Incubit Contracted by Japan Meteorological Agency to Further Research and Develop the “AI based Tornado and Heavy Localized Rain Prediction and Information Distribution System”

Incubit closed the contract with the Japan Meteorological Agency to work alongside the Meteorological Research Institute on the research and development of the AI based tornado detection model, the core technology of a system that predicts tornadoes and heavy localized rain.
The Meteorological Research Institute has been working on the system since 2018. Incubit Inc. (hereafter: Incubit), which undertakes practical implementation of image recognition by utilizing deep learning technologies, commenced development of the system’s tornado detection AI component in June 2020.
An average of 20 tornadoes occur in Japan each year. Although infrequent, the sudden and fierce nature of this disaster requires the implementation of stronger preventive measures. Therefore the Meteorological Research Institute, initiated the project with the aim to immediately detect the occurrence of tornadoes, and automatically alert stakeholders in the predicted tornado’s possible pathway.
Incubit has experience developing AI projects with meaningful social impact, such as a system that makes use of topological maps to predict mudflows, facilitating stakeholders’ preemptive response abilities, or a system that detects nodules in thyroids from ultrasound images and assists in potential cancer diagnostics. The knowledge and expertise fostered from these and many other past projects that make use of cutting-edge technologies to solve problems in real life situations, will be employed to drive this project to success.
The Meteorological Research Institute has been using data collected by Doppler radars to predict tornadoes. By making use of particular tornadic vortex patterns present in the Doppler velocity field, a model was trained to detect the occurrence of tornadoes. These patterns, however, change depending on season and the topography of the radar location, making them difficult to accurately predict.
This year, Incubit will introduce the temporal dimension to the detection equation, enabling the model to take into account the continuous evolution of the tornado patterns in its decisions. In addition to this, large-scale radar reflectivity data, environmental wind data, and topographical information surrounding the Doppler radar will be added to the model to increase the information available to the model and allow it to more accurately predict tornadoes. The tornado detection system is hoped to improve the safety of railways and other high-speed transportation methods, autonomous car systems and drones, where strong winds could be detrimental to operation.