Understanding Ensemble Probabilities and Plumes

A probability map and a plume describe different views of the same uncertainty: where a condition occurs in the ensemble, and how forecasts evolve near one location.

What a precipitation probability means here

SKC’s precipitation threshold processing counts members whose accumulated precipitation is at least the specified amount, divides by the member count and multiplies by 100. This is a model-member fraction, not an independently calibrated promise of the chance at your house.

Illustrative calculation: if 10 of 31 GEFS members reach at least 0.25 inches during a selected 12-hour period, the member fraction is about 32%. That does not mean rain lasts 32% of the period or covers 32% of your town.

Match the window to your question

SKC includes thresholds such as 0.25 inches in 12 hours and 1 inch in 24 hours. At F024, the first refers to F012–F024 and the second to F000–F024. Windows can overlap as you step forward. Adding overlapping totals double-counts precipitation, and adding probabilities does not produce a valid longer-period probability.

Mean, P90 and custom accumulation

A mean QPF map summarizes member precipitation amounts. P90 is the 90th percentile of the member amounts at each grid point; it is neither a worst-case ceiling nor the amount that has a 90% chance of being exceeded. Because statistics are calculated separately across the grid, a percentile map need not match any single physically evolving member.

For a custom accumulation, read the selected start and end forecast hours and the statistic actually offered by that model. Do not assume all models expose the same summary choices. Missing or unavailable periods must not be interpreted as dry weather.

Read a point plume

Choose a location, model and plume type. Inspect the run, units and time axis. Individual member traces show alternate evolutions; the ensemble mean smooths them. Separate groups of wet and dry members can be more informative than a moderate mean alone. SKC precipitation plumes include cumulative amounts and interval guidance; read the particular panel’s label before comparing them.

The GEFS precipitation plume also includes precipitation-type information. Liquid-equivalent precipitation is not automatically a snow-depth forecast. Model type, surface conditions and local terrain can all be wrong. EPS and AIFS precipitation plumes should not be assumed to contain GEFS’s precipitation-type detail.

Try a comparison

Use the free GEFS viewer to inspect a 12-hour precipitation probability, then open a GEFS precipitation plume near that region. Check whether the timing and amount ranges help explain the map. A point sample still represents model guidance near a location, not a street-level observation.