Carbon Footprint Benefits of Digital Twin-Enabled Food Loss Management in Fruit Supply Chains

Food loss in fresh fruit supply chains is strongly influenced by temperature fluctuations and the limitations of conventional static shelf-life management. This study integrates digital twin technology, deep learning, dynamic shelf-life prediction, and life cycle assessment (LCA) to investigate the potential for reducing fruit losses and associated greenhouse gas emissions. An experimental monitoring system comprising 32 fruit boxes and 103 temperature data loggers was used to characterize spatial temperature heterogeneity across four cold-chain stages: postharvest handling, cold storage, loading/unloading, and transportation. A bidirectional Long Short-Term Memory (BiLSTM) model with an attention mechanism was developed to reconstruct in-box temperature histories. These predictions were subsequently integrated with quality degradation kinetics to estimate remaining shelf life. The digital twin approach achieved temperature deviations of 0.12–0.29°C and estimated shelf life within 0.1–0.7 days of reference calculations based on measured temperatures. Across strawberries, fresh oranges, and apples, the evaluated scenarios indicated potential food loss reductions ranging from 5.7% to 15.5%. Estimated gross avoided greenhouse gas emissions reached up to 6.70 × 10⁻² kg CO₂-eq within the assessed scenarios. The findings demonstrate the potential of digital twin-enabled dynamic shelf-life management to improve fruit quality monitoring and support food loss prevention. However, a comprehensive evaluation of net environmental benefits requires further consideration of digital infrastructure burdens, actual waste reduction, and operational validation.