Centre to apply AI to tropical cyclone forecasting

The Planning, Fair and Exhibition Palace is devastated by Typhoon Yagi on September in the north-eastern province of Quang Ninh. (Photo: VNA)

NCHMF Director Assoc. Prof. and Dr. Mai Van Khiem made the requirement at a recent symposium on AI application to early warning systems for tropical cyclones in the East Sea.

Khiem said that the centre has already applied AI in forecasting thunderstorms within a six-hour period with initial success. AI is also being tested in hydrological forecasting.

Head of the centre’s Numerical Weather Prediction and Remote Sensing Office Du Duc Tien said that their research team is currently exploring methods for forecasting the intensity of tropical cyclones, using ensemble learning techniques with best-track data and re-analysis.

The research is focusing on building a comprehensive database on tropical cyclones and oceanic meteorological fields in the East Sea over the last decade and their impact on Vietnam.

Additionally, the research aims to develop an AI system for forecasting tropical cyclones with a focus on extending those forecasts up to three days in advance.

The study also covers the transfer of technology for forecasting tropical cyclones using AI systems.

At the seminar, engineer Nguyen Duc Long, a representative from the research group at the institute, talked about the use of deep learning (DL) models to enhance the accuracy of cyclone intensity forecasts.

The deep learning model, which used the Transformer architecture and model calibration techniques, is aimed at predicting cyclone intensity up to 24 hours in advance. Preliminary results indicated that the model significantly outperforms existing forecasting methods.

The team will continue to collaborate with the NCHMF to expand their research topics, tackling challenges in storm forecasting and rainfall prediction. They will also apply their research findings to operational forecasting, providing an additional reference source for meteorologists.

The results of this research are expected to be used in the next rainy season.

Commenting on the potential of AI models for regional cyclone intensity forecasting, Kieu Quoc Chanh from Indiana University, the US, expressed optimism about the AI model's potential for forecasting storms in Vietnam.

However, he noted that the model has not yet reached the level of comparison with the global and regional models currently used at the NCHMF./.