Framework Balances Low-Carbon Concrete Sourcing With Supply Chain Disruption Risks

Researchers have introduced an evidence-aligned supply allocation framework designed to help green construction teams balance embodied carbon reductions against concrete delivery logistics and supply chain disruption risks. In a study published in Frontiers in Materials, investigators Xin Yang, Yong Xiang, and Yao Wei examined how construction managers can defensibly convert product-specific environmental product declarations into operational procurement decisions. Conventional workflows that link machine-learning predictions directly to mathematical optimization often blur a vital evidence boundary between verified declarations and statistical estimates. The authors demonstrated that while verified environmental product declaration values are sufficiently grounded to support formal supplier comparisons, machine-learning estimations demand rigorous decision-level validation before they can be safely utilized to allocate materials. To address this challenge, the study established a structured decision pipeline that clearly distinguishes between observed environmental product declaration attributes, model-estimated screening outputs, scenario-derived transport emissions, and assumption-based operational disruption stresses, assigning a specific and dedicated function to each component.

To construct and evaluate the decision system, the research team assembled an extensive dataset comprising 45,739 harmonized and deduplicated environmental product declaration records collected from 663 concrete manufacturing plants. The investigators subjected their predictive screening model to transportability audits across multiple structural boundaries, testing model performance against plant, company, temporal, and geographic holdouts. These audits specifically evaluated how well the model performed when faced with facilities, geographic markets, and time periods excluded from initial training. Rather than relying solely on abstract statistical benchmarks, the auditing protocol included direct empirical comparisons between allocations and supplier rankings generated by model predictions and those generated from observed environmental product declaration values. This verification step ensured that predictive screening tools could be evaluated on their practical decision fidelity before being integrated into broader procurement workflows.

The transportability of the screening architecture was further validated through a prospective external cohort containing 22 concrete declarations obtained from 14 plants across seven corporate entities completely unseen during model derivation. Evaluated against this prospective cohort, the frozen predictive model reduced record-level mean absolute error by 6.6 percent relative to the frozen training median. Despite these predictive accuracy gains, the system’s supplier-ranking gate preserved the environmental product declaration-grounded pathway for final selection rather than relying on algorithmic estimates. This architectural safeguard verified that model predictions were confined strictly to preliminary screening tasks, thereby preventing unverified statistical approximations from directly determining supplier contracts or displacing certified manufacturer declarations in high-stakes structural procurement decisions.

For formal procurement decisions, the framework implemented strength-constrained mathematical optimization to minimize observed environmental product declaration impacts across lifecycle stages A1 through A3 alongside scenario-derived A4 transportation impacts, subject to plant-share distribution caps. Across nine representative city-strength benchmark cases, imposing a 40 percent ceiling on any single plant’s share of total project supply systematically produced three-plant allocations with a mean Herfindahl–Hirschman Index of 0.36. This multi-supplier distribution prevented project teams from over-concentrating material delivery within a single manufacturing facility. In exchange for this diversified operational resilience, the multi-plant configurations incurred modest carbon trade-offs across the combined A1 to A4 lifecycle stages, exhibiting a mean embodied carbon premium of 14.57 percent and a median premium of 7.76 percent compared to single-source minimum-carbon baselines.

Stress testing demonstrated that multi-plant diversification delivers substantial protections against supply failure under independent and correlated disruption events. Under a baseline stress scenario featuring an independent 10 percent plant disruption rate, capping plant allocation at 40 percent reduced the probability of losing at least half of the total concrete supply from 0.100 down to 0.027. Sensitivity analyses showed that these supply reliability benefits remained measurable across simulated plant disruption rates spanning from 5 percent to 30 percent. Furthermore, the diversification strategy retained its defensive value when subjected to correlated supply failures, maintaining measurable resilience advantages under pairwise plant disruption correlations ranging between 0 and 0.50. These stress tests confirmed that diversified supply allocations insulate projects against both isolated plant shutdowns and broader regional supply chain shocks.

To verify logistical assumptions underlying scenario-based transport emissions, the authors conducted an OpenStreetMap routing validation covering 79 distinct plant-to-project delivery pairs. The spatial routing audit identified a median transport distance error of 6.39 percent relative to assumed haul routes. When these verified routing networks were incorporated into the allocation models, 17 of the 18 evaluated allocation decisions remained completely unchanged. For the single case where routing differences altered the optimal configuration, the resulting supply decision incurred a route-evaluated objective regret of only 0.026 percent. The authors concluded that establishing a transparent carbon-concentration frontier coupled with an empirical decision-fidelity gate allows construction teams to utilize machine-learning models for early supplier screening while firmly anchoring final allocations in verified environmental product declarations, thereby preserving defensible boundaries between planning scenarios and real-world concrete procurement.

Subject of Research: Low-carbon concrete supply chain allocation and disruption resilience

Article Title: EPD-grounded carbon-resilience allocation for low-carbon concrete supply: decision fidelity, route validation, and correlated-disruption stress testing

Article References: Yang, X., Xiang, Y., & Wei, Y. (2026). EPD-grounded carbon-resilience allocation for low-carbon concrete supply: decision fidelity, route validation, and correlated-disruption stress testing. Frontiers in Materials, 13, Article 1950801. https://doi.org/10.3389/fmats.2026.1950801

Image Credits: AI Generated

DOI: 10.3389/fmats.2026.1950801

Keywords: concrete procurement, environmental product declarations, low-carbon concrete, supply chain resilience, embodied carbon, decision fidelity, logistics routing, supplier diversification, EPD-grounded, carbon-resilience, allocation, low-carbon