Shakeel et al. (2025) Advancing climate modeling: a multi-framework methodology for evaluating GCMs and predicting drought trends

Identification

• Journal: Acta Geophysica
• Year: 2025
• Date: 2025-10-02
• Authors: Muhammad Shakeel, Hussnain Abbas, Zulfiqar Ali, Maysaa Elmahi Abd Elwahab, Ehtesham Asharf, Amna Nazeer
• DOI: 10.1007/s11600-025-01701-7

Research Groups

• College of Statistical Sciences, University of the Punjab, Lahore, Pakistan
• Department of Mathematical Sciences, College of Science, Princess Nourah bint Abdulrahman University, Riyadh, Saudi Arabia
• Department of Mathematics & Statistics, University of Sialkot, Sialkot, Pakistan
• Department of Mathematics, COMSATS University Islamabad, Islamabad, Pakistan

Short Summary

This study develops a multi-framework methodology to evaluate 22 Global Climate Models (GCMs) for historical precipitation simulation in Sindh, Pakistan, and proposes a novel Hybrid Framework-Gaussian Climate Drought Index (HF-GCDI) for future drought prediction under Shared Socioeconomic Pathways (SSPs), finding increased severe drought likelihood under higher emission scenarios.

Objective

• To enhance future drought assessment in the Sindh region of Pakistan by identifying high-performing GCMs, reducing projection uncertainties through optimized ensemble techniques, and introducing a novel drought index.

Study Configuration

• Spatial Scale: Sindh, Pakistan (23° to 30° north latitude and 66° to 71° east longitude), across 21 grid points with a spatial resolution of 0.5° × 0.5° (approximately 50 km × 50 km).
• Temporal Scale: Historical period: 1950–2014; Future period: 2015–2100. Drought assessment at multiple timescales (1, 3, 6, 9, 12, 24, and 48 months).

Methodology and Data

• Models used:
  • 22 Global Climate Models (GCMs) from CMIP6.
  • GCM ranking frameworks: Support Vector Regression (SVR), Bayesian Model Averaging (BMA), Normalized Joint Mutual Information Maximization (NJMIM).
  • Ensemble methods: Simple Average Ensemble (SAE), Constrained Least Squares Ensemble (CLSE), Lp-norm Ensemble (LpNE), Trimmed Eigenvector Ensemble (TEVE), Trimmed Bias-Corrected Eigenvector Ensemble (TBCEVE), Brouta Regression Ensemble (BRE).
  • Drought Index: Hybrid Framework-Gaussian Climate Drought Index (HF-GCDI) based on a K-Component Gaussian Mixture Model (K-CGMM).
  • Evaluation metrics: Comprehensive Rating Metric (CRM), Majority Judgment Majority Rule (MJMR), Kling-Gupta Efficiency with knowable moments (KGEkm), Bayesian Information Criterion (BIC).
  • Trend analysis: Markov chain steady-state probabilities.
• Data sources:
  • CMIP6 monthly precipitation data (TS v4.04).
  • Observed precipitation data for the historical period.
  • Future climate projections under Shared Socioeconomic Pathways (SSPs): SSP1-2.6, SSP2-4.5, and SSP5-8.5.
  • Data obtained from the open-source climate data store (https://climate.copernicus.eu).

Main Results

• The GCMs IPSL-CM6A-LR, BCC-CSM2-MR, INM-CM5-0, CNRM-CM6-1-HR, and MPI-ESM1-2-LR were identified as the most suitable high-performing models for the Sindh region based on the Majority Judgment Majority Rule (MJMR) score.
• The Lp-norm Ensemble (LpNE) emerged as the most effective Multi-Model Ensemble (MME) framework, achieving the highest correlation coefficient, minimal bias, and a good variability ratio (KGEkm = 0.547) in capturing historical precipitation trends.
• The K-Component Gaussian Mixture Model (K-CGMM) precisely captured multi-modal precipitation trends across various timescales (1, 3, 6, 9, 12, 24, and 48 months) for all SSPs, outperforming traditional univariate probability models (demonstrating lower BIC values).
• Long-term drought trend analysis using Markov chain steady-state probabilities indicates a significant increase in the likelihood of severe and prolonged droughts under the high-emission scenario (SSP5-8.5) at longer timescales (24 and 48 months). Normal Drought (ND) conditions are expected to prevail with the highest probabilities (0.677 to 0.686) across all scenarios.

Contributions

• Proposes a robust multi-framework methodology for GCM evaluation and selection that addresses multicollinearity, nonlinear discrepancies, and model dependency.
• Introduces a novel regional aggregation framework to enhance ensemble accuracy, particularly for extreme precipitation events.
• Develops a novel drought index, the Hybrid Framework-Gaussian Climate Drought Index (HF-GCDI), based on a K-Component Gaussian Mixture Model, offering higher flexibility and accuracy in fitting multi-modal precipitation distributions and capturing heavy tails compared to traditional indices.
• Utilizes advanced statistical and machine learning approaches (SVR, BMA, NJMIM) for GCM ranking and selection, and evaluates six diverse MMEs (geometric, regression, ML) using the KGEkm metric for robust uncertainty reduction.
• Provides a comprehensive assessment of future drought trends in Sindh, Pakistan, under three SSPs using steady-state probabilities of Markov chains, emphasizing the importance of careful GCM selection and robust ensemble modeling for regional climate risk assessments.

Funding

• Princess Nourah bint Abdulrahman University Researchers Supporting Project number (PNURSP2025R913), Princess Nourah bint Abdulrahman University, Riyadh, Saudi Arabia.

Citation

@article{Shakeel2025Advancing,
  author = {Shakeel, Muhammad and Abbas, Hussnain and Ali, Zulfiqar and Elwahab, Maysaa Elmahi Abd and Asharf, Ehtesham and Nazeer, Amna},
  title = {Advancing climate modeling: a multi-framework methodology for evaluating GCMs and predicting drought trends},
  journal = {Acta Geophysica},
  year = {2025},
  doi = {10.1007/s11600-025-01701-7},
  url = {https://doi.org/10.1007/s11600-025-01701-7}
}