Two Minutes in the Rain: Giving Weather Warnings a Hierarchy That Survives a Phone Screen

A warning rarely arrives at a good moment. It appears as a banner while someone is pulling on a coat, checking a train time, or deciding whether to move a car away from a known flood line. The Met Office has built a disciplined system for expressing risk — a matrix of likelihood against impact, a colour scale from yellow through amber to red, a polygon drawn over the affected area — but that system was designed to be understood in full, not glimpsed in passing. The gap between the two is where communication quietly fails.

What the warning is actually saying

Behind every warning sits a compact argument: this hazard, at this likelihood, with this impact, in these places, during these hours. The colour is a summary of the argument, not a substitute for it. Yellow signals that people should be aware and keep an eye on conditions; amber that they should be prepared to act; red that action is required now. Those distinctions matter, and they are easily lost when the only thing a reader retains is a shade of ochre.

The visual problem is one of compression. A shaded polygon placed over a regional map does three jobs at once — it carries severity through colour, extent through shape, and urgency through the accompanying text — and on a 320-pixel-wide screen the first of those dominates while the other two dissolve.

Why the layered map struggles on a handset

Three failures recur in usability testing of public warning graphics.

• The legend disappears. Many mobile renderings crop or stack the key beneath the fold, leaving a colour with no scale attached to it.
• Colour alone carries the risk. Readers with colour vision deficiency, and readers in glare or on a dimmed screen, lose the primary channel entirely.
• Geography becomes abstract. A polygon that spans four counties tells a commuter almost nothing about whether their own route is affected.

There is also a memory effect worth noting. Asked a day later, people tend to recall that a warning existed and roughly how serious it felt, but rarely the likelihood band or the end time. The graphic has therefore done its emotional work and none of its practical work.

A three-line hierarchy for the glance

A clearer treatment does not require a new risk model. It requires a fixed reading order that the eye can follow without instruction. We suggest a compact stack beneath the headline, in this sequence:

1. What — the hazard named in plain words, with the likelihood and impact stated as a short phrase rather than a matrix coordinate.
2. Where — the nearest named places and the main transport corridors inside the polygon, not merely the polygon itself.
3. When — the valid period, followed by the window in which conditions are expected to be worst.

The official colour stays as a persistent chip beside the hazard name, never as the sole carrier of meaning. It should be reinforced with a word — Yellow, Amber, Red — and, where possible, with a border treatment or pattern so the distinction survives greyscale, night mode and a screen reader.

Designing for the thumb, not the poster

Progressive disclosure is the practical mechanism. The glance layer answers the three questions above; a single tap opens the full polygon, the impact table and the official guidance. Two rules keep that arrangement honest. First, the collapsed state must always display the hazard, the colour word and the time window — never a bare icon. Second, nothing in the collapsed state may reduce the stated severity, even for space. If the layout will not accommodate 'Amber', the layout is wrong, not the warning.

Contrast deserves specific attention. Amber is a difficult hue on white; yellow is worse. A dark outline and a text label cost almost nothing and rescue both. The temptation to 'soften' a warning palette for elegance should be resisted outright: the colour is part of the official message, and altering it is an editorial act, not a design one.

What should not change

Clarity work of this kind carries a risk of drift. Three constraints hold it steady:

• Do not invent a summary score. A single 'danger index' replaces a calibrated judgement with an unvalidated one.
• Do not average across a region. A county-wide mean flattens the coastal strip that will flood and the valley that will not.
• Do not paraphrase the official wording into something friendlier. The published text carries legal and operational weight.

A short test before publication

Any revised warning layout should pass a simple set of checks. Show it for no more than two minutes on a handset at arm's length in daylight, then ask the viewer what is happening, where, and when it ends. If any of the three answers is missing, the hierarchy has failed. Repeat the exercise with a screen reader, in greyscale, and with a reader who has deuteranopia. Finally, confirm that the most severe element on the screen is still the official risk level — not a decorative header, not a sponsor mark, not an illustration of a cloud.

A weather warning is judged by what people do after seeing it. The graphic's task is not to be admired, but to survive the two minutes between the notification and the door.