A tracer is a single fish carrying a PIT tag — a rice-grain transponder — that crossed the Bonneville ladder and was detected there. If that same tag later appears on an antenna in the Lochsa, or the Imnaha, or at Lower Granite Dam, we know one real fish made that exact journey. No modelling, no assumption: one fish, two places, a date at each.
Everything on this site that says where the run is going rests on these fish. So it is worth being blunt about how many there are and how good a sample they make.
About half the tagged fish that pass Bonneville are seen again at Lower Granite, which by itself tells us they are Snake-bound. Fewer are ever detected on a tributary array that names an actual river — and that count includes every region, not just the Snake: the Methow and the Okanogan up the Columbia, the Deschutes and the John Day in between. The rest simply vanish from our instruments: they spawn in a stream with no antenna, or drop back downstream, or die.
Of the tracers we do resolve in a recent year, 52% ended up in the Snake, 13% in the Upper Columbia and 35% in the mid-Columbia. Each one of those is a real fish with a date at Bonneville and a date on a home-river antenna.
Read this as the composition of the fish we can follow, not of the run. The regions are not equally well instrumented, so a resolved-tracer share is not the same thing as a run share — which is exactly what the next chart is about.
This is the question that decides whether the tracers can be trusted to speak for the whole run. Different rivers run at different times: the large B-run fish bound for the Clearwater arrive later than the A-run fish headed up the Salmon. So if tagging is concentrated in certain weeks, the tracer cohort over-represents whatever was running those weeks.
Measured: the sampling rate runs from 16.4 to 49.4 tags per 1,000 fish across the season, against an average of 27.3, peaking in the week of Jun 4. That is a weighted spread of 18% about the mean. The cohort is not a flat sample of the run, and any composition read from it inherits that tilt.
Two very different clocks. A fish that crosses Bonneville is at Lower Granite in a median of 36 days — so within the season we can say with confidence how much of the run is Snake-bound. But being placed in a particular river takes a median of 158 days, because most tributary antennas only detect a fish when it moves up to spawn the following spring.
That gap is structural, and no amount of extra tagging closes it. It is why this site will tell you in October that the Snake run is strong, and make you wait until spring to learn whether it was the Clearwater or the Salmon that carried it.