Robustness and uncertainty in drought trend projections across dryland and non-dryland regions under greenhouse warming

Droughts strongly affect ecosystems, hydrology, agrifood production, and their interactions, and understanding how drought trends may evolve under global warming is critical for climate adaptation. However, the robustness and sources of uncertainty in projected drought trends remain insufficiently quantified. Our results show that future drying signals differ systematically between dryland and non-dryland regions depending on the drought metric used. The Standardized Precipitation Evapotranspiration Index (SPEI) indicates stronger drying in drylands, where the area with significant decreases reaches 48-59% under SSP245 and 69-77% under SSP585, compared with 20-33% and 40-53% in non-drylands. In contrast, the Aridity Index (AI) indicates broader drying in non-dryland regions, where significantly affected areas reach 26-41% under SSP245 and 55-68% under SSP585, compared with 17-24% and 40-53% in drylands. Uncertainty is dominated by inter-model differences: GCMs explain 37.5% of the total variance, exceeding the contributions of emission scenario (18.4%) and drought-index choice (12.8%), indicating that model structural differences are the primary source of spread in projected drought trends. These findings show that future drought evolution depends strongly on both region type and drought metric, and they highlight the need to better constrain model-related uncertainty for more reliable drought-risk assessment under global warming.