PREDICTION OF THE ANTARCTIC MARGINAL ICE ZONE EXTENT BASED UPON ITS MULTIFRACTAL PROPERTY

The marginal ice zone (MIZ) around Antarctica is an important air-ice-ocean-wave interaction area and a crucial habitat for marine life. Given its dynamic nature, it is essential to understand and quantify both the short- and long-term changes of its extent. In this study, we investigated the MIZ extent time series using multifractal spectrum and R/S analysis, examining Antarctica as a whole and per sub-region. The MIZ extent was predicted using a back-propagation (BP) algorithm. The results show that the wide multifractal spectrum width of the MIZ extent time series has strong multifractal features for the entire Antarctic and for the five sub-regions. By comparing the spectrum width of the original time series with those of a shuffled series, we found that multifractality is related to long-range correlations in the time series. The Hurst exponent obtained using the R/S analysis indicates that the long-range correlation over six months is strongest for all time intervals. Compared to the autoregressive integrated moving average model, the mean absolute percentage error obtained using the BP algorithm was lower by 4-15%. We conclude that the BP algorithm combined with the multifractal property is well suited to predict the Antarctic MIZ extent.