Preprints
https://doi.org/10.5194/egusphere-2022-1216
https://doi.org/10.5194/egusphere-2022-1216
 
21 Nov 2022
21 Nov 2022
Status: this preprint is open for discussion.

Forecasting tropical cyclone tracks in the Northwest Pacific based on a deep-learning model

Liang Wang1,2, Bingcheng Wan3, Shaohui Zhou3, Haofei Sun1,2, and Zhiqiu Gao1,3 Liang Wang et al.
  • 1State Key Laboratory of Atmospheric Boundary Layer Physics and Atmospheric Chemistry, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing, 100029, China
  • 2University of Chinese Academy of Sciences, Beijing, 100049, China
  • 3School of Atmospheric Physics, Nanjing University of Information Science and Technology, Nanjing, 210044, China

Abstract. Tropical cyclones (TCs) are one of the most severe meteorological disasters, making rapid and accurate track forecasts crucial for disaster prevention and mitigation. Because TC tracks are affected by various factors (the steering flow, thermal structure of the underlying surface, and atmospheric circulation), their trajectories present highly complex nonlinear behavior. Deep learning has many advantages in simulating nonlinear systems. In this paper, we explore the movement of TCs in the Northwest Pacific from 1979 to 2021 based on deep-learning technology, divided into training (1979–2014), validation (2015–2018), and test sets (2019–2021), and create 6–72 h TC track forecasts. Only historical trajectory data are used as input for evaluating the forecasts of the three recurrent neural networks utilized: recurrent neural network (RNN), long short-term memory (LSTM), and gated recurrent unit (GRU) models. The GRU approach performed best; to further improve forecast accuracy, a model combining GRU and a convolutional neural network (CNN) called GRU_CNN is proposed to capture the characteristics varying with time. By adding reanalysis data of the steering flow, sea-surface temperatures, and geopotential height around the cyclone, we can extract sufficient information on the historical trajectory features and three-dimensional spatial features. The results show that GRU_CNN outperforms other deep-learning models without CNN layers. Furthermore, by analyzing three additional environmental factors through control experiments, it can be concluded that the historical steering flow of TCs plays a key role, especially for short-term predictions within 24 h, while sea-surface temperatures and geopotential height can gradually improve the 24–72-h forecast. The average distance errors at 6 h and 12 h are 17.22 km and 43.90 km, respectively. Compared with the forecast results of the Central Meteorological Observatory, the model proposed herein is suitable for short-term forecasting of TC tracks.

Liang Wang et al.

Status: open (until 16 Jan 2023)

Comment types: AC – author | RC – referee | CC – community | EC – editor | CEC – chief editor | : Report abuse

Liang Wang et al.

Viewed

Total article views: 110 (including HTML, PDF, and XML)
HTML PDF XML Total Supplement BibTeX EndNote
81 24 5 110 16 2 2
  • HTML: 81
  • PDF: 24
  • XML: 5
  • Total: 110
  • Supplement: 16
  • BibTeX: 2
  • EndNote: 2
Views and downloads (calculated since 21 Nov 2022)
Cumulative views and downloads (calculated since 21 Nov 2022)

Viewed (geographical distribution)

Total article views: 105 (including HTML, PDF, and XML) Thereof 105 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
1
 
 
 
 
Latest update: 06 Dec 2022
Download
Short summary
The past 24-h TC trajectory and meteorological field data were used to forecast TC tracks in the Northwest Pacific from hours 6–72 based on GRU_CNN we proposed in this paper, which has better prediction results than traditional single deep-learning methods. The historical steering flow of cyclones has a significant effect on improving the accuracy of short-term forecasting, while, in long-term forecasting, the SST and geopotential height will have a particular impact.