Forecast Arctic Sea Ice Extent over the next 20 years using time series analysis methods.
- Source: National Snow and Ice Data Center (NSIDC)
- Data: Monthly Sea Ice Extent (1979-2025)
- File:
N_03_extent_v3.0.csv
- Polynomial Regression
- ARIMA (1,1,1) Model
- Prophet Model
- Identified a clear declining trend in sea ice extent over the years.
- Recent years show a sharper and faster decline.
| Model | RMSE | R² | Comments |
|---|---|---|---|
| Polynomial Regression | 0.250 | 0.810 | Great historical fit, unrealistic future |
| ARIMA (1,1,1) | 2.402 | -16.510 | Poor fit, flat future |
| Prophet | 0.251 | 0.808 | Good fit, realistic trend with clipping |
- Prophet Model (with clipping)
- Forecasts a realistic slow decline in sea ice extent.
data/ # Dataset
notebooks/ # EDA + Modeling
outputs/ # Saved Plots
README.md # Project report
requirements.txt # Libraries used
- Python
- Pandas
- Matplotlib / Seaborn
- Scikit-learn
- Statsmodels (ARIMA)
- Prophet (Facebook)
- Numpy
- Importance of stationarity in time series modeling.
- Why polynomial models can overfit.
- How Prophet automatically handles trend and uncertainty.
- Importance of applying real-world constraints (e.g., clipping).
Vinit Singh Pathir
LinkedIn
Feel free to reach out if you want to collaborate on remote sensing, data science, or climate projects!
This project is open-source and free to use under the MIT License.