methodology

pow-finder answers one question: where in South America is the skiing best right now. Incumbents publish depth and snowfall totals. We publish a quality ranking. Everything below is deterministic. Change a constant and the scores move. There is no editorial override.

data source

Weather comes from Open-Meteo's free forecast API (models=best_match). Each resort is queried three times — base, mid, and top elevation — so temperature is downscaled to the band. One call returns 7 days of analysis plus 8 days of forecast, hourly. We recompute snowfall from precipitation and wet-bulb phase; Open-Meteo's own snowfall field uses a fixed 7:1 ratio and is stored only for comparison.

derived meteorology

Wet-bulb temperature uses the Stull (2011) approximation. Precipitation is partitioned by wet-bulb: fully snow at Tw ≤ 0°C, fully rain at Tw ≥ 1°C, linear between. Fresh snow density follows air temperature at fall:

Snowfall cm = SWE mm × 100 / density. Positive degree-hours accumulate as max(T, 0) each hour. A melt hour is T > +1°C, or shortwave > 500 W/m² with T > −3°C. A melt run followed by a 6-hour window with mean T < −2°C sets a crust flag, cleared after 10 cm of new snow.

season state

Settled SWE is a running sum of hourly SWE minus melt at 0.125 mm per positive degree-hour (3 mm/°C·day). Modeled settled depth assumes 350 kg/m³: settled_base_cm = settled_swe_mm / 3.5. A refresh event is ≥ 10 cm in a rolling 24 hours. This is a weather-derived proxy, not a resort snow-depth measurement. It also only covers weather retained since ingestion began, so it may omit older seasonal snow. Treat base adequacy as the least confident component.

daily snow ratings

There is no single blended score. For each local calendar day from today through the next seven days we compute a 0–100 rating on the mid band, then rank resorts by the median of those eight numbers. Base and top metrics are shown on the resort page but do not enter the ratings.

Each day is scored as-of that day (today: as-of now; future days: as-of the last forecast hour that day). Three components sum to a raw max of 75, then scale to 0–100:

• Freshness (0–35). Effective fresh snow over the prior 7 days relative to that day, decayed by age (τ = 96 h) and warmth (τ = 18 PDH), then min(35, round(35 × (eff_cm/40)^0.7)).
• Preservation (0–25). Start at 25. Subtract rain-on-snow (1.5 pts/mm, cap 15), unburied crust (10), windy snowfall > 25% of last 72 h at gusts > 60 km/h (8), icy snowpack since last snow — day above freezing, night below (8), slush — above freezing day and night (14). Yesterday max T > +5°C (5) only when icy/slush did not already apply. Floor 0.
• Base adequacy (0–15). min(15, round(15 × settled_base_cm / base_needed_cm)).

Grades: ≥ 80 exceptional, 65–79 very good, 50–64 good, 35–49 fair, < 35 poor. A day at 100 means excellent fresh snow, clean preservation, and adequate base that day — not a week-long average.

temperature regimes

Day hours are local 12:00–17:00; night hours are local 00:00–06:00. Since the last snowfall we classify the snowpack as cold (stays frozen), icy (day thaws, night refreezes), or slush (above freezing day and night). Each forecast day gets the same label; a day with ≥ 3 cm of snow while cold is tagged snow. Regimes feed preservation and appear next to each daily rating on the resort page.

limitations

Be blunt about what this is not. Precipitation amount is grid-cell scale; Open-Meteo cannot elevate-correct it, so orographic enhancement in the Andes is underestimated at some resorts. The settled base is a water-balance proxy with no ground truth and no wind transport. We do not ingest resort-reported depths, webcams, or satellite wet-snow flags. Forecast skill is being collected in forecast_archive for later bias correction; v1 only accumulates that archive. If a band fetch fails, the resort stays on the board with a stale badge and the previous score.

Open-Meteo's free tier is non-commercial. So is pow-finder v1.