Risk-Averse Green Hub Location Under Multi-Source Uncertainty: A CVaR-Based Model With Scenario Reduction

This study presents a novel framework for designing resilient and environmentally sustainable logistics networks under deep uncertainty. We formulate a Risk-Averse Stochastic Multiple Allocation Green Hub Location Problem (RASMAGHLP) that simultaneously addresses three critical sources of uncertainty—demand, transportation costs, and carbon taxes—within a Conditional Value-at-Risk (CVaR) optimization framework. To overcome the computational intractability of large-scale stochastic models, we introduce a systematic scenario reduction technique based on K-means clustering, enabling tractable yet statistically representative analysis. Unlike prior models that are either risk-neutral or limited to single uncertainty sources, our approach integrates risk aversion, environmental objectives, and capacity constraints into a unified two-stage stochastic programming structure. Through extensive computational experiments on 2,340 instances, we demonstrate that increasing risk aversion leads to strategic shifts in hub network topology, including decentralization and redundancy, which enhance both economic robustness and carbon efficiency. The proposed model offers a tunable decision-support tool for managers and policymakers seeking to balance cost, resilience, and sustainability in logistics network design.