Deep Residual Convolutional Neural Network Combining Dropout and Transfer Learning for ENSO Forecasting. Issue 24 (15th December 2021)

• Record Type: Journal Article
• Authors: Hu, Jie; Weng, Bin; Huang, Tianqiang; Gao, Jianyun; Ye, Feng; You, Lijun
• Is Part Of: Geophysical research letters. Volume 48:Issue 24(2021)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2021-12-15
• Subjects: ENSO -- forecasting -- deep learning -- residual neural network -- transfer learning -- dropout; Geophysics -- Periodicals; Planets -- Periodicals; Lunar geology -- Periodicals; 550
• DOI: 10.1029/2021GL093531
• Languages: English
• ISSNs: 0094-8276

Abstract

Abstract: To improve EI Niño‐Southern Oscillation (ENSO) amplitude and type forecast, we propose a model based on a deep residual convolutional neural network with few parameters. We leverage dropout and transfer learning to overcome the challenge of insufficient data in model training process. By applying the dropout technique, the model effectively predicts the Niño3.4 Index at a lead time of 20 months during the 1984–2017 evaluation period, which is three more months than that by the existing optimal model. Moreover, with homogeneous transfer learning this model precisely predicts the Oceanic Niño Index up to 18 months in advance. Using heterogeneous transfer learning this model achieved 83.3% accuracy for forecasting the 12‐month‐lead EI Niño type. These results suggest that our proposed model can enhance the ENSO prediction performance. Plain Language Summary: El Niño‐Southern Oscillation (ENSO) is an irregular periodic variation along with complex tropical atmosphere‐ocean interaction. It impacts interannually human lives globally and locally. Hence, we contribute, the first time as we know, a deep learning model that can effectively predict EI Niño strength and type. The model can transfer the knowledge learned from Niño3.4 Index prediction to Oceanic Niño Index and type prediction, respectively. We find that our proposed model has a high correlation skill and a good precision for predicting strength and type respectively in relation to an evaluation between 1984 and 2017. Moreover, our model requires smaller‐sized storage against the existing deep learning model. Key Points: Deep residual convolutional neural network is designed to forecast the amplitude and type of ENSO The prediction skill is improved by applying dropout and transfer learning Our method can successfully predict 20 months in advance for the period between 1984 and 2017