Read the mountain.

Fire does two things to a hillside: it removes the plants that slow water down, and intense heat can make the top layer of soil water-repellent. The next hard rain doesn’t soak in — it runs off, fast, picking up ash, gravel, and then boulders as it converges into the canyons (the arroyos this tool is named for). That slurry is a debris flow. It moves faster than a person and arrives with little local warning, which is why the whole system runs on rainfall thresholds rather than on watching the channel.

Post-fire debris flows are triggered by short bursts of intense rain, not by storm totals. The USGS therefore publishes triggers as a 15-minute intensity, written as mm/h: rain falling at 24 mm/h for fifteen minutes is 6 mm of water — about a quarter inch — in that window. A long gentle storm can drop far more total water than a ten-minute cloudburst and pose far less debris-flow danger. When you hear a forecast “rate,” it is this kind of number.

In the first winter after a fire, practice keyed to the USGS assessment uses the rainfall that gives a 50% modeled likelihood (P50) as the trigger. As slopes begin to recover, the second winter conventionally moves to P75 — more rain is needed for the same concern. On the M1 model’s logistic curve, that shift is the same multiplier for every basin:

R75 / R50 = 1 + ln(3) / 3.63 = 1.3026…

so a year-1 trigger of 20 mm/h becomes 26.1 mm/h in year 2. Arroyo applies the factor to the trigger — never to the rain — and if the USGS publishes an updated Eaton assessment, those published numbers replace this convention outright. Current USGS research (Graber et al., 2026) is moving toward recovery-based updates of exactly this kind.

In the first winter after the Eaton Fire, mid-February storms sent mud and debris out of the western canyons into Altadena streets, most memorably along Tanoble Drive below Rubio Canyon. Nobody had to guess whether burned basins could reach houses: the flows demonstrated it, in year one, at rain rates within the range the assessments contemplate. The second winter — the one ahead — is the season the year-2 convention above exists for.

Arroyo never issues a warning and never clears one; no tier here means “safe.” Its tiers are computed only from rainfall numbers against published thresholds — photos, reports, and hunches never move them. Every tier links out to Alert LA County, the National Weather Service, and Genasys Protect, because those are the products to act on. This page collects no personal data: no accounts, no location permission, no analytics, no cookies.

In live mode, every trigger on this page traces to a named field in the USGS post-fire debris-flow hazard assessment for the Eaton Fire (King et al., 2025, DOI 10.5066/P14EWYME), which conforms to the PWFDF data specification (DOI 10.5066/P13KUWCO). The provenance badge on the basin card shows the dataset version and retrieval date. When the badge instead reads TRAINING, the structure is real but the numbers are synthetic teaching examples — the maintainer loads the official dataset before the site is used publicly, and the banner at the top of the page disappears only then.