Yandex Weather calculates precipitation forecasts based on data from weather radars and satellites. Information from weather stations is not suitable for this, as it isn't updated frequently enough and isn't always accurate. The stations cover the areas unevenly and are located at a great distance from each other.
Stage 1. Getting data from weather radars and satellites
Weather radars make a 3D image of the atmosphere within a radius of 200-250 km every 10 minutes.
A satellite registers the position of clouds over Europe and part of Asia every 15 minutes.
Stage 2. Image processing
Radar and satellite images are fed into two convolutional neural networks. Both networks predict precipitation independently:
They analyze a sequence of images and integrate the results with the weather forecast calculated using the Meteum technology.
They forecast where precipitation areas will move and how their shape will change.
They create precipitation maps for the next two hours.
Stage 3. Making the final forecast
The service combines the maps received from the neural networks and displays the final forecast on the weather map layer.
In contrast to other weather maps, the forecast on the precipitation map is made based on the current data received from weather radars and the satellite. The map shows the wind direction, precipitation forecast near your home, and when the snowfall will end.
The precipitation map covers Europe, Asia, North and South America.
For more information about forecast results in the service, see the section Precipitation map.
How often data is updated
The data on the map is updated every 10 minutes (as soon as the service receives information from the radars).