Extended record — reconstructed, not measured

This gage's own record starts in 1925. A retired gage nearby, RIO GRANDE AT EMBUDO, NM., ran earlier and overlapped it, so its measurements can be translated onto this gage's scale to extend the picture back roughly 27.7 years. The result before 1925 is a reconstruction. It is not what this gage measured, because this gage did not exist.

Donor gage	RIO GRANDE AT EMBUDO, NM., 19.4 km away
Overlap used to fit	36,887 days, 1925-10-01 to 2026-09-27
Agreement (held out)	r² 0.986
Typical error (held out)	4.7%, bias -2.1%
Grade	strong

How this was built, and what it cannot tell you

• The two gages had to overlap. Without shared days there is nothing to measure the relationship against, and no extension is offered at all — that rule rejected the overwhelming majority of candidate pairs.
• The relationship is fitted in log space, because the ratio between two gages changes with flow. On Clear Creek the downstream gage reads 95% of the upstream one at low flow but 88% at high flow, because ditches in between take more when there is more to take. A single scale factor would get both ends wrong.
• It is scored out of sample: fitted on half the overlap, tested on the other half. The numbers above are from the half it never saw.
• Reconstruction never replaces a real observation. Every day in the series is labelled measured or reconstructed, and the overlap itself stays measured.
• The trends published on this site still use the measured record only. This extension is offered for reading the longer shape, not as an input to the statistics.
• DWR's published record is the authority. A reconstruction is an estimate of what this location would have read, not a measurement of it.

What the longer record says

Record used	Years	Trend	p	Change / decade
Measured only (what the site publishes)	101 1926–2026	no trend	0.07115	-17.513 cfs
RIOEMBNM alone (measured, no reconstruction)	127 1890–2026	decreasing	2e-05	-37.766 cfs
Extended (with reconstruction)	127 1890–2026	decreasing	7e-05	-31.469 cfs

The strongest evidence needs no reconstruction at all. RIOEMBNM shows decreasing across its own 127 measured years (1890–2026, p = 2e-05) — that is a measurement of this reach, not an estimate. The extended record mostly inherits it, which is why the long view differs from the 101-year modern one.

The longer record reaches a different conclusion, and the reconstruction's bias does not explain it: a -2.1% bias would tend to manufacture a increasing trend, which is not what appears. The likeliest reading is statistical power — 101 years cannot detect what 127 can.

Extending changes the conclusion for roughly half the gages that have a reconstruction — which is why both rows are here and neither replaces the other. Every trend published elsewhere on this site uses the measured record only.

The record

Readings come from the weekly-refreshed cache, never fetched live — the date beside the value says exactly how fresh it is. The live gage is one click up.

Provenance

DWR station abbreviation RIOTAONM; recorded station name “RIO GRANDE BLW TAOS JUNCTION BRIDGE NEAR TAOS, NM”; operated by USGS. Daily flows from the DWR HydroBase snapshot plus CDSS REST updates, refreshed on the weekly cycle. Trend statistics are Mann-Kendall (Hamed-Rao corrected) with Sen's slope in native cfs, the same numbers as the basin Trends tab.