Why the Way We Count Fish Is Getting a Long-Overdue Upgrade

Fish counting has never been the glamorous part of fisheries management. It doesn't attract the public attention that fish passage does or the research investment that population modeling receives. But counting fish accurately, in real time, at species-level resolution, is the foundation everything else rests on. Management decisions that should be responsive to what's happening in a river this season can only be as good as the data describing what's happening in that river this season.

The fish scanning equipment currently in use at most monitoring facilities was designed around what was technically achievable and economically feasible at the time those programs were built. Sonar transects. PIT tag arrays. Scheduled visual sampling by trained staff. Each method works within its limits. Each also leaves meaningful gaps in the data record that affect what managers can know and, therefore, what they can act on.

Whooshh Innovations' updated FishL Recognition system takes a different approach to the scanning problem. 

Three cameras capture nine images per fish as each individual passes through the imaging zone. Machine learning algorithms classify species, measure fork length and girth, detect adipose fin status for wild versus hatchery origin determination, assess injury, and log environmental data, all in milliseconds, for every fish. Not a sample. Every fish.

The design uses a V-shaped diffusion housing that creates consistent, controlled illumination within the imaging zone independently of ambient conditions outside it. That lighting control is what enables the system to perform as reliably at 3 AM as it does at noon, and in post-storm turbidity as in clear spring water. Automated fish sorting through the GateKeeper gate integrates directly with this scanning output, routing each fish based on the classification result without any human intervention.

What Does Nine Images Per Fish Actually Enable?

The nine-image capture set gives the classification algorithm meaningful redundancy. Fish arrive at the imaging zone at different angles, speeds, and orientations. Individual frames in any scanning sequence will vary in quality depending on those variables. A nine-image set taken across the passage event allows the algorithm to classify from the clearest and most informative frames in the set rather than being constrained by the worst one.

The practical consequence is more consistent classification accuracy across the full range of fish behavior during a migration event, including at peak migration hours when fish density is highest and individual passage conditions are most variable.

How Does the Data Output Change What Managers Can Do In-Season?

Every fish record from the FishL system is timestamped, tagged with environmental data including water temperature and flow rate, and uploaded automatically to cloud storage. Authorized users access the full dataset remotely. No end-of-season compilation. No manual entry from technician notes. Data that was generated this morning is accessible this morning.

For managers tracking run timing, species composition, or wild-to-hatchery ratios against management thresholds, that data latency reduction is the difference between responding to what's happening in the current run and responding to what happened in the previous one.

Where Has This Changed the Data Picture at Real Facilities?

At Bonneville Dam on the Columbia River, a FishL deployment generated over 220,000 classified fish records from 12 species in a single season. That volume of structured, verified data from a single facility, produced without expanding the monitoring staff, represents a substantial change in the data available to the National Marine Fisheries Service for population assessment and management decisions.

Here's what the scanning system delivers that conventional approaches at the same site couldn't:

• Complete individual-level records rather than sampled estimates for every fish that passes

• Automated species classification consistent across 24/7 operation regardless of staffing or visibility conditions

• Verified image files attached to every record for post-hoc audit and verification

• Wild versus hatchery origin data for every fish without requiring prior tagging enrollment

• Real-time data availability for within-season management decisions

Visit the FishL product page, explore Whooshh Innovations full product range, and contact us to talk through what automated monitoring could add to your program.