Salini et al. (2026) Performance assessment of general circulation models for meteorological droughts: application of complex network theory

Identification

• Journal: Theoretical and Applied Climatology
• Year: 2026
• Date: 2026-04-10
• Authors: Devika Chandrababu Salini, Bellie Sivakumar
• DOI: 10.1007/s00704-026-06208-9

Research Groups

Department of Civil Engineering, Indian Institute of Technology Bombay, Powai, Mumbai, Maharashtra, India

Short Summary

This study applies complex network theory, specifically node efficiency, to assess the performance of 53 CMIP6 General Circulation Models (GCMs) in simulating meteorological droughts in India using the Standardized Precipitation Index (SPI). The research reveals significant variability in model performance across different timescales, identifying NorESM2-MM, CESM2-FV2, and KACE-1-0-G as consistently top-performing GCMs for drought-related studies.

Objective

• To assess the ability of GCMs to reproduce Standardized Precipitation Index (SPI) values (derived from GCM-simulated rainfall) across India.
• To construct SPI-based complex networks and compute node efficiency for different timescales (1, 3, 6, and 12 months).
• To compare GCM-simulated SPI networks with observed SPI networks.
• To rank the GCMs using a Group Decision Making (GDM) approach (for grids across India as a whole) combined with a comprehensive rating metric (RM) across different timescales.

Study Configuration

• Spatial Scale: Mainland India, covering 288 grids at a 1° × 1° spatial resolution (between 8.5° N to 35.5° N and 69° E to 97° E).
• Temporal Scale: Data spanning the period 1961–2014, analyzed at four different SPI timescales: 1 month, 3 months, 6 months, and 12 months.

Methodology and Data

• Models used: 53 General Circulation Models (GCMs) from the Coupled Model Intercomparison Project Phase 6 (CMIP6).
• Data sources:
  • Observed rainfall data from the India Meteorological Department (IMD), originally at 0.25° × 0.25° spatial resolution, regridded to 1° × 1° using bilinear interpolation.
  • GCM-simulated rainfall outputs from the Earth System Grid Federation (ESGF) portal (https://esgf-node.llnl.gov/search/cmip6/), regridded to 1° × 1°.
  • Standardized Precipitation Index (SPI) values calculated from both observed and GCM-simulated rainfall data.
  • Complex network theory applied using node efficiency as the primary metric for performance assessment.
  • Group Decision Making (GDM) approach for ranking GCMs across grids.
  • Comprehensive Rating Metric (RM) to consolidate rankings across different timescales.

Main Results

• Significant variability in GCM performance was observed across the four SPI timescales (1, 3, 6, and 12 months).
• NorESM2-MM consistently emerged as the top performer with the highest overall Rating Metric (RM) value of 0.925 (final rank 1), demonstrating robust performance across all timescales (ranks 1, 6, 6, and 3 for 1-, 3-, 6-, and 12-month SPI, respectively).
• CESM2-FV2 showed strong overall performance with an RM value of 0.910 (final rank 2), excelling at 1-month (rank 2), 3-month (rank 2), and 12-month (rank 1) timescales, despite a slightly diminished performance at the 6-month timescale (rank 14).
• KACE-1-0-G ranked 3rd overall, performing exceptionally well at the 6-month timescale (rank 1) and well at 1-month (rank 3) and 3-month (rank 3) timescales, but showing a drop at the 12-month timescale (rank 15).
• Models like MPI-ESM1-2-LR (RM value 0.057, rank 53) and FGOALS-g3 (RM value 0.057, rank 52) exhibited the poorest overall performance in simulating SPI.
• A comparison of rankings based on raw rainfall versus SPI values revealed that IPSL-CM5A2-INCA, NorESM2-LM, and CMCC-ESM2 performed consistently well for both, while FGOALS-g3, CanESM5, and GISS-E2-2-G performed poorly for both.
• Notably, models like KACE-1-0-G (ranked 22nd for raw rainfall, 3rd for SPI) and CESM2-FV2 (ranked 36th for raw rainfall, 2nd for SPI) showed significant improvement in ranking when assessed by SPI, indicating their superior ability to capture the statistical distribution and spatial connectivity of rainfall anomalies relevant to drought, despite potential biases in raw rainfall magnitude.

Contributions

• First study to evaluate GCM performance for drought-related studies using a drought index (SPI) within the context of complex network theory.
• Introduces a novel application of complex network theory, specifically node efficiency, for assessing GCM performance in replicating drought characteristics in India.
• Provides a standardized and functionally meaningful assessment framework for GCMs, moving beyond raw climate variables to impact-relevant indices.
• Establishes a robust methodology for future climate model evaluations using drought indices, complementing traditional methods.
• Identifies specific CMIP6 GCMs (NorESM2-MM, CESM2-FV2, CMCC-ESM2) as consistently reliable for drought-related studies in India across various timescales, offering guidance for model selection in regional and temporal contexts.

Funding

• Scheme for Transformational and Advanced Research in Sciences (STARS) (STARS-2/2023–0704) managed by Indian Institute of Science (IISc), Bangalore, and funded by Ministry of Education (MoE), India.

Citation

@article{Salini2026Performance,
  author = {Salini, Devika Chandrababu and Sivakumar, Bellie},
  title = {Performance assessment of general circulation models for meteorological droughts: application of complex network theory},
  journal = {Theoretical and Applied Climatology},
  year = {2026},
  doi = {10.1007/s00704-026-06208-9},
  url = {https://doi.org/10.1007/s00704-026-06208-9}
}