Perisher — interval timeline, 0–3 days

The model explorer's matrix, cut by window instead of by day. Inside three days the timing is the forecast: fourteen millimetres in one overnight burst and fourteen spread across a weekend are the same daily figure and completely different events, and the snow level moves further inside a day than a daily number can show. Every column is a fixed window; every row is one model's reading of it, phase-split at the elevation you choose.

Beyond three days this resolution would be a drawing rather than a forecast, which is why the page stops at 72 hours and the daily view keeps the rest. The vertical-structure pages answer the other half of the same question — what shape the column has at one instant, rather than how much falls through a window.

Resort Elevation

Window Rows

Models

Precipitation and phase by window

Model rows against fixed windows, AEST — totals are liquid-equivalent millimetres unless the row says otherwise

Snow share of precipitation 0% all rain all snow 100% The Precip row carries this ramp: hue is the snow share of the precipitation whose phase could be called at the selected elevation, and fill strength is the amount, so a heavy window states its split loudly and a trace states the same split quietly. Phase-uncertain and unclassified millimetres are held out of the ratio and damp the colour instead. Under 0.2 mm nothing is shaded — the split would be rounding. The Snow, SWE and Snow depth rows carry amount only; a dashed cell is data this source did not supply, never a zero.

What a window here does and does not claim

A blank cell is missing input, not a dry window. A model that stops before the end of the horizon leaves gaps; it has not forecast a dry Sunday.
Phase-uncertain precipitation is neither snow nor rain. It is precipitation falling within the snow level's own uncertainty band, shown beside the snow figure as +u and never folded into either side. Counting it as snow flatters the forecast; counting it as rain hides real snow.
Three-hour windows can be finer than the model. Several centres publish native three-hourly or coarser accumulations that the provider spreads evenly across hourly slots. Where that is measurable in the hours shown, the model's row says so — a three-hour column is then the provider's disaggregation, not independent model detail.
Provider SWE and snow depth belong to model-grid terrain. A source badge says whether SWE is supplied natively by the forecasting centre or derived by Open-Meteo. Both sit at whatever height the grid puts under this resort, which is not the elevation the phase split is evaluated at. They are a cross-check on that split, not a rival estimate of it, and the gap between them is mostly elevation. The depth figure's snow-to-liquid conversion is the provider's, not this project's. The fixed 0.7 cm/mm depth conversion does not alter the SWE millimetres.
The snow level is a range, not a track. Min and max across the window's hours. ≤ marks hours where the wet-bulb was already sub-zero at the model's own terrain, so the level is only an upper bound there; ~ marks hours estimated below that terrain by lapse rate rather than bracketed between supplied levels.
No snow depth is derived here. This project converts no liquid equivalent into centimetres anywhere, and does not start on this page.