Data-driven stochastic programming for sustainable agricultural supply chain design: Toward the circular economy

Agricultural supply chains face significant challenges in achieving food security and sustainability, particularly due to climate change and waste production. Effectively managing these supply chains, especially in the context of uncertainties, is crucial for optimizing resource use and minimizing waste. This research develops a multi-objective optimization for designing a sustainable and responsive closed-loop agricultural supply chain network, focusing on jujube products under uncertain conditions. This model optimizes economic costs, social benefits, and responsiveness through reliability modeling, taking into account decisions related to facility locations, inter-facility flow, and inventory and shortage management. To achieve the principles of a circular economy, we utilize the thermochemical conversion of agricultural waste into biochar. Our main contribution is a data-driven multi-stage stochastic programming approach that integrates a Fuzzy C-means machine learning algorithm to generate scenarios from historical data patterns, addressing the dynamic and interrelated uncertainties in jujube supply and purchase prices. We examine environmental sustainability by establishing strict upper-bound constraints on CO₂ emissions and water consumption across all network operations. Additionally, the augmented ε-constraint method is used to solve the multi-objective model, balancing trade-offs between conflicting objectives. Finally, a real-world case study in Iran validates the model, demonstrating its efficacy in terms of applicability and managing uncertainty. The results of simultaneous optimization of objectives demonstrate that a 3 % increase in costs can lead to a 25 % improvement in social benefits and a 24 % increase in responsibility. This framework guides managers toward a data-driven stochastic approach that addresses uncertainty, advancing the circular agricultural supply chain.