🔍 Developed by Nishant
Evaluating rainfall prediction accuracy of the Weather Research and Forecasting (WRF) model during Super Cyclone Kyarr (2019) using Control (CNTL) and Data Assimilation (DA) experiments.
This open-source project performs both day-wise and threshold-wise skill analysis using GPM and TRMM observations, producing visual analytics and an automated.docxreport.
✅ Day-wise & threshold-wise skill evaluation (POD, ETS, HSS, RMSE, Bias, Correlation)
✅ Comparison between CNTL & DA runs against GPM and TRMM observations
✅ Automatic report generation (.docx format with plots and tables)
✅ Clean, reproducible Python/Colab workflow
✅ Includes sample data for quick testing
| Metric | Description |
|---|---|
| POD | Probability of Detection — model’s success rate in detecting rainfall |
| ETS | Equitable Threat Score — accuracy after removing random hits |
| HSS | Heidke Skill Score — overall model skill against chance |
| RMSE | Root Mean Square Error — deviation from observed rainfall |
| Bias | Mean difference between model and observation |
| r | Pearson correlation — linear agreement strength |
- Model Outputs: CNTL (Control run) & DA (Data Assimilation run)
- Observations: GPM & TRMM satellite rainfall data
- Event: Super Cyclone Kyarr (October 2019)
- Analysis Period: Days 1–5 + Combined case (24–30 Oct 2019)
📊 Plots:
- Day-wise skill metrics (POD, ETS, HSS, RMSE, Bias, r)
- Threshold-wise performance curves
- Comparison bar charts across GPM & TRMM
📑 Report:
Final_Super_Cyclone_Kyarr_Rainfall_Analysis_Report.docx
If you use this work, please cite as:
Nishant (2025). WRF Model Rainfall Skill Evaluation (CNTL vs DA). GitHub Repository.
⭐ Star this repo if you find it useful!
👩🔬 Contributions and collaborations welcome!