Ensemble forecast of tropical cyclone tracks based on deep neural networks

Abstract

A nonlinear artificial intelligence ensemble forecast model has been developed in this paper for predicting tropical cyclone (TC) tracks based on the deep neural network (DNN) by using the 24-h forecast data from the China Meteorological Administration (CMA), Japan Meteorological Agency (JMA) and Joint Typhoon Warning Center (JTWC). Data from a total of 287 TC cases over the Northwest Pacific Ocean from 2004 to 2015 were used to train and validate the DNN based ensemble forecast (DNNEF) model. The comparison of model results with Best Track data of TCs shows that the DNNEF model has a higher accuracy than any individual forecast center or the traditional ensemble forecast model. The average 24-h forecast error of 82 TCs from 2016 to 2018 is 63 km, which has been reduced by 17.1%, 16.0%, 20.3%, and 4.6%, respectively, compared with that of CMA, JMA, JTWC, and the error-estimation based ensemble method. The results indicate that the nonlinear DNNEF model has the capability of adjusting the model parameter dynamically and automatically, thus improving the accuracy and stability of TC prediction.

Graphical abstract

Keywords

tropical cyclone track / deep neural network / ensemble forecast

References

[1] BuckinghamC, MarchokT, GinisI, RothsteinL, RoweD. ( 2010). Short- and medium-range prediction of tropical and transitioning cyclone tracks within the NCEP global ensemble forecasting system. Weather Forecast, 25( 6): 1736– 1754

[2] BuizzaR, MilleerM, PalmerT N. ( 1999). Stochastic representation of model uncertainties in the ECMWF ensemble prediction system. Q J R Meteorol Soc, 125( 560): 2887– 2908

[3] ChenG M, YuH, CaoQ, ZengZ. ( 2013). The performance of global models in TC track forecasting over the Western North Pacific from 2010 to 2012. Trop Cyclone Res Rev, 2( 3): 149– 158

[4] DengL LiJ Huang J T YaoK YuD Seide F SeltzerM ZweigG HeX Williams J GongY AceroA (2013). Recent advances in deep learning for speech research at Microsoft. In: Proceedings of International Conference on Acoustics, Speech, and Signal. Vancouver: 8604– 8608

[5] ElsberryR L. ( 1995). Recent advancements in dynamical tropical cyclone track predictions. Meteorol Atmos Phys, 56( 1−2): 81– 99

[6] EstevaA, KuprelB, NovoaR A, KoJ, SwetterS M, BlauH M, ThrunS. ( 2017). Dermatologist-level classification of skin cancer with deep neural networks. Nature, 542( 7639): 115– 118

[7] EpsteinE S. ( 1969). Stochastic dynamic prediction. Tellus, 21( 6): 739– 759

[8] GaoS, ZhaoP, PanB, LiY, ZhouM, XuJ, ZhongS, ShiZ. ( 2018). A nowcasting model for the prediction of typhoon tracks based on a long short term memory neural network. Acta Oceanol Sin, 37( 5): 8– 12

[9] HamillT M, WhitakerJ S, FiorinoM, BenjaminS G. ( 2011). Global ensemble predictions of 2009’s tropical cyclones initialized with an ensemble kalman filter. Mon Weather Rev, 139( 2): 668– 688

[10] JeffriesR A SampsonC R CarrL E ChuJ (1993). Tropical cyclone forecasters reference guide numerical track forecast guidance. Tech. Rep. No. NRL/PU/7515–93–0011

[11] JinJ, LiM, JinL. ( 2015). Data normalization to accelerate training for linear neural net to predict tropical cyclone tracks. Math Probl Eng, 2015: 931629

[12] KnaffJ A, SampsonC R, DeMariaM, MarchokT P, GrossJ M, McAdieC J. ( 2007). Statistical tropical cyclone wind radii prediction using climatology and persistence. Weather Forecast, 22( 4): 781– 791

[13] KrishnamurtiT N, KishtawalC M, LaRowT E, BachiochiD R, ZhangZ, WillifordC E, GadgilS, SurendranS. ( 1999). Improved weather and seasonal climate forecasts from multimodel superensemble. Science, 285( 5433): 1548– 1550

[14] KrizhevskyA, SutskeverI, HintonG E. ( 2012). ImageNet classification with deep convolutional neural networks. In: Proceedings of the 25th International Conference on Neural Information Processing Systems, 1097– 1105

[15] LandseaC W CangialosiJ P ( 2018). Have we reached the limits of predictability for tropical cyclone track forecasting? Bull Am Meteorol Soc, 99(11): 2237–2243

[16] LeithC E. ( 1974). Theoretical skill of Monte Carlo forecasts. Mon Weather Rev, 102( 6): 409– 418

[17] LeutbecherM, PalmerT N. ( 2008). Ensemble forecasting. J Comput Phys, 227( 7): 3515– 3539

[18] LiH, LuoJ, XuM. ( 2019). Ensemble data assimilation and prediction of typhoon and associated hazards using TEDAPS: Evaluation for 2015–18 Seasons. Front Earth Sci, 13( 4): 733– 743

[19] NeumannC J, LawrenceM B. ( 1975). An operational experiment in the satistical-dynamical prediction of tropical cyclone motion. Mon Weather Rev, 103( 8): 665– 673

[20] PengX, FeiJ, HuangX, ChengX. ( 2017). Evaluation and error analysis of official forecasts of tropical cyclones during 2005–14 over the Western North Pacific. Part I: storm Tracks. Weather Forecast, 32( 2): 689– 712

[21] PluM. ( 2011). A new assessment of the predictability of tropical cyclone tracks. Mon Weather Rev, 139( 11): 3600– 3608

[22] RaddawayB (2012). Newsletter No.130-Winter 2011/12. Available at ECMWF website

[23] RappaportE N, FranklinJ L, AvilaL A, BaigS R, BevenJ L II, BlakeE S, BurrC A, JiingJ G, JuckinsC A, KnabbR D, LandseaC W, MainelliM, MayfieldM, McAdieC J, PaschR J, SiskoC, StewartS R, TribbleA N. ( 2009). Advances and challenges at the National Hurricane Center. Weather Forecast, 24( 2): 395– 419

[24] SahaS, MoorthiS, WuX, WangJ, NadigaS, TrippP, BehringerD, HouY T, ChuangH, IredellM, EkM, MengJ, YangR, MendezM P, van den DoolH, ZhangQ, WangW, ChenM, BeckerE. ( 2014). The NCEP climate forecast system version2. J Clim, 27( 6): 2185– 2208

[25] SilverD, HuangA, MaddisonC J, GuezA, SifreL, van den DriesscheG, SchrittwieserJ, AntonoglouI, PanneershelvamV, LanctotM, DielemanS, GreweD, NhamJ, KalchbrennerN, SutskeverI, LillicrapT, LeachM, KavukcuogluK, GraepelT, HassabisD. ( 2016). Mastering the game of Go with deep neural networks and tree search. Nature, 529( 7587): 484– 489

[26] SzeV, ChenY, YangT, EmerJ S. ( 2017). Efficient processing of deep neural networks: a tutorial and survey. Proc IEEE, 105( 12): 2295– 2329

[27] TothZ, KalnayE. ( 1997). Ensemble forecasting at NCEP and the breeding method. Mon Weather Rev, 125( 12): 3297– 3319

[28] VeigasK W (1996). The development of statistical-physical hurricane prediction model. Final Report, U.S.W.B. Contract Cwb 10966, Travellers Weather Research Center, Hartford, CT, 1996. 19

[29] WuL, ZongH, LiangJ. ( 2011). Observational analysis of sudden tropical cyclone track changes in the vicinity of the East China Sea. J Atmos Sci, 68( 12): 3012– 3031

[30] YingM, ZhangW, YuH, LuX, FengJ, FanY, ZhuY, ChenD. ( 2014). An overview of the China Meteorological Administration tropical cyclone database. J Atmos Ocean Technol, 31( 2): 287– 301

[31] YuanJ ChenY PanY Dong J LuoY (2017). Improvement of ensemble forecast of typhoon track in the Northwestern Pacific. Marine Forecasts, 34(2): 37– 42 (in Chinese)

[32] ZhangR H, ShenX S. ( 2008). On the development of GRAPES-A new generation of the national operational NWP system in China. Chin Sci Bull, 53: 3429– 3432

[33] ZhiX, ZhangL, BaiY. ( 2011). Application of the multi-model ensemble forecast in the QPF. In: Proceedings of International Conference on Information Science and Technology, 657– 660

[34] ZhuL, JinJ, CannonA J, HsiehW W. ( 2016). Bayesian neural networks based bootstrap aggregating for tropical cyclone tracks prediction in South China Sea. In: Proceeding of International Conference on Neural Information Processing, 475– 482

Acknowledgments

This study was supported by the National Key Project of Research and Development Plan of China (No. 2016YFC1401905), the National Natural Science Foundation of China (Grant Nos. 41976163 and 41575107), the Key Special Project for Introduced Talents Team of Southern Marine Science and Engineering Guangdong Laboratory (Guangzhou) (No. GML2019ZD0302), and the Guangdong Special Fund Program for Marine Economy Development (No. GDNRC[2020]050).

RIGHTS & PERMISSIONS

2022 Higher Education Press