Research on multivariate Yellow Sea SST week prediction method based on encoder-decoder LSTM

SUN Jiyu1,2ZHU Zemin1,2SONG Jun1,2GUO Junru1,2CAI Yu1,2FU Yanzhao1,2Wang Linhui1,2, Polonsky A.3

 1Dalian Ocean University, School of Ocean Technology and Environment,  Dalian, Liaoning 116023

2Operational Oceanographic Institution, Dalian Ocean University,  Dalian, Liaoning 116023

3Institute of Natural and Technical Systems, RF, Sevastopol, 28 Lenin St. 299011


DOI: 10.33075/2220-5861-2022-1-5-14

UDC 551.465 (262.5)


In order to further improve the accuracy and stability of Yellow Sea surface temperature forecasting, this paper uses the 25-year data of OISST V2.0 and OAFlux. The following factors (determining a sea surface temperature – SST – variability) are considered: radiation and total heat fluxes, wind speed, air temperature and air specific humidity. By controlling the variables and selecting the best model parameters, a multivariate weekly SST prediction model based on the Encoder-Decoder LSTM (Long Short-Term Memory) was constructed for the first time. It is shown that the model can effectively track the daily SST changes and describe its fluctuation changes to achieve relatively accurate prediction.  Taking 2008 as an example the following result is obtained: the daily absolute errors of the test set within a week are 0.3836, 0.4523, 0.5276, 0.5905, 0.6362, 0.6644, and 0.6827, and the overall standard error (RMSE) is 0.7594. It is concluded, that further research is needed on the optimization of predictors and the applicability of single-point forecasting of the model in the SST prediction.


To quote:

Full text in PDF(RUS)