Weather labs

One event, twenty possible outcomes

HOW TO READ THIS EXPERIMENT

The physical idea

First define the event: rainfall at least equal to your selected threshold, at one fictional point, over one fixed six-hour period. Each bar is an alternative outcome for that same place and period. It is not another location and does not represent another hour. The ensemble fraction is the number of members meeting the event divided by twenty. A 40% result therefore means eight of these twenty outcomes meet the event.

Both ensembles use the same fixed, symmetric list of twenty anomalies. The amount is max(0, location parameter + spread multiplier × anomaly). Assumption B adds 3 mm to the location parameter before applying the floor at zero. This creates a reproducible controlled comparison: the pattern of perturbations is unchanged while one wetness assumption shifts. Because zero clipping can alter the average, the control is intentionally called a location parameter rather than an ensemble mean.

Increase the threshold and watch the fraction fall or remain unchanged. Then increase spread: the fraction need not move the same way at every threshold. This is why a single probability without its event definition is incomplete. Operational services use specific thresholds, accumulation periods, calibration and processing; for example, the US NWS definition of measurable precipitation uses at least 0.01 inch in the stated period at a point. This lab’s chosen threshold and six-hour period are its own explicit product definition, not a universal PoP formula.

Basis: Met Office [1] · Met Office [2] · NOAA / National Weather Service [3]

What the model assumes

  • Twenty equally weighted, deterministic hypothetical members; no random resampling.
  • One fictional point and a fixed six-hour accumulation period.
  • Event is amount ≥ selected threshold, not percentage of area or time.
  • B changes only the location parameter by +3 mm. Both sets share fixed anomalies.

Where the explanation stops

  • The result is an ensemble fraction, not a calibrated real-world probability.
  • Model bias, member dependence and missing outcomes are not estimated.
  • Twenty members produce 5-percentage-point steps; 0% and 100% do not establish impossibility or certainty outside this constructed set.
FOLLOW THE EVIDENCE

Sources & provenance

  1. Using ensemble forecasts in decision-making Met Office · Publication date not statedReference only; original explanations, calculations and diagrams. No source media reproduced.
  2. Understanding ensemble forecasting: How the Met Office predicts uncertainty Met Office · 2025Reference only; original explanations, calculations and diagrams. No source media reproduced.
  3. FAQ — What is the Meaning of PoP NOAA / National Weather Service · Publication date not statedReference only; original explanations, calculations and diagrams. No source media reproduced.

Explanations and diagrams are original educational interpretations. Source institutions have not endorsed or reviewed this publication. How we distinguish evidence from models.