Oloniyo et al. (2026) Assessment of Heat Stress Hazards in Africa Using CMIP6 and NEX-GDDP Datasets

Identification

• Journal: Earth Systems and Environment
• Year: 2026
• Date: 2026-02-18
• Authors: Olumuyiwa A. Oloniyo, Babatunde J. Abiodun, Romaric C. Odoulami, AKINTUNDE Israel MAKINDE, Abayomi A. Abatan, Gbenga J. Abiodun
• DOI: 10.1007/s41748-026-01028-3

Research Groups

• Climate System and Analysis Group, Department of Environmental and Geographical Sciences, University of Cape Town, Cape Town, South Africa
• Nansen-Tutu Research Centre for Marine Environment, Department of Oceanography, University of Cape Town, Cape Town, South Africa
• African Climate and Development Initiative, University of Cape Town, Cape Town, South Africa
• Rural Development Foundation, California, USA
• Mathematics Department, Southern Methodist University, Texas, USA

Short Summary

This study assesses the ability of CMIP6 and NEX-GDDP datasets to reproduce heat stress characteristics across Africa, focusing on heat stress hazard. It finds that while CMIP6 simulations exhibit substantial biases, the NEX-GDDP dataset significantly improves the representation of heat stress hazards by correcting many of these biases, though some overcorrection occurs.

Objective

• To assess the ability of CMIP6 and NEX-GDDP datasets to characterize heat stress across Africa, with a focus on heat stress hazard.
• To evaluate the extent to which bias correction of CMIP6 simulations, implemented through the NEX-GDDP framework, enhances the representation of heat stress hazards over Africa.
• To propose and utilize annual cumulative heat stress as a novel indicator for quantifying heat stress hazards across Africa.

Study Configuration

• Spatial Scale: Africa, extending from latitudes 35°S to 37°N and longitudes 20°W to 60°E. The reference and NEX-GDDP datasets have a spatial resolution of 0.25° × 0.25° (approximately 25 km). CMIP6 GCMs have varying native resolutions (e.g., 1.0° × 1.0° to 2.8° × 2.8°).
• Temporal Scale: Daily resolution data for the period 1970–2010 for analysis. The reference GMFD dataset covers 1948–2016.

Methodology and Data

• Models used:
  • Coupled Model Intercomparison Project Phase 6 (CMIP6) simulations (17 Global Climate Models).
  • NASA Earth Exchange Global Daily Downscaled CMIP6 Projections (NEX-GDDP), which are bias-corrected and statistically downscaled CMIP6 simulations using the Bias Correction Spatial Disaggregation (BCSD) method.
• Data sources:
  • Global Meteorological Forcing Data (GMFD) dataset (Sheffield et al. 2006) served as the reference.
  • Heat Index (HI) was used to quantify heat stress, calculated using the US National Weather Service (NWS) algorithm, combining air temperature and relative humidity.
  • Annual Cumulative Heat Stress (ACHS) was introduced as a novel metric to quantify heat stress hazard, defined as the cumulative magnitude of HI exceeding a specified threshold (e.g., 27 °C for "Caution" level).
  • Self-Organising Map (SOM) technique was employed for clustering simulations based on HI biases and identifying dominant patterns of interannual variability in heat stress hazard.
  • Standard evaluation metrics included Pearson's correlation coefficient (r), Mean Bias (MBias), Percentage Bias (PBias), and Root Mean Square Error (RMSE).

Main Results

• Both CMIP6 and NEX-GDDP datasets generally agree with GMFD in capturing the spatial and seasonal patterns of heat stress across Africa, identifying tropical regions as hotspots and acknowledging the modulating roles of elevation and relative humidity.
• CMIP6 simulations exhibit substantial biases, underestimating annual HI by more than 4 °C in tropical regions and overestimating it by up to 5 °C over elevated areas (e.g., East African Mountain Range, Atlas Mountains). They also overestimate "safe" heat stress coverage and misrepresent higher heat stress categories.
• NEX-GDDP significantly reduces CMIP6 biases, improving the correlation between simulated and observed variables (r = 1 for HI) and reducing RMSE to less than 0.2 °C and PBias to less than 0.2% for annual HI. Errors in the NEX dataset are generally less than 1 °C for annual HI and temperature, and less than 6% for relative humidity.
• While NEX-GDDP effectively removes most biases in heat stress categories and frequency, it sometimes overcorrects, leading to slight positive biases (approximately 3 days per year) in heat stress frequency south of 20°N.
• CMIP6 underestimates annual heat stress hazard across Africa, particularly in the tropics (up to 5 × 10² °C-days per year for Caution+ hazard level), but NEX-GDDP corrections reduce these biases to less than 1 × 10² °C-days per year.
• SOM analysis revealed dominant patterns of interannual variability in heat stress hazard, but one extreme pattern appeared as an artifact common to both CMIP6 and NEX-GDDP datasets, suggesting an exaggeration of interannual variability peaks.

Contributions

• Provided a comprehensive evaluation of the effectiveness of bias correction (BCSD method) in the NEX-GDDP dataset for improving CMIP6 simulations of heat stress hazards over Africa.
• Introduced and validated "annual cumulative heat stress" (ACHS) as a novel, more holistic metric for quantifying heat stress hazard, which integrates both intensity and duration, offering more actionable insights for public health and policy planning compared to traditional frequency-based metrics.
• Demonstrated the utility of Self-Organizing Maps (SOM) as a diagnostic tool for evaluating climate simulations, enabling the grouping of models based on bias patterns and revealing spatial coherence of errors, thereby complementing conventional statistical metrics.
• Highlighted the limitations of raw global climate simulations in accurately capturing heat stress variability across Africa and underscored the significant improvements offered by downscaled and bias-corrected datasets for regional impact assessments.

Funding

• International Development Research Centre (IDRC)
• Foreign, Commonwealth & Development Office (FCDO)
• Pan-African and Transdisciplinary Lens in the Margins: Tackling the Risks of Extreme Events (PALM-TREEs)
• National Research Foundation of South Africa (Reference Number: RPOAM231218201502)
• Degree Initiative
• Water Research Commission South Africa
• University of Cape Town (Open access funding)

Citation

@article{Oloniyo2026Assessment,
  author = {Oloniyo, Olumuyiwa A. and Abiodun, Babatunde J. and Odoulami, Romaric C. and MAKINDE, AKINTUNDE Israel and Abatan, Abayomi A. and Abiodun, Gbenga J.},
  title = {Assessment of Heat Stress Hazards in Africa Using CMIP6 and NEX-GDDP Datasets},
  journal = {Earth Systems and Environment},
  year = {2026},
  doi = {10.1007/s41748-026-01028-3},
  url = {https://doi.org/10.1007/s41748-026-01028-3}
}