BC salmon · watch 45 years of returns hold & fall

Every population is indexed three ways at once, because the single biggest attack on a chart like this is "you picked a baseline to manufacture a crisis." So you pick it — and see all three. Early-record = mean of a population's first up-to-10 surveyed years (most forgiving; reads the 1980s near 100). Historical-best = its best sustained 4-year level (most demanding; the decline-revealing default). 75th percentile = a high-but-typical level, in between. For each watershed-year we take the median across populations counted, lightly 3-year-smoothed. The shaded band on the timeline is the spread between the three: narrow band = the methods agree (a robust call you can defend in any room); wide band = they disagree (honest uncertainty — flag it, don't bury it). "Robust collapse" counts watersheds below 50 on all three; "contested" counts those where the baselines straddle the line. Toggle the baseline and watch the coral line jump within the band — same fish, three defensible framings. The one thing this does not fix is the survivorship problem the blue line shows: when fewer populations are counted, the index leans on whoever's still monitored. Different problem, handled separately, still flagged.