Choudhary et al. (2025) Comprehensive Evaluation of Precipitation Reanalysis Products and CMIP6 Models Using Statistical and Machine Learning Techniques With Nature‐Inspired Optimization

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: International Journal of Climatology
• Year: 2025
• Date: 2025-10-17
• Authors: Sourav Choudhary, Santosh Murlidhar Pingale, Deepak Khare, Ruchir Patidar, Radha Krishan
• DOI: 10.1002/joc.70159

Research Groups

[Not specified in the abstract]

Short Summary

This study developed a comprehensive strategy combining reanalysis products, trend analysis, and optimized machine learning models to improve precipitation forecasts and evaluate hydroclimatic variability in the Upper Godavari Sub-basin, India, finding MERRA2 reanalysis and the RF-HHO model to be most accurate for prediction.

Objective

• To improve water resource forecasts and evaluate hydroclimatic variability in the Upper Godavari Sub-basin (UGSB), India, by effectively combining trend analysis, machine learning, and climate model review, including the evaluation of CMIP6 and reanalysis datasets and the optimization of ML models for precipitation prediction.

Study Configuration

• Spatial Scale: Upper Godavari Sub-basin area (UGSB), India.
• Temporal Scale: Long-term variability, with identified shifts towards higher frequencies after 2000.

Methodology and Data

• Models used: CMIP6 models, Mann-Kendall test, Pettitt's test, Van Neumann ratio (VNR), Innovative Trend Analysis (ITA), Continuous Wavelet Transform (CWT), Random Forest (RF), Multi-Layer Perceptron (MLP), Long Short-Term Memory (LSTM), XGBoost, Harris Hawks Optimization (HHO) algorithm.
• Data sources: MERRA2 reanalysis datasets, JRA-55 reanalysis datasets, CMIP6 climate model outputs.

Main Results

• MERRA2 reanalysis datasets demonstrated the highest accuracy for precipitation forecasting among reanalysis products, achieving a Probability of Detection (POD) of 0.82 and a Critical Success Index (CSI) of 0.71.
• JRA-55 reanalysis closely followed MERRA2 with a CSI of 0.69.
• CMIP6 models exhibited overestimation tendencies, with a mean False Alarm Ratio (FAR) of 0.34, indicating limitations in capturing precipitation extremes.
• Trend analysis revealed significant spatial variability of precipitation within the UGSB, showing both increasing (1.5–2.3 mm/year) and decreasing rates at various stations.
• Continuous Wavelet Transform (CWT) analysis identified shifts in hydroclimatic variability towards higher frequencies after 2000.
• The Harris Hawks Optimization (HHO) algorithm successfully optimized various machine learning models, with the Random Forest (RF) model optimized by HHO (RF-HHO) performing best, reducing the Root Mean Square Error (RMSE) to 4.92 mm at Ambajogai, 4.81 mm at Bodhegaon, and 5.21 mm at Ranjni.

Contributions

• Developed a comprehensive strategy integrating trend analysis, machine learning, and climate model review for improved water resource forecasts.
• Provided a detailed evaluation of CMIP6 and reanalysis datasets in the Upper Godavari Sub-basin using various categorical and continuous performance metrics.
• Identified the most accurate reanalysis product (MERRA2) for precipitation forecasting in the study area.
• Applied and optimized multiple machine learning models using the Harris Hawks Optimization algorithm for enhanced precipitation prediction accuracy.
• Demonstrated the importance of combining reanalysis products, trend analysis, and optimized ML models for effective water resource management and future precipitation predictions.

Funding

[Not specified in the abstract]

Citation

@article{Choudhary2025Comprehensive,
  author = {Choudhary, Sourav and Pingale, Santosh Murlidhar and Khare, Deepak and Patidar, Ruchir and Krishan, Radha},
  title = {Comprehensive Evaluation of Precipitation Reanalysis Products and <scp>CMIP6</scp> Models Using Statistical and Machine Learning Techniques With Nature‐Inspired Optimization},
  journal = {International Journal of Climatology},
  year = {2025},
  doi = {10.1002/joc.70159},
  url = {https://doi.org/10.1002/joc.70159}
}