Development of a next-generation weather prediction system ～Realization of "micro-meteorology" prediction～
For social applications, the integrated use of software and hardware (infrastructure) technologies is essential to provide forecast information on demand and in real time. In the so-called "super-smart society," we believe that a mechanism that enables all IoT devices to access prediction information at the individual level (several-meter mesh) will be necessary. In this laboratory, we are working on the realization of real-time prediction of "micro-meteorology" in urban districts.
Micrometeorology refers to meteorological phenomena from the earth's surface to an altitude of about 100 meters, which are strongly influenced by buildings, human activities, and other factors. For real-time prediction, it is necessary to complete, for example, a 30-minute time integration in a few minutes. Real-time prediction of such micrometeorological phenomena cannot be achieved even with today's top-performance supercomputers. Furthermore, in order to capture local observation data and provide information to end users in real time, it is desirable that the prediction calculation is also carried out on a PC (edge server) close to the local area. In order to generate micro-meteorological prediction information of several meter mesh in real time on a PC, for example, we are developing a prediction simulation system using deep learning, that is, AI-fusion prediction simulation.
A part of this research has been published as an article in the November 2019 issue of JST News (Japanese).
Fig.1 : A future society in which micro-meteorology prediction is realized
This project has been adopted as a Grant-in-Aid for Transformative Research on Area (B) in the field of micrometeorological control.