Backcountry Water Oracle

Current read

Every source in the database, most reliable first, as of 2026-08-06. Sources within 25 km of each other lend one another statistical strength where their rainfall responses agree.

SourceReportsDryBest windowr*Last seenRead
Chilson Spring583%90d+0.454 months agoProbably has water (light flow / pools)
Big Kahuna Falls - Mazatzal Wilderness16014%30d+0.551 month agoLikely flowing well
Castersen SeepWeak confidence1533%60d+0.451 month agoProbably has water (light flow / pools)

r* is the season-controlled rain correlation, pooled toward nearby sources where they agree — the trustworthy number. Counter-intuitively, a low r* on a rarely-dry source is good news: it means groundwater keeps it running regardless of the weather.

Carry your water. These are base rates from historical correlation, not measurements of what is there today. The precipitation model also smooths over isolated summer storms, so warm-season reads are the least reliable exactly when they matter most — cross-check radar before a summer go/no-go.

Chilson Spring

34.08587, -111.49097 · ~19″/yr

Reported within a yearlast reported 4 months ago

Probably has water (light flow / pools)

As of 2026-07-31 · 90d rain = 5.58″ · nearest historical analogs averaged 0.36 (Trickle)

Reports
58
Moderate confidence
Ever dry
3%
of all reports
Average flow
0.50
Moderate
Type
Reliable

Dry only 3% of the time, and rainfall barely moves it. That decoupling is exactly why it is dependable — it is fed by groundwater, not by last month's weather.

Warm-season read. The rainfall model averages over roughly a 9–11 km grid, so it smooths out the isolated monsoon storms that actually fill tanks. This read is least reliable right now, which is exactly when a dry spring costs the most. Check radar before committing.

How well rain predicts this source

WindowRawSeason-controlledStrength
90d←+0.51+0.46
60d+0.50+0.42
30d+0.48+0.44
270d+0.42+0.35
180d+0.41+0.27
365d+0.12+0.17

Rain and season are entangled — more reports get filed in cool, wet months. The season-controlled column removes that day-of-year cycle, so it is the trustworthy one. A big gap between the two columns means a correlation was mostly calendar, not rainfall. survives -> genuine rain response

24% of this correlation is borrowed from 2 nearby sources that respond to rain the same way.

Average flow by month

JFMAMJJASOND

Dashed = no reports that month, which is not the same as dry.

Big Kahuna Falls - Mazatzal Wilderness

34.08716, -111.45293 · ~19″/yr

Reported this seasonlast reported 1 month ago

Likely flowing well

As of 2026-07-31 · 30d rain = 4.61″ · nearest historical analogs averaged 0.72 (Strong)

Reports
160
Moderate confidence
Ever dry
14%
of all reports
Average flow
0.42
Trickle
Type
Flashy

Driven by recent rain: the last 30 days predict it better than any longer window (strongly). It turns on fast after storms and off just as fast, and it is dry 14% of the time.

Warm-season read. The rainfall model averages over roughly a 9–11 km grid, so it smooths out the isolated monsoon storms that actually fill tanks. This read is least reliable right now, which is exactly when a dry spring costs the most. Check radar before committing.

How well rain predicts this source

WindowRawSeason-controlledStrength
90d+0.62+0.50
60d+0.61+0.56
30d←+0.57+0.56
180d+0.57+0.45
270d+0.42+0.44
365d+0.27+0.30

Rain and season are entangled — more reports get filed in cool, wet months. The season-controlled column removes that day-of-year cycle, so it is the trustworthy one. A big gap between the two columns means a correlation was mostly calendar, not rainfall. survives -> genuine rain response

Has enough of its own data to stand alone — only 9% of this correlation is borrowed from 2 nearby sources.

Average flow by month

JFMAMJJASOND

Dashed = no reports that month, which is not the same as dry.

Castersen Seep

34.09059, -111.46653 · ~19″/yr

Reported this seasonlast reported 1 month ago

Probably has water (light flow / pools)

As of 2026-07-31 · 60d rain = 4.90″ · nearest historical analogs averaged 0.48 (Trickle)

Reports
15
Weak confidence
Ever dry
33%
of all reports
Average flow
0.41
Trickle
Type
Flashy

Driven by recent rain: the last 60 days predict it better than any longer window (moderately). It turns on fast after storms and off just as fast, and it is dry 33% of the time.

Warm-season read. The rainfall model averages over roughly a 9–11 km grid, so it smooths out the isolated monsoon storms that actually fill tanks. This read is least reliable right now, which is exactly when a dry spring costs the most. Check radar before committing.

How well rain predicts this source

WindowRawSeason-controlledStrength
180d+0.72+0.09
90d+0.71+0.07
60d←+0.67+0.40
30d+0.62+0.14
365d-0.21-0.08
270d+0.14+0.00

Rain and season are entangled — more reports get filed in cool, wet months. The season-controlled column removes that day-of-year cycle, so it is the trustworthy one. A big gap between the two columns means a correlation was mostly calendar, not rainfall. partly seasonal, real signal remains

65% of this correlation is borrowed from 2 nearby sources. On its own the record is too thin to trust, so the reading leans on neighbors that demonstrably behave alike.

Average flow by month

JFMAMJJASOND

Dashed = no reports that month, which is not the same as dry.