Backcountry Water Oracle
How the read is built

Method

Every number on this site comes from one source’s own report history, correlated against the rain that fell on its exact coordinates. This is how that works, and where it stops working.

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.

0.0
Dry
No water at all — not even pools
0.2
Pools or dripping
Water present but not flowing; stagnant pools, a drip, a full spring box with no outflow
0.4
Trickle
Light but filterable — you can fill a bottle, slowly
0.6
Moderate
Steady flow, roughly a quart per minute
0.8
Strong
Roughly a gallon per minute
1.0
Raging
Gallon-plus, loud, no question about it

Score the water you could actually use. If the seep itself has stopped but the rock tanks below it are holding, that is water — score the tanks.

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 10 reports this site issues no read at all: the engine will still produce a number and that number is noise. Between 10 and 25 you get a read labelled weak — suggestive, not solid. We would rather show 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.

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.

Next

Go look something up