legitWeather, and How to Read a Forecast Properly
Current conditions, hourly and daily forecasts, a map and saved locations — plus what a 30% chance of rain actually means.
legitWeather is a weather app for iOS and Android: current conditions, an hourly view, a daily view, a map, and as many saved locations as you want. Links to both stores are in the navigation and on the home page.
Since the app's job is presenting forecast data, here is what that data actually is — which makes every weather app easier to read, this one included.
What a percentage chance of rain means
Almost universally misread. A 30% chance of rain does not mean light rain, or rain across 30% of the area, or rain for 30% of the day.
In the US convention it is the probability of precipitation: the forecaster's confidence that rain occurs, multiplied by the fraction of the forecast area expected to get it. A 30% reading can mean "certain rain over 30% of the region" or "a 30% chance of rain everywhere" — the published number does not distinguish them.
The practical reading is simply: three days out of ten that look like this, you get wet. It says nothing about how much.
Why hourly and daily disagree
They are often different models. Short-range forecasts come from high-resolution models run frequently over a small area; multi-day forecasts come from global models on a coarser grid run less often. A daily summary may also be generated from a different aggregation than the hourly figures it sits above.
So an hourly view showing rain at 3pm and a daily view showing a dry day are not necessarily a bug. They are two models, and the shorter-range one is usually the better bet inside about 12 hours.
Forecast skill has a real horizon
Weather forecasting has improved enormously — a modern five-day forecast is about as good as a three-day forecast was in the 1980s. The limit is not computing power but chaos: tiny errors in the initial state grow until the forecast is no better than climatology, and that happens somewhere around ten to fourteen days.
A specific temperature quoted for a date sixteen days out is not a forecast in any meaningful sense. Anything beyond about a week should be read as a trend, not a plan.
Why apps disagree with each other
Most consumer apps do not run their own models. They consume output from a handful of national services — NOAA's GFS, the ECMWF model, and regional equivalents — and differ in which they weight, how they blend them, and how they interpolate a grid to your exact coordinates.
That interpolation matters more than people expect. Forecast grids are kilometres across, and if you live near a coast or in hills, your actual conditions can differ substantially from the grid cell average. Two apps disagreeing usually means they resolved that cell differently, not that one is broken.
The map is not a forecast
Radar shows precipitation happening now, detected by bouncing microwaves off it. Most "future radar" is a short extrapolation of recent motion — good for the next hour or so, and increasingly unreliable after that because it assumes systems keep moving as they have been and does not model formation or decay.
Radar also reports what is aloft, not what reaches you. Virga — rain evaporating before it lands — shows clearly on radar and never arrives.
What to take from all this
- Trust the hourly view inside 12 hours and the daily view beyond it.
- Read precipitation probability as frequency, not intensity.
- Treat anything past a week as a trend.
- Expect more local error if you are near a coast, a lake or hills.
- Save the locations you actually check, so you are comparing the same places over time.