Long-Term Typhoon Trajectory Prediction: A Physics-Conditioned Approach Without Reanalysis Data

Young-Jae Park,Minseok Seo,Doyi Kim,Hyeri Kim,Sanghoon Choi,Beomkyu Choi,Jeongwon Ryu,Sohee Son,Hae-Gon Jeon,Yeji Choi

2024-01-29

Abstract:In the face of escalating climate changes, typhoon intensities and their ensuing damage have surged. Accurate trajectory prediction is crucial for effective damage control. Traditional physics-based models, while comprehensive, are computationally intensive and rely heavily on the expertise of forecasters. Contemporary data-driven methods often rely on reanalysis data, which can be considered to be the closest to the true representation of weather conditions. However, reanalysis data is not produced in real-time and requires time for adjustment because prediction models are calibrated with observational data. This reanalysis data, such as ERA5, falls short in challenging real-world situations. Optimal preparedness necessitates predictions at least 72 hours in advance, beyond the capabilities of standard physics models. In response to these constraints, we present an approach that harnesses real-time Unified Model (UM) data, sidestepping the limitations of reanalysis data. Our model provides predictions at 6-hour intervals for up to 72 hours in advance and outperforms both state-of-the-art data-driven methods and numerical weather prediction models. In line with our efforts to mitigate adversities inflicted by \rthree{typhoons}, we release our preprocessed \textit{PHYSICS TRACK} dataset, which includes ERA5 reanalysis data, typhoon best-track, and UM forecast data.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

This paper focuses on the problem of long-term typhoon trajectory prediction. With climate change, the intensity of typhoons is increasing, making accurate prediction of typhoon paths essential for effective disaster control. Although traditional physically-based models are comprehensive, they require large computational resources and rely on the expertise of forecasters. On the other hand, modern data-driven methods often rely on reanalysis data, such as ERA5, which are close to real weather conditions but not generated in real-time, making them unsuitable for real-time tasks like 72-hour advance prediction. To address these issues, the paper proposes a new method called LT3P (Long-Term Typhoon Trajectory Prediction) that utilizes real-time Unified Model (UM) data without the need for reanalysis data, enabling predictions every 6 hours with a maximum lead time of 72 hours. LT3P consists of two main components: a physical condition encoder and a trajectory predictor. The encoder processes key variables that affect the trajectory and intensity of typhoons, while the predictor receives the coordinates of the typhoon center and predicts the trajectory for the next 72 hours. Through this approach, LT3P not only achieves real-time 72-hour typhoon trajectory prediction without relying on reanalysis data but also surpasses physically-based models and data-driven methods in prediction accuracy. Additionally, the paper provides a pre-processed PHYSICS TRACK dataset, including ERA5, best tracks, and UM forecast data, to support research. In summary, the paper addresses the limitations of existing methods in real-time capability and long-term prediction accuracy by proposing a new data-driven model for typhoon trajectory prediction that combines physical conditions, and provides relevant data resources to facilitate future research.

Long-Term Typhoon Trajectory Prediction: A Physics-Conditioned Approach Without Reanalysis Data

Development of a prediction system for precipitation- and wind-causing typhoons affecting the Korean peninsula using observational data

Deep Learning Approaches in Predicting Tropical Cyclone Tracks: an Analysis Focused on the Northwest Pacific Region

A Statistics and Physics-Based Tropical Cyclone Full Track Model for Catastrophe Risk Modeling

Strategy for the Prediction of Typhoon Wind and Storm Surge Height Using the Parametric Typhoon Model: Case Study for Hinnamnor in 2022

A Study on the Improvement of Wave and Storm Surge Predictions Using a Forecasting Model and Parametric Model: a Case Study on Typhoon Chaba

A Hybrid Machine Learning/Physics‐Based Modeling Framework for 2‐Week Extended Prediction of Tropical Cyclones

Improving Typhoon Predictions by Integrating Data-Driven Machine Learning Models with Physics Models Based on the Spectral Nudging and Data Assimilation

Typhoon Tracks Prediction with ConvLSTM Fused Reanalysis Data

An Adaptive Nested Dynamic Downscaling Strategy of Wind-Field for Real-Time Risk Forecast of Power Transmission Systems During Tropical Cyclones

Ensemble data assimilation and prediction of typhoon and associated hazards using TEDAPS: evaluation for 2015–2018 seasons

Typhoon Track, Intensity, and Structure: from Theory to Prediction

A Study on the Pre-Survey and Plan for the Establishment of the Korean Typhoon Impact-Based Forecast

TROPHY: A Topologically Robust Physics-Informed Tracking Framework for Tropical Cyclones

TC-SINDy: Improving physics-based deterministic tropical cyclone track and intensity model via data-driven sparse identification of Nonlinear Dynamics

A New Prediction Model of Daily Weather Elements in Hainan Province under the Typhoon Weather

Comparative study of very short-term flood forecasting using physics-based numerical model and data-driven prediction model

Typhoon disaster emergency forecasting method based on big data

A Physics-informed Deep-learning Intensity Prediction Scheme for Tropical Cyclones over the Western North Pacific

AM-ConvGRU: a spatio-temporal model for typhoon path prediction

Tropical Cyclone Track Forecasting Using Fused Deep Learning From Aligned Reanalysis Data