ABSTRACT Reliable assessment of drought risks under climate change requires robust evaluation of climate models, redundancy‐resilient ensemble construction, and accurate quantification of extremes. Existing approaches often fail to account for non‐linear dependencies, inter‐model redundancy, and spatial performance, limiting the reliability of drought projections. This study introduces Integrated Information‐theoretic redundancy‐resilient ensemble and drought assessment (IIREDA), a multi‐criteria framework that integrates information‐theoretic dependence measures, redundancy adjustment, spatial performance consistency, and advanced ensemble techniques. Applying IIREDA to 22 CMIP6 global climate models (GCMs) across 29 grid stations in drought‐prone Balochistan (1950–2014; projected SSP1–2.6, SSP2–4.5, SSP5–8.5 for 2015–2100), we identify high‐performing models (e.g., CNRM‐CM6‐1‐HR, MIROC‐ES2L) and highlight underperforming models (e.g., CanESM5‐CanOE). Using the novel HC‐GDM ranking, an optimal subset of 11 GCMs is selected, ensuring robust ensemble construction. Ensemble experiments demonstrate that constrained least squares ensembles (CLSE) outperform equal‐weighted, machine learning and probabilistic methods by reducing bias and effectively managing outliers. For drought characterisation, K‐component Gaussian mixture modelling (K‐CGMM) captures extremes more accurately than traditional univariate distributions. Drought assessment using the proposed IW‐GSPI reveals stable conditions under SSP1–2.6, intensification under SSP2–4.5, and persistent multi‐year severe droughts under SSP5–8.5, further supported by Markov chain analysis. In summation, IIREDA advances methodological rigour in climate risk assessment, providing redundancy‐resilient, probabilistic drought projections to inform mitigation and adaptation strategies.
Abbas et al. (Sun,) studied this question.