A cloud inversion is the view from a hilltop when the valleys below are full of cloud and the sky above is clear. Photographers and walkers plan early starts around them, and most of those starts end in fog. I built the Cloud Inversion Checker to make that decision with more than a hunch.
The short version:
- It reads the forecast temperature and humidity at several heights above a place, not just the ground-level forecast.
- It looks for three things at once: a warm layer acting as a lid, cloud trapped under it, and dry air above it.
- It suggests nearby high ground and how high to climb.
Why a normal forecast is not enough
A standard forecast tells you whether it will be foggy where you are. It does not tell you whether the fog has a top you can climb above. That depends on what the air is doing a few hundred metres up.
Air normally gets colder with height. On a clear, calm night the ground loses heat, the air beside it cools, and you can end up with colder air at the bottom and warmer air above. That warmer layer is a temperature inversion. It acts as a lid: damp air and fog below it cannot mix upwards, so the cloud stays in the valley.
Comparing ground-level temperatures between a valley and a summit does not show this reliably. Those figures are adjusted for height, so the inversion can disappear from the numbers.
Reading the atmosphere above a hill
The checker uses the free Open-Meteo forecast API, which provides model data at standard pressure levels: 1000, 975, 950, 925, 900, 850, 800 and 700 hPa. Together they cover roughly the lowest three kilometres of the atmosphere. Each level comes with its real height, temperature, humidity, wind and cloud cover.
For each hour around sunrise, the checker builds a vertical profile from the ground upwards, ignoring levels that sit below or too close to the model's ground. It then looks for the strongest layer where temperature rises with height, joining neighbouring warm layers so one lid spread across two levels is counted once.
Turning three ingredients into a score
Each morning gets a score from five parts:
- The lid. A fraction of a degree of warming is weak; three degrees or more is strong.
- Cloud under the lid. Near-saturated air, model cloud at those levels, a small gap between temperature and dew point, or fog-level visibility.
- Clear air above. Humidity falling away above the lid, and little mid or high cloud to spoil the view.
- Wind. Calm valleys hold cold air; stronger winds mix the fog out.
- Dry weather. Rain at dawn usually means a front, not a cloud sea.
The lid and the cloud beneath it are both essential, so those scores multiply: if either is missing, the result is close to zero. The other three can lower the score but cannot cancel a good set-up on their own. A cloud sea under a grey sky still happens.
The overnight picture then nudges the result by up to about a quarter. High pressure, a long clear night, light wind and damp ground from earlier rain push it up. Pressure falling fast pushes it down. These rules come from published research on valley cold pools and radiation fog, and from mountain forecasters' guidance, all noted in the source code.
The checker scores each dawn from about an hour before sunrise to two hours after, because valley fog often lifts into a low cloud layer after sunrise that a summit still looks down on.
Where to stand
A score on its own is not much use if you live in a city. When the checker can estimate the cloud top, it suggests a height to aim for a little above it, because cloud tops undulate and rise as the morning warms.
For searches away from the hills, it finds named peaks within 35 km using OpenStreetMap data, adds heights from Open-Meteo's terrain model, and lists the ones likely to be above the cloud. Six upland areas have their own pages with chosen viewpoints: the Lake District, the Peak District, Eryri (Snowdonia), the Scottish Highlands, Bannau Brycheiniog (Brecon Beacons) and the Yorkshire Dales.
What it does not do yet
The checker does not yet compare its forecasts with what people actually saw, so I cannot say how often it is right. A weather model also smooths the terrain, so a deep, narrow valley can hold cold air the model does not see; the checker adds a small allowance for that on calm, clear nights, but it remains an estimate.
The obvious next step is a simple way for walkers to report what they found at the top, so the scoring can be checked against real mornings rather than theory alone.
How it is built
It is a React application hosted on GitHub Pages. A build step renders every page to static HTML with its own title, description and structured data, so search engines can read it without running JavaScript. All weather data comes from Open-Meteo under its CC BY 4.0 licence, and the browser calls it directly, so there is no server to run or pay for.
I started the project in July 2020 while learning React, and rebuilt the forecast logic in November 2026 from the research above.