Business-as-usual will lead to super and ultra-extreme heatwaves in the Middle East and North Africa

Human-induced global warming is projected to strongly intensify summer heat extremes in the Mediterranean region, including MENA. Warming is seasonally asymmetric, with much stronger increases in summer temperatures, implying more frequent, longer, and more intense heatwaves. These changes pose serious risks to human health, agriculture, water and energy systems, and broader socioeconomic stability, including labor productivity losses and potential links to conflict and migration. Parts of the Middle East may approach or exceed human adaptability thresholds due to combined temperature and humidity effects. Rapid population growth and urbanization will likely exacerbate heat exposure via urban heat island effects. Existing studies have relied largely on coarse global models or incomplete regional modeling that do not fully capture MENA’s climatic particularities. This study uses a dedicated regional downscaling ensemble to project future heatwaves across MENA and assess potential population exposure under business-as-usual conditions.

The paper notes extensive prior evidence of increasing heat extremes in MENA and the Mediterranean, including observed trends since the 1960s–1980s and projections of intensified heatwaves. Previous work highlights strong summertime warming linked to regional circulation features (e.g., Persian trough, Saharan thermal low), land–atmosphere interactions, and soil moisture deficits. Studies have reported health impacts, labor productivity losses, and vulnerability of sectors such as agriculture and livestock. However, much of the literature relies on global models with coarse resolution or regional studies that do not adequately represent MENA’s diverse regimes. Recent CORDEX initiatives, including MENA-CORDEX and EURO-CORDEX, have improved regional downscaling capability, but comprehensive MENA-focused assessments of future heatwave magnitude, duration, and population exposure have remained limited. This study addresses that gap using a multi-model regional ensemble and a standardized heatwave magnitude index.

Climate data and projections: The study analyzes a multi-model MENA-CORDEX ensemble of 10 simulations combining six global Earth system models and four regional climate models, following CORDEX guidelines. Future projections use RCP8.5 (2006–2100), with additional RCP4.5 analysis in the Supplementary Information. Horizontal resolution is 0.44° (~50 km). The focus region excludes areas south of 10°N and grid points extending across eastern and western boundaries. Time span generally covers 1951–2100, with the control period defined as 1981–2010. Two future 30-year periods are used: near future (2021–2050, 21C1) and end-century (2071–2100, 21C2). Evaluation and variables: The analysis focuses on warm-season (May–September) daytime maximum temperature (tasmax). Simulated means and absolute maxima are compared against gridded observations: CRU v4.04 (monthly, 0.5°) and Berkeley Earth (daily, 1°). Reference datasets were remapped to the model grid (bilinear interpolation; land-only) and assessed via spatial bias and other metrics. Model biases are typical for dynamical downscaling; about 70% of MENA land shows ensemble mean bias within ±3 °C. Driving GCMs also tend to underestimate tasmax, implying the study’s heat extremes are conservative. Statistical testing: Student’s t-tests (95% confidence) assess significance of changes in warm-season average and maximum tasmax. Significant differences emerge by 2021–2050 across most of MENA, and by 2071–2100 for almost all areas. Probability density functions illustrate shifts toward hotter distributions, with future coolest summers comparable to historical hottest summers and increased occurrence of seasonal extremes. Heatwave definition and index: Heatwaves are defined (following Russo et al.) as at least three consecutive days with daily maximum temperature above a daily threshold. The study applies a percentile-based approach using the 95th percentile of daily maxima (centered 31-day window) from the 1981–2010 reference period to accommodate leptokurtic temperature distributions in MENA and minimize model bias effects. Heatwave intensity is quantified by the daily Heat Wave Magnitude Index (HWMId), which integrates both duration and temperature anomalies above a normalized threshold. The daily magnitude Md(Td) is computed as (Td − T30y,25p)/(T30y,50p − T30y,25p), where Td is the daily maximum temperature anomaly, and T30y,25p and T30y,50p are the 25th and 50th percentiles for each calendar day over 1981–2010. Categories of events range from normal/moderate/severe to extreme/very extreme and unprecedented super- and ultra-extreme levels based on HWMId thresholds. Population exposure: Future exposure is estimated by combining annual HWMId fields with gridded global population projections under SSPs (focus on SSP5, which yields high urbanization). Original population data at 0.1° (~1 km²) for 2000–2100 are spatially aggregated to a grid comparable to MENA-CORDEX, then resampled via nearest neighbor to match the model grid, conserving totals. Linear interpolation provides annual values for 2010–2100. Exposed populations (total and urban) are computed for each HWMId category annually. Implementation: Analyses were performed in R; HWMId calculations used the extRemes/extrends-based tooling as noted. Visualizations and diagnostics include spatial maps (mean and max HWMId, durations, amplitudes, frequencies), PDFs of tasmax, and time series of area and population exposure.

- Validation and biases: The ensemble reproduces spatial patterns of warm-season tasmax. About 70% of MENA land shows mean bias within ±3 °C versus observations; driving GCMs also underestimate tasmax, implying conservative heatwave intensity estimates. - Warming signal: By 2021–2050, warm-season average and maximum tasmax changes are already statistically significant for most of MENA; by 2071–2100 nearly all areas show significant warming. PDFs shift to hotter distributions; future coolest summers ≈ historical hottest summers. - Heatwave magnitude (HWMId): Under RCP8.5, average conditions transition from normal/moderate to severe, extreme, and very extreme by 2050–2070. By late century, unprecedented super- and ultra-extreme heatwaves become common. - Spatial extent: Before ~2000, ~20% of MENA land experienced normal/moderate heatwaves annually. After 2020, severe/very extreme phases appear. After mid-century, the entire MENA region experiences at least one moderate/severe/very extreme event per year, with super-extreme events emerging. By 2100, super- and ultra-extreme heatwaves affect ~60% of MENA land annually. - Duration: Control period heatwaves last on average 4–6 days; high-impact historical events up to ~2 weeks. For 2021–2050, average durations exceed 10 days in places; longest events >1 month in North Africa and the Arabian Peninsula. By 2071–2100, events last several weeks on average; Mediterranean coasts ~3 weeks; warmer interior regions often 1–2 months or longer, especially in exceptionally warm years. - Amplitude: 2021–2050 heatwave amplitudes increase by ~0.5–2 °C on average; hottest events during this period are ~3 °C warmer than in 1981–2010. By 2071–2100, maximum temperatures during heatwaves exceed 56 °C in some locations (notably near the Arabian Gulf). Given model underestimation of tasmax and the absence of UHI, these are conservative. - Frequency: Historical period shows 6–15 normal/moderate events per 30 years in many areas, with 1–3 severe/extreme/very extreme events; no super/ultra events. By 2021–2050, normal/moderate events increase to >60 per 30 years in coastal regions; inland areas see 15–20 severe/extreme/very extreme events; super/ultra events begin to appear (rare). By 2071–2100, inland regions experience severe/extreme/very extreme at least biennially or more often; the hottest regions approach near-annual occurrence of very high-magnitude events. - Population exposure: MENA population is ~1 billion today, ~60% urban, rising to >90% urban by late century per SSPs. Super- and ultra-extreme heatwaves begin affecting >100 million people by ~2060; by 2100, ~600 million (~50% of MENA population) may be exposed annually to super- and ultra-extreme events, with an additional ~400 million exposed annually to severe/extreme/very extreme events. >90% of exposed populations are urban. - Scenario comparison: Under RCP4.5, end-century HWMId and exposure are comparable to mid-century RCP8.5. Mid-century conditions under RCP4.5 and RCP8.5 are only marginally different.

The ensemble indicates a marked transition toward unprecedented heatwave regimes across MENA under RCP8.5, with concurrent increases in frequency, duration, and amplitude. The physical drivers include enhanced greenhouse forcing, regional circulation changes (e.g., westward expansion and coupling of the Persian trough with the Saharan thermal low), and land–atmosphere feedbacks such as soil moisture deficits. Model-derived peak temperatures surpass 56 °C in several locations by late century, which is life-threatening to humans and exceeds survivability for even heat-tolerant animals in prolonged events. Urban areas, where most of the population will reside, are likely to experience even higher temperatures due to urban heat island (UHI) effects that are not represented in these simulations; accounting for moderate UHI intensities suggests some megacities could reach or exceed 60 °C during super- and ultra-extreme heatwaves. Although RCP4.5 reduces the severity relative to RCP8.5, it still yields substantial public health and societal impacts. The results underscore the urgency of mitigation to limit greenhouse forcing and robust adaptation, particularly urban heat management, to reduce exposure and risk.

This study provides a comprehensive, MENA-focused regional downscaling assessment projecting that, under business-as-usual emissions (RCP8.5), super- and ultra-extreme heatwaves will emerge by mid-century and become commonplace by 2100. Heatwaves will last longer (weeks to months), occur more frequently, and reach higher peak temperatures (>56 °C, conservatively), exposing roughly half the region’s population—predominantly urban dwellers—to dangerous conditions annually. Key contributions include robust multi-model regional projections using a standardized heatwave magnitude metric and integration with SSP-based population exposure estimates. Future work should: (1) expand scenario coverage to include stronger mitigation pathways (e.g., RCP2.6/SSP combinations) as more simulations become available; (2) incorporate urban climate processes and UHI explicitly via higher-resolution urbanized climate modeling; (3) refine impact-relevant metrics (e.g., wet-bulb temperatures, heat stress indices) and sectoral risk assessments; and (4) further constrain uncertainties via larger ensembles and improved observational benchmarks.

- Model biases: Both regional and driving global models tend to underestimate tasmax, leading to conservative estimates of heatwave magnitude and peak temperatures. - Spatial resolution and urban processes: The 0.44° (~50 km) resolution does not resolve city-scale processes; UHI effects are not included, likely underestimating urban heat exposure and peak temperatures. - Observational references: CRU provides monthly data and Berkeley Earth has coarser spatial resolution (1°), limiting direct daily-scale validation; remapping and differing temporal resolutions introduce uncertainty. - Downscaling uncertainty: Inter-model spread in regional simulations remains similar to that of GCMs, indicating some uncertainties are not reduced by dynamical downscaling. - Scenario and data availability: The primary focus is RCP8.5; fewer simulations exist for lower-emission scenarios (e.g., RCP2.6), limiting comparison. Population exposure relies on SSP projections (notably SSP5), which carry their own uncertainties in urbanization and demographics. - Heatwave definition: While percentile-based thresholds mitigate bias and allow regional consistency, lack of a universal heatwave definition and omission of humidity or other variables may underrepresent heat stress in some contexts.