Deep learning and the weather forecasting problem: Precipitation nowcasting

Precipitation nowcasting refers to the forecasting of rainfall and other types of precipitation up to 6 hours ahead (as defined by the World Meteorological Organization)1. Since rainfall can be localized and highly changeable, users of precipitation nowcast typically demand to know the exact time, location and intensity of rainfall. It is therefore necessary to make very high resolution, both spatially and temporally, precipitation nowcast products in a timely manner, typically in the order of minutes. The most important use of precipitation nowcast is to support the operations of rainstorm warning systems managed by meteorological services around the world. Rainstorm warning systems provide early alerts to the public, disaster risk reduction agencies, government departments in particular those related to public security and works, as well as managers of infrastructures and facilities. Upon the issuance of rainstorm warnings, these parties take actions according to their own standard operating procedures with a view to saving lives and protecting properties. It has tremendous impact on various areas from aviation service and public safety to people’s daily life.