How well a snowpack predicts the year its river had

If you know how deep the snow got, do you know what the river will do? And where does that stop being true?

In the American West more water arrives as snow than as rain, and it arrives months later, so the winter snowpack is the closest thing the region has to a forecast. Whether it is a good one has been unanswerable here until now, because Hydrionis held snow water equivalent for hundreds of stations and flow for thousands of gauges and could not say which of them stood above which — the strongest relationship the geography could establish between a pillow and a river was that both were American. Every station now carries its publisher’s own subbasin code, so the question can be asked in every basin at once rather than argued from one.

1. Each row is one basin: the snow station with the deepest seasonal record against the gauge with the deepest water-year record. One row per basin, so a basin holding four of each cannot dominate the distribution with sixteen.
2. The coefficient is between the fullest the snowpack got and the mean flow of the water year that contained the melt. Both are one value per water year, so this is a season set against its outcome rather than two series sharing a calendar.
3. The spread is the finding, and the figures above it are computed from the data rather than written here. A basin near the top of the range is what a snowmelt-dominated catchment looks like when the pack is genuinely the year’s water budget; one near the middle is a catchment where the snow is a large contribution among several.
4. Read the negative rows last and read them carefully. They are not noise: they are basins where something other than the local snowpack decides the flow.

4 of the 140 basins return a negative coefficient: the years of deepest snow were the years of lowest flow. Nothing in the hydrology makes that happen on its own, so each is a signal that the gauge is measuring a river whose flow is decided somewhere other than the snowfield above it — an inter-basin transfer, a reservoir operated to a schedule, or a catchment where the snowpack was never the main supply. They are listed at the foot of the table rather than dropped, because a survey that reports only the rows agreeing with it has stopped being a survey.

Ordered from strongest to weakest. A low coefficient is not a failed basin — it is a basin where the local snowpack explains less of the river's year, which is itself the finding about that place. Basin names link where the basin has a profile; roughly a quarter hold enough record to answer this test without holding enough to earn a page, and those are named without a link.

Counted rather than quietly dropped: a survey reporting only what qualified has hidden its own selection.

Eligibility is decided by the pairing registry, not by this page, so a comparison shown here is one Hydrionis would allow anywhere else. Hydrionis has no code path that promotes a correlation to a cause.

Every number on this page was computed from the observations Hydrionis holds at the moment the page was produced, using methods declared in the methodology register. None of them is typed into this page, so none of them can disagree with the data behind it. The underlying series are in the dataset catalogue and reachable through the API.