Linking predictive modelling of site-level resource use with life cycle assessment of construction-phase emissions for managing carbon limits

Accurate, time-resolved data on resource use and waste generation are essential for assessing the environmental impacts of construction activities. Construction phases (Modules A4–A5 in life cycle assessment) contribute substantially to upfront emissions but are often simplified in current assessments due to a lack of granular predictive tools. This study develops a modular modelling approach using random forest regression to forecast time-distributed flows of electricity, heating, fuel, material transport, and construction waste. The model is trained on empirical data from Danish construction projects, incorporating temporal, project-specific, and contextual features to simulate consumption patterns over project timelines. Results show that the model captures variations across tasks and material categories, offering more detailed inputs for dynamic life cycle assessment and scenario analysis. By linking construction logistics to environmental performance, this approach facilitates more accurate emissions budgeting and informed decision-making at the early stages of sustainable construction planning. • Modular ML framework predicts time-based resource use in construction • Separate models forecast waste, energy, and transport emissions • Trained on data from 63 Danish construction projects • Enables life cycle assessment for A4-A5 stages with temporal resolution • Supports early-stage planning, emissions budgeting, and decision-making