•  
  •  
 

Abstract

To investigate the impact of climatic factors on skipjack tuna catch, this study collected and analyzed annual global skipjack tuna catch data spanning from 1960 to 2021, along with corresponding monthly data for the key indicator of El Niño events (Niño1+2, Niño3, Niño4, and Niño3.4), the Southern Oscillation Index (SOI), the North Atlantic Oscillation (NAO), the Pacific Decadal Oscillation (PDO), the North Pacific Index (NPI), and the global sea-air temperature anomaly index (dT). The optimal lag years of climatic factors were identified by selecting the maximum absolute crosscorrelation coefficients. Subsequently, the predictive performance of LSTM, BiLSTM, and CNN-BiLSTM models was evaluated using climatic factors as input variables. The relative contributions of individual climatic factors to catch prediction were further examined using the best-performing model. The results identified the optimal lag time for each climatic variable, with Niño1+2, dT, SOI, NPI, NAO, and PDO corresponding to 12, 14, 2, 15, 13, and 0 years, respectively. In terms of prediction accuracy, the three deep learning models performed in descending order: CNN-BiLSTM, BiLSTM, and LSTM. The CNN-BiLSTM model exhibited a fitting degree of 0.919 between predicted and actual values, with a trend that was largely consistent, suggesting excellent model fitting. The importance indices, ranked in descending order, are dT,

NPI, NAO, Niño1+2, SOI, and PDO. These findings underscore the high degree of accuracy that the CNN-BiLSTM model possesses in predicting skipjack tuna catch.

Download
COinS