How this is built

The three sources

USGS water services. Instantaneous gauge readings for water temperature and discharge, refreshed every four hours, plus twenty years of daily values used to work out whether today is normal. A reading older than six hours is dropped rather than shown: one gauge in this set was still listed while returning a value from sixteen years ago, and presenting that as current would send somebody to a river on a number from another decade.

New York State DEC open data. Stocking records from 2011 onward, and the state's own list of recommended fishing waters, which is where the set of waters comes from.

NOAA National Weather Service. The forecast for the coordinates of the nearest listed access point, not for the county or the nearest town.

Matching a gauge to a water

This is the part that is easy to get wrong and we did. A first version picked the closest gauge by distance, which gave Conewango Creek a gauge on the Allegheny River and Black Creek a gauge on the Erie Canal. Distance is not evidence: 42 of the 62 gauges that rule selected were measuring a different water.

A gauge is now used only when its own name contains every distinguishing word of the water's name. Where no such gauge exists within range, the page says so instead of showing a number from somewhere else.

Telling waters apart

New York has a Beaver Kill, a Beaver Brook, a Beaver Creek and a Beaver Lake. It has a Sandy Creek in Jefferson County and another in Monroe, 128 miles apart, and a Mill Creek in Steuben and another in Warren, 261 miles apart. An early version of this site merged them and published a page claiming 324,053 fish stocked, some of them released 150 miles away.

Waters are now keyed by name and county together, and the county is in every page title and URL for the same reason.

What a percentile means here

A reading is ranked against every reading that same gauge took within seven days of the same calendar date, across every year on record. An 80th percentile flow means four out of five readings from this time of year were lower.

No percentile is offered on fewer than five years or thirty samples. A percentile from four years is a number with false precision, and this project has already published one of those.

High and low are positions, not verdicts. High water and low water are each good for some things and bad for others, which is why nothing here is coloured green or red.

What this site will not tell you

Whether fish are biting. Whether there is parking. Whether the season is open, what the limits are, or whether the stretch you are looking at is legal to fish. Check the NY DEC regulations.

An earlier version carried a "worth a trip today" verdict driven by a table of feeding temperatures written from general knowledge and labelled as state fisheries guidance. The attribution was false and the table had no source, so it was removed rather than patched. It will come back only with a citation per species.

Stocking records say what was released and when. They are not a count of what is in the water now, and they do not indicate when fishing is good.