Four ways this goes
Each card is one storyline with its odds and middle outcome. Tap a card to follow it through the page.
Every forecast, every hour
One row per ensemble member, grouped by storyline. Color is the precipitation type that hour; stronger color is heavier precipitation. Blank is dry.
Rows inside each storyline are sorted by when snow starts. Hover a cell for that member and hour.
How it piles up
Thin lines are individual forecasts; thick lines are each storyline’s middle forecast, hour by hour.
Temperature writes the storyline
Surface wet-bulb temperature, which sets rain versus snow better than air temperature. Above the band is usually rain, inside it a sleet and wet-snow mix, below it snow; a heavy burst can drag snow down a degree or two above it.
When it starts, switches and ends
For each storyline: the window in which 80% of its forecasts start, turn to snow and end. The dot is the middle forecast.
The fingerprints
Left: each forecast traced across the five numbers that define it. Storylines separate where their lines bundle apart. Right: the single biggest driver, how close the low passes to the coast.
Dry forecasts are left out because they never start. “Never” on the snow-start axis means the forecast stayed rain or sleet.
How cleanly do they split?
Left: the forecasts in two dimensions (principal components of the four clustering features), with each storyline’s outline and the direction each feature pulls. Right: silhouette scores, where 1 means a forecast sits squarely in its storyline and below 0 means it fits a neighbor better.
Storyline odds over four runs
Earlier ensemble runs, sorted into today’s storylines by nearest match. A storyline gaining runs in a row is a trend worth trading on; one that flips back and forth is noise.