Nowcasting Trajektori dan Intensitas Siklon Tropis Seroja Menggunakan Temporal Convolutional Network

Pertiwi Setyowati; Yosafat Donni Haryanto; Adi Mulsandi; Latifah Nurul Qomariyatuzzamzami

doi:10.25077/jfu.15.5.497-504.2026

Authors

Pertiwi Setyowati Sekolah Tinggi Meteorologi Klimatologi dan Geofisika
Yosafat Donni Haryanto Sekolah Tinggi Meteorologi Klimatologi dan Geofisika
Adi Mulsandi Sekolah Tinggi Meteorologi Klimatologi dan Geofisika
Latifah Nurul Qomariyatuzzamzami Sekolah Tinggi Meteorologi Klimatologi dan Geofisika

DOI:

https://doi.org/10.25077/jfu.15.5.497-504.2026

Keywords:

Tropical stroom, Deep Learning, Prediction, TCN

Abstract

Tropical cyclones have an extremely short detection-to-peak-impact lead time, as exemplified by Tropical Cyclone Seroja, which caused catastrophic damage in East Nusa Tenggara in April 2021, while conventional methods such as the Dvorak technique, the Advanced Dvorak Technique, and Numerical Weather Prediction remain limited in update speed and objectivity. This study applies the Temporal Convolutional Network (TCN) for nowcasting the track and intensity of Tropical Cyclone Seroja as a faster, computationally lighter alternative capable of capturing nonlinear cyclone dynamics. The study uses historical best-track data from the Bureau of Meteorology (BOM) Australia covering 1973–2026 (longitude, latitude, central pressure, wind speed), with the model trained via a sliding-window approach for predictions at 6–24-hour horizons and tested on Cyclone Seroja, evaluated using RMSE, MAPE, and R². Results show TCN performs best in track prediction (longitude and latitude RMSE of 1.01–2.97 and 0.40–1.88, consistently high R² up to 24 hours), whereas intensity performance declines sharply (central-pressure RMSE of 4.01–10.40 hPa, wind-speed RMSE of 2.87–7.22 m/s, wind-speed MAPE up to 23.77%, negative R²), associated with the interaction between Cyclone Seroja and Tropical Cyclone Odette through the Fujiwhara effect. TCN has potential as an alternative to conventional methods for tropical cyclone early-warning systems in Indonesia, particularly for track prediction, with further development needed to improve intensity-prediction accuracy.

References

Alemany, S., Beltran, J., Perez, A., & Ganzfried, S. (2019). Predicting Hurricane Trajectories Using a Recurrent Neural Network. Proceedings of the AAAI Conference on Artificial Intelligence, 33(01), 468–475. https://doi.org/10.1609/aaai.v33i01.3301468

Bai, S., Kolter, J. Z., & Koltun, V. (2018). An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling (arXiv:1803.01271). arXiv. https://doi.org/10.48550/arXiv.1803.01271

Gan, S. L., Fu, J. Y., Zhao, G. F., Chan, P. W., & He, Y. C. (2024). Short-term prediction of tropical cyclone track and intensity via four mainstream deep learning techniques. Journal of Wind Engineering and Industrial Aerodynamics, 244, 105633. https://doi.org/10.1016/j.jweia.2023.105633

Knaff, J. A., Brown, D. P., Courtney, J., Gallina, G. M., & Beven, J. L. (2010). An Evaluation of Dvorak Technique–Based Tropical Cyclone Intensity Estimates. Weather and Forecasting, 25(5), 1362–1379. https://doi.org/10.1175/2010WAF2222375.1

Kurniawan, R., Harsa, H., Nurrahmat, M. H., Sasmito, A., Florida, N., Makmur, E. E. S., Swarinoto, Y. S., Habibie, M. N., Hutapea, T. F., Sudewi, R. S., Fitria, W., Praja, A. S., & Adrianita, F. (2021). The Impact of Tropical Cyclone Seroja to The Rainfall and Sea Wave Height in East Nusa Tenggara. IOP Conference Series: Earth and Environmental Science, 925(1), 012049. https://doi.org/10.1088/1755-1315/925/1/012049

Latos, B., Peyrillé, P., Lefort, T., Baranowski, D. B., Flatau, M. K., Flatau, P. J., Riama, N. F., Permana, D. S., Rydbeck, A. V., & Matthews, A. J. (2023). The role of tropical waves in the genesis of Tropical Cyclone Seroja in the Maritime Continent. Nature Communications, 14(1), 856. https://doi.org/10.1038/s41467-023-36498-w

Mulyana, E., Prayoga, M. B. R., Yananto, A., Wirahma, S., Aldrian, E., Harsoyo, B., Seto, T. H., & Sunarya, Y. (2018). Tropical cyclones characteristic in southern Indonesia and the impact on extreme rainfall event. MATEC Web of Conferences, 229, 02007. https://doi.org/10.1051/matecconf/201822902007

Olander, T. L., & Velden, C. S. (2019). The Advanced Dvorak Technique (ADT) for Estimating Tropical Cyclone Intensity: Update and New Capabilities. Weather and Forecasting, 34(4), 905–922. https://doi.org/10.1175/WAF-D-19-0007.1

Pu, J., Mu, M., Feng, J., Zhong, X., & Li, H. (2025). A fast physics-based perturbation generator of machine learning weather model for efficient ensemble forecasts of tropical cyclone track. Npj Climate and Atmospheric Science, 8(1), 128. https://doi.org/10.1038/s41612-025-01009-9

Safwan Mahmood Al-Selwi, Mohd Fadzil Hassan, Said Jadid Abdulkadir, & Amgad Muneer. (2023). LSTM Inefficiency in Long-Term Dependencies Regression Problems. Journal of Advanced Research in Applied Sciences and Engineering Technology, 30(3), 16–31. https://doi.org/10.37934/araset.30.3.1631

Setiawan, M. A., Winastuti, R., Hayat, D. M., Christanto, N., Wahyu Rahmadana, A. D., Meilinarti, Ngurah, I., Ndapareda, E., Swastanto, G. A., & Amri, I. (2025). Tropical cyclone Seroja’s aftermath: Flash flood inundation modeling and recovery efforts in rural coastal Lembata, East Nusa Tenggara Province, Indonesia. Tropical Cyclone Research and Review, 14(4), 323–339. https://doi.org/10.1016/j.tcrr.2025.11.010

Trenggono, M., Berlianty, D., Priyono, B., Wei, Z., Li, S., & Xu, T. (2024). Impact of the Fujiwhara effect from tropical cyclones Seroja and Odette on ocean dynamic in Southern Indonesia: Insights from argo data and model analysis. Regional Studies in Marine Science, 80, 103877. https://doi.org/10.1016/j.rsma.2024.103877

Wang, L., Wan, B., Zhou, S., Sun, H., & Gao, Z. (2023). Forecasting tropical cyclone tracks in the northwestern Pacific based on a deep-learning model. Geoscientific Model Development, 16(8), 2167–2179. https://doi.org/10.5194/gmd-16-2167-2023

Yuan, S., Wang, C., Mu, B., Zhou, F., & Duan, W. (2021). Typhoon Intensity Forecasting Based on LSTM Using the Rolling Forecast Method. Algorithms, 14(3), 83. https://doi.org/10.3390/a14030083.