Great Lakes Acoustic Telemetry Observation System

The pulse of Lake Erie

Between 2013 and 2019, fisheries biologists tagged 860 walleye with acoustic transmitters at opposite ends of Lake Erie, then tracked them on a network of underwater receivers. Through to August 2025 those fish were detected roughly 29.7 million times. Lake Erie holds two walleye populations that management agencies treat as separate entities. What the detections trace is mostly a western basin story: those fish slide east through summer and autumn, overlapping with the eastern population, then pull back west again.

—
detections analysed
—
tagged walleye
—
Lake Erie receiver sites
—
tags recovered by fishers
2013–2025
years of detections

How a lake gets wired for sound

Each fish carries a small transmitter that emits a coded ultrasonic ping every couple of minutes for the life of its battery — often five years or more. Moored receivers record the tag's ID and the moment it passed. A detection is therefore not a position fix but a proximity event: this fish, this receiver, this second. String enough of them together and a track emerges.

A receiver only hears a few hundred metres. Lake Erie covers about 25,000 square kilometres, so at any moment the network is listening to a very small fraction of it. Between one detection and the next, a fish could be anywhere. Nothing on this page is interpolated or smoothed: every line joins two places a fish genuinely was, and says nothing about the route it took in between.

Two separate tagging programmes are combined here. They share no fish and no tags, so together they give an unusually clean read on both ends of the lake at once.

The centrepiece

One year, on repeat

Detections folded into a single composite year. Each coloured circle is a receiver that heard fish that week; its size is the number of distinct individuals heard — not raw pings, which would mostly measure how long a fish loitered near a mooring. The small grey dots are the receivers actually in the water that week, which is not a constant: the network grew from nine dozen receivers in 2013 to roughly six hundred by 2017.

Because of that growth the default view uses 2017 onward, when coverage was broadly stable. Switching to all years adds the 2013–2016 fish, but their apparent distribution partly reflects where receivers happened to be rather than where fish were.

Week 14
— Lake Erie receivers listening in a typical year, this week
Grey dots are receiver sites deployed in that week of the year, from GLATOS deployment and recovery records; each is faded by how many years of the selected range it was actually in the water, so a faint dot means intermittent coverage. Hover a coloured circle for detail. Lake outline: Natural Earth (public domain).

The migration, as one line

The same data reduced to a single measure: the average longitude of every fish heard that week, weighted by individual. When the line climbs, the population's centre of gravity is moving east; when it falls, the fish are heading back west. Both cohorts trace the same annual shape from opposite starting points.

View as table

Where in the lake, week by week

Lake Erie divides naturally into three basins along a strong environmental gradient. The western basin is shallow, warm and highly productive; the central basin is deeper, cooler and productive but less so; the eastern basin is the deepest, coldest and least productive. Walleye spawn on reefs and in rivers throughout all three. This is the share of each week's detected fish found in each — the clearest view of the seasonal handover.

Twelve fish, followed

Population averages hide the individuals. These are the widest-ranging fish in the two cohorts, each reconstructed from its own detections. Pick one and play its life back, or run all twelve together on a shared calendar timeline.

—
Grey dots move with the date: they are the receivers in the water on the day shown, so gaps in a track may mean nothing was listening rather than that the fish stopped moving. — deployed on this date.

The one that went over the falls

Fish 19899001, a 51 cm male, was tagged at Smokes Creek on the Buffalo shoreline on 9 May 2016. For a fortnight it behaved like any other eastern-basin walleye, racking up thousands of detections around Buffalo Harbour. On 24 May it was still in Lake Erie. On 25 May it was detected 35 km north — below Niagara Falls.

It never came back. The lower Niagara receivers picked it up for another three years, most often at the river mouth where it opens into Lake Ontario, with a final detection on 14 May 2019. Its tag was never recovered. Select it above to watch the trip.

How to read this honestly

Detections are filtered. Acoustic tags collide and receivers mis-decode, producing phantom detections. They are not always far away — a false detection is simply most obvious when it puts a fish somewhere it could not possibly be. Everything here passes a short-interval test: a detection counts only if the same tag was heard again on the same receiver within an hour. That removes — records.

One hour is a simplification. The usual rule of thumb scales the threshold to the tag — thirty times its nominal transmission delay — and these fish were not all carrying tags on the same schedule. The detection intervals point to at least two programming classes: most transmitters ping a minute or two apart, while roughly a third sit closer to ten minutes. A single one-hour cut is therefore stricter than the rule for the slow tags and looser for the fast ones. Re-running with a per-tag threshold moves 0.7% of detections across the line and retains 0.2% more overall — far too little to change anything visible on these maps, but worth stating plainly.

Effort is not constant. Inside this map's viewport the network grew from about 90 receivers in 2013 to roughly 600 by 2017, and has kept growing since. The grey dots on both maps therefore change with time rather than sitting still, and the composite map defaults to 2017 onward so that the spatial pattern is not dominated by where receivers were added. Tag batteries also die, so later years carry fewer detections regardless of how many fish were alive: the maps show the shape of the seasonal cycle, not abundance.

The network is sparse, and silence is not absence. Each receiver hears only a few hundred metres of a very large lake. A fish that stops being detected may have died, been caught, moved beyond the network, or simply outlived its battery — but most often it has just swum somewhere nothing is listening. Roughly half of the tags in both cohorts were physically recovered, mostly through the lake's commercial and recreational walleye fisheries.