A Multi-Year Forecast Dataset of Arctic Sea Ice Concentration Based on Spectral Analysis and Modeling

1. Data Description The data include both codes, forecasted and observed Arctic SIC. The forecasted data, contains the predicted SIC from January, 1, 2020 to December, 31, 2029. The observed data is SIC from 2013 to 2024 is obtained by the University of Bremen (https://seaice.uni-bremen.de/sea-ice-concentration/amsre-amsr2/). 2. SIC data Observed Data: Temporal Coverage: January 1, 2013, to December 31, 2024. Data Formats: NetCDF, format. Naming Convention: Files are named according to the date in the format asi-AMSR2-n6250-YYYYMMDD-v5.4.nc. Forecasted Data: Temporal Coverage: January 1, 2020, to December 31, 2029. Data Formats: GeoTIFF format. Projection System: Northern Polar Stereographic. Spatial Resolution: 6.25 km. 3. code This repository provides a complete pipeline for long-term daily Arctic sea ice concentration (SIC) prediction based on time series modeling. The included code consists of four major functional components: 🔧 1. Data Preprocessing preprocess.py: This script processes raw NetCDF-format Sea Ice Concentration (SIC) data from the Bremen University dataset. The preprocessed data is also provided as input_data for direct use in 2Time Series Forecasting. 📈 2. Time Series Forecasting forecast.py: Main forecasting script that uses a hybrid model (Least Squares + AutoRegressive). ls.py: Performs least squares fitting using a quadratic trend and dual seasonal (365.2 and 182.6 days) components. ar.py: Applies an AutoRegressive model to residuals for further prediction refinement. ls_ar.py: Combines LS and AR forecasts. The output is a predicted .npy file containing daily SIC forecasts for a full year. 🗺 3. GeoTIFF Export npy2tif.py: Converts the predicted SIC .npy data into daily GeoTIFF images using georeferencing from a sample NetCDF file and a North Polar Stereographic projection.