Yandex has launched Meteum 2.0, a new machine learning-based weather forecasting technology. This is the world’s first weather forecast technology, the algorithms of which are trained not only on the data of instruments and weather stations but also on the messages of users. As a result, the short-term precipitation forecast on Yandex.Pogoda became 20% more accurate.
Every day, users send over a million messages about precipitation, and sometimes more than two million. These data helped improve short-term precipitation forecasts for the entire country, and especially in the Urals, Siberia, and the Far East, where there are no weather radars.
You can view the short-term forecast on the precipitation map. It shows how the situation will develop in the next two hours. The user can look at the map and decide when is the best time to take the dog for a walk or, say, run to the store.
It is easy to send a message about precipitation through the Yandex or Yandex.Weather application – this can be done in one click. Yandex marks such messages with umbrellas on the precipitation map. And the more messages users send, the more reliable the forecast.
The weather forecast for the week is also more accurate compared to the first version of Meteum. Now, not three, but five different forecast models are used, including our own. They also added satellite imagery and radar data. The information is processed using a CatBoost-based machine learning model and neural networks. It allows you to take into account not only wind speed, air temperature, and other weather parameters but also additional factors – for example, distance from a reservoir or the height of the sun above the horizon.
This is a further advancement of Yandex’s weather reporting systems, following the precipitation addition to Maps in July 2021.