Analysis and forecasting of sea ice concentration for Antarctic using time series

This study employs time-series modeling to forecast the amount of sea ice concentration (SIC) in the Antarctic sea ice region. One of the main factors preventing ships from moving through sea ice is the concentration of sea ice. Its goal is to predict the concentration of sea ice in the Antarctic between the coasts of Bharati and Maitri (Indian Research Stations in Antarctica). Six grids from the study region were chosen for this investigation. Sea ice metrics obtained from satellite remote sensing data (NSIDC sea ice concentration) have been utilized as inputs for ConvLSTM, XGBoost and linear regression to forecast SIC. After training our models with data from the previous ten years (January 1 to April 30), we were able to forecast the observations of the next ten days with high accuracy. This suggests that our model can accurately forecast sea ice concentration values for each node with an accuracy of 90% or greater. The models’ excellent forecasting accuracy can reduce the risks posed by sea ice and enhance navigation safety and logistics for ships operating in Antarctic region.