Reading a runoff forecast

Odds of at least, not odds of rain

A weather forecast says “70% chance of rain” — odds that a thing happens. A runoff forecast says “90% chance the volume is at least 42 kAF” — odds that the river clears a bar. Five bars are published, and they read like this:

So when a forecast table on this site reads “42 · 48 · 52 · 56 · 63” across the 90/70/50/30/10 columns, the plain-English sentence is: “almost certainly at least 42; probably not more than 63; call it 52.”

Why April, and why it moves

Forecasts are issued monthly, January through June. Each issue also shrinks its window to the season still ahead, so the raw volumes fall through spring even in a normal year — which is why this site charts them as percent of normal for the window each issue covered. April is the benchmark: the snowpack is essentially set, the season has not started, and it is the number Colorado's planning has run on for a century. January’s issue is a first guess with half the winter still unwritten; by June the river is already answering the question itself.

When to trust it less

This site grades every archived April forecast against what the river then delivered — fifty years and 69 forecast points. Two verified habits worth carrying:

Dry years run high. In each river’s own driest third of years, the April even-odds number has typically promised about a third more water than arrived. The drier the year shaping up, the more a careful reader leans toward the 90% end of the range.

The stated odds lean wet everywhere. Pooled over ~1,900 forecast-years, values stated with “90% confidence” were reached about 75% of the time, and the even-odds number only about a third of the time. Read the whole range one notch drier than it says. The full accounting — every point ranked, mapped and charted — is the forecast report card; your own river’s version lives on its district’s Snow & runoff tab.

None of which makes the forecasts bad — an April guess at a summer’s runoff within ~20% is real skill on hard rivers. It makes them instruments with a known reading error, which is the only kind worth using.

What this does not let you conclude

A forecast is a statement about a river’s volume, not about your right to divert any of it — a wet forecast does not move you up the priority line, and a dry one does not curtail you by itself (the call does that; see anatomy of a call). Nor does a verified bias let you “correct” a single year’s forecast with confidence — the dry-year lean is a tendency over decades, not an adjustment factor for this April. And at points below reservoirs and big diversions, the forecast may describe computed natural flow while the gage records what was released — the two can disagree without either being wrong.