Physics-guided deep learning for global sea surface temperature forecasting: Balancing accuracy and stability across timescales

Abstract

Accurate sea surface temperature (SST) forecasting across multiple timescales remains challenging. Daily forecasting frequently relies on autoregressive models prone to instability and over-smoothing, whereas monthly forecasting suffers from sparse data and the complex dynamics of ocean systems. Existing deep learning methods struggle to address these diverse challenges simultaneously. We introduce SSTFormer, a novel physics-guided deep learning framework that achieves leading results, with root mean squared error of 0.17 °C for daily forecasts and 0.60 °C for monthly forecasts, yielding lower bias and improved spatial coherence. The model’s core innovation is its unified and flexible architecture. For multi-step daily forecasts (1-15 days), it deploys as a ‘‘two-phase sequential ensemble” that replaces conventional autoregression and uses ocean current to solve instability and mitigate error accumulation. For single-step monthly forecasts, it is used in a direct forecasting configuration, proving effective at handling ‘‘sparse data” and ‘‘complex ocean dynamics.” SSTFormer demonstrates how a single architecture, through flexible deployment, can address the unique challenges of multi-scale SST forecasting, highlighting its potential as a unified and robust framework.

Keywords

Global sea surface temperature / Prediction / Ocean currents / Deep learning / Physical guidance

Data a vailability

Optimum Interpolation Sea Surface Temperature (OISST) data (Reynolds et al., 2007; NOAA National Centers for Environmental Information, 2023) were obtained from the NOAA National Centers for Environmental Information (available at https://www.ncei.noaa.gov/products/optimum-interpolation-sst). The Multi Observation Global Ocean 3D Temperature Salinity Height Geostrophic Current and Mixed Layer Depth dataset (Copernicus Marine Service, 2023) was obtained from the Copernicus Marine Service data portal (available at https://data.marine.copernicus.eu/product/MULTIOBS_GLO_PHY_TSUV_3D_MYNRT_015_012/download), and the Global Ocean Gridded Level 4 Sea Surface Heights and Derived Variables dataset (Copernicus Marine Service, 2023) was obtained from the same portal (available at https://data.marine.copernicus.eu/product/SEALEVEL_GLO_PHY_CLIMATE_L4_MY_008_057/download). The Niño 3.4 index (NOAA Physical Sciences Laboratory, 2023) was obtained from the NOAA Physical Sciences Laboratory (available at https://psl.noaa.gov/data/timeseries/month/Nino34_CPC).

CRediT authorship contribution statement

Shiji Dong: Writing - original draft, Methodology, Investigation. Yan Li: Validation, Investigation. Xiaobin Yin: Supervision, Methodology, Conceptualization. Qing Xu: Writing - review & editing. Peng Mao: Writing - original draft, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements

This work was supported by the National Natural Science Foun-dation of China (Grant No. 42406175), the Marine S&T Fund of Shandong Province for Pilot National Laboratory for Marine Science and Technology (Qingdao) (No.2022QNLM050301-1), the Hainan Key Research and Development Program (ZDYF2023SHFZ089).

References

[1] Cane, M.A., Clement, A.C., Kaplan, A., Kushnir, Y., Pozdnyakov, D., Seager, R., Zebiak, S.E., Murtugudde, R., 1997. Twentieth-century sea surface temperature trends. Science 275, 957-960. https://doi.org/10.1126/science.275.5302.957.

[2] Chen, X., Li, Q., Zeng, X., Zhang, C., Xu, G., Wang, G., 2022. A hybrid ARIMA-GABP model for predicting sea surface temperature. Electronics 11, 2359. https://doi.org/10.3390/electronics11152359.

[3] Dai, H., He, Z., Wei, G., Lei, F., Zhang, X., Zhang, W., Shang, S., 2024. Long-term prediction of sea surface temperature by temporal embedding transformer with attention distilling and partial stacked connection. IEEE J. Sel. Topics Appl. Earth Observ. Remote Sens. 17, 4280-4293. https://doi.org/10.1109/JSTARS.2024.3357191.

[4] De Mattos Neto, P.S.G., Cavalcanti, G.D.C., Santos, D.O., Júnior, D.S., Silva, E.G., 2022. Hybrid systems using residual modeling for sea surface temperature forecasting. Sci. Rep. 12, 487. https://doi.org/10.1038/s41598-021-04238-z.

[5] Deser, C., Alexander, M.A., Xie, S.-P., Phillips, A.S., 2010. Sea surface temperature variability: Patterns and mechanisms. Annu. Rev. Mar. Sci. 2, 115-143. https://doi.org/10.1146/annurev-marine-120408-151453.

[6] Fei, T., Huang, B., Wang, X., Zhu, J., Chen, Y., Wang, H., Zhang, W., 2022. A hybrid deep learning model for the bias correction of SST numerical forecast products using satellite data. Remote Sens. 14, 1339. https://doi.org/10.3390/rs14061339.

[7] Feng, Y., Sun, T., Li, C., 2021. Study on long term sea surface temperature (SST) prediction based on temporal convolutional network (TCN) method. In: ACM Turing Award Celebration Conference - China (ACM TURC 2021), ACM TURC ’21. Association for Computing Machinery, New York, USA, pp. 28-32. https://doi.org/10.1145/3472634.3472641

[8] Fiedler, P.C., Lavín, M.F., 2017. Oceanographic conditions of the eastern tropical Pacific. In: Glynn P.W., Manzello D.P., Enochs I.C. (Eds.), Coral Reefs of the Eastern Tropical Pacific: Persistence and Loss in a Dynamic Environment. Springer, Netherlands, Dordrecht, pp. 59-83. https://doi.org/10.1007/978-94-017-7499-4_3.

[9] Goswami, B.B., An, S.-I., 2023. An assessment of the ENSO-monsoon teleconnection in a warming climate. npj Clim. Atmos. Sci. 6, 82. https://doi.org/10.1038/s41612-023-00411-5.

[10] Ham, Y.-G., Kim, J.-H., Luo, J.-J., 2019. Deep learning for multi-year ENSO forecasts. Nature 573, 568-572. https://doi.org/10.1038/s41586-019-1559-7.

[11] Hou, S., Li, W., Liu, T., Zhou, S., Guan, J., Qin, R., Wang, Z., 2022. MIMO: A unified spatio-temporal model for multi-scale sea surface temperature prediction. Remote Sens. 14, 2371. https://doi.org/10.3390/rs14102371.

[12] Hu, D., Wu, L., Cai, W., Gupta, A.S., Ganachaud, A., Qiu, B., Gordon, A.L., Lin, X., Chen, Z., Hu, S., Wang, G., Wang, Q., Sprintall, J., Qu, T., Kashino, Y., Wang, F., Kessler, W.S., 2015. Pacific western boundary currents and their roles in climate. Nature 522, 299-308. https://doi.org/10.1038/nature14504.

[13] Krishnamurti, T.N., Chakraborty, A., Krishnamurti, R., Dewar, W.K., Clayson, C.A., 2006. Seasonal prediction of sea surface temperature anomalies using a suite of 13 coupled atmosphere-ocean models. J. Climate 19, 6069-6088. https://doi.org/10.1175/JCLI3938.1.

[14] Li, X., Guo, X., Jiang, H., Lu, X., Zhang, Z., Ma, J., Hu, S., 2024. Long-term variability of the Kuroshio since 1788 based on coral records. Glob. Planet. Change 243, 104611. https://doi.org/10.1016/j.gloplacha.2024.104611.

[15] Ma, Y., Xie, B., Feng, Z., Sun, G., Zhang, C., Yang, S., 2025. A deep learning-based hybrid model for improved SST prediction in the tropical Pacific Ocean. J. Oceanol. Limnol. 43, 1709-1725. https://doi.org/10.1007/s00343-025-4333-8.

[16] Needham, H.F., Keim, B.D., Sathiaraj, D., 2015. A review of tropical cyclone-generated storm surges: Global data sources, observations, and impacts. Rev. Geophys. 53, 545-591. https://doi.org/10.1002/2014RG000477.

[17] Odena, A., Dumoulin, V., Olah, C., 2016. Deconvolution and checkerboard artifacts. Distill 1 https://doi.org/10.23915/distill.00003.

[18] Pan, X., Jiang, T., Sun, W., Xie, J., Wu, P., Zhang, Z., Cui, T., 2024. Effective attention model for global sea surface temperature prediction. Expert Systems with Applications 254, 124411. https://doi.org/10.1016/j.eswa.2024.124411.

[19] Patil, K., Deo, M.C., Ravichandran, M., 2016. Prediction of sea surface temperature by combining numerical and neural techniques. J. Atmos. Oceanic Technol. 33, 1715-1726. https://doi.org/10.1175/JTECH-D-15-0213.1.

[20] Srinivas, G., Vialard, J., Lengaigne, M., Izumo, T., Guilyardi, E., 2022. Relative contributions of sea surface temperature and atmospheric nonlinearities to ENSO asymmetrical rainfall response. J. Climate 35, 3725-3745. https://doi.org/10.1175/JCLI-D-21-0257.1.

[21] Stockdale, T.N., Balmaseda, M.A., Vidard, A., 2006. Tropical atlantic SST prediction with coupled ocean-atmosphere GCMs. J. Climate 19, 6047-6061. https://doi.org/10.1175/JCLI3947.1.

[22] Timmermann, A., An, S.-I., Kug, J.-S., Jin, F.-F., Cai, W., Capotondi, A., Cobb, K.M., Lengaigne, M., McPhaden, M.J., Stuecker, M.F., Stein, K., Wittenberg, A.T., Yun, K.-S., Bayr, T., Chen, H.-C., Chikamoto, Y., Dewitte, B., Dommenget, D., Grothe, P., Guilyardi, E., Ham, Y.-G., Hayashi, M., Ineson, S., Kang, D., Kim, S., Kim, W., Lee, J.-Y., Li, T., Luo, J.-J., McGregor, S., Planton, Y., Power, S., Rashid, H., Ren, H.-L., Santoso, A., Takahashi, K., Todd, A., Wang, G., Wang, G., Xie, R., Yang, W.-H., Yeh, S.-W., Yoon, J., Zeller, E., Zhang, X., 2018. El Niño-Southern Oscillation complexity. Nature 559, 535-545. https://doi.org/10.1038/s41586-018-0252-6.

[23] Tippett, M.K., Trenary, L., DelSole, T., Pegion, K., L’Heureux, M.L., 2018. Sources of bias in the monthly CFSv2 forecast climatology. J. Appl. Meteorol. Clim. 57, 1111-1122. https://doi.org/10.1175/JAMC-D-17-0299.1.

[24] Tochimoto, E., Iizuka, S., 2023. Impact of warm sea surface temperature over a Kuroshio large meander on extreme heavy rainfall caused by an extratropical cyclone. Atmos. Sci. Lett. 24, e1135.

[25] Verdy, A., Marshall, J., Czaja, A., 2006. Sea surface temperature variability along the path of the Antarctic circumpolar current. J. Phys. Oceanogr. 36, 1317-1331. https://doi.org/10.1175/JPO2913.1.

[26] Wan Mohamad Nawi, W.I.A., Lola, M.S., Zakariya, R., Zainuddin, N.H., K. Abd Hamid, Abd.A., Aruchunan, E., 2021. Improved of forecasting sea surface temperature based on hybrid ARIMA and support vector machines models. Malays. J. Fundam. Appl. Sci. 17, 609-620. https://doi.org/10.11113/mjfas.v17n5.2356.

[27] Wang, H., Hu, S., Guan, C., Li, X., 2024. The role of sea surface salinity in ENSO forecasting in the 21st century. npj Clim. Atmos. Sci. 7, 206. https://doi.org/10.1038/s41612-024-00763-6.

[28] Webster, P.J., Yang, S., 1992. Monsoon and Enso: Selectively interactive systems. Q. J. R. Meteorol. Soc. 118, 877-926. https://doi.org/10.1002/qj.49711850705.

[29] Wentz, F.J., Gentemann, C., Smith, D., Chelton, D., 2000. Satellite measurements of sea surface temperature through clouds. Science 288, 847-850. https://doi.org/10.1126/science.288.5467.847.

[30] Xiao, C., Chen, N., Hu, C., Wang, K., Gong, J., Chen, Z., 2019. Short and mid-term sea surface temperature prediction using time-series satellite data and LSTM-AdaBoost combination approach. Remote Sens. Environ. 233, 111358. https://doi.org/10.1016/j.rse.2019.111358.

[31] Xu, Z., Ji, F., Liu, B., Feng, T., Gao, Y., He, Y., Chang, F., 2021. Long-term evolution of global sea surface temperature trend. Int. J. Climatol. 41, 4494-4508. https://doi.org/10.1002/joc.7082.

[32] Yamagami, Y., Tatebe, H., Kohyama, T., Kido, S., Okajima, S., 2025. Gulf Stream drives Kuroshio behind the recent abnormal ocean warming. arXiv, 2503.01117. https://doi.org/10.48550/arXiv.2503.01117.

[33] Zhao, X., Qi, J., Yu, Y., Zhou, L., 2024. Deep learning for ocean temperature forecasting: a survey. Intell. Mar. Technol. Syst. 2, 28. https://doi.org/10.1007/s44295-024-00042-3.

[34] Zheng, G., Li, X., Zhang, R.-H., Liu, B., 2020. Purely satellite data-driven deep learning forecast of complicated tropical instability waves. Sci. Adv. 6, eaba1482. https://doi.org/10.1126/sciadv.aba1482.