Backcountry Water OracleSource on GitHub

Early build — the report corpus is small and growing

Will that seep
be running?

We correlate a backcountry water source’s historical field reports against nearly two decades of daily precipitation for its exact coordinates, and give you a read on whether it’s likely to have water.

You also get the more interesting answer: what kind of source it is — flashy runoff that lives and dies by recent rain, or buffered groundwater that barely notices the weather.

Method

How the read is built

  1. 01

    Field reports

    Dated observations from people who were actually standing there, each scored from bone dry to raging. A source with a long report history is one we can say something about; a source with three reports is not.

  2. 02

    Precipitation

    Daily rainfall back to 2007 for that source's exact coordinates, then antecedent totals over the 30, 60, 90, 180, 270 and 365 days before every single report.

  3. 03

    Correlation

    How tightly flow tracks rain in each of those windows, with the seasonal cycle statistically removed. The strongest window is that source's memory — how far back its water remembers the weather.

The more reliable a water source is, the less precipitation predicts it.
It sounds backwards and it is the whole point. A true groundwater spring is decoupled from last month’s weather — that decoupling is why you can count on it. A runoff-fed falls is a slave to it: on fast after a storm, off just as fast. So a weak rain correlation is not a failed prediction. It is the finding.

Scoring

Every report becomes one number

Trail reports are prose — “good flow at the box,” “bone dry, tanks held.” The engine needs a number, so every observation is mapped onto one scale. Report what is usable: if the seep proper is dry but the rock tanks are holding, that is water.

0.0
Dry
no water, no pools
0.2
Not flowing
dripping, stagnant pools
0.4
Trickle
light but filterable
0.6
Moderate
about a quart per minute
0.8
Strong
about a gallon per minute
1.0
Raging
gallon-plus, loud

Honesty

What this cannot tell you

It is a base rate, not a measurement

Nobody is watching the spring. This is what the historical record says about conditions like today's — genuinely useful for planning, and no substitute for the report of someone who walked past yesterday.

No reports, no read

The whole method rests on a source's own report history. For an unreported spring, the most we can honestly offer is how this year's rain compares to normal at that spot — which is not the same as saying there is water in it.

Summer storms are the weak spot

The precipitation model averages over roughly a 9–11 km grid, so it smooths out isolated monsoon cells — least reliable in exactly the season when a desert water call matters most. For a summer go/no-go, cross-check radar.

Thin data stays thin

Under about 25 reports, a correlation is suggestive at best. We would rather show you an honest shrug than a confident number built on four observations.

Carry your water. This is a planning aid built from historical correlation, not a promise that anything is wet. Never carry less because of what you read here.

Status

What is here, and what is coming

The forecasting engine is built, open source, and has been used before real trips. This site is the front door for it, and it is being assembled now — source lookup, report entry, and the comparison table are the next things to land.

The honest constraint: the model is free, but the reports are the hard part. Every dated observation someone contributes sharpens the source it belongs to and the ones around it, because nearby sources lend each other statistical strength. If you keep trip notes on water, you already have the valuable thing.