Tayyaba et al. (2026) Review of global climate models used for drought assessment and forecasting

Identification

• Journal: Discover Water
• Year: 2026
• Date: 2026-01-30
• Authors: Syed Tayyaba, Harish Puppala, Manoj Kumar Arora
• DOI: 10.1007/s43832-026-00352-z

Research Groups

• Department of Civil Engineering, SRM University-AP, Amaravati, Andhra Pradesh, India
• Centre for Geospatial Technology, SRM University-AP, Amaravati, Andhra Pradesh, India

Short Summary

This paper critically reviews the current state of drought assessment and forecasting using Global Climate Models (GCMs), highlighting advancements in post-processing techniques and multi-model ensembles, while identifying existing gaps and proposing future research directions for more precise predictions.

Objective

• To critically review the current state of drought assessment and forecasting, with a focus on the use, precision, and constraints of Global Climate Models (GCMs), and to identify future research directions for enhancing drought modeling performance and dependability, particularly for regional-scale assessments.

Study Configuration

• Spatial Scale: Global to regional scales, covering continents (Asia, Africa, Europe, America, Australia) and specific countries/basins (e.g., India, Pakistan, Morocco, Bangladesh, Hanjiang River Basin).
• Temporal Scale: Historical periods (early 20th century to present) and future projections (up to 2100), including short-term, near-future (e.g., 2020–2060), and far-future (e.g., 2061–2100) scenarios.

Methodology and Data

• Models used:
  • Global Climate Models (GCMs): CMIP5, CMIP6 (SSP126, SSP245, SSP370, SSP585 / RCP2.6, RCP4.5, RCP7.0, RCP8.5), NASA NEX-GDDP.
  • Regional Climate Models (RCMs): CORDEX domains.
  • Hydrological and Land Surface Models.
  • Statistical Models: Time-Series Models (ARIMA, SARIMA), Regression Models (Multiple Linear Regression), Markov Chain Models, Copula-based joint probability models.
  • Machine Learning (ML) Models: Bagging, Random Subspace, Random Tree, Random Forest, Support Vector Regression (SVR), Artificial Neural Networks (ANN, ANN-MLP), Adaptive Neuro-Fuzzy Inference Systems (ANFIS), Wavelet-enhanced models, Supervised Neural Networks, Long Short-Term Memory (LSTM), Decision Tree Regressor, Extra Tree Regressor (ETR), eXtreme Gradient Boosting (XGBoost), Convolutional Neural Networks (CNNs), Time-Lagged Feedforward Networks (TLFN), Recurrent Neural Networks (RNNs), Generative Adversarial Networks (GANs), Deep Convolutional Neural Networks (DCNNs), Transformers, Light Gradient Boosting Machine, Gradient Boosting Regressor, K-Nearest Neighbors (KNN), Gradient Boosting Decision Tree (GBDT).
• Data sources:
  • GCM outputs (CMIP5, CMIP6).
  • Observational data (in-situ, station-level, historical hydroclimatic records).
  • Remote sensing data.
  • Reanalysis products.
  • Hydroclimatic variables (precipitation, temperature, soil moisture, runoff, wind speed, humidity, latent heat, sea surface temperature, sea level pressure, radiation fluxes, evapotranspiration, streamflow, groundwater levels, vegetation cover).
  • Drought indices (e.g., Standardized Precipitation Index (SPI), Palmer Drought Severity Index (PDSI), Standardized Precipitation Evapotranspiration Index (SPEI), Standardized Streamflow Index (SSI), Composite Drought Index (CDI), China-Z Index (CZI), Effective Drought Index (EDI), Multivariate Standardized Drought Index (MSDI), Reconnaissance Drought Index (RDI), Adaptive Multi-model Standardized Drought Index (AMSDI)).
  • Socioeconomic data (population, Gross Domestic Product (GDP), land cover, water resources).

Main Results

• Droughts are complex natural hazards whose frequency, severity, duration, and spatial extent are increasing due to climate change.
• Global Climate Models (GCMs) are crucial for drought assessment and forecasting but suffer from coarse spatial resolution and systematic biases.
• Post-processing techniques, including bias correction (e.g., Quantile Mapping, Detrended Quantile Mapping) and downscaling (Statistical Downscaling, Dynamical Downscaling), are essential to refine GCM outputs for regional applications.
• Machine learning (ML) algorithms are increasingly vital for improving bias correction, downscaling, and Multi-Model Ensemble (MME) weighting, demonstrating superior performance in capturing complex, nonlinear relationships in climate data.
• Multi-Model Ensembles (MMEs) significantly reduce uncertainties and enhance the reliability of drought projections compared to individual GCMs, with differential-weighted (especially ML-based) MMEs often outperforming equal-weighted approaches.
• Despite advancements, significant gaps remain in GCMs regarding biases, accurate regional-scale behavior, and the physical representation of key climate processes such as land-atmosphere coupling and soil moisture memory.
• Future research should focus on integrated frameworks combining improved climate process models, observational data, ensemble techniques, and advanced ML/Deep Learning methods for more precise and reliable drought assessment and prediction.

Contributions

• Provides a comprehensive critical review of the current state of drought assessment and forecasting using GCMs, encompassing prediction methods, bias correction, downscaling, and Multi-Model Ensembles (MMEs).
• Systematically highlights the growing importance and application of machine learning tools in enhancing the accuracy and reliability of GCM-based drought measurements.
• Identifies critical existing gaps in GCM capabilities, particularly concerning systematic biases, regional-scale model behavior, and the physical representation of major climate processes.
• Proposes clear future research directions, advocating for combined frameworks that integrate advanced climate process models, observational data, ensemble techniques, and machine learning for more precise and actionable drought predictions.
• Offers a systematic comparative analysis of regional coverage, GCM selection, climate scenarios, drought indices, and bias-correction methods across a wide range of existing literature.

Funding

No financial support was received for this research work.

Citation

@article{Tayyaba2026Review,
  author = {Tayyaba, Syed and Puppala, Harish and Arora, Manoj Kumar},
  title = {Review of global climate models used for drought assessment and forecasting},
  journal = {Discover Water},
  year = {2026},
  doi = {10.1007/s43832-026-00352-z},
  url = {https://doi.org/10.1007/s43832-026-00352-z}
}