products support more informed, collaborative decision-making by land managers and policymakers. This methodology is broadly transferable to other regions or taxa where synthesis of planning documents could improve conservation coordination and outcome tracking. |
| 4:20 PM | Modernizing Spawning Ground Surveys with Mobile Technologies: Joint Insights from Oregon and Idaho on Survey123 Design, Deployment, and Field Application | Kasey Bliesner, Joseph Feldhaus, Melody Feden, Joe Dittmer, Alex Woolen, Erik Suring, Briana Sounhein, Matt Weeber (ODFW) |
| Evan Brown, Megan Merkley, Bryan Boson, Chris Harrington, Brock Lipple (IDFG) |
| Mobile data collection tools continue to reshape aquatic monitoring by improving accuracy, efficiency, and workflow integration across field programs. This joint presentation from the Oregon Department of Fish and Wildlife (ODFW) and Idaho Department of Fish and Game (IDFG) highlights the development and application of Survey123 mobile data collection data systems supporting salmonid spawning ground survey efforts across both states. |
| 4:40 PM | Additional Q&A | |
| 4:50 PM | Break | |
| 5:00 PM | Poster Session and Exhibitor Spotlight |
| Reception in Stevenson Ballroom with appetizers and no-host bar |
| |
| Poster | Otolith stable isotopes infer partial migration in herring from the Strait of Georgia, Canada | Jessica M. Qualley, Will Duguid, and Francis Juanes (Pacific Salmon Foundation) |
| Herring in the Strait of Georgia (SoG) are the largest population along the coast of British Columbia, Canada, currently managed as a single migratory stock. Otolith δ¹³C and δ¹⁸O stable isotope analyses reveal evidence of an alternative non-migratory life-history strategy among adult SoG spring spawners collected from the DFO Seine Roe Herring Test Fishery between 2020 and 2022. Migratory (West Coast Vancouver Island offshore) and non-migratory (SoG inshore) individuals were detected within spawning aggregations targeted by the commercial seine herring fishery. A logistic regression model based on whole otolith δ¹³C and δ¹⁸O values estimated that 30–33% of spring spawners were non-migratory. Growth analyses suggest that divergent migration strategies are not driven by differences in early juvenile growth trajectories as no significant differences were observed in otolith increment distances from the core to the first winter growth zone in age-4 herring. Migration type was associated with differences in adult morphology; migratory individuals were larger and heavier yet had lower condition indices than non-migratory counterparts. Divergent strategies likely have ecological consequences for marine food web dynamics, particularly given that adult Chinook salmon prey heavily on non-migratory herring in summer in the northern Strait of Georgia. This study demonstrates that natural tags based on otolith stable isotopes provide a powerful tool for resolving sub-stock structure, assessing fisheries impacts on alternative migration strategies, and identifying the ecological and evolutionary processes maintaining partial migration in SoG herring and other migratory fish species. |
| Poster | Pacific Salmon Data Discovery Tool: Enhancing Marine Ecosystem Data Integration and Collaborative Science | Katie Barnas, Damon Holzer, Mary Hunsicker, Chris Jordan, Michelle Rub, Dawn Urycki, Eric Ward (NOAA) |
| Monica Diaz, Nancy Leonard, Greg Williams, Mari Williams (PSMFC) |
| The Northwest Fisheries Science Center and Pacific States Marine Fisheries Commission are developing an emerging tool that enhances access to data relevant to salmon in marine ecosystems. The Pacific Salmon Data Discovery Tool demonstrates new approaches for data connectivity and cross-ecosystem synthesis by providing annually updated summaries and visualizations of Pacific salmon data (e.g., population abundance estimates, productivity trends, age data) alongside environmental drivers in marine ecosystems (e.g., marine heat waves, abundance and survival rates of predator and prey species). This web-based platform reduces barriers to information discovery across agencies, regions, and knowledge systems by centralizing datasets that currently exist in disparate agency reports, databases, and research repositories maintained by various agencies and tribes. By serving as a data repository that publicly houses salmon-relevant data across ecosystems (freshwater, estuary, marine) and data types (documents, data files, derived data products, spatial data, GitHub links), this shared digital infrastructure accelerates collaborative science and improves understanding of ocean drivers of salmon survival. Our focus will be to serve as an information clearinghouse for a broad user community - from casual data exploration through standardized data visuals, to comprehensive data downloading supporting regional harvest and conservation management analysis. |
| Poster | Leveraging historical imagery to assess floating kelp canopy changes over 30 years in northeastern Puget Sound | Danielle Claar, Gary McKenna (Washington State Department of Natural Resources) |
| Kelp forests form critical nearshore habitats in Washington, providing structural and energetic support for a diversity of culturally, ecologically, and economically significant species. In the southern Salish Sea, kelp mapping, monitoring, and historical research efforts have shown that kelp forests are stable in some regions but have substantially declined in others. In one region, northeastern Puget Sound, current data are insufficient to identify whether kelp forests have changed over time. To fill this data gap, we conducted a new analysis of a historical data source; color infrared aerial film imagery of Whatcom and Skagit County shorelines collected during summer low tides in 1995 and 1996. Film was scanned, digitized, aerially, triangulated, and orthomosaicked for analysis in a modern GIS format. Using ArcGIS Pro and ArcPy-based tools for spectral indexing, segmentation, and iso clustering, we classified kelp canopy area, or the area of the kelp at or near the surface, and delineated kelp bed area, which represents kelp forest habitat, in the historical imagery. We compared the historical distribution of kelp forests with data from our modern kelp aerial monitoring program, which began in 2022, and collects 0.5 ft 4-band imagery of over 750,000 acres of the shoreline. We developed accuracy-informed area estimations of kelp canopy and evaluated changes in kelp bed area to identify loss, gain, and stability of kelp forests in North Puget Sound across three decades. The results of this analysis will help inform ongoing efforts to manage, conserve, and restore kelp forests in this region and across Washington state. |
| Poster | Automated Field Assessment of Cyanobacterial Harmful Algal Bloom Intensity in the Sacramento–San Joaquin Delta Using Transfer Learning | George Batten (Environmental Science Associates) |
| Cyanobacterial harmful algal blooms (CHABs) degrade water quality and produce cyanotoxins that threaten public health, and the Sacramento–San Joaquin Delta has seen recurring CHABs — at times above recreational advisory thresholds — since 1999. As the Delta builds toward sustained, coordinated CHAB monitoring, bloom severity is a core indicator, yet it is usually captured by visually estimating surface coverage — an approach that is subjective, varies between observers, and depends on trained staff, limiting the consistency and reach of the resulting data. We are developing an image classifier that converts a single photograph into a repeatable CHAB intensity score. Building on a percent-coverage scale defined by aquatic biologists, it classifies images into five ordered intensity levels, from no visible algae to near-complete coverage. Because the levels are ordered, we assess models with ordinal-aware metrics such as within-one-level agreement, which treat a near-miss as better than a far-off error, alongside exact accuracy. Our primary approach is transfer learning: fine-tuning CNN backbones (e.g., ResNet) pretrained on large image datasets, a strategy suited to the modest, domain-specific datasets typical of environmental monitoring. We compare architectures and benchmark them against classical baselines built on engineered color and texture features, and we are establishing a standardized image-capture protocol and expert labeling scheme attentive to Delta conditions — glare, surface scum, and turbidity — that complicate color-based discrimination. We will present preliminary results. By replacing subjective judgment with a reproducible score, the tool aims to reduce variability between observers, lower the expertise barrier so more field staff and volunteers can contribute comparable data, and extend coverage to more sites and more frequent sampling — strengthening coordinated CHAB monitoring and supporting timely management and public-health decisions. |
| Poster | Modernizing Fisheries Data in the Klamath Basin: A Collaborative Leap Forward | Monica Diaz, Teddy Peterschmidt, Madeleine Kopf-Petterson, Nancy Leonard (PSMFC) |
| The Klamath Basin Fisheries Collaborative PIT tag monitoring and database project is a Klamath Basin fish tracking infrastructure and associated fisheries monitoring collaboration with more than 30 entities. The PIT tag project supports an ambitious effort by many partners to monitor and evaluate Klamath River restoration opportunities in the face of an ecological system in crisis and several imperiled fish species requiring Endangered Species Act (ESA) protections. Several years of PIT tag research—led by multiple entities across the Basin—catalyzed the formation of KBFC and informed many of the goals and strategies now driving collaboration, data sharing and database development work. Key collaborators include Tribes, federal and state agencies, non-governmental organizations and the Pacific States Marine Fisheries Commission. These data collected focuses on native Klamath fish data that will be shared through the KBFC PIT tag database and will be used to inform fisheries management and ecosystem restoration in the Basin. |
| Poster | Juvenile Fish Trap Discovery Tool | Sam Cimino (PSMFC), Russell Scranton (BPA) |
| The Juvenile Fish Trap Discovery Tool’s origination came out of a need to have one consolidated product that allows fisheries managers and biologists to quickly, and visually (geospatially), inventory the Columbia River Basin’s juvenile fish trapping efforts. After the development of a rotary screw trap (RST) dashboard in 2022, fisheries manager’s appetite for the provided information and ease of which to access the information grew. In the past few years, the RST dashboard evolved into the broader Juvenile Fish Trap Data Discovery Tool, and interest in this tool has expanded it beyond the Columbia River Basin. This product not only allows fisheries professionals the opportunity to observe current monitoring efforts, but it also allows managers to identify potential gaps in monitoring efforts. |
| Exhibitor | Biomark/Merck | Matt Bower, Richie Carmichael, and Nicholas Porter |
| Exhibitor | Hinchinbrook | Keith van den Broek and Stine Griep |
| Exhibitor | Blue Data Technologies | Doug Bonham |
| Exhibitor | Lotek | Matt Knoff |
| 7:00 PM | Adjourn | |
|
| TIME | DAY 2 – OCTOBER 27, 2026 | PRESENTERS |
| 7:00 AM | Breakfast | |
| 8:00 AM | Orientation for Day 2 | |
| | Tagging and Trapping Tech | |
| 8:10 AM | A Self-Powered Acoustic Transmitter for Long-Term Fish Movement Monitoring | Daniel Deng, Huidong Li, Jun Lu, Hyunjun Jung, Zhaocheng Lu, Bingbin Wu, Stephanie Larson (Pacific Northwest National Laboratory) |
| Acoustic telemetry is a widely used method for monitoring fish movement, migration, and interactions with hydropower systems. It works by implanting small acoustic transmitters in target animals and detecting their signals with underwater receivers. However, transmitter battery life remains a major limitation, especially for long-term monitoring of long-lived species and basin-scale movement. In many current tags, the battery accounts for about half of the tag’s weight and volume, yet the device may operate for only several months. |
| We are developing a self-powered, modular acoustic telemetry system designed to extend monitoring duration and improve data collection across large spatial and temporal scales. The system includes an energy-harvesting unit, acoustic transmitter, sensing unit with edge computing, rechargeable microbattery, receiver, and a cloud-based interface for system control and data visualization. The transmitter is powered by mechanical energy generated from fish swimming movements using a piezoelectric energy-harvesting design. |
| Benchtop testing showed that the energy-harvesting unit can generate sufficient power without fatigue degradation. We have also begun evaluating the system through implantation studies with American eels and white sturgeons, and pilot field evaluations are planned for 2026. |
| This technology could improve long-term ecological monitoring by reducing battery-life constraints and supporting more continuous fish movement and migration datasets. It has potential applications in hydropower permitting, fish passage evaluation, and basin-scale assessments of river connectivity. By extending the life and functionality of acoustic telemetry tags, this approach can provide monitoring professionals, data managers, and decision-makers with more complete information to support aquatic resource management. |
| 8:30 AM | Yakima Basin Flexible PEX Antennas | Zack Mays, Ceiteag Hennis (Yakama Nation Fisheries) |
| Flexible PEX pass-through antennas are an effective tool that Yakama Nation Fisheries uses to monitor fish movement throughout the Yakima River Basin, especially for detecting juvenile salmonids that can be missed by flat-plate antennas. Pass-through antennas are relatively inexpensive, require less anchoring, and can be tuned in the field, making them specifically useful in areas with surrounding ferrous material. At the Sunnyside irrigation dam, pass-through antennas mounted to the backside of the trash rack bays have provided important data on juveniles entrained in the canal and have demonstrated the advantages of mounting these systems to existing man-made structures. Future expansions at Sunnyside and Wapato dams will further increase detection coverage through floating, bypass, and sluiceway antenna systems. We discuss where and why flexible PEX antennas are being used throughout the Yakima Basin, their advantages and disadvantages, and takeaways from deploying these antennas in a variety of field settings. |
| 8:50 AM | Evaluating the Effects of PIT Tag Size and Release Timing on Fall Chinook Survival and Detection | Shubha Pandit, Melinda Goudy, Mark Johnston, Joe Blodgett (Yakama Nation Fisheries) |
| Passive Integrated Transponder (PIT) tagging is widely used to monitor fish movement and survival, but current protocols limit tagging of smaller juvenile fish. Current protocols utilized since 2011 specify tagging of Parr/smolt 50-60mm Fork Length (FL) using 9mm PIT tags, and >60mm FL using 12mm PIT tags. Waiting for fish to reach the required size may delay release and increase exposure to warmer water temperatures and predation. We evaluated whether smaller Fall Chinook juveniles could be tagged and released earlier using smaller PIT tags. In 2021 and 2022, an early group of smaller fish was tagged and released in early April with 9-/10-mm tags, followed approximately two weeks later by a group of larger fish tagged with 12-mm tags. We compared PIT tag retention, dam detection, travel time, downstream survival, and adult returns between the two groups. Preliminary results showed slightly lower detection rates and longer travel times to Bonneville Dam for the early-release group, but neither detection nor downstream survival differed significantly between groups. These results suggest that earlier tagging and release using smaller PIT tags may be a potential strategy for reducing exposure to warmer river conditions and other environmental stressors. Adult returns will provide additional information on longer-term effects of release timing and PIT tag size. |
| 9:10 AM | Tracking Ocean Migration Routes of Steelhead Using Pop Up Satellite Archival Tags (PSATs) | Matthew Thornton, Ryan Solee (The Wilderness Calling Conservation Initiative) |
| We aim to track ocean migration routes of steelhead off Oregon’s North Coast. We are the first folks to have successfully deployed PSAT tags and gotten data back on exact migration routes of steelhead. We use the Mini PSAT tag from Wildlife Computers. This tag transmits data to the Argos satellite system. We are compiling data and creating a map of exact migration routes of steelhead. |
| 9:30 AM | SMART Dart(TM): Development and Feasibility Testing of a Novel RFID Fish Tagging System for Enhanced Detection and Monitoring | Keith van den Broek, Stine Griep (Hinchinbrook, Inc.) |
| Emerging advances in radio frequency identification (RFID) technology present new opportunities to improve fisheries monitoring, tag detection efficiency, and data acquisition in challenging field environments. This presentation will introduce the Serialized Marker & RFID Transponder "SMART" Dart(TM) Tag, a newly developed RFID fish tag designed to support enhanced individual identification and automated detection capabilities across a range of fisheries applications. |
| The SMART Dart(TM) platform integrates a hybrid RFID-enabled extension of a proven tag architecture with a field-deployable autonomous reader system, intended to expand the utility of passive fish tagging technologies beyond conventional approaches. Initial development has focused on evaluating detection performance, deployment practicality, retention and durability considerations, and operational feasibility under controlled and semi-natural conditions. Ongoing feasibility testing will quantify system efficacy in a real-world application as an autonomous virtual fishery observer for rapid, non-visual identification of tagged fish within a high volume commercial fishery. |
| This presentation will provide a high level overview of the technology development process, study design, and preliminary findings from ongoing feasibility evaluations. Discussion will emphasize potential applications for fisheries research, stock assessment, harvest monitoring, aquaculture management, and monitoring programs where efficient individual identification and automated data collection are desirable. Broader implications for next-generation fish tagging technologies and future integration opportunities within fisheries monitoring systems will also be discussed. |
| 9:50 AM | Tiny Tags, Big Results: Using FRyID™ RFID Technology to Study Early Life-Stage Salmonids | Stine Griep, Keith van den Broek (Hinchinbrook, Inc.) |
| Individual identification of small and early life-stage fishes remains a major challenge in fisheries research due to the size limitations and potential physiological impacts associated with conventional tagging technologies. FRyID™ is a miniaturized RFID tagging system developed to enable individual identification of fish below the practical size threshold for standard PIT and external tags. |
| FRyID utilizes a flexible parylene-coated ultra-high frequency RFID tag (size: 6 mm × 0.5 mm × 0.25 mm, weight: 1.4 mg, volume: <1 µL). This is more than 16 times smaller than the standard 12 mm glass PIT tag. Each tag contains a permanently encoded 24-character unique identifier and is implanted subcutaneously using sterile preloaded needle cartridges designed to minimize handling time and tissue damage. |
| Initial evaluations have demonstrated successful implantation in multiple species, including Chinook salmon, steelhead trout, shiner, goldfish, and zebrafish ranging from 27–80 mm fork length. This presentation summarizes ongoing hatchery studies evaluating tag retention and survivorship in salmon fry, as well as completed mark–recapture studies of Chinook and steelhead fry conducted within floodplain habitats across three Upper Columbia River tributaries. |
| Preliminary results indicate that FRyID enables reliable individual identification at body sizes previously difficult to study using existing tagging methods. Floodplain mark–recapture studies provide insights into fry growth, movement patterns, survival, and floodplain habitat utilization, demonstrating the potential for improved understanding of early life-stage ecology. Results also suggest that FRyID may expand opportunities for longitudinal studies of juvenile fish in hatchery and natural systems while reducing handling constraints associated with larger tagging approaches. With no interference with conventional PIT tags, FRyID is being proven as a reliable secondary mark for PIT tag retention and similar studies. |
| 10:10 AM | Additional Q&A | |
| 10:20 AM | Break | |
| | Predictive Modeling and Best Practices | |
| 10:40 AM | Proof of Concept: Stocking YY-males to Eradicate Brook Trout in a Closed Population | Brian Davis, Jen Poirier (US FWS) |
| Introducing YY-male brook trout (fish with two Y chromosomes) is a biological control strategy designed to eradicate a population by skewing the sex ratio over time to all male (Schill et al. 2016). In 2018, the Columbia River Fish and Wildlife Conservation Office (FWCO) and partners, initiated a proof of concept study to assess the feasibility of using the YY-male approach to eradicate nonnative Brook Trout from Tyee Springs. Tyee Springs is a spring fed stream in SW Washington that flows less than one kilometer before being screened and treated as the water source for fish reared at Carson National Fish Hatchery. In 2019 we developed a stochastic population model to estimate the number of years it would take to eradicate the Tyee Springs Brook Trout population. Based on annual stocking and removal rates, the model predicted eradication in about 11 years. Demographic data used to populate model parameters were estimated from three mark/recapture events. We began annual suppression and stocking efforts in 2020, and the project is now in its sixth year. This presentation will provide an overview of the project, monitoring results, model updates, genetic results, and highlight lessons learned. |
| 11:00 AM | Evolving Ecosystem Diagnosis & Treatment: Integrating Monitoring Data to Advance Adaptive Salmon Recovery Planning | Laura McMullen (ICF) |
| Across the Pacific Northwest, decades of monitoring have produced an increasingly detailed picture of aquatic ecosystems, yet translating those data into clear, actionable decisions for salmon recovery remains a persistent challenge. Managers are asked to make high-stakes choices under uncertainty, where consequences directly affect the trajectory of salmon populations over time. For more than two decades, the Ecosystem Diagnosis & Treatment (EDT) model has helped bridge this gap by providing a flexible framework to synthesize and interpret monitoring data through the lens of salmon habitat needs. EDT predicts how environmental conditions influence survival, productivity, and capacity across life stages of salmonids; translating complex monitoring datasets into biologically meaningful metrics. EDT supports prioritization of restoration actions and evaluation of alternative management scenarios relevant to salmon recovery. This presentation highlights EDT’s enduring role as a tool for adaptive watershed planning, emphasizing its strength in assembling and using monitoring data to generate decision-relevant insights grounded in population response. Examples from current applications demonstrate how EDT continues to integrate diverse monitoring inputs to evaluate tradeoffs among restoration strategies and inform recovery pathways for salmon populations. Our team is investing in modernization to ensure continuity of use while evolving the tool to meet current expectations for data integration and accessibility. Efforts are focused on transitioning EDT to a cloud-based platform, improving workflow efficiency, and expanding visualization capabilities to deliver outputs through interactive maps and dashboards. By maintaining a consistent scientific foundation while enhancing usability and communication, EDT is positioned to remain a trusted platform for connecting monitoring data to decisions—supporting more transparent, adaptive, and effective salmon recovery planning. |
| 11:20 AM | Integrating Ecosystem Models with Long‐Term Monitoring to Support Salmon Recovery | Eric Doyle (Confluence Environmental Company), John Arterburn, Ryan Klett (Confederated Tribes of the Colville Reservation) |
| The Ecosystem Diagnosis and Treatment model (EDT) is a deterministic, life cycle‐based habitat model developed to support the conservation and recovery of declining Pacific salmon Oncorhynchus spp. and steelhead Oncorhynchus mykiss in the Pacific Northwest. Originally conceived in the 1990s, the current generation of EDT is proving its value as a data synthesis and analysis platform, capable of transforming complex environmental data into useful quantitative metrics to guide decision making. Here we describe the integration of EDT with long‐term research, monitoring, and evaluation in the Okanogan River in the state of Washington to support the ongoing conservation and recovery of steelhead listed under the Endangered Species Act. The lessons learned in this important Columbia River subbasin demonstrate the value of EDT as an adaptive management tool that is both effective and transferable. Modeling tools like EDT are one of many technological advances that will help resource managers identify priority habitats for conservation and restoration. |
| 11:40 AM | Investigating precision and bias of commonly used salmonid abundance survey methods | Brian McGreal, William Jaeger, Mark Scheuerell (University of Washington) |
| The effectiveness of fisheries management and conservation policies depends on surveys designed to estimate the abundance of target species over time. A variety of survey methods are employed for these purposes, each characterized by some level of precision and the potential for directional bias. While low levels of precision in survey methods lead to low confidence in point estimates of population abundance or forecasts, directional bias presents a more pernicious problem as it may lead to the mismanagement of imperiled species, invasive species, and commercially valuable populations, due to over- or under-estimation of their numbers. This study estimates levels of precision and bias associated with 28 survey methods commonly used by the Oregon Department of Fish and Wildlife to estimate the abundance of 68 populations of Chinook (Oncorhynchus tshawytscha) and coho salmon (O. kisutch), and steelhead trout (O. mykiss), from 1980 to 2022. Using multivariate autoregressive state-space models (MARSS), we distinguish observation (sampling) error from process (environmental) error and estimate directional bias at the level of individual methods. We find that, while some survey methods are generally more precise than others (dam counts, weir counts, and redd counts are highly precise), the same survey methods can have very different levels of precision and biases when applied to different species. Additionally, we investigate the extent to which individual populations are subject to inherent imprecision and bias as a result of the survey methods historically employed in monitoring efforts. |
| 12:00 PM | Lunch | |
| | Remote Sensing | |
| 1:30 PM | Advancing Habitat Monitoring with Bathymetric LiDAR in the Methow Basin | Ryan Klett (Confederated Tribes of the Colville Reservation) |
| The Okanogan/Methow Basin Monitoring and Evaluation Program (OBMEP) has historically relied on field-based habitat surveys to support Ecosystem Diagnosis and Treatment (EDT) modeling, evaluate habitat status and trends, and guide restoration investments throughout the Methow Basin. As advances in remote sensing technologies continue to expand monitoring capabilities, OBMEP is exploring how these tools can complement or replace traditional field methods in larger rivers and streams. |
| This presentation will describe the ongoing development of a new monitoring framework that integrates remotely sensed datasets derived from bathymetric (green) LiDAR. We will discuss our progress as OBMEP seeks to transform this dataset into habitat metrics such as habitat unit type, substrate composition, submerged woody debris and other EDT and USFS Level II habitat attributes in more than 100 kilometers of the Methow River. |
| 1:50 PM | Wetland Vegetation Modeling with UAV Lidar & Multispectral Data for Habitat Monitoring | Katarina Lunde (Lower Columbia Estuary Partnership) |
| Monitoring shifts in plant community composition is critical to understanding the condition and trajectory of wetland habitats in both natural and restored wetlands. For restored sites, plant community composition can be an indicator of habitat quality for juvenile salmonids and may also demonstrate desirable changes in soil and hydrology resulting from restoration actions, towards a more natural and resilient state. In recent years, the Lower Columbia Estuary Partnership has incorporated predictive modeling with UAV-collected lidar and multispectral imagery to characterize and track changes in plant community composition in monitored wetlands. Our modeling methods use digital elevation and canopy height models from classified lidar, co-registered high resolution multispectral imagery (RGB, red-edge and near infrared wavelengths), and on-the-ground plant species data to train supervised machine-learning classification algorithms. From the resulting plant community maps, we can generate habitat metrics and track changes over time at a site-wide scale, capturing more within-site variation compared to traditional on-the-ground sampling methods. We discuss collecting, processing and modeling with these data types in wetland contexts and considerations for using modeled vegetation to supplement traditional methods of plant community monitoring. |
| 2:10 PM | Leveraging UAV Lidar Data in Wetland Habitat Monitoring | Derek Marquis (Lower Columbia Estuary Partnership) |
| The use of drone (UAV) data has become more common in habitat monitoring, offering new methods for understanding critical ecosystems such as wetlands. The Lower Columbia Estuary Partnership Monitoring Program stewards a large body of monitoring data focused on wetland habitat for juvenile salmonids in the lower Columbia River. Since 2023, the Estuary Partnership has been collecting lidar data and co-acquired multispectral data as part of our action effectiveness monitoring in restored wetlands. Lidar has been an exciting addition to previously available data types, yielding classified wetland point clouds, bare earth digital elevation models (DEMs), and canopy height models, with flight timing and extents tailored to the specific needs of the Estuary Partnership and our partners. These data products can be used on their own, or in combination with other data types such as multispectral imagery, site hydrology and vegetation data to monitor wetland ecosystem metrics. This technology has also introduced a sizeable learning curve as we navigate the challenges specific to collecting and processing high quality lidar data in wetland sites, develop new workflows and data management strategies, and explore the best use-cases for these UAV lidar data products in pre- and post-restoration monitoring. |
| 2:30 PM | 3D Characterization of Streambank Complexity from UAS Surveys | Jen O'Neal, Shawn Higgins (Natural Systems Design) |
| Structure from motion (SfM) photogrammetry using imagery from Unmanned Aircraft Systems (UAS) streamlines data collection for characterization of large, alluvial rivers. We apply SfM methods to analyze topographic complexity of streambanks as part of a study with the Federal Highway Administration’s Western Federal Lands Division comparing the biological and physical habitat characteristics of engineered bank stabilization treatments with nearby control sites consisting of natural, forested banks. UAS surveys were completed for two sites (Skagit River and Upper Hoh River) at which bank stabilization treatments have been implemented using Engineered Log Jams (ELJs) composed of large wood and concrete Dolosse. Oblique images collected from UAS operations were processed using SfM methods to generate high resolution, three-dimensional (3D) point clouds for both treatment and control reaches. The study highlights the utility of UAS surveys for characterization of roughness as a metric of streambank complexity. Roughness was derived as the distance between each point in the 3D point cloud and a best fit plane representing the generalized streambank topography at that location. Results for the Skagit River site showed that roughness was 31% higher at the treatment reach stabilized with ELJs compared to a nearby control segment. Streambank roughness at the Upper Hoh River site was 82% higher at the treatment reach compared to the control. The 3D characterization of streambank complexity from UAS surveys was combined with additional physical and biological sampling methods to evaluate the habitat characteristics of streambanks treated with these types of bank stabilization projects. |
| 2:50 PM | Integrating UAV and Ground Surveys to Monitor Redd Superimposition of Tule Fall Chinook in the White Salmon River, WA | Justin Baker, Rikeem Sholes, David Hand (US FWS) |
| Interactions between hatchery origin upriver bright (URB) fall Chinook and ESA listed tule fall Chinook in the White Salmon River may reduce tule productivity through hybridization and redd superimposition. Because tule fall Chinook spawn earlier (Sept–Oct), their redds are vulnerable to URB superimposition later in the season. Initial 2022 surveys showed that 71% of tule redds were superimposed. To improve the efficiency and accuracy of monitoring these interactions, we integrated weekly ground surveys with UAV based aerial imagery during 2023–2025. Aerial imagery was reviewed by multiple trained observers, while ground crews mapped redds using high accuracy GNSS equipment to ensure spatial precision. Observer training and a structured review phase reduced redd misidentification by 45% and improved consistency among observers, lowering variability from 26% to 12%. Spatial agreement between methods was strong, with 66–77% of ground mapped redd area overlapping aerial delineations. Superimposition estimates differed between methods (aerial: 31–88%; ground: 97%), but a combined approach produced estimates closely matching ground surveys (91%). By refining methods for assessing hatchery impacts on natural populations, this study advances monitoring efforts and highlights how emerging technologies can enhance evaluations of biological outcomes critical to salmon recovery and hatchery management. The approach is transferable to other salmon populations and programs, supporting regional efforts to improve monitoring and management. |
| 3:10 PM | Additional Q&A | |
| 3:20 PM | Break | |
| | Genetics (eDNA, PBT, etc.) | |
| 3:40 PM | Implementation of a mobile DNA sequencing laboratory for real-time assessment of Columbia River basin fisheries | Lanie Galland, Shawn Narum (Columbia River Inter-Tribal Fish Commission) |
| Estimating the migration timing, stock-specific abundance, and ancestry proportions of native salmonids is essential for effective conservation and management in the Columbia River basin. Here, we present results from the inaugural year of an innovative monitoring program intended to provide real-time genetic stock assessment of salmonids from priority fisheries during critical migration, spawning, and harvest periods. Throughout 2025, we developed the mobile genetics laboratory in the form of a custom 44’ trailer where receipt of tissues, completed genotypes, and final PBT and GSI analyses were intended to be completed within a 24-hour period, providing the most up-to-date genetic assignments for monitoring trends across stocks in the Columbia River and its tributaries. We extensively designed, tested, and demonstrated protocols that support our selected DNA sequencing technology platform, achieving >99% concordance between the permanent laboratory in Hagerman, ID, and the mobile laboratory. After further equipping the trailer with all necessary DNA extraction and sequencing library preparation equipment, we successfully deployed the mobile laboratory on a total of four occasions to address priority needs including in-season stock identification, broodstock screening, and verification of source stocks for reintroduction programs. Across all deployments, genotyping results were generated within 16-24 hours of sample receipt, achieving >97% genotyping success, including genotyping of degraded samples. Overall, we highly successfully demonstrated the utility of the mobile genetics laboratory, providing critical information to managers in real time. With the unprecedented success of the laboratory in its inaugural year, we aim to increase deployments to additional sites of spawning, migration, and harvest significance while increasing our capacity for processing greater numbers of samples. |
| 4:00 PM | Can Metabarcoding of Aquatic eDNA Replace Traditional Biological Assessments in Streams and Rivers? | Daren Carlisle, Karen Beaulieu, Yer Lor, David Pilliod, Stephen Spear, Paul Frandsen, Chuck Hawkins (USGS) |
| Genomic-based monitoring technologies have expanded rapidly in recent years, but it remains unclear whether metabarcoding of environmental DNA (eDNA) can fully replace traditional assessments of biological condition. Our objective was to evaluate whether aquatic eDNA could be used to assess the condition of streams and rivers in the Upper Colorado River Basin. We collected eDNA by filtering water at 191 streams and rivers having minimal environmental disturbance and distributed over a wide range of natural environmental gradients. DNA was amplified using two published metabarcoding primers of the COI gene region: one targeting aquatic invertebrates and the other targeting all metazoans. We applied three different bioinformatic pipelines on the resulting Amplicon Sequence Variants (ASVs) to derive taxonomic classifications. For each pipeline, we evaluated how well machine-learning models calibrated with over 100 local- and watershed-scale environmental attributes predicted the occurrence of each taxon. We then applied the ensemble of taxon models to hold-out reference sites to quantify model precision and accuracy. We report our results to date and offer a prospectus for future application and improvement of eDNA tools for biological assessments. |
| 4:20 PM | Environmental DNA as a Tool for Aquatic Pathogen Surveillance: From Detection to Population Health Inference | Corbin J. Schuster, Bailey Morgan, Tamsen Polley, Michael Kent (Oregon State University) |
| Prespawn mortality (PSM) represents a significant conservation threat to spring Chinook salmon (Oncorhynchus tshawytscha) throughout the Pacific Northwest, with mortality rates exceeding 90% in some Oregon river systems. Adult Salmon Enteritis (ASE), a severe ulcerative enteritis characterized by extensive intestinal epithelial loss and inflammation, is associated with PSM and has recently been demonstrated to have an infectious etiology. Enterocytozoon schreckii, a microsporidian parasite of sexually mature Chinook salmon, is consistently associated with ASE-affected fish. Through the development of sensitive molecular assays, our laboratory has demonstrated that E. schreckii is more prevalent and widely distributed than previously recognized, revealing a substantial reservoir of infection that conventional histopathological surveillance has missed. |
| We present the development and validation of a qPCR and digital PCR (dPCR) assay for E. schreckii as a non-lethal, water-based biomarker for population-level PSM risk. The parasite resides within intestinal enterocytes, the same cells that slough into the water column as ASE progresses, potentially carrying intact spores. We hypothesize that environmental E. schreckii concentration, quantified by dPCR concurrent with hatchery broodstock histopathological assessments, is positively correlated with population-level ASE severity. Critically, fish harboring infection but held in cold water with intact epithelium produce no detectable environmental signal, demonstrating the eDNA signal is coupled to disease state rather than infection alone, a finding that both constrains and validates our framework. We describe the analytical validation, paired sampling design, and statistical modeling used to test this hypothesis within a tiered molecular surveillance strategy for real-time PSM management. |
| 4:40 PM | Mapping Washington's Aquatic Biodiversity Using Environmental DNA | Allison Ying (WDFW) |
| The Washington Department of Fish and Wildlife’s Aquatic Biodiversity Study, led by the Native Aquatic Species Unit, documents species presence across Washington’s river systems using environmental DNA (eDNA). eDNA sampling identifies organisms by collecting genetic material shed into the environment, without the need for direct observation or capture. Samples are analyzed using a DNA sequencing method called metabarcoding, which allows detection of multiple identifiable species from a single sample. Results from this work are compiled and presented through an ArcGIS Online Web Experience. The interactive map displays detections of genetic material from a subset of aquatic species including freshwater fish, shellfish, and crayfish across the state. Data is aggregated and displayed at the HUC12 watershed scale, providing a consistent spatial framework for viewing species distributions and biodiversity patterns. |
| This presentation includes a preview of the new mapping system. Sampling is ongoing statewide, and the dataset will continue to expand over time. The Aquatic Biodiversity Map is designed to support resource management, planning, and public access to eDNA datasets. |
| 5:00 PM | Adjourn | |
| | | |
| TIME | DAY 3 – OCTOBER 28, 2026 | PRESENTERS |
| 7:30 AM | Breakfast | |
| 8:30 AM | Orientation for Day 3 | |
| | Best Practices – Survival and Competition | |
| 8:40 AM | Incremental improvements in RFID technology leads to major gains | Nicholas J. Porter (Biomark) |
| Radio Frequency Identification (RFID) has played a major part in fisheries science over the past 50 years. Currently, there are hundreds of fixed detection sites in the Columbia Basin with many more on the global landscape from Antarctica to Australia in various terrestrial and aquatic environments. These environments, typically wild and beautiful, pose challenges for consistent system and site performance due to their harsh and remote nature. However, incremental improvements over the decades has allowed for higher, more consistent performance of these sites. Here, we discuss how improvements in tag performance, off grid power, and antenna design result in improved tag detection range and greater run time, even in noisy environments. We will examine the step by step improvements to an instream detection site on the lower South Fork Salmon River, Idaho as an example of incorporating system improvements over time with comparisons of read range pre and post improvements. Incremental improvements have resulted in a more effective system that allows for increased site reliability and improved detection range. These improvements, while seeming minor on their own, combine to result in better detection and therefore more accurate estimates with higher confidence. |
| 9:00 AM | Incorporating angler perspectives and ecological evidence to inform management of non-native fish interactions with native Yellowstone Cutthroat Trout in the Teton River Basin, Idaho | Lucas Ellingson-Cosenza, Colden Baxter, Morey Burnham (Idaho State University) |
| Management actions involving non-native species can create tension between fisheries managers and anglers, which poses challenges in fostering collaborative efforts and leads to a lack of angler buy-in to management. These are circumstances where both ecological and social science are needed, and may require viewing management as a collaborative effort where angler’s attitudes and anecdotal experiences are considered in decision-making. One such context occurs in the Teton River drainage, where Yellowstone Cutthroat Trout (YCT) (Oncorhynchus clarkii bouvieri) are a native trout species imperiled (reduced to half their historical range) due to habitat degradation and negative interactions with invasive, nonnative trout. In particular, Rainbow Trout (RBT) (Oncorhynchus mykiss) are considered a major threat due to their hybridization with YCT which decreases genetic diversity and quantity of pure YCT populations, but competitive interactions with nonnative Brook Trout (Salvelinus fontinalis) are also a concern. To address the status of YCT, we are integrating social research methods with an ecological field study. I conducted semi-structured interviews with Teton River anglers and fisheries managers to explore their values and perceptions of YCT conservation and management approaches to mitigate invasive species impacts. My interviews were coded into network mental models of both group’s varying perspectives. The ecological component of my project focused on Badger Creek, a tributary of Teton River that received a rotenone treatment in fall 2025 to remove its high RBT numbers. I evaluated ecological assumptions underpinning this removal action and that may help in predicting its ecological outcomes. I also drew on a drainage-wide, 20-year time series of e-fishing monitoring and a species distribution snapshot (based on eDNA) to determine if there is evidence for additional locations that might be priorities for future removal efforts. |
| 9:20 AM | Finding the Bottlenecks: Stage-Specific Survival of Juvenile Chinook Across Eastern Vancouver Island | Sam James, Jamieson Atkinson, Will Duguid, Nicole Hill, Joe Thorley, Katie Innes, Thomas Negrin, Cypress Hunder-Rookes, Isobel Pearsall (Pacific Salmon Foundation) |
| Studying anadromous species across their full life cycle can be challenging, making it difficult to determine where and when mortality is concentrated. We know that the first year of marine life plays a central role in regulating productivity, however, partitioning mortality among discrete life stages remains difficult. To address this, the Bottlenecks to Survival Project, a partnership between the Pacific Salmon Foundation and the British Columbia Conservation Foundation, established an extensive Passive Integrated Transponder (PIT) antenna network and captured and PIT-tagged juvenile Chinook across six river systems on the east coast of Vancouver Island from 2020 to 2025. A total of 224,009 Chinook tagged at successive life stages (hatchery, river, estuary, and first marine winter) over the five years were used to develop an analytical framework to estimate stage-specific survival. We used two hierarchical Bayesian Cormack–Jolly–Seber (CJS) models that jointly estimate stage-specific survival and recapture (detection) probabilities from field capture and multi-array PIT detection data. For computational efficiency, a multinomial (m-array) formulation was used for adult return detections, which was linked to a marginalized, individual-based model for juvenile stage-specific survival using Bayesian sequential updating. By partitioning salmon survival into successive stages, this framework aims to identify not only where survival is lowest, but where it varies most and thus where that variation can have the greatest impact on adult returns. This tells us where interventions could have the greatest impact on survival and could allow us to better direct recovery efforts. |
| 9:40 AM | Additional Q & A | |
| 9:50 AM | Break | |
| 10:10 AM | Replacing the Clipboard: Computer Vision for Scalable, Continuous Recreational Salmon Catch Monitoring in the Strait of Georgia | Jamieson Atkinson, Sami Ma, Philip Lemp, Sam James, Riya Shaju, Matthew Clarke, Jiangchuan Liu (Pacific Salmon Foundation) |
| Recreational fisheries are an important economic driver in coastal communities throughout British Columbia. They are also a significant source of mortality for Chinook and Coho salmon in the Strait of Georgia, making accurate monitoring of catch and effort essential for sustainable fisheries management. Conventional creel surveys are costly, labour-intensive, limited in temporal coverage, and can be subject to bias; these constraints undermine the timely, robust catch estimates managers need. As part of the Bottlenecks to Survival Project, four high-use public cleaning tables at three locations on the east coast of Vancouver Island were outfitted with overhead, motion-triggered video cameras and integrated tabletop PIT antennas. This infrastructure allows each fish placed on a table to be passively scanned for a PIT tag and imaged for species identification, hatchery origin (adipose-clip status), size, and participation in the coded-wire-tag head-recovery program, 24 hours per day across the full fishing season. To date, however, footage has been reviewed manually, a process that is slow, costly, and frequently extends well beyond the season, limiting the scalability of the program. Here we present an automated image-analysis approach developed in collaboration with Simon Fraser University. Using an open-source computer-vision framework trained on existing, manually annotated cleaning-table imagery, our objectives are to derive catch numbers, species composition, mark rates, and fork-length estimates. These catch data can then be integrated with complementary camera-based estimates of fishing effort to provide continuous, year-round, lower-cost monitoring of catch-per-unit-effort that is more robust to budget variability, providing managers, the recreational community, and First Nations with timely, high-quality data to assess the outcomes of management actions. |
| 10:30 AM | Applying Machine Learning to Learn from Fish Returning to the Klamath River After Dam Removal | Michael Hobley, Damon Goodman (CalTech and CalTrout) |
| The removal of Iron Gate, Copco 1, Copco 2, and J.C. Boyle dams reopened more than 600 river kilometers of habitat in the Klamath River basin, creating an unprecedented opportunity to monitor salmonid recolonization at broad spatial and temporal scales. We deployed adaptive imaging sonar near the former Iron Gate Dam to estimate salmonid abundance above the previous limit of anadromy and evaluate early recolonization patterns following dam removal. |
| We used an ARIS imaging sonar to enumerate upstream and downstream fish passage during the fall migration season, paired with carcass surveys, video weirs, telemetry, and length-based species apportionment models to estimate Chinook Salmon, coho salmon, and steelhead/rainbow trout passage. Manual review of sonar echograms provided the foundation for abundance estimates, while emerging AI and machine-learning tools are being developed and applied to improve processing efficiency, consistency, and scalability. |
| This presentation will describe how imaging sonar, automated detection tools, field validation data, and basin-wide distribution surveys are being integrated into a monitoring framework for large-scale river restoration. We will discuss lessons learned from the first two years of post-dam-removal monitoring, including data QA/QC, species apportionment, AI-assisted review, and the use of sentinel monitoring locations to track abundance, distribution, and life-history diversity. The Klamath experience provides a timely case study for applying emerging technologies to evaluate restoration outcomes in large, complex river systems. |
| | AI and Machine Learning | |
| 10:50 AM | Advancing AI-Assisted Fish Counting in the Columbia River Basin Using BlueFish | Elliot Koontz (Spheros Environmental) |
| Hydropower facilities in the Columbia River Basin rely on fish counting programs mandated under the U.S. Endangered Species Act to monitor adult salmonid passage and inform conservation and management decisions. Traditional counting methods (e.g., manual video review or “live” on site observers) are labor-intensive and vulnerable to errors in detection and identification from observer fatigue and fast passage. This presentation describes a computer vision camera system jointly developed by Spheros Environmental and MarineSitu to automate fish detection, classification, and enumeration. The system has been trained, calibrated, and deployed at multiple sites in the Columbia River Basin, where it has reduced observer workload while improving consistency and maintaining accuracy of count data. An edge-based computing platform performs real time species level classification and securely uploads imagery and metadata to the cloud. Observers then review and refine these automated classifications through the BlueFish software to efficiently generate validated count summaries for regulatory reporting. A continuous maintenance and retraining model enables the system to adapt to dynamic river conditions, including turbidity changes, lighting variation, and changes in seasonal species distribution. Case studies from recent deployments show species level classification accuracies approaching 90% for Pacific salmonids in certain assemblages. |
| In this talk, we will present results from validation studies at Priest Rapids and Wanapum Dams on the Mid-Columbia, as well as results of a species-specific detection model at Bonneville Dam operating in Summer 2026. We will also present on on-going development supported by the U.S. Department of Energy, including counting with an underwater camera, stereo based length estimation, and expanded juvenile monitoring. |
| 11:10 PM | AI-Enabled, Uncertainty-Aware 3D Tracking for Acoustic Telemetry in Ecological Monitoring | Daniel Deng, Wenqian Chen, Tao Fu, Jayson Martinez, Adam Hall (Pacific Northwest National Laboratory) |
| Accurate three-dimensional fish tracking is important for understanding dam passage, fish behavior, and habitat use, but field acoustic telemetry data are often affected by noise, signal reflections, and timing errors. These challenges can reduce the accuracy and completeness of traditional tracking methods. |
| We present an AI-based fish tracking approach that uses physics-informed neural networks to estimate fish movement from acoustic receiver detections. The method combines time-of-arrival data with realistic assumptions about fish movement to reconstruct continuous trajectories and estimate uncertainty. It is designed for complex acoustic environments such as hydropower dams. |
| We evaluated the method during a season-long field deployment at Little Goose Dam on the Snake River in Washington. Before the fish passage season, controlled tests used a remotely operated boat carrying acoustic transmitters, with high-accuracy GPS as ground truth. In these tests, the AI-based method achieved tracking errors of 0.48–0.69 m with 99% tracking efficiency, compared with 2.14–4.79 m and 85% efficiency for a traditional Approximate Maximum Likelihood solver. |
| During the field season, the method was also used to track receiver-mounted reference beacons and acoustically tagged fish released upstream of the dam. Across the season, 3,587 tagged fish were released, and the AI-based solver successfully tracked 3,516 fish, producing more than 2.7 million position estimates. Compared with the traditional solver, the AI approach produced more complete tracks and more positions per fish. |
| These results show that physics-informed AI can improve the accuracy, efficiency, and scalability of acoustic fish tracking. This approach offers a practical tool for monitoring professionals, data managers, and decision-makers who need high-resolution movement data for ecological research, dam passage evaluation, and fisheries management. |
| 11:30 AM | Additional Q&A | |
| 11:40 AM | Closing Remarks | |
| 11:50 AM | Lunch/Adjourn | |
| 1:30 PM | Field Trip | Lanie Galland, Shawn Narum (Columbia River Inter-Tribal Fish Commission) |
| Mobile Genetics Lab - Oxbow Fish Hatchery in Cascade Locks, Oregon |
