Tomorrow Weather logo Tomorrow Weather

Consensus Weather Forecasting

6 Weathermen in one - Averaged.

Forecast Accuracy and Safety

How the forecast is produced, where its honest limits are, and why it must not be used for safety decisions.

Last updated 08 October 2026

How the forecast is made

There is no single weather forecast on this site. Six independent weather models are asked for the same place and time: ICON from the German weather service, ECMWF from the European centre, GFS from the American service, AROME from Meteo-France, the UK Met Office model, and GEM from the Canadian service. Their answers are averaged, and each model's vote is weighted by how accurate it has actually been at the weather station nearest to that location, measured against real observations over the past six months.

Why the disagreement is printed

The number on its own hides the most useful part. If all six models agree, the forecast is worth planning around. If they are three or four degrees apart, the shape of the day will probably hold but the numbers will not, and the spread printed beside the temperature tells you which of those you are looking at. We also remove a model's long run bias when it is larger than one degree at that station, because a model that always runs two degrees warm there is not bringing new information by being warm again.

Where the honest limits are

Do not use this for safety decisions

This site is not a warning service and is not a substitute for one. Do not rely on it for aviation, marine, mountain or fire decisions, for emergency response, for anything where life, health or property is at risk, or for a decision you would not make if the forecast turned out wrong.

In South Africa the official authority for weather warnings is the South African Weather Service at weathersa.co.za. If a warning from them disagrees with anything on this site, their warning is the one that counts.

When we get it wrong

Models do fail, and when several fail together we will be wrong with them. If a forecast here was badly wrong for your area, tell us the location and the date. We keep the station observations, so we can check what the models said against what actually happened, and we will tell you what we find.

Back to the forecast